Neuronal Control of Leech Behavior William B

Progress in Neurobiology 76 (2005) 279–327 www.elsevier.com/locate/pneurobio

Neuronal control of leech behavior William B. Kristan Jr.a, Ronald L. Calabrese b, W. Otto Friesen c,* a Section of Neurobiology, Division of Biological Sciences, 9500 Gilman Dr., University of California, San Diego, La Jolla, CA 92093-0357, USA b Department of Biology, Emory University, 1510 Clifton Road, Atlanta, GA 30322, USA c Department of Biology, Gilmer Hall, University of Virginia, P.O. Box 400328, Charlottesville, VA 22904-4328, USA Received 7 April 2005; received in revised form 23 August 2005; accepted 26 September 2005

Abstract The medicinal leech has served as an important experimental preparation for neuroscience research since the late 19th century. Initial anatomical and developmental studies dating back more than 100 years ago were followed by behavioral and electrophysiological investigations in the first half of the 20th century. More recently, intense studies of the neuronal mechanisms underlying leech movements have resulted in detailed descriptions of six behaviors described in this review; namely, heartbeat, local bending, shortening, swimming, crawling, and feeding. Neuroethological studies in leeches are particularly tractable because the CNS is distributed and metameric, with only 400 identifiable, mostly paired neurons in segmental ganglia. An interesting, yet limited, set of discrete movements allows students of leech behavior not only to describe the underlying neuronal circuits, but also interactions among circuits and behaviors. This review provides descriptions of six behaviors including their origins within neuronal circuits, their modification by feedback loops and neuromodulators, and interactions between circuits underlying with these behaviors. # 2005 Elsevier Ltd. All rights reserved.

Keywords: Elemental oscillators; Interneurons; Serotonin

Contents

1. Introduction ...... 280 1.1. Anatomy and electrophysiology ...... 282 1.2. The hydroskeleton and behaviors ...... 284 2. Circulation and heartbeat ...... 285 2.1. The heartbeat neural control system of the leech ...... 286 2.2. The elemental oscillators...... 286 2.3. Mechanisms of oscillation in an elemental half-center oscillator ...... 286 2.4. Coordination in the beat timing network ...... 287 2.5. Heartbeat motor pattern switching by switch interneurons ...... 288 2.6. Gaps in our current knowledge ...... 290 3. Overt behaviors ...... 290 3.1. Introduction ...... 290 3.2. Local bending ...... 290 3.2.1. Motor neurons ...... 290 3.2.2. Mechanosensory neurons that produce local bending ...... 293 3.2.3. Local bend interneurons ...... 293 3.2.4. The local bend response as a directed behavior ...... 295 3.2.5. Gaps in our current knowledge ...... 295

* Corresponding author. Tel.: +1 434 982 5493; fax: +1 434 982 5626. E-mail address: [email protected] (W.O. Friesen).

3.3. Shortening ...... 296 3.3.1. Whole-body shortening ...... 296 3.3.2. Local shortening...... 296 3.3.3. Gaps in our knowledge ...... 297 3.4. Swimming ...... 298 3.4.1. History: reﬂex chain versus central pattern generator ...... 298 3.4.2. Swimming behavior and motor control ...... 298 3.4.3. Central oscillator circuits...... 300 3.4.4. Control of swimming activity ...... 300 3.4.5. Neuromodulatory control: serotonin and other biogenic amines ...... 302 3.4.6. Role of sensory feedback ...... 304 3.4.7. Functional aspects of the central oscillator ...... 306 3.4.8. Gaps in our current knowledge ...... 308 3.5. Vermiform crawling ...... 308 3.5.1. Behavior ...... 308 3.5.2. Kinematics ...... 309 3.5.3. Motor neuron activity ...... 309 3.5.4. Sensory input...... 309 3.5.5. Models ...... 311 3.5.6. Initiation of crawling ...... 312 3.5.7. Gaps in our current knowledge ...... 313 3.6. Feeding ...... 313 3.6.1. Behavior ...... 313 3.6.2. Chemosensation ...... 314 3.6.3. Motor patterns ...... 314 3.6.4. Regulation and plasticity ...... 314 3.6.5. Gaps in our knowledge ...... 314 3.7. Interactions among behaviors...... 315 3.8. Methodologies and approaches for further research ...... 316 3.8.1. Functional indicator dyes ...... 316 3.8.2. Modeling...... 318 3.8.3. Plasticity ...... 318 3.8.4. Development ...... 319 4. Conclusion ...... 319 Acknowledgements ...... 320 References ...... 320

1. Introduction Teshiba et al., 2001; Prinz et al., 2003; Beenhakker et al., 2004), insects (Sasaki and Burrows, 2003; Wang et al., 2003; The major goal of neurobiology is to understand how Riley et al., 2003; Wilson et al., 2004; Daly et al., 2004), the brain works: how it senses the external world and internal amphibians (Roberts et al., 1999; Combes et al., 2004), fish states, how it processes this sensory input, how it evaluates (Higashijima et al., 2003; Grillner, 2003), and rodents different inputs to select an appropriate motor act, and how it (Sekirnjak et al., 2003; Kiehn and Butt, 2003; Yvert et al., generates that behavior. Oneapproachtostudyingthese 2004). questions is to study the function of a particular neural Rhythmic movements such as chewing, respiratory structure (e.g. the superior colliculus or the habenular movements, locomotory movements, and, in some animals, nucleus) in a complex brain and ask how it works. Another heartbeat are of particular interest because of their approach is to select a behavior and ask how the properties combination of complex dynamics and relative stereotypy of neurons and their interconnections produce that behavior. (Marder and Calabrese, 1996; Stein et al., 1997; Orlovsky The latter approach is the more direct, but is possible only et al., 1999). Oscillatory networks of central neurons in animals with relatively simple nervous systems, or in are important components of most such motor pattern- selected parts of complex nervous systems with neurons that generating networks. The anatomical wiring and synaptic are identifiable from animal to animal. For example, connectivity within a network is the backbone on which behavioral circuits have been described in a number of such intrinsic and synaptic properties of component neurons animals: mollusks (Arshavsky et al., 1998; Satterlie et al., operate to produce network dynamics. The states of these 2000; Brembs et al., 2002; Dembrow et al., 2003; Sakurai intrinsic and synaptic properties are themselves dynamic, and Katz, 2003; Jing and Gillette, 2003; Staras et al., 1999; being subject to modulation through a multiplicity of Bristol et al., 2004), crustaceans (Selverston et al., 2000; sensory inputs provided by neurons extrinsic and intrinsic W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 281 to the network (Katz, 1995; Harris-Warrick et al., 1997; Nusbaum et al., 1997). A very useful animal for establishing the neuronal bases of behaviors has been the leech, particularly the European medicinal leech, Hirudo medicinalis. In this animal, more behaviors have been studied in neuronal terms than in any other. This review provides an overview of all the behaviors that have been studied, and an update of less comprehensive but more detailed reviews that have appeared elsewhere (Brodfuehrer et al., 1995b; Calabrese et al., 1995; Kristan et al., 1995). There is always the possibility that the neuronal mechanisms found in a particular animal will be unique to that animal, due to its specific evolutionary history and individual peculiarities of anatomy and biomechanics. We, however, believe the opposite: that there are general strategies for producing behaviors that will be found in all animals with a central nervous system (CNS). There are many technical reasons why the medicinal leech is an auspicious animal for identifying behaviorally relevant neuronal systems. Some of the reasons are generally true of simple animals, and others are true of the leech in particular. It is worth enumerating the list to indicate why the medicinal leech has been so useful in studying the neuronal bases of behaviors:

1. The leech nervous system is relatively simple (Fig. 1A) and readily accessible even while the animal is behaving in a variety of semi-intact preparations (Fig. 1B), making it possible to relate motor patterns directly to behaviors. 2. Quite accurate representations of all the behaviors, or at least their rudiments, can be elicited in isolated nerve cords Fig. 1. Anatomy of the medicinal leech and its nervous system. (A) Sche- (Fig. 1B), where intracellular and optical recording is more matic diagram of the leech, showing the major features of its nervous system. favorable. There are 21 segmentally homologous midbody ganglia, numbered M1–M21. The anterior brain (inset) consists of a supraesophageal ganglion (sup.) that is 3. The neurons are easily seen and readily identiﬁed, based on part of the prostomium, plus a subesophageal ganglion (sub.), which forms the location of their somata (Fig. 1C), morphology (Fig. 1D), from the coalescence of the four most anterior embryonic ganglia that are and physiological properties. visible in the adult brain as neuromeres 1–4. (B) Types of preparations used to 4. Intracellular neuronal activity can be recorded readily study the neuronal bases of leech behaviors. The kinematics of all behaviors because the somata are relatively large (10–80 mm) and have been characterized in intact animals (top panel) by measuring the distances between markers placed on the external body wall in successive every soma is visible in segmental ganglia. These properties frames of a movie or video. A variety of semi-intact preparations (example in also make optical recording feasible. middle panel) have provided intracellular and extracellular recordings during 5. Long, easily accessible peripheral nerves allow for stimula- each of the behaviors, thereby revealing the underlying motor neuronal ﬁring tion of selected neurons and monitoring of neuronal activity patterns. The isolated nervous system (bottom panel), in its entirety or in with extracellular electrodes. pieces, produces motor patterns that are distinguishable as the neuronal substrates of each of the behaviors. Such preparations are the most useful for 6. Most relevant electrical parameters can be measured. electrophysiological characterizations of neuronal properties and synaptic Intracellular recordings from somata reveal relatively large, connections. (C) Schematic view of the ventral surface of a midbody gang- individual synaptic potentials, which are not greatly lion, indicating the arbitrary numbering scheme used to identify ganglionic attenuated from their origins in the neuropil, and attenuated neurons. Most midbody ganglia have the same neurons and locations of the action potentials. soma. The dotted lines indicate the packet margins formed by the six giant glial cells, each of which encapsulates characteristic clusters of neuronal 7. The nervous system is iterated, with homologous neurons somata. The scale bar (200 mm) indicates the size of a midbody ganglion in a found in most, if not all, 21 segmental ganglia (Fig. 1). So mature leech weighing 2–5 g. The functions of about a third of the neurons despite having more than 10,000 neurons, the functional unit are known. (D) Structure of a single neuron. Dye, injected into the soma, (i.e. the number of different kinds of neurons) of the leech diffused into the processes in the center of the ganglion, where all synaptic CNS is relatively small. For instance, there are only 400 contacts are made. This region is termed the neuropil. This neuron, cell 208, has extensive, bilaterally symmetric branches and sends a single axon neurons per segmental ganglion (Macagno, 1980), and most posteriorly down one of the two lateral connectives (the small, medial of these are paired. Thus, in essence, the segmental nerve connective is called ‘‘Faivre’s nerve’’). Based upon the location of its soma cord (roughly corresponding to the spinal cord in verte- and its branching pattern, each neuron has a distinctive morphology. The brates) consists of 42 copies (one on each side of 21 scale bar in D represents 100 mm. segments) of a basic unit of 200 neurons. 282 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

8. Most neurons in the CNS are unique rather than members of ganglia are called neuromeres. Together, the supraesophageal functionally identical clusters, hence activating or ablating and subesophageal ganglia form the anterior brain (sometimes single neurons (irreversibly by killing or reversibly by called the head brain). Similarly, the last seven ganglia in the hyperpolarization) often has behaviorally detectable con- chain fuse embryonically to form the posterior brain (also sequences. called the tail brain). The neuromeres in the anterior brain are denoted as R1–4 (rostral neuromeres 1–4), those in the Because of these favorable features, it is possible, in posterior brain are C1–7 (caudal neuromeres 1–7), and the principle, to identify every neuron that contributes substantially individual, mid-body ganglia are labeled M1–M21. to any leech behavior. In practice, this task is far from trivial, so Neuronal somata within the CNS are roughly spherical, and that no single behavior has yet been completely characterized. are located on the surface of segmental ganglia and terminal In no other system, however, have so many behaviors been brains. In midbody ganglia, somata are in 10 clusters – four on investigated and described at the neuronal circuit level. This the dorsal surface and six on the ventral surface – delineated by review is intended to provide a brief overview of the state of giant glial cells that effectively engulf the somata of dozens of what is known, and what remains to be known about the most neurons. In fact, modern-day characterization of the function of completely described behaviors. To understand these descrip- leech neurons began with a series of elegant studies by Stephen tions, background information concerning basic leech anatomy Kuffler and his colleagues using these giant glial cells to study and electrophysiology is essential. After this introduction, the the electrical properties and potential functions of glia (Kuffler neuronal circuits underlying six different behaviors are and Potter, 1964; Nicholls and Kuffler, 1964). They concluded discussed individually, followed by a discussion of how these that the membranes of these glial cells are nearly perfect K+ circuits interact. A final section provides a vision of how the electrodes, and that their contributions to the electrical function leech may prove useful for future research. of the nervous system is to sequester K+ ions released by active neurons in order to buffer the effects of local release of the K+. 1.1. Anatomy and electrophysiology These giant glial cells, therefore, gather the neuronal somata into packets, with the lateral edges serving as packet margins Leeches are annelids, all of which are segmented worms that provide useful markers for identifying neurons. The ventral (Fig. 1A). Unlike most other annelids, leeches have a fixed surface of a typical midbody ganglion is shown in Fig. 1C. Most number of segments – 32 – plus an anterior non-segmental or all of the leech central neurons are identifiable from animal- region called the prostomium. The segments form as a repeated to-animal and segment-to-segment on the bases of the size and iteration of divisions of the same stem cells, whereas the location of their somata within a cluster, as well as their prostomium is derived from a different set of stem cells (Stent characteristic electrophysiological properties and morphologi- et al., 1992). The prostomium and the most anterior four cal features (Muller et al., 1981). segments form the head and the most posterior seven segments All neurons in the leech CNS are monopolar: a single form the tail. There are a variety of specializations in the head process extends from each soma. Typically, this process gives and tail, the most striking of which are the suckers. The mouth rise to one or more axons that leave the ganglion, via nerves to is in the middle of the front sucker, whereas the anus is located the periphery in the case of sensory and motor neurons (MNs), in the body wall anterior and dorsal to the posterior sucker. At and via the connectives in the case of interneurons (INs) and rest, the posterior sucker is usually attached to the substrate. some sensory and secretory neurons. Secondary branches The anterior end is used to explore the environment, so that the emerge from the main process; these side branches may anterior sucker is typically attached only when the leech is subdivide to generate many orders of branching (Fig. 1D). crawling or feeding. The body is a tube formed by epidermis Synaptic connections are made primarily on these fine and muscles, which encases the internal organs: the gut and branches. intestines, the nephridia and urinary sacs, the reproductive All the ganglionic MNs that send axons via segmental nerves organs, and the blood vessels (Fig. 2A). The circulatory system to muscles in the body wall (Stuart, 1970; Ort et al., 1974; of a leech is closed, with four major longitudinal blood vessels Norris and Calabrese, 1987) and to the lateral heart tubes that run the length of the leech and a mesh of circumferentially (Thompson and Stent, 1976a) are located within the CNS. directed vessels connecting them. The dorsal and ventral The muscles used by the leech to make overt movements are of longitudinal vessels are passive (they function as veins) and the four types: longitudinal, circular, oblique, and dorsoventral lateral tubes are contractile (they function as hearts). (Fig. 2B). Contractions of each of these muscles produce The leech CNS consists of a ventral nerve cord with a brain characteristic types of movements: longitudinal muscle at each end (Fig. 1A). Each segment contains a single ganglion, contractions produce shortening, circular muscle contractions which communicates with the adjacent anterior and posterior produce a reduction in cross-section and elongation, oblique ganglia via three connectives (a pair of large lateral connectives muscle contractions cause stiffening at an intermediate body and a smaller medial connective, known as Faivre’s Nerve). The length, and dorsoventral muscle contractions cause a flattening four anterior ganglia fuse during embryogenesis to form a of the body and contribute to elongation. Each MN connects to subesophageal ganglion, and a supraesophageal ganglion forms a single muscle type, and only to muscle fibers on either the left within the prostomium. The borders of these individual ganglia or right side of its own segment, and then only to a regional are visible in the adult. The neuronal compartments of the four subset of muscle fibers. For instance, MNs – both excitatory and W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 283

Fig. 2. Schematic views of a leech midbody segment. (A) Cut-away view of the middle of a leech, showing the location of the central nervous system, major peripheral nerves, blood vessels, viscera, and musculature. Each midbody ganglion connects to adjacent ganglia via connectives and to the periphery through characteristic nerves. The four major longitudinal blood vessels (heart tubes) connect to one another via circumferential blood vessels. The lateral longitudinal blood vessels are contractile and serve as hearts. The ventral nerve cord is suspended in the ventral blood vessel, which, like the dorsal vessel, is passive. A rich capillary vascularization of the skin provides for gas exchange in these aquatic animals, so that the skin effectively functions as a gill. (B) A simplified schematic diagram, emphasizing the geometric relationships of the muscle groups used to produce the behaviors shown in Fig. 3. Contractions of circular muscles produce elongation, longitudinal muscles produce shortening, and dorsoventral (DV) muscles produce flattening. Oblique muscles stiffen the animal at a length intermediate between maximal contraction and maximal elongation. Not shown are annulus erector muscles, located in the skin, that cause the individual annuli (five per segment in the midbody) to form peaked ridges around the animal. inhibitory – that project to longitudinal muscles, innervate either (Elliott, 1987); stretch receptors, embedded in the dorsal, lateral, dorsal, dorsolateral, lateral, ventrolateral, ventral, or dorsolater- and ventral body wall of each segment (Blackshaw et al., 1982; oventral regions (Stuart, 1970). The excitatory neuromuscular Blackshaw and Thompson, 1988; Blackshaw, 1993; Cang et al., transmitter is ACh (Sargent, 1977) and the inhibitory transmitter 2001); and mechanoreceptors of different sorts. There are is GABA (Cline, 1986). Activity of various combinations of ciliated mechano-receptive neurons whose somata are in the these MNs in different temporal patterns produce the behaviors sensilla (DeRosa and Friesen, 1981; Phillips and Friesen, 1982) described in subsequent sections. and respond to movements of the water (Brodfuehrer and Leeches have a variety of sensory receptors. For instance, Friesen, 1984). There are also mechanoreceptors with somata in there are light-sensitive receptors in the sensilla located in each the CNS (Nicholls and Baylor, 1968) that have free nerve endings segment and in the eyes (a pair of expanded sensilla located on in the skin (Blackshaw, 1981) and respond to different intensities the lateral edge of each of the first five segments) (Kretz et al., of stimulation to the skin: light touch (T cells), pressure (P cells), 1976); chemoreceptors, in placodes on the upper lip and noxious stimuli (N cells). These neurons have primary 284 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 receptive fields within their own segment, and secondary ones Nicholls lab identified mechanosensory neurons (Nicholls and (via axons through connectives) in the adjacent ganglia both Baylor, 1968) and MNs (Stuart, 1970) in stereotyped locations anterior and posterior. There are three pairs of T cells and two within each segmental ganglion. They showed that these pairs of P cells. The receptive fields of these cells divide up the sensory neurons made both electrical and chemical synaptic body wall into roughly equal, overlapping receptive fields around connections onto the MNs, and also established a number the circumference of the animal. There are two pairs of N cells in of physiological techniques to suggest strongly that these each ganglion, each of which innervates half the ipsilateral body connections were direct, monosynaptic contacts, without any wall. All the N cells respond to noxious mechanical stimuli, intervening INs (Nicholls and Purves, 1970). Subsequently, but the two on each side differ in their responses to other stimuli, several other MNs have been identified (Ort et al., 1974; such as heat and low pH (Pastor et al., 1996). Thompson and Stent, 1976a; Sawada et al., 1976; Norris The great majority of the neuronal somata within the leech and Calabrese, 1987). Many of these MNs are inhibitory CNS, as in other animals, are neither sensory nor MNs; rather (Stuart, 1970; Sawada et al., 1976; Ort et al., 1974), releasing they are INs without direct connection to the periphery. These GABA onto muscle fibers to hyperpolarize them and cause their INs were identified largely by methodically searching for relaxation (Cline, 1986). Surprisingly, at least some of the neurons associated with specific behaviors. For example, inhibitory MNs also make strong central connections, with both specific neurons were identified when intracellular current excitatory and inhibitory MNs (Ort et al., 1974; Granzow et al., injection evoked (or terminated) a behavior in either semi-intact 1985; Granzow and Kristan, 1986; Friesen, 1989a), and with preparations, or fictive behavior in the isolated CNS. Using this INs (Friesen, 1989b). technique, INs were found that participate in seven behaviors: heartbeat (Thompson and Stent, 1976b), local bend (Lockery 1.2. The hydroskeleton and behaviors and Kristan, 1990b), shortening (Shaw and Kristan, 1995), swimming (Friesen et al., 1978; Weeks, 1982a,b,c; Friesen, Leeches perform a variety of distinguishable behaviors by 1985, 1989b; Brodfuehrer and Friesen, 1986a,b,e), crawling combinations of lengthening, shortening, and bending. Fig. 3 (Eisenhart et al., 2000), reproduction (Zipser, 1979) and shows five of these behaviors: local bending, swimming, feeding (Zhang et al., 2000). It is largely true that homologs of whole-body shortening, crawling, and feeding. Each behavior each of the neurons found in one ganglion can be found in the is produced by a characteristic temporal and spatial pattern of remaining 20 segments. There are exceptions to this general muscle contractions. The nature of these motor patterns is rule, which are pointed out in the sections below. In addition, discussed below, in individual sections devoted to each of the homologs of the Retzius neurons (Lent, 1977) and several behaviors. mechanosensory neurons (Yau, 1976) have been found in the Leeches have no hard, fixed skeleton. Instead, they use a neuromeres of the subesophageal ganglion, although many of muscular arrangement that has been termed a ‘‘muscular- the INs in the subesophageal ganglion do not appear to have hydrostat’’(Kier and Smith, 1985) or a ‘‘hydroskeleton’’(Kristan homologs in the segmental ganglia (Brodfuehrer and Friesen, et al., 2000). The leech body is a tube whose shape is controlled 1986a,b,c,e). This ability to identify a particular neuron in by muscles in each segment. To a first approximation, each segment after segment and in animal after animal has greatly segment is a cylinder with an ovoid cross-section (Fig. 2B); a aided the characterization of neuronal circuits. In addition, this segment maintains roughly the same volume during all stereotypy has led to the notion that all neurons within the leech behaviors. The muscles used to produce the behaviors shown CNS are unique (with the possible exception of the PE cells in Fig. 3 are the longitudinal and circular layers in the body wall, (Baptista and Macagno, 1988), neurons that develop post- plus the dorsoventral muscles that span the body cavity. embryonically in the ganglia of segments 5 and 6, which are the Contraction of one muscle type changes the shape of the reproductive segments of leeches). cylinder by increasing internal pressure, stretching the other The leech was developed as a neurophysiological prepara- muscle types. Hence, each of these three muscle types is tion in the 1930’s by Gray et al. (1938), who studied the potentially an antagonist for the other two sets of muscles. The neuronal bases of leech swimming and crawling. The first body stiffness, which acts like a skeleton, is caused by co- intracellular recordings were accomplished in the early 1960’s, contraction of antagonistic muscles or, for some behaviors, by when Hagiwara and Morita (1962) and Eckert (1963) recorded contraction of a fourth set of muscles, the oblique muscles. These intracellularly from the somata of the paired Retzius neurons in latter, thin muscles, which lie between the longitudinal and segmental ganglia of Hirudo. They both showed convincingly circular muscles, are oriented obliquely, so that their contraction that these neurons are strongly electrically coupled. They also stiffens the body at an intermediate body length – the posture seen demonstrated that at least two of the neurons in each ganglion in a leech at rest. In effect, leeches have a degree of freedom not were identifiable by the location and size of their somata, as available to skeletized animals: they can control the stiffness of well as their electrophysiological properties. The identification their skeleton dynamically during a behavior. A single segment and characterization of leech neurons advanced greatly when can perform four types of basic movements: John Nicholls chose to identify neurons that had been used in the laboratory of Stephen Kuffler to characterize the electro- (1) bending, by contracting the longitudinal muscles on one physiological properties of glial cells and their effects on the side; the longitudinal muscles on the opposite side may also electrical function of neurons (Kuffler and Potter, 1964). The relax; W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 285

(2) shortening, by contracting all longitudinal muscles at the same time; (3) elongation, by contracting the circular muscles and (4) ﬂattening, by contracting the dorsoventral muscles. (Note that ﬂattening also produces elongation).

The first four behaviors illustrated in Fig. 3 are caused by different temporal and spatial patterns of these segmental movement units. Local bending (Fig. 3A) occurs in a small number of adjacent segments (1). Whole-body shortening (Fig. 3C) takes place in the whole animal almost simultaneously (2). In swimming (Fig. 3B), flattening of the whole body (4) is maintained throughout swimming episodes, and serves to stiffen the body and present a wide surface to the water. During swimming, each segment alternately bends dorsally and ventrally, with each segment producing the same movement as its more anterior neighbor at a phase delay of about 5%. This produces a repeated up-and-down undulation with about one peak and one trough in the body at any given time (Kristan et al., 1974a). Crawling (Fig. 3D) is also an oscillatory locomotory pattern, but in this case shortening (2) alternates with elongation (3) in each segment. Again, whatever occurs in a given segment is repeated in the next segment with a delay of about 5% of the cycle period. Compared to swimming, Fig. 3. Examples of leech behaviors. (A) Local bending: pressing on the skin at crawling cycles are slow: swim cycles are 0.4–2.0 s in duration any location around the leech’s surface (e.g. dorsal, ventral, lateral) causes (Kristan et al., 1974a,b; Kristan and Calabrese, 1976), whereas contraction of the longitudinal muscles at the site of the touch, and relaxation of crawling cycles are 2–20 s in duration (Stern-Tomlinson et al., the longitudinal muscles on the opposite side. (B) Swimming: successive frames 1986; Cacciatore et al., 2000). The action of the suckers is of a film of a swimming leech, from the side, taken at 50 ms intervals. A quasi- important in shortening and crawling, but very little is known sinusoidal wave of dorsoventral contractions moves from the front of the leech (at left) to the back. The trough of the wave is produced by local contractions of about their muscular or neuronal control. The typical feeding the dorsal longitudinal muscles, and the crest is produced by contractions of the posture (Fig. 3E) is with one or both suckers attached. The jaws ventral longitudinal muscles. The whole body is strongly flattened throughout evert through the front sucker, rasp a hole in the skin of the prey, the swim cycle. In a single segment, swimming consists of repeating alterna- and blood is sucked through the oral opening in this sucker. tions between dorsal and ventral longitudinal contractions. The rearward Specialized internal muscles in the pharynx produce the suction progression of the traveling waves results from an intersegmental delay in longitudinal muscle contractions. This delay varies in proportion to the cycle that brings the blood into the body, but longitudinal muscles period to maintain approximately one waveform in the body at all cycle periods produce a peristalsis that moves blood into the various chamber (the normal range being 0.4–2 s). The white dots at four locations along the of the gut (Wilson and Kleinhaus, 2000). animal’s body are white beads sewn onto the lateral edge of the leech at specific locations. The 12 frames shown constitute one swim cycle (note that the body 2. Circulation and heartbeat shape in the last frame is very similar to that seen in the first frame). (C) Whole- body shortening: when the front end of the leech is touched (top frame), it pulls back rapidly (bottom frame) by contracting the longitudinal muscles in all body Heartbeat is an autonomic function that is rhythmic and segments simultaneously. (There is a short intersegmental delay, caused by continuous in Hirudo medicinalis. The circulatory system is a spike conduction between segments, but this delay is much shorter than the closed network comprising four longitudinal vessels – one delays seen in swimming.) This response is greatest when the animal is fully dorsal, one ventral, and two lateral – that run the length of the elongated. (D) Crawling: schematic drawings of six characteristic stages of vermiform crawling. In the top trace, the animal has shortened fully and animal, communicating in every segment by a series of attached both suckers. The step begins with release of the front sucker (to branched vessels (Fig. 2A). Rhythmic constrictions of the the right) and the beginning of a wave of elongation at the front end produced by muscular lateral vessels (the heart tubes) drive the flow of blood contraction of the circular muscles (second frame). The circular muscle-induced through the closed circulatory system (Thompson and Stent, elongation moves back along the animal until the body is fully extended; at this 1976a). The heart muscle can generate a myogenic rhythm of point, the front sucker attaches (third frame). With both suckers attached, the elongation at the front end is replaced by contraction (fourth frame), caused by contractions, which are normally entrained by patterned relaxation of the circular muscles and contraction of the longitudinal muscles. rhythmic motor outflow from segmental ganglia. The hearts This contraction moves back along the animal, pulling on the posterior end and are coordinated so that one beats in a rear-to-front progression lengthening it. As the contraction wave moves through the back half of the animal, the back sucker releases (fifth frame). The cycle is completed when the anterior end is held in a characteristic, rigid posture with pulsations at 2–4 Hz contraction wave reaches the rear end and the back sucker is reattached (sixth reflecting the sucking movements made by the pharyngeal muscles in the frame). (E) Feeding: shown is a typical posture during the consummatory phase anterior end. The rest of the body produces slower movements, either undula- of feeding. The back sucker can be attached to the host or, as shown here, it can tions or peristaltic movements (see Fig. 25) that move the ingested blood into be floating free. The front sucker is tightly attached to the skin surface and the the gut pouches (see Fig. 2A). Scale bar, 2.5 cm. 286 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

(peristaltically), whereas the other one beats synchronously along most of its length. The peristaltic side produces high systolic pressure and forward ﬂow of blood in the heart tubes, whereas the synchronous side produces low systolic pressure and blood ﬂow into the peripheral circulation (Krahl and Zerbst-Boroffka, 1983; Hildebrandt, 1988; Wenning et al., 2004a,b).

2.1. The heartbeat neural control system of the leech

The hearts are innervated in each segment by heart excitatory (HE) MNs, a bilateral pair of neurons found in each of the third through 18th segmental ganglia (i.e. M3–M18) (Thompson and Stent, 1976a,b). Each HE MN contacts heart muscle directly and forms conventional cholinergic neuromuscular junctions (Maranto and Calabrese, 1984a,b; Calabr- ese and Maranto, 1986). The heart MNs are rhythmically active (Fig. 4C); they entrain the myogenic rhythm of the heart through rhythmic excitation. Thus the spatio-temporal activity pattern of the segmental heart MNs (the motor pattern) determines the constriction pattern of the hearts. This motor pattern is organized by a central pattern generator (CPG): the isolated ventral nerve cord continuously produces a heartbeat pattern that is similar in period and spatial pattern to the heart constrictions seen in the animal (Fig. 4C) (Calabrese and Peterson, 1983). The rhythmic activity pattern of the heart MNs derives from the cyclic inhibition that they receive from this CPG (Fig. 4). When these inhibitory inputs to the heart MNs are Fig. 4. Activity and synaptic connectivity of heart excitatory (HE) MNs and blocked by bicuculline, the MNs fire at a steady rate (Schmidt heart interneurons (HN INs). (A) Circuit diagram showing the inhibitory and Calabrese, 1992). synaptic connections from identified HN INs to HE MNs. (B) Circuit diagram showing the inhibitory synaptic connections among all the identified HN INs. There is an asymmetry in the heart constriction pattern that Neurons with the same input and output connections are lumped together. In all arises from an asymmetry in the motor pattern (Thompson circuit diagrams, large unfilled circles represent neurons (each identified by the and Stent, 1976b). That is, the same coordination modes – number of its ganglion) and the lines represent major neurites or axons. Small peristaltic and synchronous – observed in the two hearts are filled circles represent inhibitory chemical synapses. (C) Simultaneous intra- also seen in the HE MNs. The HE MNs on one side are active in cellular recordings showing the normal rhythmic activity of two reciprocally inhibitory oscillator heart INs of midbody ganglion 4, HN(R,4) and HN(L,4), a rear-to-front progression, while the HE MNs on the other are and a heart MN HE(R,5) postsynaptic to HN(R,4) in an isolated nerve cord active nearly synchronously along most of the nerve cord preparation. Dashed lines indicate a membrane potential of À50 mV. In all (Calabrese and Peterson, 1983; Wenning et al., 2004b), and the figures, the body side and ganglion number are indicated as in the following coordination of the HE MNs along the two sides switches example: cell HN(L,1). approximately every 20–40 heartbeat cycles (Thompson and Stent, 1976b; Wenning et al., 2004a,b). Because switching (Fig. 4B). The other three pairs of heart INs are followers of between coordination states is produced by the isolated nerve these anterior pairs. Two foci of oscillation in this beat timing cord, a CPG must produce this switching, too (Gramoll et al., network have been identified in M3 and M4 (Peterson, 1983a). 1994; Wenning et al., 2004b). Reciprocally inhibitory synapses between the bilateral pairs of The pattern generator comprises seven bilateral pairs of heart INs in these ganglia (Figs. 4B and 5A), combined with the identified heart INs (HNs) that occur in the first seven intrinsic membrane properties of these neurons pace the segmental ganglia (Thompson and Stent, 1976c; Calabrese and oscillation (Peterson, 1983a,b; Cymbalyuk et al., 2002). Thus Peterson, 1983). The connections made by heart INs are largely each of these two reciprocally inhibitory heart interneuronal inhibitory; they inhibit each other, and those in M3, M4, M6, pairs is an elemental half-center oscillator (Figs. 4B and 5A) and M7 inhibit heart MNs (Fig. 4A and B). and the M3 and M4 heart INs are called oscillator INs.

2.2. The elemental oscillators 2.3. Mechanisms of oscillation in an elemental half-center oscillator Because passing current into any one of the ﬁrst four pairs of heart INs can reset and entrain the rhythm of the entire network Voltage-clamp studies have identiﬁed several intrinsic of INs (Peterson and Calabrese, 1982), the heart INs in M1–M4 currents that contribute to the oscillatory activity of oscillator constitute the timing network of the heartbeat pattern generator INs. These include: W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 287

oscillator IN. Release from inhibition results from a waning of the depolarization in the active oscillator IN due to the slow inactivation of its ICaS, which slows its spike rate and thereby reduces its spike-mediated inhibition of the contralateral oscillator IN. A model of the half-center interactions was tested by voltage clamping oscillator INs using waveforms that mimic the slow wave of oscillation (Olsen and Calabrese, 1996). Quantitative estimates of Ih, IP, and the low-threshold Ca2+ currents, as well as their trajectories during oscillation, demonstrated that Ih has a negative feedback relationship with period. Perturbations that produce a longer cycle period increase Ih, thereby shortening the cycle period, whereas perturbations that produce a shorter period decrease Ih, which increases the cycle period. Thus Ih homeostatically regulates the heartbeat cycle period. Fig. 5. Circuit diagram and electrical activity of the heartbeat-timing network. (A) The timing network consists of paired heart INs (HN) in the first four 2.4. Coordination in the beat timing network midbody ganglia (M1–M4). The first and second ganglia are represented as a single ganglion for simplicity. Large unfilled circles represent neurons (each identified by the number of its ganglion of origin) and the lines represent major The heart INs in M1 and M2 act as coordinating INs, serving neurites or axons, squares are distal sites of spike initiation, and small filled to couple the two elemental oscillators (Fig. 5A). Together with circles are inhibitory synapses. (B) Activity of a segmental oscillator. Simulta- the HN cells in M3 and M4, these INs form a beat timing neous extracellular recordings of the paired oscillator INs in M3 made with a network that paces the pattern generator and establishes the suction electrode placed snugly over the soma of each neuron. The two INs underpinnings of intersegmental coordination (Fig. 5A; Peter- alternate in their activity. (C) Coordinated activity of the timing network. Simultaneous extracellular recordings of the coordinated activity of ipsilateral son, 1983a,b; Masino and Calabrese, 2002a). The INs in M1 heart INs in M2–M4 made with suction electrodes. The oscillator INs in M3 and and M2 do not initiate spikes in their own ganglion; instead they M4 are active nearly in-phase but with a perceptible phase lag of about 10% of have two spike initiating sites, one in M3 and the other in M4. the cycle period, while the coordinating IN is active in anti-phase (after Masino Normally, the great majority (>85%) of spikes in the and Calabrese, 2002a). Please note that the literature on heartbeat uses an older coordinating neurons are initiated in M4, but under certain designation for the midbody segmental ganglia: G1–G21, rather than M1–M21. conditions spikes can also be initiated in M3 (Peterson, 1983a,b; Masino and Calabrese, 2002a). For example, when a fast Na+ current that mediates spikes; M3 and M4 are separated by cutting the connective, the two low-threshold Ca2+ currents [one rapidly inactivating processes of the coordinating neurons in M3 will initiate spikes. (ICaF) and one slowly inactivating (ICaS); Angstadt and Thus, isolated M3 and M4 each contain a segmental oscillator Calabrese, 1991]; that consists of a pair of reciprocally inhibitory oscillator INs + three outward currents [a fast transient K current (IA) and (elemental half-center) and the active stumps of processes from + two delayed rectifier-like K currents, one inactivating (IK1), the coordinating neurons, which provide additional inhibition and one persistent (IK2); Simon et al., 1992]; (Hill et al., 2001). (Invertebrate axons may have neuritic a hyperpolarization-activated inward current (Ih), a mixed branches (input and output) at several sites within the nervous Na+/K+ current with a reversal potential of À20 mV system and thus have multiple, electrotonically distant sites of (Angstadt and Calabrese, 1989); synaptic input, spike initiation, and synaptic output (Perrins and + a low-threshold persistent Na current (IP)(Opdyke and Weiss, 1998; Coleman et al., 1995).) The coupling between the Calabrese, 1994). M3 and M4 segmental oscillators causes the M3 and the M4 oscillator INs on the same side to be active roughly in phase The inhibition between oscillator INs consists of a graded (Fig. 5C). component that is associated with the low-threshold Ca2+ In an isolated beat timing network (M1–M4), the phase currents (Angstadt and Calabrese, 1991) and a spike-mediated relationship between the oscillators is constant (i.e. their component associated with high-threshold Ca2+ current (Simon activity is phase-locked) but varies among preparations and et al., 1994; Lu et al., 1997; Ivanov and Calabrese, 2000). whether the preparation is exposed to modulator substances. In Spike-mediated transmission is sustained during normal the majority of unmodulated preparations, the M4 oscillator bursting (Nicholls and Wallace, 1978; Ivanov and Calabrese, leads the M3 oscillator (Fig. 5C; Peterson, 1983a,b; Masino and 2003), while graded transmission wanes during a burst because Calabrese, 2002a). Because both the M3 and M4 oscillator INs the low-threshold Ca2+ currents inactivate (Angstadt and provide inhibitory input to heart MNs, the phase relationships Calabrese, 1991). Modeling studies based on these biophysical between the M3 and M4 segmental oscillators is important in measurements indicate that oscillation in an elemental half determining the HE MN activity pattern. When the M3 and M4 center oscillator is a subtle mix of escape and release (Fig. 6; segmental oscillators were reversibly uncoupled by blocking Nadim et al., 1995; Olsen et al., 1995; Hill et al., 2001). Escape axonal conduction in the connectives between M3 and M4 with from inhibition is due to the slow activation of Ih in the inhibited a sucrose solution, the ‘‘intact’’ phase difference proved to be 288 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

Fig. 6. Synaptic conductances and some major intrinsic currents that are active during a single cycle of a two cell (half-center) heart IN oscillator model (Hill et al.,

2001). (A) Biological neurons. (B) Model neurons. The graded synaptic conductance (gSynG) is shown at the same scale as the total synaptic conductance (gSynTotal), which is the sum of the graded and spike-mediated conductances. The slow calcium current (ICaS), the hyperpolarization-activated current (Ih), and the persistent sodium current (IP) are shown to the same scale. Note that IP is active throughout the entire cycle period. determined by the difference between the ‘‘isolated’’ periods in lization has been shown in modeling studies (Hill et al., 2002, the segmental oscillators: the faster oscillator leads in phase, 2003; Jezzini et al., 2004). and the phase difference is nearly linearly related to the period difference (Masino and Calabrese, 2002c). Both the period 2.5. Heartbeat motor pattern switching by switch difference between the segmental oscillators as well as their interneurons phase differences can be changed using neuromodulators and pharmacological agents (Masino and Calabrese, 2002b). Again, Switching between the peristaltic and synchronous modes in these experiments, the period of the intact beat timing (Fig. 7A) is accomplished by a pair of switch INs whose somata network was the same as the period of the faster segmental are in M5. The M3 and M4 oscillator INs on one side inhibit the oscillator (Masino and Calabrese, 2002a,b; Hill et al., 2002). switch heart IN on the same side (Figs. 4B and 7B)(Thompson This rate dominance by the faster oscillator is likely to result and Stent, 1976c; Calabrese, 1977). These switch INs inhibit from the faster oscillator beginning to ﬁre earlier in each cycle, both heart INs in M6 and M7. Only one of the switch INs thereby silencing the ipsilateral coordinating INs and removing produces impulse bursts during any given heartbeat cycle; the their inhibition onto the slower oscillator, thus speeding up the other switch IN is silent, although it receives rhythmic slower oscillator’s rhythm. The feasibility of this conceptua- inhibition from the beat timing oscillator (Fig. 7B; Calabrese W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 289

Fig. 7. Switches in coordination state in the central motor program for heartbeat. (A) Continuous extracellular records from the vascular nerves (VN) of three segmental ganglia of an isolated nerve cord. The records are indexed for segment and body side as indicated in Fig. 4. The bursts of impulses recorded on the vascular nerves result from activity in the axons of the heart HE MNs. Small arrowheads at the beginning and end of the record indicate the starts of bursts. The record begins with the right side coordinated peristaltically (note rear-to-front progression in the start of the heart MN bursts) and the left side coordinated synchronously. At the large arrowheads a switch in coordination state occurs so that at the end of the record the right side is coordinated synchronously and the left is coordinated peristaltically (after Calabrese and Peterson, 1983). (B) Simultaneous intracellular recordings from two switch heart INs of ganglion M5 show that only one is rhythmically active at a time; the other is inactive. The large arrowheads indicate a spontaneous reciprocal switch of their activity states (after Lu et al., 1999). (C) Circuit diagrams of the heartbeat pattern generator and the phase relations among the heart INs (HN cells) before and after the switch illustrated in B. Circuit diagrams are like that of Fig. 4B. Neurons that are stippled fire approximately in phase with one another and in antiphase with those that are not stippled. The system has two different coordination states, depending on which of the two switch heart INs is active and which is inactive (dashed line) (Calabrese, 1977; Gramoll et al., 1994; Lu et al., 1999). The coordination of MNs anterior to ganglion M7 is also controlled by the switch IN, which unilaterally drives an unidentified premotor IN in a nearly one-to-one fashion (Calabrese, 1977). The small phase differences (5–10%) between the HN(3) and HN(4) neurons and the between the HN(6) and HN(7) INs are not illustrated. and Peterson, 1983; Gramoll et al., 1994; Lu et al., 1999). With synchronous activity mode (Fig. 7B). The switch INs link a period approximately 20–30 times longer than the period (8 s) the timing oscillator in M1–M4 to the M6 and M7 heart INs. of the heartbeat cycle, the silent switch IN is activated and the Because only one switch IN is active at any given time, there is previously active one is silenced. The activity of the switch INs an inherent asymmetry in the coordination of the heart INs on determines which side is in the peristaltic versus the the two sides: the ipsilateral M3, M4, M6, and M7 heart INs are 290 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 active roughly in phase on the side of the active switch IN, intensity of the touch. Such studies have used a calibrated whereas the ipsilateral M3 and M4 INs are active out of phase touching device (Lewis and Kristan, 1998c), electrical with the M6 and M7 INs on the side of the silent switch IN stimulation of the skin (Kristan et al., 1982), and intracellular (Fig. 7C). Hence, the heart MNs are coordinated synchronously depolarization (Nicholls and Baylor, 1968) to activate on the side of the rhythmically active switch IN, whereas the mechanoreceptive T (touch), P (pressure), and N (nociceptive) MNs are coordinated peristaltically on the side of the silent cells. One generality from these studies is that activating switch IN. The observed switches in the coordination state of T cells elicits the same behaviors as activating P cells, but the heart MNs, therefore, reflect switches in the activity state of P cells are more effective than T cells. That is, fewer P cells the switch INs (Fig. 7B; Thompson and Stent, 1976c; need to be activated at lower frequencies to produce the same Calabrese, 1977). behavioral responses as T cells. N cell activation also evokes Is the switch in the activity states of the switch INs swimming behavior, but can also cause qualitatively different controlled by an independent switch timing network or is the responses, such as writhing and flailing, which have yet to be switching mechanism inherent to these neurons themselves? studied. There appears to be no synaptic connections between the switch At a threshold level of mechanosensory stimulation, the INs, because injecting current to change the activity in one predominant response elicited depends upon the location of the switch IN does not influence the activity state of the other, even stimulus: stimulating the anterior end produces shortening, though spontaneous switches in the activity state are always stimulating the posterior end produces crawling or swimming, reciprocal (Lu et al., 1999). Voltage-clamp studies showed that and stimulating midbody sites produces local bending (Kristan in the silent state, switch neurons have a persistent outward et al., 1982). Over some range, increasing the stimulus intensity current that is not voltage-sensitive and reverses around merely increases the intensity of the response. At high À60 mV (Gramoll et al., 1994). This current turns off in a stimulation frequencies (in the range of 20–40 Hz), however, switch to the active state. Thus, in its silent state, the switch IN the responses change qualitatively. Strong stimuli to the front is inhibited by a persistent leak current. The dynamic-clamp and back often produce sequential responses, with a violent (Sharp et al., 1993) was used to simulate the turn-on and turn- version of the primary response followed by a vigorous off of such a leak, which was found to be sufficient to change secondary response. For instance, stimulating strongly the front the activity state of the manipulated switch IN (Gramoll et al., end often produces a vigorous shortening followed by a rapid 1994; Lu et al., 1999). These results argue that switching is elongation that might lead to crawling or swimming (Kristan controlled by an independent timing network, extrinsic to the et al., 1982). Increasing the intensity of stimulation to a middle switch neurons, that alternately imposes a tonic inhibitory leak segment produces progressively stronger and more extensive alternately on one of the two switch INs. This network remains local bending, in which more remote segments anterior and unidentified. posterior to the one stimulated stiffen, so that the bend of the segment being stimulated becomes more exaggerated and 2.6. Gaps in our current knowledge widespread (Wittenberg and Kristan, 1992a). At even higher stimulation intensities, the behavior switches to a whole-body The heartbeat CPG can be conceptualized as two timing response, such as curling or twisting (Kristan et al., 1982). networks: a beat timing network comprising the first four pairs The following sections summarize what is known about each of heart INs (two oscillator pairs and two coordinating pairs) of the neuronal circuits underlying the five individual and a switch timing network that governs the activity of the behaviors. Local bending and shortening are purely reflexive, switch INs. The two timing networks converge on the switch episodic responses to a well-defined stimulus. Swimming, INs, and together with the heart INs in M6 and M7, make up the crawling, and feeding are rhythmic behaviors that, like the heartbeat CPG (CPG). The output of the CPG is configured into heartbeat system described above, are produced by CPGs that two coordination states of heart MNs by the alternating activity can also produce the underlying motor pattern in the isolated states of the two switch INs. Although much remains to be done nervous system (Fig. 1B). This part includes a section on to understand the dynamics of the activity pattern of the INs in interactions among the behaviors, emphasizing how individual the heartbeat CPG and how these dynamics translate into the neurons are used in more than one behavior. HE MN output pattern, the problems presented by such a synthesis seem tractable. 3.2. Local bending

3. Overt behaviors 3.2.1. Motor neurons In response to a light touch to the skin of a segment in the 3.1. Introduction middle of a leech (Fig. 8A), that segment produces shortening on the side touched and lengthening of the side away from the Four of the leech behaviors – local bending, whole-body touch (Fig. 3A). A segment produces this behavior by shortening, crawling, and swimming – can be elicited by tactile contracting longitudinal muscles on the side of the touch stimulation of the leech. The ﬁfth – feeding – is elicited by and elongating those on the opposite side, after initially chemical, tactile, and thermal stimuli. Which behavior is contracting both sides (Fig. 8B; Kristan, 1982). (The elicited by tactile stimulation depends upon the location and contribution of circular muscles is discussed below.) Mono- W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 291

Fig. 9. Local bend circuitry. (A) Simpliﬁed schematic summary of connections. Pressure-sensitive mechanoreceptors innervating dorsal and ventral body wall

(PD and PV cells, respectively), with overlapping receptive fields, sense pressure to the skin. They transmit this activity, via excitatory synapses, to a layer of INs that excites longitudinal MNs. The inhibitory MNs, in addition to inhibiting muscles, also inhibit the excitors of the same muscles. This central inhibition provides the only known inhibitory input to the MNs during local bending. (B) Schematic summary of the central connections. All four P cells connect to all 17 Fig. 8. Behavior and neuronal bases of local bending. (A) The semi-intact known INs, although the strengths of these connections produce cosine-shaped preparation used to characterize local bending. One lateral half of a leech’s body receptive fields for the INs. The connections from the cell immediately above wall, from dorsal midline to ventral midline, is removed and pinned out on one (in the same vertical column) are strong, and the strength diminishes with lateral end. Two tension transducers are attached to the other end to measure tension distance. For instance, cell 115 receives strong synaptic input from the generated by the dorsal and ventral longitudinal muscles. The dark dots along ipsilateral P cells but very weak connections from the contralateral P cell. three of the annuli indicate the sites of the sensilla, located along the middle d v Note that inhibitory MNs receive inputs that are the same as for the excitatory annulus in each segment. One ganglion is left attached to the piece of body wall MNs of the opposite sign. For instance, DI (in this figure, called ‘‘iMNd’’) by its nerves; it is pinned out to allow recording from individual MNs. A receives the same input as the DE MNs (‘‘eMNd’’). This connectivity produces stimulating electrode is placed on the skin to excite the terminals of the responses seen in Fig. 8C. In both diagrams, inhibitory connections are mechanosensory neurons innervating the mid-dorsal region of the body wall indicated by filled circles and excitatory connections by lines. All excitatory in the middle segment. (B) Tension produced by dorsal and ventral longitudinal connections are feed-forward, from P cells to INs and from INs to MNs. muscles in response to stimulating the dorsal skin. The stimulus activated sensory neurons for a short time, indicated by the bar (stim.) under the tension traces. Initially, both dorsal and ventral longitudinal muscles contracted, but the ventral muscle started to relax in less than 1 s, while the dorsal longitudinal synaptic connections from the T and P mechanosensory cells muscle continued to contract for several more seconds. If the ventral skin had onto the L MNs (each of which activates all the longitudinal been stimulated the pattern would be reversed. (C) Recordings from excitatory and inhibitory MNs to longitudinal muscles during a local bend response. The muscles in one side of the body wall) cause the initial co- same kind of electrical stimulus was given to the dorsal skin while recording contraction. Polysynaptic connections onto MNs with more from four different MNs. (The recordings were made successively; they are restricted motor fields (e.g. dorsal excitor neuron cell 3 (DE-3) lined up relative to the stimulus to indicate their activity during a single excites a band of dorsal longitudinal muscles and ventral response.) DE is an excitor MN to the dorsal longitudinal muscles; VE is an excitor of the ventral longitudinal muscles; DI is an inhibitor of the dorsal of the dorsal and ventral longitudinal muscles, in fact, is produced largely by the longitudinal muscles, and VI is an inhibitor of the ventral longitudinal muscles. activation of the L cells, MNs that cause strong contraction in all the long- Note that stimulating the dorsal skin activates DE and inhibits VE, as expected itudinal muscles on one side of a segment. The L cells are activated only during from the tension recordings. The inhibitors, DI and VI, show the opposite the stimulus, so that the initial response to the stimulus is, as previously response. The motor neuronal responses are entirely consistent with the tension surmised from neuronal connections (Nicholls and Baylor, 1968), a shortening recordings except that VE shows only inhibition; it does not show the initial response. Activity in the more localized MNs (i.e. DE and VE) persists far after excitation apparent in the Ventral tension trace in part B. This initial excitation the stimulus; this persistent activity is the cause of the local bend. 292 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

Fig. 10. Summary of receptive fields of mechanosensitive P cells, INs, and MNs. The left column (panels A, C, and E) shows data used to determine the receptive fields obtained from recordings of the different neurons. The maximal response location was calculated for all the neurons by fitting a cosine function to their stimulus response magnitudes at several locations within their receptive fields. The data were plotted relative to the peak location (08). Receptive fields for the P cells were determined from preparations illustrated in Fig. 8A. Stimulating each of the four P cells and plotting the amplitude of the response, as a function of the locations of the middle of the P cell receptive fields, determined receptive fields for INs and MNs. In all cases, the data were normalized to the maximal response for each neuron. The panels in the right column (B, D, and F) are idealized summaries of all the receptive fields for all the neurons of each kind, which were used to simulate the function of the system. (A) Touch (T) and pressure (P) cell receptive fields. The curve is the best-fit cosine function to all data. (B) The best-fit cosine functions for each of the four P cells, each centered on the location of the maximal response. (C) Local bend IN receptive fields, with the best-fit cosine curve. Normalized data are shown for four W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 293 excitor neuron cell 4 (VE-4) excites a ventral longitudinal band) 3.2.3. Local bend interneurons produce the more prolonged excitatory response on the side of INs involved in local bending, the LBIs, were located by stimulation (Fig. 8C). The same stimulation also activates the recording intracellularly from many of the 400 neurons in inhibitory MNs on the side away from the touch, thereby individual segmental ganglia, searching for neurons that (1) inhibiting both longitudinal muscles and the excitatory MNs to received strong input from a P cell with a dorsal receptive field the longitudinal muscles during the contraction on the side of (PD), and (2) activated the dorsal MN, DE-3 (Lockery and the touch (Lockery and Kristan, 1990a). Kristan, 1990b). Seventeen such INs were found, six of which This inhibition could be produced either by direct inhibitory were bilaterally paired and one of which was unpaired. By connections from the local bending INs (LBIs), or by excitatory stimulating each of the P cells individually and in pairs connections onto appropriate inhibitory MNs. Tests showed (Lockery and Kristan, 1990b), the receptive fields of each IN that when the inhibitory MNs synapsing onto DE-3 were were inferred (Lewis, 1999). The receptive fields of each LBI hyperpolarized during ventral stimulation, the input onto DE-3 could be nicely fit by a cosine function, and, because each LBI was excitatory (Lockery and Kristan, 1990b). This experiment receives input from all the P cells, their receptive fields span the suggests that LBIs make only excitatory connections onto MNs, whole circumference of the body (Fig. 10C), that is, twice the and that the only inhibition in the pathway is the one made by width of the P-cell receptive fields. The center of the receptive the inhibitory MNs onto the excitatory MNs (and muscles; field for each identified LBI is characteristic for that IN, and the Fig. 9A). Because there are neither feedback connections (e.g. locations of the centers are distributed fairly evenly around the from INs to sensory cells, or MNs back onto INs) nor any dorsal and lateral body wall (Fig. 10D). In the search for LBIs, functionally important lateral connections at the sensory or none were found to have ventral receptive fields (Lockery and interneuronal level, the first two layers of the local bending Kristan, 1990b). This is likely the result of the search strategy, system form an excitatory, feed-forward circuit (Fig. 9B). The which was directed specifically to find INs with dorsal inputs third layer provides the only lateral connections, namely the and outputs. Because LBIs with ventral fields would be only inhibitory connections made by the inhibitory MNs onto the weakly activated by PD inputs and would not activate DE MNs, antagonistic excitors. These inhibitory connections provide the they could easily have been missed in the original search. If the reciprocity of the response; inhibitory MNs to a given search was thorough, another three or four pairs of ventral LBIs longitudinal muscle (e.g. the left dorsal) receive the sensory might yet be identified. input that is identical to that for the excitatory MNs on the The connections from LBIs to MNs were inferred from the opposite side (i.e. the right ventral) and relax the muscles on the responses of MNs to stimulating P cells, individually and in side contralateral to the touch. adjacent pairs. Because each MN receives input from every P cell, their receptive fields – like those of the LBIs – span the 3.2.2. Mechanosensory neurons that produce local bending entire circumference of the body wall (Fig. 10E). Also like the Two lines of evidence indicate that the P cells are the most LBIs, the receptive fields of the MNs are well characterized by a important mechanosensory neurons for producing the local cosine function. Unlike the LBIs, however, the receptive fields bend response. First, monitoring responses of T and P cells to have an inhibitory component from the P cells whose receptive tactile stimuli of different intensities showed that the amplitude fields are on the side opposite to fields activated by each MN. of the local bend tracked the responses of P cells but not T or N This inhibition ensures that the side opposite to the location of a cells (Lewis and Kristan, 1998c). Second, stimulating a single P touch relaxes, allowing the bend to be more effective in moving cell produces a response that is essentially the same as the body away from an object touching it. As discussed above, stimulating the skin in the middle of that P cell’s receptive field the direct LBI input onto the MNs appears to be exclusively (Kristan, 1982; Lewis and Kristan, 1998a). Hence, to determine excitatory, but this excitation is overridden by indirect the neuronal bases of local bending, the activity of the P cells is inhibition via the inhibitory MNs (Fig. 9B). The receptive the most important. The receptive field of each P cell is the top fields of the MNs, like those of the P cells and INs, are of a cosine function, cut off at Æ908 (Fig. 10A), taking the body distributed uniformly over the body wall. In fact, the tactile circumference as 3608. The scatter of the points around this receptive fields of the MNs are very similar to their motor units. curve shows that there is considerable variability in the For instance, excitatory dorsal longitudinal MNs have receptive responses at different locations even when the site and fields centered on the dorsal surface and excitatory ventral MNs magnitude of the stimulus is kept the same (Lewis and Kristan, have receptive fields centered on the ventral surface. It is this 1998c; Lewis, 1999). Adjacent P cells form overlapping coherence of sensory and motor fields, in fact, that provides the receptive fields that span about half the body circumference, localization of the local bend. The relaxation of the side with peak responsiveness in the middle of the dorsal and ventral opposite to the touch results from the disjunction in the sensory quadrants on each side (Fig. 10B). and motor fields of the inhibitory MNs. For example, the

LBIs (cells 115, 125, 212, and 218) in response to stimulating each of the four P cells. (D) The idealized receptive fields for all the known and hypothesized INs. (E) Motor neuron receptive fields. Normalized synaptic responses from the four P cells for both excitors (cell 3 is a DE, cell 4 is a VE) and inhibitors (cell 1 is a DI, cell 2 is a VI) are shown, with the cosine function that best fits the data for all the motor neurons. (F) Idealized receptive fields for the longitudinal muscle MNs, with the peaks lined up by the locations of their maximal responses. (G) The approximated motor fields of the longitudinal MNs used in modeling the function of the local bending circuit. All connections are excitatory to the three or four sections of longitudinal muscles (from Lewis and Kristan, 1998b). These are approximations of the innervation fields of the MNs (Stuart, 1970). 294 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

Fig. 11. Vector model matches local bending behavior. (A1) The direction of 16 local bending responses to identical stimuli delivered to the mid-dorsal skin, 458 from the dorsal midline (arrow). The dashed line indicates the average response to the 16 stimuli. The angular difference between the arrow and the dashed line is the error of the response. (A2) The locations of 12 responses to repeated bursts of intracellular stimulation delivered to the right ventral P cell that were similar to the train of impulses seen when the middle of the receptive field of this neuron was touched. These stimulus trains produced an average response (dashed line) very close to the middle of the receptive field of this neuron, which is the expected response to activation of a single P cell. (From Fig. 10B, the only location at which a touch would W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 295 sensory receptive fields of the dorsal inhibitors are centered on contribulte significantly to the local bend (Zoccolan and Torre, the ventral surface of the opposite side, and the ventral 2002). Interestingly, a linear combination of the contributions inhibitors are most strongly activated by touching the dorsal of circular and longitudinal MNs matched the observed skin on the opposite side. responses to mechanosensory stimulation very well. In addition, applying this much finer-grained analysis to 3.2.4. The local bend response as a directed behavior stimulating individual T, P, and N cells confirmed that the The characteristics of the local bend circuit (broadly tuned major contribution to local bending is from the P cells. On a neurons with overlapping receptive fields) provide an ideal way more global scale, this analysis showed that, despite a large for a neuronal population to perform a vector estimation of the variability in the individual MN responses, the contraction location of a stimulus (Salinas and Abbott, 1995) or for phase of local bending was quite reproducible, because the calculating the direction of a movement controlled by those INs variability between MNs was independent (Zoccolan et al., (Georgopoulos et al., 1988). The schematic summary diagrams 2002). The relaxation phase of the response, however, was in Fig. 10B, D, F, and G emphasize the fact that touching a much more variable, probably due to variabilty in the duration particular location produces shortening centered at that of the responses of the INs and in the biomechanics of the location. Behavioral tests showed that the response was muscles (Garcia-Perez et al., 2004). centered to within 8% of the circumference of the stimulated By using principle component analysis on optic flow data, segment (Lewis and Kristan, 1998b), despite the fact that the the local bend response was shown to discriminate the location responses of only four sensory neurons are used to make this of two tactile stimuli as well as humans can do with their finger localization. This precision is possible by using the relative tips (Baca et al., 2005). As suggested previously (Lewis and firing rates of overlapping sensors that are evenly distributed Kristan, 1998c), this analysis also showed that the location of over the circumference. This population code is maintained the stimulus is coded within the first 200 ms of the stimulus through the LBI layer to the MNs. A simple model of this (before the movement actually starts), so that more prolonged network, using the amount of variability seen in the responses at stimuli did not produce better localization (Baca et al., 2005). the sensory level, showed that this network was sufficient to Longer stimuli did, however, produce larger responses, explain the precision shown by the animal’s localized bending suggesting that response location and response magnitude response to local stimulation (Lewis and Kristan, 1998a). In are coded somewhat differently. Because its neuronal circuitry addition, the model matched the accuracy of the response to is so well defined, local bending is an ideal behavior for testing stimulation at two locations (Fig. 11). This experiment ideas about sensory coding. confirmed that individual P cells contribute strongly to the Neural net models of local bending behavior have been used direction of the response as well as showing that a vector to investigate both the feasibility of using a distributed network population code is important in determining response direction. to perform local bending (Lockery et al., 1989; Kristan, 2000) Further analyses made three additional points: and the influence of synaptic plasticity on local bending (Lockery and Sejnowski, 1992). A very interesting, although (1) Very few P cell spikes – less than five typically – were somewhat troubling, conclusion from these neural net models is required to determine the location of the response (Lewis that synaptic changes large enough to produce significant and Kristan, 1998c). behavioral changes could be undetectable electrophysiologi- (2) The response direction was determined tens of milliseconds cally, if they were distributed over many neurons (Lockery and before the response began. This shows that sensory Sejnowski, 1993). A more optimistic outcome of these models feedback was not required for determining response is that population coding schemes make it possible for a small location (Lewis and Kristan, 1998c). number of neurons to produce several different behaviors, (3) The number of INs required to determine the response provided that they are activated by different sensory neurons or direction depends critically upon the noise in the system. INs (Lockery and Sejnowski, 1992; Kristan, 2000). A more The number of INs used – about 24 – nicely compensates physiologically realistic model (Baca and Kristan, 2001) for the level of noise found in the system (Lewis and showed that the inhibition provided by the inhibitory MNs Kristan, 1998a). could both produce lateral inhibition and smooth out the responses in the coarse representation of location by the P cells. More recent studies have used an optic flow algorithm developed for video surveillance to plot local bending 3.2.5. Gaps in our current knowledge movements very precisely (Zoccolan et al., 2001). A vector The local bend response arises from a relatively simple, field analysis of optic flow data, showed that circular muscles mostly feed-forward system that pulls the body away from the activate only one P cell is in the middle of its receptive field, i.e. at 458.) The responses had less variability than did stimulation of the body wall (e.g. the left panel) because the variability of the P-cell response was eliminated by the electrical stimulation. (B1) Data obtained when the stimuli shown in A1 and A2 were delivered simultaneously, showing the locations of the responses to 16 repetitions of this stimulus pairing. The average response (dashed line) is nearly half-way between the average of the two individual responses. (B2) Directions of the responses predicted from the individual responses of the P cells. The tighter clustering of the responses indicates that much of the variability in the responses to touch was due to the variability of the P cell responses. (C1 and C2) Responses of the model from Fig. 10 to the same P cell activations shown in B1. These responses show the same general features as the real data, suggesting that the neuronal network, as characterized, is essentially complete (from Lewis and Kristan, 1998a). 296 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 site of touch. The orderly overlap in the receptive fields of the ganglion) to every ganglion in the nerve cord (Frank et al., neurons at each level provides a simple network to calculate a 1975). The S cells respond to mechanosensory and photic population vector. Such a system has been proposed for a inputs (Frank et al., 1975; Bagnoli et al., 1975). In each variety of other behaviors: eye movements in mammals (Sparks segment, S cells make electrical connections with L cells, the et al., 1997), arm movement in primates (Georgopoulos et al., MNs that, as stated earlier, cause all longitudinal muscles in one 1988), escape responses in crickets (Theunissen et al., 1996) hemisegment to contract. The FCS is strongly activated during and cockroaches (Camhi and Levy, 1989), and flight orientation whole-body shortening (Magni and Pellegrino, 1978; Mistick, in insects (Douglass and Strausfeld, 2000). In all these systems, 1978). In the Amazonian leech, Haementeria ghilianii, all the a common theme is the mapping of location from a sensory connections are sufficiently effective so that activating the S space onto motor space to produce an accurate response with a cell at moderate rates produces a significant shortening minimal number of neurons. response (Kramer, 1981). In Hirudo, however, the S-to-L The most obvious gaps in our knowledge of the local connection is so weak that stimulating the S cell even at rates bending system are the ‘‘missing’’ LBIs (i.e. the ones with higher than those observed during shortening behavior, ventral receptive fields, and determining how the LBIs connect produces only a very weak motor response (Shaw and Kristan, to circular muscle motor neurons). But even without this 1999). The strongest, most behaviorally relevant pathway is knowledge, the neuronal circuitry for local bending is so well from mostly unidentified INs, which have much slower characterized, that it provides a very useful system for testing conduction velocities than the FCS, but which produce a ideas about coding. Previous studies have suggested that the strong and prolonged activation of the MNs with more limited location of local bending is encoded in the relative number of motor fields than the L cell. It appears that the FCS provides a impulses in adjacent P cells. This can be tested by recording fast but subthreshold activation of the longitudinal muscles, intracellularly from two P cells with adjacent receptive fields upon which the slower system produces full-blown shortening and imposing variations in their normal firing patterns by (Fig. 12B and C). stimulating them in variations of their normal patterns. In other These interneuronal pathways activate all the excitatory words, the convincing way to determine the validity of a code is MNs to the longitudinal muscles in all segments of the body to see whether the system can in fact decode the hypothesized with short latencies. The S cell excitation spreads at 5–6 ms per code. In addition, it would be useful to know the function(s) of segment, and the other pathways are significantly slower, at 15– the inhibitory connections among the MNs. Are they simply for 17 ms per segment (Shaw and Kristan, 1999). In addition, the lateral inhibition, or do they contribute significantly to the inhibitory MNs are strongly inhibited during whole-body calculation of the population vector? How these calculations shortening (Shaw and Kristan, 1995). There is significant are performed, as well as the basis for the independence of variability in the response of individual MNs during whole- variations in MN responses (Zoccolan et al., 2002), can be body shortening, but the ensemble average of the MN response traced out in the activity of the LBIs. is much less variable, because the responses of each MN is statistically independent from all the others (Pinato et al., 2000; 3.3. Shortening Arisi et al., 2001).

Shortening in the leech is caused by a co-contraction of all 3.3.2. Local shortening longitudinal muscles in body segments. Contractions can occur Mechanical stimulation of the skin in mid-body segments in all segments (whole-body shortening) or only a few segments elicits a local shortening response (Fig. 13A; Wittenberg and (local shortening). Kristan, 1992a). As is true for whole-body shortening, P cells are most effective in producing local shortening; in both 3.3.1. Whole-body shortening behaviors all the longitudinal muscle excitors are excited and Moderate activation of mechanosensory neurons – particu- all the inhibitors are inhibited. What distinguishes the two types larly P cells – in the anterior end of the leech reliably produces of shortening is that local shortening is restricted to the whole-body shortening (Fig. 3C). Such stimulation activates segments adjacent to the stimulated segment rather than two parallel interneuronal pathways: a fast, weak one and a involving the whole body. The number of segments involved in slower, more prolonged one (Fig. 12A). The S cells are the local shortening increases with increasing stimulus intensity, central elements in the fast pathway. There is a single S cell in until, at the extreme, the whole body can be involved. Even at each ganglion (Frank et al., 1975) that sends an axon both this extreme, however, local shortening differs from whole- rostrally and caudally into the medial connective (Faivre’s body shortening in that the segment being stimulated produces nerve). The axons of the S cells are among the largest in the a bend rather than a shortening (i.e. the longitudinal muscles on leech CNS, and have the fastest conduction velocity. The axons the side of the segment being stimulated contract and those on from S cells in adjacent ganglia meet midway between the the other side relax). The consequence of this radiating ganglia and make highly effective electrical junctions that shortening on either side of the site is that the degree of bending allow action potentials to pass without fail in either direction. about the site of stimulation increases with increasing stimulus These interconnected S cells form a reliable fast-conducting intensity. Interestingly, a subset of the very same INs system (FCS, Bagnoli et al., 1975) that carries action potentials responsible for local bending send processes out of their from the site of spike initiation (which can be in the S cell in any own ganglion into several adjacent ones, where they make W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 297

Fig. 12. Neuronal bases of whole-body shortening. (A) Schematic summary diagram of what is known about the neuronal circuit underlying whole-body shortening. Touch (T) and pressure (P) neurons in the anterior end (from the brain down to about segmental ganglion 7) activate both the fast, but weak S cell network,andaslower, stronger network (other pathways); both drive the excitatory MNs (DE, VE, and L) to longitudinal muscles and inhibit the inhibitors (DI, VI) to these same muscles in all segmental ganglia. (B) Synaptic input onto excitatory MNs during whole-body shortening in a semi-intact preparation (diagram on the left). Activating mechanosensory neurons in segments 3 and 4 (stimulus) excited both the L cell in segment 11 for the duration of the stimulus train and the dorsal excitor cell DE-3 for a much longer time. Extracellular recordings in the sameganglion showed the spikes from the L cell (marked by dots over the DP:B1 recording—branch 1 of the dorsal posterior nerve) and cell 3 (the large spikes in the DP:B2 recording). The tension recordings show that the response is somewhat larger and faster in the anterior end than in the posterior end. (C) The whole-body nature of the shortening response. Stimulating the skin in segment 4 activates the P cell in that segment, as shown in the Pd intracellular recording from that segment (top trace). The extracellular recordings from nerves DP:B1 and DP:B2 in segments 7 and 12 (bottom four recordings) show that the L cell and cell 3 in these segments were also activated during the shortening response. The stimulus marker provides a time calibration, which is 500 ms long. connections to MNs that produce shortening rather than overpowering behavior; it appears to override all behaviors bending (Fig. 13B; Wittenberg and Kristan, 1992b). It will be but feeding (Shaw and Kristan, 1997), and it activates the L interesting to ﬁnd out whether the two kinds of shortening share cells, which cause the fastest and most powerful contraction of neuronal circuitry in the individual segmental ganglia. any longitudinal MNs (Stuart, 1970; Mason and Kristan, 1982). Some of the INs that trigger swimming are also active during 3.3.3. Gaps in our knowledge shortening (Shaw and Kristan, 1997), so it will be interesting to For both whole-body and local shortening, there are more determine how these very different behaviors share crucial gaps than knowledge. Whole-body shortening is a very interneurons. In addition, the function of the two parallel 298 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

underlying rhythmic neuronal activity patterns. Nearly con- currently, Brown (1911) proposed that central neuronal oscillators generate these neuronal substrates via reciprocal inhibition between neurons within the spinal cord. Ironically, both researchers studied the same behavior – stepping – in mammals. Despite the demonstration by Adrian that there is an inherent rhythmicity in the CNS of the beetle Dytiscus (Adrian, 1931), Gray and coworker’s seminal work on swimming in leeches (Gray et al., 1938), and even the demonstration that crustacean swimmeret movements are generated by neuronal oscillators (Ikeda and Wiersma, 1964), the chain- of-reflexes view prevailed for more than 50 years, leading to Wilson’s critical research on flight in deafferented locusts (Wilson, 1961). Finally, research on the cardiac ganglion and stomatogastric systems in lobster (Hartline, 1967; Mulloney and Selverston, 1974), leech swimming (Kristan and Calabrese, 1976), and lamprey swimming (Cohen and Walleń, 1980), as well as numerous additional preparations (Selverston, 1985; Marder and Calabrese, 1996), clinched the argument unequi- vocally in favor of the central oscillator theory (Delcomyn, 1980); or so it seemed (Pearson, 2000).

3.4.2. Swimming behavior and motor control This review of leech swimming begins with a brief overview of the circuits and mechanisms by which muscles, MNs, sensory neurons and INs control leech swimming movements. The older results have been reviewed many times (Stent et al., 1978; Kristan, 1974, 1980, 1983; Kristan and Weeks, 1983; Friesen, 1989d; Brodfuehrer et al., 1995b). Thus weight is given to recent experiments on the role of sensory feedback in Fig. 13. Neuronal bases of local shortening. (A) A single P cell located in the setting intersegmental phase lags and cycle period, and an middle of the animal (PD, segment 10) was stimulated intracellularly (bottom examination of functional aspects of the central oscillator and trace). The response was measured in DP nerve recordings from segments intersegmental coordination. throughout the nerve cord. This stimulus elicited the largest response in the The quasi-sinusoidal undulations that characterize swim- same segment stimulated, with a smaller response in adjacent segments, both anterior (9,L) and posterior (11,L). The excitation did not spread to either the ming leeches (Fig. 14A) are a consequence of tension and most anterior segment recorded (4,L) or the most posterior one (18,L). (B) The relaxation cycles in two types of segmental muscles. First, the response that spreads to other ganglia activates both dorsal and ventral leech body, is flattened to form an elongated ribbon by tonic excitatory MNs. Stimulating a PD neuron in ganglion 8 (bottom trace) activated contractions of dorsoventral muscles. Second, the alternating cell 3, a dorsal excitor; cell 4, a ventral excitor; as well as the L cell (which contraction and relaxation of dorsal and ventral longitudinal contracts both dorsal and ventral longitudinal muscles) in ganglion 10. muscles (DLM and VLM, respectively) act against internal pressures, with a period of about 0.3–1.0 s, to generate pathways (Fig. 12A) is provocative. It may be that the fast rhythmic bending in body segments (Fig. 14B; Uexku¨ll, 1905; pathway is to ready the animal to behave and the slower Gray et al., 1938; Kristan et al., 1974a; Wilson et al., 1996b). pathway actually produces the response. As discussed in Two major technical advances were critical for research into the Section 3.8.3 below, the S cell is very important for plasticity of neurobiological substrates of leech swimming movements: one the shortening response. Perhaps the INs that are active in was the advent of extracellular recordings, the other the shortening and other behaviors have a special role in plasticity. development of a semi-intact preparation in which neuronal activity and body wall movements are available simultaneously 3.4. Swimming for detailed analyses (Gray et al., 1938). These techniques, together with intracellular recording, revealed that the 3.4.1. History: reflex chain versus central pattern antiphasic contraction and relaxation of the DLM and VLM generator are commanded, respectively, by excitatory (six pairs/segment) Early studies on the neuronal bases of animal locomotion and inhibitory (five pairs/segment) motoneurons (MNs; Stuart, spawned two opposing theories. The first, effectively promul- 1970; Ort et al., 1974; Norris and Calabrese, 1987). gated by Philippson (Philippson, 1905, cited in Gray, 1950) and Intersegmental phase lags of approximately 208/segment in Sherrington (1906) at the beginning of the 20th century, was MN activity ensure the generation of the rearward-traveling that coordinated chains of sensory reflexes generate the body wave (Kristan et al., 1974a,b). These MNs are W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 299

Fig. 14. Expression of leech swimming movements. (A) Swimming leeches generate a quasi-sinusoidal wave of approximately one full wavelength. These swim proﬁles are essentially identical in three different leeches of very different length and weight. (B) Swimming movements result from antiphasic contractions in dorsal and ventral longitudinal muscles. The inset depicts the nerve cord–body wall preparation used to obtain the extracellular records [DP(R,7) and DP(R,14)] and the tension measurements (dorsal and ventral tension, obtained from the proximal ventral and the distal dorsal body wall strips). A single swim episode was elicited by nerve stimulation (artifacts in the DP nerve traces). The inset at the bottom shows the four traces at higher temporal resolution to illustrate the intersegmental phase lags of MN activity and the antiphasic tension rhythms in the muscles. interconnected by inhibitory chemical synapses from the dorsal (comprising eight or more of the 21 midbody ganglia) are inhibitor (DI) to dorsal excitatory (DE) MNs, and from ventral capable of generating rhythmic bursting in MNs that is an inhibitor (VI) to ventral excitor (VE) MNs (Ort et al., 1974; excellent facsimile of the overtly expressed rhythm (Kristan Granzow and Kristan, 1986). The DIs also inhibit the VIs, an and Calabrese, 1976). This ﬁctive swimming has a longer cycle interaction that, unlike most other MNs, crosses the ganglion period (0.5–2 s) and reduced intersegmental phase lags (about midline (Friesen, 1989a). 108/segment) from that seen in intact animals (Pearce and Once thought to be an exception to the rule that rhythmic Friesen, 1984). Experiments on semi-intact preparations also movements in animals are generated by neuronal oscillators revealed that, although sensory feedback is not necessary for located within the CNS (Kristan and Stent, 1975), further generating the basic swim rhythm, feedback loops, perhaps research revealed that isolated nerve cord preparations acting via the DI, are indeed present (Kristan and Calabrese, 300 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

1976). These seminal results laid a solid foundation for studies ignored (Fig. 15B). Reciprocal interactions, both direct and via of the central oscillator that generates the swimming rhythm, electrical coupling, also link the DI and VI MNs to the INs. including studies on the central and sensory mechanisms that Because of these interactions and because their depolarization control initiation and termination of swim episodes, the can shift the phase of swimming activity, these inhibitory MNs mechanisms for chemical modulation of swim expression, are candidate members of the swim oscillator circuit (Kristan and most recently, the role of sensory feedback in interseg- and Calabrese, 1976; Friesen, 1989b; Mangan et al., 1994b). mental coordination. The unpaired cell 208 makes only excitatory connections (Weeks, 1982b; Nusbaum et al., 1987). Tests to determine 3.4.3. Central oscillator circuits whether intrasegmental interactions in the swim circuits are The swimming rhythm is not generated by a MN circuit. monosynaptic have proven difficult, in large part because many Although MNs are phasically active and interact with each of these interactions are not spike-mediated. Consequently, other, this network is not sufficiently complex, nor do MNs individual synaptic potentials arising from synaptic connec- have intrinsic oscillatory properties, to generate the swim tions within ganglia are rarely observed and action potentials oscillations. Subsequent searches among the considerably are not required for synaptic interactions (Granzow et al., 1985; smaller INs identified thirteen candidate swim oscillator INs in Friesen, 1985). Intersegmental interactions, however, occur via most midbody ganglia (at least from M2 to M16) by the criteria spike-mediated synapses. Those connections made by rostrally that: (a) their membrane potential is phase-locked to the projecting INs are repeats their intrasegmental connections, swimming rhythms, (b) current injection into the IN somata – whereas synaptic outputs by caudally projecting neurons are either depolarization or hyperpolarization – shifts the phase of more diverse (Fig. 15C). Not shown in Fig. 15 are the very swimming activity, and (c) there are synaptic interactions with extensive intra- and intersegmental connections by which the other candidate oscillator INs (Friesen et al., 1976, 1978; INs and inhibitory MNs control the excitatory MNs (Poon et al., Weeks, 1982b; Friesen, 1985, 1989b). A set of INs that satisfies 1978; Weeks, 1982b). For completeness, it is essential to point all three criteria is found on both the dorsal and ventral aspects out that inhibitory inputs to cells 60 and 33 are unknown and to of the midbody ganglia (Fig. 15). With somata ranging from mention that two additional neurons, cell 18 (Nusbaum, 1986) about 10–15 mm in diameter, these INs are inhibitory and and 42 (Poon, 1976), are candidate, but poorly characterized, bilaterally paired (cells 27, 28, 33, 60, 115, and 123), except for oscillator neurons. (Discussion of the mechanisms that generate an unpaired medium-sized excitatory IN (cell 208). Without swim oscillations is found below, in Section 3.4.7). exception, their axons project either anteriorly or posteriorly in lateral nerve cord connectives (Poon et al., 1978; Weeks, 3.4.4. Control of swimming activity 1982b; Friesen and Hocker, 2001). However, definitive Leech swimming activity is largely an episodic, rather than information about maximal projections for individual oscillator continuous, behavior, whether evoked in the intact animal, in INs remains elusive. The current understanding is that INs semi-intact preparations, or in preparations of the isolated leech project about six segments, but not more than seven segments cord. In intact animals, swimming may be initiated by a variety (Pearce and Friesen, 1985b; Friesen and Hocker, 2001), except of sensory inputs including tactile stimulation of the body wall for cell 208, which may project as far as 10 segments (Weeks, and water movements. In the isolated nerve cord, bouts of 1982b). fictive swimming are readily obtained in response to An important characteristic of swim-related neurons is the stimulation of a peripheral nerve, intracellular stimulation of phase of their activity (Fig. 15B). The phase reference mark, 08, single neurons, or even without stimulation, e.g. as ‘sponta- for each segment is the median impulse in DE-3 MN bursts neous’ episodic events following application of serotonin (5- recorded with extracellular electrodes (Kristan et al., 1974a). HT) or other neuromodulators. The neuronal network consists The activity phases of other MNs are either 08 (DE and VI) or of at least three layers in addition to the two (oscillator INs and 1808 (VE and DI), except for DI-1, which has a phase of 1208 MNs) that have already been considered (Fig. 16). This section (Ort et al., 1974; Friesen, 1989d). Phases of the INs are more discusses the roles of the neurons in each of these hierarchical dispersed, spanning 08 to nearly 3008. For simplicity and as an layers. aid to modeling, IN phases may be placed into three phase Water vibrations, such as those caused by surface waves or groups: about 08 (40–508), 1208 (130–1708), and 2408 (220– due to water currents, also reliably elicit swimming in hungry, 2608)(Friesen, 1989d). Intersegmental phase lags in the quiescent leeches (Young et al., 1981). In addition, these isolated nerve cord preparations are 8–108 for extended surface waves provide directional cues causing aroused leeches preparations (Kristan and Calabrese, 1976), but increase to to swim towards the source of such waves. The anatomical about 408 if preparations are reduced to two segments (Pearce structures responsible for these responses are sensillar move- and Friesen, 1985b). ment receptors (SMRs), sensory hairs that cluster at the seven Nearly all pairs of bilaterally homologous neurons in the pairs of sensilla located on each midbody segment (DeRosa and leech swim circuit are electrically coupled. More importantly, Friesen, 1981; Phillips and Friesen, 1982). In scanning electron the swim-related oscillator INs are interconnected via microscopy (SEM), SMRs appear as 1 mm wide, 10 mm long, inhibitory synapses (Fig. 15A). Only one set of these IN single hairs; in transmission EM, SMRs have the distinct 7 + 2 interactions (between cells 27 and 28) is strictly reciprocal; microtubule structure of sensory hair cells. Physiologically, other interactions appear reciprocal if laterality of INs is responses recorded from individual sensillar nerves are W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 301

Fig. 15. Intra- and intersegmental interactions between swim-related oscillator INs and MNs. DI and VI are inhibitory MNs; 123, 115, 60, 33, 28, and 27 are inhibitory INs; and 208 is an excitatory IN. (A) Subset of the neurons showing bilateral and ipsilateral interactions. Numerous reciprocal inhibitory interactions occur between INs and MNs across the midline of individual ganglia. (B) The segmental swim oscillator circuit with laterality of neuronal interactions collapsed. (C) Intersegmental interactions between the INs (gray lines denote intrasegmental connections). In B and C neurons are arranged in columns reflecting three phase groupings (top of B). Filled circles, inhibition; T-endings, excitation; resistor, nonrectifying electrical interaction; diode, rectifying electrical interaction. Both the synaptic targets and the sources of synaptic inputs for some neurons remain unidentified. (Modified from Friesen and Hocker, 2001, Fig. 1). selective to near field stimulation and appear in extracellular ganglion; their axons project from that origin into the caudal recordings from sensory nerves as graded compound action nerve cord, with broadly distributed input and output sites in the potentials that cover a wide frequency range [about 0.1–10 Hz intervening ganglia (Brodfuehrer and Friesen, 1986a,c). Brief (Friesen, 1981; Brodfuehrer and Friesen, 1984)]. Target cells stimulation of individual trigger neurons elicits bouts of for water vibration stimulation include the CBW cells found in swimming activity, with no correlation between the duration of the anterior medial packet on the ventral side of many midbody trigger neuron activity and the ensuing swim episode ganglia (Gascoigne and McVean, 1991). (Brodfuehrer and Friesen, 1986a). Trigger neuron Tr2 is Sensory stimulation initiates swimming activity through a particularly interesting in that it acts as a toggle switch: brief cascade of interactions that eventually drive the swim oscillator stimulation can elicit swimming activity and a second stimulus, (Fig. 16). One pathway in this cascade comprises the P and N once swimming has commenced, brings that activity to an sensory neurons, which drive trigger neurons (Tr1 and Tr2) via abrupt halt (O’Gara and Friesen, 1995). Multiple, as yet monosynaptic connections (Brodfuehrer and Friesen, 1986c,e). unknown factors, determine whether swimming will occur in The somata of trigger neurons are located in the subesophageal response to any stimulus (Cellucci et al., 2000). Trigger 302 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

activity is elicited by any means, including brief trigger neuron activity, these cells depolarize and remain depolarized throughout the swim episode (Weeks and Kristan, 1978). Part of this excitation results from monosynaptic excitatory input from trigger neurons (Brodfuehrer and Friesen, 1986b), but the source of the persistent depolarization remains unidentified. Body wall stimulation, followed by Tr1 activation, elicits concurrent depolarization of all cell 204 gating neurons. These, in turn, provide nearly simultaneous excitatory drive to a subset of the oscillator INs throughout the nerve cord. (Tr2 interactions with gating cells are polysynaptic and inhibitory.) Cells 204/205, in turn, drive both INs and MNs (Weeks, 1982a,c; Nusbaum et al., 1987). Interestingly, only three identified members of the central swim oscillator receive direct input from cells 204/205: cells 115, 28, and 208. At present, the only identified target for Tr2 is cell 256, a neuron that terminates swimming activity (Taylor et al., 2003). The neurotransmitter acting at the top of the excitatory cascade is glutamate, acting through non-NMDA receptors at synapses between P cells and Tr1, between Tr1 and cell 204, and between cell 204 and cells 28, 115, and 208 (Brodfuehrer and Cohen, 1990, 1992; Thorogood and Brodfuehrer, 1995; Brodfuehrer and Thorogood, 2001). The great importance of gating neurons 204/5 in controlling swim period may be deduced from the fact that impulse frequency in these neurons, whether set Fig. 16. Cascade of information flow from mechanosensory neurons to the by the experimenter or observed during swim episodes, predicts swim-related muscles. Cell names: T, touch cell; P, pressure cell; N, nociceptive swim period with high precision (Pearce and Friesen, 1985a; cells; SMR, sensillar movement receptor; Tr1 and Tr2, trigger neurons 1 and 2; SE1, swim excitor neuron 1; SIN1, swim inhibitor neuron 1; VE and VI, ventral Debski and Friesen, 1986). Flattening of the leech prior to excitors and inhibitors; DE and DI, dorsal excitors and inhibitors. FL, flattener swimming is mediated via activation of flattener MNs by cells neuron. The small inset numbers designate the reference phase of each 204, 205, and 208 (Weeks, 1982a,b,c). oscillatory neuron. Darkly filled circles designate inhibitory neurons or those A set of serotonin-containing neurons, cells 21/61, also gate that inhibit swimming. The synaptic targets of some cells are unidentified. swimming (Nusbaum, 1986; Nusbaum and Kristan, 1986). Synaptic symbols as in Fig. 15. These neurons receive input from T, P and N cells indirectly (Gilchrist and Mesce, 1997) and make excitatory interactions neurons, which receive no feedback from the swim oscillator with the same oscillator INs as cells 204/5. Despite these neurons, may be regarded as the highest centers for the control similarities to cells 204/5, these more locally acting neurons are of swimming activity (Friesen, 1989c,d). Segmental targets of less robust in their swim initiation abilities, at least in Hirudo Tr2 have been found (Taylor et al., 2003); these neurons slow or (Nusbaum et al., 1987). stop ongoing swim motor patterns. Also located in the Control of leech swimming is obviously more complex than subesophageal ganglion, a pair of swim inhibitory INs (SIN) suggested by the linear cascade outlined in Fig. 16. For may be part of the swim-inactivating system (Brodfuehrer and example, swimming activity evoked in ventral nerve cords that Burns, 1995), whereas another set of neurons, the excitatory include the head ganglia is weaker and less regular than in SE1 cells, may act as gain control elements that determine brainless preparations, particularly in bursting recorded in the whether swimming will occur in response to some specific anterior half of the nerve cord (Brodfuehrer and Friesen, stimulus (Fig. 16; Brodfuehrer et al., 1995a). The subesopha- 1986d). Thus, the head ganglia have a marked inhibitory effect geal ganglion may also contribute directly to swim oscillations on swimming. In contrast, the tail ganglion facilitates swim via cell SRN1, a brain IN that exhibits oscillations phase-locked initiation and duration, by reversing the inhibitory action of the to the swimming rhythm and which can shift the swimming head ganglia and prolonging swim episodes (Brodfuehrer et al., phase (Brodfuehrer and Friesen, 1986d). 1993). These differing effects of the head and tail brains initiate Gating neurons, such as cell 204 and its homolog, cell 205 and reverse within seconds when impulse traffic to the brains is (Weeks, 1982a,c) occupy the third level of the swim-initiation interrupted and reinstated, hence synaptic rather than hormonal cascade. The somata of these unpaired excitatory INs are mechanisms appear to be at work here, although those that restricted to the posterior half of the nerve cord (M9–M16). Their mediate the excitatory actions of the tail ganglion are unknown. axons project to most, if not all ganglia of the nerve cord. Strong depolarization of even an individual swim gating neuron drives 3.4.5. Neuromodulatory control: serotonin and other the expression of swimming activity even in nearly isolated biogenic amines ganglia (Weeks, 1981); when repeatedly activated, swim Although neurons in the leech subesophageal ganglion exert duration does not outlast the depolarization. When swimming the highest level of control over swim initiation and W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 303 termination, the overall propensity for swimming is regulated cells, which closely neighbor the DLOS, were at first thought to by neuroactive substances, most notably by 5-HT. Leeches contain octopamine (Belanger and Orchard, 1988). Although with a high blood concentration of 5-HT swim more, and the Leydig cells are not octopaminergic or synaptically linked isolated nerve cord preparations engage in swimming activity to the DLOs (Gilchrist et al., 1995; Crisp et al., 2002), these spontaneously, when 5-HT levels are elevated (Willard, earlier studies helped to establish that octopamine is a naturally 1981). Swimming activity reaches a half-maximal level about occurring neuromodulator in the leech. Because swim- 15 min after exposure to 50 mM 5-HT and returns to control activating mechanosensory inputs (T and P cells) have been levels about 30 min after washout. The primary effect in the found to co-activate the DLOs, these octopamine and 5-HT- isolated nerve cord (without the brain) is the appearance of containing cells (Gilchrist and Mesce, 1997) may act in parallel spontaneous swim episodes, with few if any changes in cycle with the Retzius neurons and other serotonergic cells (61 and period, impulse frequencies, and the duration of episodes 21). Subsequent studies have documented that bath application (Hashemzadeh-Gargari and Friesen, 1989). Bath application of a mixture of the two amines (isolated nerve cords with of 5-HT can even induce isolated ganglia individual to brains) results in a novel non-additive suppression of express the rudiments of the swimming rhythm. These swimming, followed by a robust activation of swimming after hormonal effects, also observed when the 5-HT-containing the mixture is removed during a 30 min saline wash (Mesce Retzius cells are stimulated, standincontrasttotherapid et al., 2001). Related to the modulation of feeding behavior, 5- activation of swimming by depolarization of the other 5-HT- HT or octopamine also might be involved in the circadian containing neurons, cells 21/61 (Kristan and Nusbaum, 1983). regulation of behavior observed in a predatory leech (Angstadt In the latter case, swim initiation appears as a synaptic rather and Moore, 1997). than as a hormonal event; however, the importance of 5-HT, Further evidence that 5-HT is a critical hormone for the acting as a local hormone and neuromodulator, cannot be expression of swimming activity comes from depletion overstated. There is a negative correlation between the state of studies. Thus in isolated nerve cords acute treatment with satiety in leeches and swimming behavior—hungry leeches reserpine – which blocks the monoamine vesicle transporter – have higher levels of 5-HT and swim more readily in response eliminates all swimming activity, which is restored when 5-HT to stimulation (Young et al., 1981; Lent, 1985). Therefore it is subsequently bath-applied (Hashemzadeh-Gargari and appears likely that 5-HT levels are broadly important in Friesen, 1989). In intact animals, injection of reserpine into regulating the appetitive phase of feeding behavior in the the crop depletes all amines for many weeks. Such animals leech (Lent and Dickinson, 1984). cease normal biting behavior, their bodies become rigid, and Because focal application of 5-HT to the subesophageal sensitivity to stimulation is greatly reduced. Surprisingly, ganglion terminates swimming (Crisp and Mesce, 2003), prolonged tactile stimulation can evoke swimming, which then spatial differences in 5-HT modulation may contribute to persists far longer than in control animals (O’Gara et al., 1991). transitions from the appetitive phase of feeding (swimming) to This apparent anomaly may be a consequence of the loss the consummatory one (ingesting the blood meal), when of other amines, such as dopamine or octopamine, which are swimming is counterproductive. During the initiation of also depleted by reserpine (O’Gara et al., 1991). Because ingestion, the serotonergic Retzius neurons stop firing (Wilson dopamine has now been linked to the suppression of swimming et al., 1996; Zhang et al., 2000), in sharp contrast to an earlier (Crisp and Mesce, 2004), the reserpine-induced loss of this view in which the Retzius neurons were thought to command inhibitory factor could contribute to prolonged swimming in both phases of feeding (Lent, 1985). Although the behavioral treated animals. Finally, it was found that when 5-HT is roles of dopamine are the least understood of the amines, chronically removed from the nervous system of juveniles dopamine also may play an important role in feeding (biting) through treatment with 5,7-dihydroxytryptamine (5,7-DHT), behavior because dopamine-containing fibers have been found swimming is not expressed in the adults (Glover and Kramer, within the three accessory ganglia controlling the jaws (Crisp 1982). et al., 2002). In addition, bath application of dopamine to The swim gating neurons, cells 204/5, with their direct isolated nerve cords inhibits all ongoing and nerve-evoked involvement in swim initiation, are obvious candidates for swimming; this is correlated with the discovery that modulation by 5-HT. A comparative study of cell 204 dopaminergic neurons are directly coupled to Tr2 (Crisp and properties in normal saline and following exposure to bath- Mesce, 2004). Thus dopamine has the potential to coordinate applied 5-HT revealed that 5-HT reduces the threshold for swim biting with the termination of swimming. [As mentioned initiation via intracellular depolarization of these swim-gating earlier, Tr2 has been found to be a more potent swim terminator neurons. Moreover, elevated 5-HT converted cell 204 into a than swim activator (O’Gara and Friesen, 1995).] Octopamine, trigger cell, so that even brief stimulation of cell 204 elicited the last of the monoamines discussed here, can induce swim episodes (Fig. 17A and B; Angstadt and Friesen, 1993a). swimming when bath applied to isolated nerve cords with Concomitant with these functional alterations, cell 204 in the the brain removed (Hashemzadeh-Gargari and Friesen, 1989) presence of bath-applied 5-HT displayed enhanced postinhi- or with it attached (Crisp and Mesce, 2003). The dorsolateral bitory rebound and afterhyperpolarization, mediated by Na+- octopamine (DLO) cells have been identified as the set of dependent and Na+-independent conductances (Angstadt and segmental neurons containing and synthesizing octopamine Friesen, 1993b). The DI and VI MNs were depolarized slightly (Gilchrist et al., 1995; Crisp et al., 2002). The segmental Leydig by elevated 5-HT levels but, as for cell 204, steady state I–V 304 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

importantly from a functional viewpoint, brief pulses of injected current caused larger phase shifts when the inhibitory MNs were exposed to bath-applied 5-HT. The dynamics of the synaptic interactions of these MNs were altered by 5-HT, with increased and more rapid onset of, and recovery from, synaptic fatigue (Fig. 17C; Mangan et al., 1994b). Application of drugs that increase the intracellular concentration of cAMP mimic the effects of bath-applied 5-HT (Hashemzadeh-Gargari and Friesen, 1989), hence elevated cAMP is a likely mediator of the swim enhancing effects of 5-HT, whether through the alterations demonstrated in cell 204 and the inhibitory MNs, in oscillator INs, or in as yet unidentiﬁed swim excitor cells.

3.4.6. Role of sensory feedback In addition to the mechanosensory activity that elicits swimming, the neuronal circuit for swimming is also affected by sensory feedback during the production of swimming. Recently, the source of this sensory feedback has been described definitively as stretch receptors in the leech body wall, and the detailed function of one type of stretch receptor – those located in the ventral body wall – has been described in detail. Whole-body undulations, as seen during swimming in leeches, lamprey, or snakes is most effective when the waveform comprises about one wavelength (or slightly greater to minimize yaw; Fig. 14A; Taylor, 1952; Gray, 1958, 1968; Kristan et al., 1974a). The triumphant central oscillator theory informs us that neuronal circuits within the CNS generate the fundamental rhythms, however caveats reminding us of the importance of sensory feedback for shaping fully expressed movement rhythms have appeared repeatedly in the literature on motor systems (Kristan and Stent, 1975; Kristan and Calabrese, 1976; Pearson et al., 1983; Pearson and Ramirez, 1990). In the leech, sensory feedback plays a significant role in setting the swim cycle period, which is longer in the absence of sensory feedback (Kristan and Calabrese, 1976). More critically, intersegmental phase lags, which are about 208/ segment in swimming leeches are reduced to 8–108/segment in the isolated nerve cord (Pearce and Friesen, 1984). Feedback Fig. 17. Modification of neuronal physiology and function by 5-HT. (A and B) appears to be less critical in the expression of swimming Bath application of 5-HT transforms the swim-gating neuron, cell 204, into a behavior in lampreys and in coordination within the crayfish trigger cell in an isolated nerve cord preparation. In both panels, the upper traces swimmeret system, where the isolated CNS can generate show the amplitude and duration of intracellular currents applied to cell 204 appropriate periods and intersegmental phase lags (Walleń and (middle traces). Identical current pulses have little effect in normal saline (A), but elicit a swim episode when 5-HT is elevated (DP records in lower traces of Williams, 1984; Grillner et al., 1991; Mulloney et al., 1993; B). (C) Serotonin induces time-dependent membrane potential trajectories in Friesen and Cang, 2001). MNs. Depolarization (by current injection, not shown) of an inhibitor MN (cell Sensory feedback is, of course, a feature of all animal DI, upper traces) hyperpolarizes its postsynaptic target (the contralateral cell locomotory systems. In leeches, feedback was deduced from VI, lower traces). Bath-applied 5-HT induces presynaptic relaxation and experiments with the following results: (1) the cycle period of postsynaptic fatigue with subsequent postinhibitory rebound (parts A and B are reprinted with permission from Angstadt and Friesen, 1993a; part C is swimming leeches increases when the viscosity of the medium modified from Mangan et al., 1994a,b). is increased (Gray et al., 1938); (2) stretching the body wall in nerve cord - body wall preparations alters the intensity of MN bursts, increases cycle period, and can even terminate relationships were essentially unaltered (Mangan et al., 1994a). swimming activity (Kristan and Calabrese, 1976); (3) As observed in cell 204, 5-HT induced an enhancement of mechanical expression of body movements is essential for postinhibitory rebound and of afterhyperpolarization follow- the continuation of swimming movements in a partially ing the cessation of depolarizing current pulses. Most restrained leech (Kristan and Stent, 1975) and (4) interseg- W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 305 mental coordination between two ends of a leech with a severed and a phase of about 1408. The VSR hyperpolarizes not only nerve cord is largely unimpaired, albeit there are greater when tonic tension is increased by manually stretching the body intersegmental phase lags across the lesion site (Fig. 18A–C; wall, but also when body wall tension is increased naturally via Yu et al., 1999). This last set of experiments overturned the MN (cell VE-4) stimulation (Fig. 19A–C; Cang et al., 2001). long-held view that an intact nerve cord is essential for Finally, hyperpolarization of the VSR mimics the effects of intersegmental coordination in leech swimming (Schu¨lter, stretching the ventral body wall on impulse rates in MNs 1933; Kristan et al., 1974a). Moreover, because intersegmental (Blackshaw and Kristan, 1990). These experiments demon- phase lags generated within the nerve cord are too small, and strate that the ventral stretch receptors are tension transducers those engendered by sensory feedback alone are too large, these for longitudinal muscles. [Tactile receptors in the nerve cord experiments suggest that the shape of a swimming leech, a sheath might also contribute to sensory feedback (Smith and single cycle of a quasi-sinusoidal wave, arises from a Page, 1974)]. compromise between central and peripheral coordination (Yu Recent experiments reveal that the state of the VSR can et al., 1999). influence intersegmental phase relationships (Cang and Eight pairs of putative stretch receptors embedded in the Friesen, 2000). In these experiments, sinusoidal currents were segmental longitudinal muscles of leech body walls appear to injected into the VSR axon near a midbody ganglion of an have properties required for providing appropriate sensory isolated nerve cord preparation. Records of swimming activity, feedback during swimming. These sensory neurons have and hence phase, were obtained from dorsal posterior (DP) peripheral somata, with dendrites that insert into longitudinal nerves of this midbody ganglion and of the two adjacent ones. muscle fibers (Blackshaw and Thompson, 1988; Blackshaw, The phase of the sinusoidal current was varied over the full 3608 1993; Huang et al., 1998). These neurons respond to stretch of the swim cycle. Under these conditions, intersegmental with graded hyperpolarization, which is conveyed electro- phase lags between the ganglion of the stimulated VSR axon nically to nerve cord ganglia via huge (>20 mm in diameter; (ganglion n) and its nearest neighbors depended on the phase of Cang et al., 2001; Gerta Fleissner, personal communication) the injected current, such that at one phase (1208) the non-spiking axons. Recordings from the axon of the ventral n À (n À 1) phase lags increased by about 58 and the stretch receptor (VSR) – which innervates VLM – in nerve n À (n + 1) phase lag deceased by a similar amount. cord—body wall preparations reveals membrane potential Conversely, when the phase of injected current was set to oscillations during swim activity with amplitudes up to 10 mV 2708, the n À (n À 1) phase lags decreased by 58, and that of n À (n + 1) increased by a similar amount (Cang and Friesen, 2000). Thus, VSR activity can alter intersegmental phases in a phase-dependent manner, and hence may be critical for setting intersegmental phase lags to generate a single body wave. How is stretch receptor output conveyed to the neurons of the swim circuit? Extensive surveys with pairwise intracellular recordings revealed only weak, presumed polysynaptic interactions between the VSR and either inhibitory or excitatory MNs in the swim system. There is, however, a very strong electrical connection between the VSR and cell 33. In addition, there are inhibitory and excitatory interactions with other oscillatory INs, cells 28, 115, and 208, that are likely to be polysynaptic (Fig. 19D; Cang et al., 2001). VSR interactions with INS were functionally significant, for depolarization of the VSR greatly reduced the amplitude of swim oscillations in cells 28, 115, and 208. Also, in experiments to mimic body wall stretch, induced VSR hyperpolarization decreased the excitatory excursions in cell DE-3 and increased those of VE-4 during swimming activity expressed in isolated nerve cord preparations. The functional significance of VSR input for swim oscillations is clear: (1) the VSRs undergo rhythmic oscillations in membrane potential in preparations with the nerve cord Fig. 18. Comparison of intersegmental phase lags in normal and ‘severed nerve attached to a flap of body wall; (2) rhythmic current injection cord’ (SNC) leeches. Inset: preparations and electrode placements for intact (A) into the VSR entrains the ongoing swimming rhythm in isolated and SNC leeches (B). Traces showing swim bursts were obtained with extra- nerve cord preparations, with a phase angle that is positively cellular nerve recordings from DP(7) and DP(14). Bursting activity remains related to the frequency of current injection; (3) strong coordinated in the two halves of the leech even with the nerve cord cut. C. depolarization of the VSR shifts the swim phase (Yu and Intersegmental phase lags in three types of preparations depicted as polar plots. Numbers are means + S.D. Phase lags in intact animals lie between those Friesen, 2004). These effects of the VSR on swim expression observed in isolated nerve cords and SNC leeches. (Modified from Yu et al., must be mediated by interactions between the VSR and 1999). oscillatory INs, which then provide the pathway for effects on 306 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

the MNs, including the observed modification of intersegmental phase lags. Of the eight putative stretch receptors in midbody segments, an additional six have been identified through filling their axons with Alexa Fluor1 hydrazide dye (Fan and Friesen, 2005). Two of these have striking axon arborizations within segmental ganglia because their processes send terminals, unlike those of the VSR, also into the contralateral neuropil. Like the axon of the VSR, the giant axons of these cells do not conduct action potentials. Nevertheless, they also exhibit small, spike-like events that are generated within their central terminals. One of these axons is associated with dorsal longitudinal muscles and hence is called the dorsal stretch receptor (DSR). During fictive swimming, the DSR undergoes rhythmic oscillations that differ in phase from oscillations recorded during swimming activity in a nerve cord–body wall preparation. The interactions of the DSR and the other identified putative stretch receptors with the circuits that control swimming or other behaviors remain to be found.

3.4.7. Functional aspects of the central oscillator Questions about the origins of the oscillations that underlie leech swimming have given rise to a series of answers. One answer is that individual ganglia are capable of generating the rudiments of the swimming rhythm when either swim-gating neurons are stimulated (Weeks, 1981) or 5-HT is bath applied (Hashemzadeh-Gargari and Friesen, 1989). Neuronal circuits within these ganglia comprise a unitary oscillator rather than the paired half-centers of the vertebrate spinal cord (Brown, 1911; Friesen and Hocker, 2001), probably because of the extensive connections, electrical and synaptic, that link bilateral cells in leech ganglia (Fig. 15A). Although initial modeling of the segmental swim circuit yielded cycle periods that were unrealistically short, more recent studies with analog neuromimes support the view that the identified interactions could account for these rudimentary oscillations (Wolpert and Friesen, 2000; Wolpert et al., 2000). In fact, although highly sensitive to parameter alterations, a computer model based on the identified intrasegmental connections generated membrane potential waveforms mimicking those observed in extended nerve cords (Taylor et al., 2000). The ability of individual ganglia to generate swim-like oscillations is not uniformly distributed along the nerve cord. The strongest swimming activity, either in isolated ganglia, or pairs of ganglia, occurs in the middle third of the nerve cord, from about M7–M12 (Pearce and Friesen, 1985a; Hocker et al., 2000), although this activity is never as robust as that of nerve cords extending from M2 to the tail brain. Only weak, erratic swim-like bursting occurs in individual isolated anterior Fig. 19. Responses of the ventral stretch receptor (VSR) in a nerve cord–body ganglia, M2–M5; isolated ganglia posterior to M12 appear wall preparation (inset). (A) Depolarization of cell VE-4 by current injection unable to generate even such swim rudiments. For producing (upper trace, intracellular record) causes increased tension in VLM (lower swim oscillations, the functionality of nerve cord ganglia trace) that in turn induces hyperpolarization in the VSR (middle trace, intracellular record). (B) Relationships between MN activity and tension in ventral ming rhythm (lower trace). (D) Summary of interactions between the VSR and longitudinal muscle [tension (R,11)]. Rhythmic MN activity in cell 8 (intra- swim-related neurons. Of the several interactions between the VSR and cell 33, cellular record, upper trace; extracellular record from a DP nerve, bottom trace), only the electronic interaction between the VSR and cell 33 is likely to be causes rhythmic tension alterations in the body wall (middle trace). (C) VSR monosynaptic. Symbols as in previous figures (redrawn from Cang and Friesen, membrane potential oscillations (upper trace) are phase-locked to the swim- 2000 and Cang et al., 2001). W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 307 appears like an inverted U, low at the two ends and highest in by the identified intersegmental interactions (Zheng et al., the center. In addition, the period of swimming expressed in 2004). completely isolated ganglia, or short chains, is U-shaped; the A recent refinement and extension of the coupled phase cycle period is shortest in the center of the cord and longer at oscillator model was used to study further the role of either end (Hocker et al., 2000). Based on the weakness of interactions within the nerve cord and between the central oscillations generated in isolated ganglia, the oscillator that circuit and sensory feedback in intersegmental coordination generates swim oscillations should not be viewed as a chain of (Cang and Friesen, 2002). This model is more realistic than its weakly coupled oscillators. Instead, the system relies on strong, immediate predecessor because channels mimicking interseg- extensive intersegmental interconnections to create robust mental interactions were limited to those identified experi- oscillations from weak subunits. As stated earlier, the INs mentally (Fig. 20A). Moreover, the approach was to constrain project at least six segments in each direction (Poon et al., 1978; the model at each step in its construction by experimental Weeks, 1982b; Pearce and Friesen, 1985b; Friesen and Hocker, results followed by tests in which model output was compared 2001). Thus in the mid nerve cord, 13 oscillator subunits are to novel experimental data. As in the earlier phase model, phase connected by direct synaptic interactions – these might be shifts of segmental unit oscillators by intersegmental inputs considered to be the ‘‘unit oscillator’’. As described above, were specified by phase response curves (PRCs; Fig. 20B). sensory feedback contributes to intersegmental phase and to These curves were shaped through the use of subsidiary cycle period. This is shown most dramatically in reduced modeling. Relative PRC amplitudes were set to ensure caudal preparations of the leech, which undergo vigorous, symmetric intersegmental coupling strengths. The extended coordinated swimming movements even though the isolated model also incorporated stretch receptors, constrained by the posterior cord does not support swim oscillations (Hocker et al., phase-shift effect of the VSR described above (Cang and 2000). However, even single segments, if isolated chronically Friesen, 2000). Simulations with this extended model from the nerve cord, can acquire swimming activity (Kristan reproduced a remarkable replica of swimming activity in and Guthrie, 1977). leeches, including: (1) intersegmental phase lags of 8–108/ Intersegmental phase relationships are determined by period segment that exhibited a small positive period dependence gradients in unit oscillators, coupling strength, and asymmetry (Kristan and Calabrese, 1976; Pearce and Friesen, 1984); (2) in intersegmental interactions (Skinner and Mulloney, 1998a). increases in phase lag when the length of the nerve cord is Asymmetry in the functional strength of intersegmental reduced (Pearce and Friesen, 1985b) and (3) intersegmental coupling was assessed in the leech through experiments on coordination with increased phase lags when the nerve cord is ‘Z-cut’ preparations, in which a given midbody ganglion was severed in mid body (Yu et al., 1999). Thus the known topology driven by input of differing cycle periods through ascending of the swim circuit can account for intersegmental coordination connections on one side and descending on the other by even though not all neurons or interactions in this circuit are selective hemi-lesions of the ventral nerve cord. Correlation identified. and spectral density analyses of the rhythmicity in this Students of rhythmic motor systems are deeply interested in preparation demonstrated that effectiveness of ascending and the mechanisms that generate the cycle period (Friesen and descending interactions in the middle of the nerve cord are Stent, 1978). The initial mechanism proposed for the leech about equal (Friesen and Hocker, 2001). swim circuits, based on the four pairs of oscillatory INs then The anterior-to-posterior phase lags observed during identified (Friesen et al., 1976, 1978), focused on intra- and swimming in the leech nerve cord must arise from asymmetry intersegmental interactions that form closed loops of inhibition in the synaptic interactions between INs because: (1) coupling (recurrent cyclic inhibition, RCI; Sze´kely, 1964). Such strength among midbody ganglia is approximately symmetrical inhibitory loops can generate stable oscillations with multiple and (2) the U-shaped period gradient tends to generate intrasegmental phases and, concomitantly, intersegmental nonuniform intersegmental phase lags. Early modeling studies, phase lags. The validity of this model was supported by relying on graphical analysis or modeling with electronic graphical analysis and neuromime simulations, which yielded neuromimes suggested that the asymmetry in the identified, cycle periods and both intra- and intersegmental phase intersegmental interactions could account, at least qualitatively, relationships similar to those observed in isolated nerve cord for the observed phase lags (Friesen and Stent, 1977; Friesen preparations (Friesen and Stent, 1977). Subsequent experi- et al., 1978; Friesen and Pearce, 1993). A subsequent computer ments, which demonstrated that an isolated ganglion could study incorporated the swim circuits as a chain of phase generate swim-like oscillations (Weeks, 1981; Hashemzadeh- oscillators coupled by multiple signal channels to mimic Gargari and Friesen, 1989; Hocker et al., 2000), cast the RCI the intersegmental connections among ganglia. This model model into doubt and supported reciprocal inhibition (RI), successfully accounted quantitatively for normal intersegmen- which generates antiphasic oscillations (Brown, 1911; Friesen, tal phase delays, for increases in phase lag when the number of 1989d; Brodfuehrer et al., 1995b). The RI model is currently connected ganglia is reduced, and for alterations in phase lags invoked to account for locomotory oscillations in many species induced by lesions and temperature manipulations (Pearce and (Friesen, 1994), including lamprey. Perhaps the question of Friesen, 1988). Mostly recently, a more biophysical, systems whether RCI or RI generates the swimming rhythm in leeches approach to modeling intersegmental coordination demon- poses a false dichotomy. Weak, long-period oscillations with strated again that the observed phase relationships are predicted inappropriate phase relationships (Weeks, 1981; Hashemza- 308 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

lack identified inputs. Additionally, intersegmental targets for cells115and60remainunknown(Fig. 15). A scanning technique, in which groups of neurons are depolarized by focal increases in extracellular K+ concentrations (Friesen and Brodfuehrer, 1984), cutting-edge optical imaging (Cacciatore et al., 1999; Taylor et al., 2003; Briggman et al., 2005), and intracellular injections of dual tracer dyes (Alexa and Neurobiotin; Fan et al., 2005; Fan and Friesen, 2005) are proving helpful in locating the missing neurons. Detailed information on the strength of synaptic interactions and the biophysics of INs is needed, particularly for the construction of detailed biophysical models. Also, the physiology and links with the swim oscillator of only one segmental stretch receptor has been studied thoroughly; many more mapping studies are need to determine how the DSR and other stretch receptors provide sensory feedback to control cycle period and intersegmental phase lags. Moreover, there is limited knowledge of the biomechanics of DLM activation— contraction strengths, rates of contraction, and length-tension relationships (Miller, 1975; Mason and Kristan, 1982; Wilson et al., 1996a). At the functional level, little is known about the control of swimming in leeches. For example, how does the brief output of trigger neurons lead to prolonged excitation in swim-gating neurons (Brodfuehrer and Thorogood, 2001)? Finally, the relative importance of central oscillators versus sensory-central loops in setting cycle period and intersegmental phase relationships in intact leeches or, for that matter, Fig. 20. Modeling and overview of the leech swim system. (A) Reduced in other animals (Pearson and Ramirez, 1997) remains largely circuit diagram employed to model intersegmental coordination. Cell numbers unknown. Even prior to more complete information, it is not and phase lags as in Fig. 15. (B) Functional model: individual ganglia are premature to develop computer models of leech swimming modeled as phase oscillators that interact via signals transmitted during the like those already available for lamprey (Ekeberg and active sector of any given neuron (shaded arcs). The signals advance or delay Grillner, 1999), which encompass the central neuronal the phase of the local oscillator in accordance with physiologically realistic oscillator, muscle mechanics, and sensory feedback. Such a phase response curves. This simple model can account well for coordination in isolated nerve cords. (C) Schematic model of the complete swim system. complete model would provide a quantitative description of Tension in the body wall, caused both by MN output and the environment, the leech swimming system presented schematically in provides sensory feedback to the central swim oscillator circuits, thereby Fig. 20C. modulating both cycle period and phase in the intact leech. Symbols: SR, stretch receptor; , central oscillator (with intra- and intersegmental compo- 3.5. Vermiform crawling nents). With the addition of sensory feedback and mechanical coupling between segments, the phase model generates a realistic simulation of swimming in the intact leech. Parts A and B are modified from Cang and Friesen, 3.5.1. Behavior 2002;partCismodifiedfromYu, 2001. Vermiform crawling is a rhythmic behavior with two major phases – extension and contraction – that are typically coordinated with attachment and release of the front and back deh-Gargari and Friesen, 1989) might be generated by RI, suckers (Fig. 21). A single cycle starts with a leech fully whereas the robust multiphasic oscillations recorded from contracted and with both suckers attached. The front sucker preparations comprising multiple segments of the nerve cord, releases and the animal starts to extend by contracting circular might be the result of known and other, as yet unidentified, muscles. The extension starts at the front end and moves intersegmental inhibitory loops. Given the weakness of progressively more posterior. When the body is fully extended, segmental oscillators and the strength and broad extent of the front sucker reattaches and the front end begins to contract. intersegmental interactions, the conclusion that intersegmental The contraction wave progresses posteriorly, putting tension on interactions are essential components of the swim oscillator the posterior end of the body. When the contraction wave seems inescapable. reaches about two-thirds of the way to the tail end, the back sucker actively releases. When the contraction wave reaches the 3.4.8. Gaps in our current knowledge posterior end, the rear sucker reattaches, thus completing one One obvious gap in the description of the leech swim circuit step cycle. is that additional oscillatory INs and their interactions await In intact leeches, a step takes 3–10 s (Stern-Tomlinson et al., identification. Two of the identified neurons – cells 60 and 33 – 1986), although it can take up to 20 s in dissected leeches W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 309

(Baader and Kristan, 1995; Cacciatore et al., 2000). This type worming while recording from its nervous system. (Please of crawl step is one of two distinct modes of crawling note that all further references to ‘‘crawling’’ in this review displayedbyleeches(Stern-Tomlinson et al., 1986). The refer to vermiform crawling.) other form, called ‘‘inch-worm crawl’’ or ‘‘looping’’ is faster Elongation of a segment is produced by a contraction of (1–3 s) and the suckers are brought adjacent to one another at the circular muscles in a segment, and contraction is produced the end of contraction. Inch-worm crawling is seen only when by the co-contraction of all the longitudinal muscles in a leech is under water (it tends to fall on its side when it that segment. For both elongation and contraction, the attempts to inch-worm in air), and when it is strongly behavior begins at the front end and moves smoothly stimulated (e.g. when pinched or very hungry). Inch-worming posteriorly as a wave. Hence, in a single segment, crawling is more efﬁcient because the leech progresses by nearly a appears as an alternation of bursting in the circular and fully-extended body length with each inch-worm step, and longitudinal MNs. This pattern repeats itself, with a delay, in much less (the body length at full extension minus the length successively more posterior segments, thereby producing the at full contraction) during a vermiform step. Most studies intersegmental progression of the elongation and contraction have focused on vermiform steps, because it is difﬁcult to waves. This pattern has been recorded in semi-intact (Baader provide a leech with the appropriate conditions for inch- and Kristan, 1992), and isolated nervous systems (Eisenhart et al., 2000).

3.5.2. Kinematics To characterize the movement patterns during crawling, pieces of white sutures were knotted onto the skin of many segments and the animals were video-taped as they crawled (Fig. 22A; Stern-Tomlinson et al., 1986; Cacciatore et al., 2000). These analyses showed that the elongation phase was significantly more prolonged in slower crawl cycles, and that the contraction phase was short and varied significantly less with step cycle duration (Fig. 22B). In addition to variations in the elongation phase, step cycles also got longer because the delay from rear sucker placement and front sucker release increased with step cycle period. The average contraction and elongation duration diminishes as the wave gets to the posterior end of the animal (Fig. 22C(i)), reflecting the fact that the front of the animal stays contracted (or elongated) as the wave passes to the posterior end.

3.5.3. Motor neuron activity The motor neuronal activity recorded in both semi-intact and isolated nerve cord preparations is fully consonant with the kinematic data; circular muscle MNs produce impulse bursts in each segment that alternate with bursts in MNs that innervate longitudinal muscle, and the bursts appear progressively later in more posterior segments in both semi-intact preparations (Fig. 22C(ii)) and isolated nerve cords (Fig. 22C(iii)). Although the cycle periods in these two preparations are much longer than in intact animals, the coordination is the same in all three preparations; data for intersegmental travel time (ISTT) plotted against step duration for both dissected preparations fall on the regression lines calculated from the data for the intact animals (Fig. 22B; Cacciatore et al., 2000).

3.5.4. Sensory input Although the basic crawling motor pattern produced in the Fig. 21. Schematic diagram of crawling behavior. This is a more detailed isolated nervous system has the essential features seen in more drawing of vermiform crawling than was provided in Fig. 3D. This ﬁgure intact preparations, there are several differences: emphasizes two features: (1) both the elongation and the contraction move slowly from front-to-back along the body and (2) there is a phase in the middle of the step when the back end of the animal is elongating while the front end is A. The period is much longer than in intact animals (as long as beginning to contract (from Cacciatore et al., 2000). 30 s); 310 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

Fig. 22. Kinematics of crawling. (A) Segmental progression of the crawling step. Threads were sewn onto the skin of a leech (image on the left), which was video- taped as it crawled. Distances were measured from the head end to segment 2 (H-2), from segment 18 to the tail end (18-T), and between all other adjacent segments. Three plots of length vs. time are shown for the front end (H-2), back end (18-T), and one segment in the middle (10). Both the elongations and the contractions move from front-to-back, with contractions appearing to move through the animal faster than elongations. (B) Plot of the rate at which elongations and contractions move through the animal, the intersegmental travel time (ISTT) as a function of the duration of step cycles. The lines are the linear regressions for the elongation (open circles) and contractions (closed circles). The slopes of the two lines are signiﬁcantly different, showing that elongations move through the leech’s body more slowly W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 311

B. There is much more variability in the period than in either fully intact or semi-intact preparations and C. The duration of the bursts is greater than in semi-intact preparations. The bursts are so long that the circular motor bursts (elongation) overlap with the longitudinal motor bursts (contraction), particularly in the rear end of the nerve cord because the contractions occur relatively closer to the elongations in the rear of the animal (Fig. 22C).

These differences are likely due to the lack of sensory input in the isolated nerve cord, which in the intact animal supplies tonic excitation as well as cycle-by-cycle feedback, and regularizes and shortens the motor bursts. The source of the sensory feedback is not known. In addition to the effects of sensory input on the basic motor pattern, other, mostly anecdotal observations suggest that crawling is a highly variable behavior. For instance, mechanosensory stimuli delivered to the rear of the animal during Fig. 23. Neuronal model of crawling. (A) The simplest model that produced a extension will often cause it to release its rear sucker and swim crawling step. The connections within each segment produce an elongation followed by a contraction (represented as an E/C unit) when the front segment is away. In fact, leeches sometimes make this transition from provided with tonic excitatory input (is). Positive feedback is needed in each E/ elongation to swimming spontaneously, with no obvious C unit to match the observation that the MNs fire at a frequency independent of stimulation of any sort. Identical stimuli delivered to a the step duration. Shown to the right are the firings of just the circular MNs contracting leech will speed up the contraction, but never (elongation units) in 10 of the segments from front-to-back. Note that increasing evoke swimming. (In fact, it seems impossible to elicit the strength of is (the serial input) increases the MN firing rate slightly, but did not change the rate of progression along the cord; i.e. the ISTT is relatively swimming in a leech with its front sucker attached.) constant. (B) The same circuit, with the addition of a parallel input, Ip, to each segmental E/C unit. With varying strength of Ip, the ISTT varies over a range 3.5.5. Models comparable to that seen in the intact animal (figures from Cacciatore et al., Two kinds of models have been proposed to explain leech 2000). crawling. The first is a neuronal model that simulates the MN patterns along the nerve cord, and the second is a biomechanical model that explains how the motor patterns lesions of nerves and connectives (Stern-Tomlinson et al., 1986; produce movements. Baader and Kristan, 1995; Cacciatore et al., 2000), but it awaits In the neuronal model (Fig. 23), there is a single oscillator at identification of any of the elements of the pattern generators. the anterior end of the nerve cord that drives a motor controller A biomechanical model of leech movements has been for elongation, and a second one for contraction in the first body constructed (Skierczynski et al., 1996) based upon the anatomy segment. Each segment has its own set of elongation and of the leech, passive mechanical properties of the body wall contraction motor controllers that are driven by the homologous (Wilson et al., 1996a), contractile properties of the muscles controller in the previous segment. To produce a realistic motor (Mason and Kristan, 1982; Wilson et al., 1996a), and pressures output, the motor controllers must have positive feedback measured in the body lumen as leeches crawl, swim, and within a segment; to produce the full range of cycle periods shorten (Wilson et al., 1996b). The model incorporates three observed, there must be parallel input to every segment that muscle layers: longitudinal, circular, and dorsoventral (flat- provides the same level of tonic input to all segments. The tener). These features are shown in Fig. 24A. The model magnitude of this tonic input controls the cycle period (the assumes that each segment maintains a constant volume at all greater the tonic input, the shorter the cycle period), although times, and that MNs cause muscle shortening in only their own positive feedback is required for the tonic input to exert its segment. Because of this geometry, longitudinal muscles effect. Without positive feedback, increasing the tonic input has (which shorten a segment) are antagonists to both the circular very little effect on the period. This model is constrained by a and the flattener muscles (both of which elongate a segment by large set of kinematic data and recordings from MNs in semi- decreasing its cross-sectional area, although they produce very intact (Baader and Kristan, 1995) and isolated nerve cords different shapes). To produce movements, bursts of MN spikes (Eisenhart et al., 2000), including the effects of a variety of cause muscle tension and shortening. Both are important: a low than contractions. (C) Summary of the MN activity patterns measured from intact animals (i), semi-intact preparations (ii), and isolated nerve cord preparations (iii). Open boxes show the average time spent in contraction and shaded boxes show the average time spent in contraction. Data for the intact animals (i) were estimated from the durations of the contractions and relaxations of the individual segments, from data like those in A. For semi-intact (ii) and isolated nerve cord (iii) preparations, durations of bursts in longitudinal (contraction) and circular (elongation) MNs were collected. Two features stand out: (1) the more intact the preparation, the shorter the step cycle duration, and (2) contractions and elongations were distinct in intact and semi-intact animals, whereas they overlapped significantly in isolated nerve cord preparations (from Cacciatore et al., 2000). 312 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

energy was calculated (Skierczynski et al., 1996). Sixteen segments were modeled (14 midbody segments were identical; the most anterior three segments were lumped with the head into one larger segment, and the most posterior two segments were lumped with the tail); the shape of the model approximates a leech fairly well (Skierczynski et al., 1996; Kristan et al., 2000). When the motor bursts estimated from the kinematics were used, the movements looked very much like a crawling leech (Fig. 24B). This is not surprising because the model was tuned to produce these movements. What was surprising was that although the pressures generated by the model were in no way constrained, they produced amplitudes that were very close to those measured in a crawling leech. In addition, the motor bursts recorded in semi-intact animals produced a slower and less efficient crawl, but the modeled animal appeared to be crawling. When motor bursts from the isolated nerve cord were used to drive the model, however, the behavior was very wrong in at least two ways: (1) the internal pressures in the model were much larger than any ever recorded from a real leech; (2) parts of the animal were contracting while other parts were elongating, so that the crawl steps were extremely inefficient—very little progress was made with each step. These results suggest why MN burst patterns during crawling closely match the biomechanics in which they operate. For instance, the large pressures seen in (1) occur because circular MNs overlap somewhat with longitudinal bursts, a condition never seen in an intact or semi-intact animal. Fig. 24. Biomechanical model of crawling. (A) Features used in the model of a single segment. The cross-section of the leech is ovoidal, with each of the four This suggests that sensory feedback can be used to tune a CPG semi-axes of the ellipse (a1, a2, b1, b2) independently variable, determined by to work in a biomechanically efficient range of its capability. the circular muscle activity in each of the four quadrants. The length (L)ofeach Other biomechanical models have been used to simulate segment is controlled directly by the activity of four longitudinal muscle bands leech crawling (Wadepuhl and Beyn, 1989; Alscher and Beyn, attached to the top, bottom, and sides of the ellipses. (B) Extension phases of 1998) and swimming (Jordan, 1998). These models use very the crawling movements predicted by a 16-segment model of the leech. The data used to drive the segments are those measured from kinematics (i) and simplified geometries, but more sophisticated mathematics for from motor neuronal bursts in semi-intact (ii) and isolated nerve cord pre- the crawling model (Alscher and Beyn, 1998), and a realistic parations (iii), as shown in Fig. 22C. Three frames are shown for each model of animal/substrate coupling in the swimming model condition, representing the beginning of the extension (top image), the midst (Jordan, 1998). All of these models will be aided by more of the elongation (middle image), and the end of the elongation (bottom information about the neuronal circuits underlying the image). The times to achieve these states varied in the three cases (from Kristan et al., 2000). behaviors.

3.5.6. Initiation of crawling level of sustained tension in antagonistic muscles is required to Intracellular activation of several identified neurons in the produce a rigidity that serves as a hydroskeleton for muscles to anterior brain of a leech can elicit overt behavior in semi-intact produce movements. (Contraction of one muscle in an and fictive behavior in isolated nerve cord preparations. otherwise flaccid segment produces very little movement of Examples include swim initiation by trigger neurons and SE anything but itself; if the segment is rigid, a single muscle can cells. Another such neuron, cell R3b1, which is located in the move the whole segment). third neuromere of the subesophageal ganglion, can initiate the The biomechanical model was applied first to crawling. crawling motor pattern in both semi-intact and isolated nerve Motor neuronal bursts from isolated nerve cords (Eisenhart cord preparations (Esch et al., 2002). It is effectively a ‘‘crawl et al., 2000) and from semi-intact preparations (Baader and gating’’ neuron because crawling stops when the depolarization Kristan, 1992) were used to make estimates of the motor bursts is released. During a crawl episode initiated in another way (e.g. from the kinematic studies. In each time bin (corresponding to by stimulation of the posterior end), R3b1 oscillates above its 0.6 s), the firing of each MN excited its appropriate muscle, resting potential. Surprisingly, depolarizing R3b1 can also which generated tension until that segment achieved its steady- produce swimming, or a combination of swimming and state shape (Mason and Kristan, 1982). The only mechanical crawling. During the latter, hybrid behavior, swimming bursts interaction among segments was imposed by the constraint that occur during the extension phase of crawling; this odd behavior adjacent segments share a cross-sectional face. To find a steady- is sometimes seen even in intact animals when they are in very state shape for the whole animal, the minimum total potential shallow water. W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 313

3.5.7. Gaps in our current knowledge Crawling behavior is well characterized kinematically, but little is known about the neuronal basis of its CPG. Because the crawling pattern can be elicited in isolated nerve cords (Eisenhart et al., 2000; Briggman et al., 2005), identifying the CPG should be a relatively straight-forward task. The neuronal model (Cacciatore et al., 2000) suggests a simple explanation for intersegmental coordination, but is agnostic about the location and nature of the underlying neuronal circuits. An appealing feature of the isolated nerve cord preparation for studying crawling is the possibility of studying the context- dependence of behavioral choice. For example, in both intact animals and isolated nerve cords, appropriate stimulation reliably triggers swimming, whereas the same stimuli delivered during the contraction phase of crawling speeds up the contraction but never leads to swimming (Eisenhart et al., 2000).

3.6. Feeding

All leech species feed by ingesting prey or blood through their anterior sucker (Sawyer, 1986). The anterior sucker is attached to the host or prey, and a muscular pharynx just posterior to the esophagus produces suction by contraction of extrinsic muscles that pull open the lumen of the pharynx. A major distinction among leech families is whether they ingest their prey whole or suck its blood or other internal fluids. The medicinal leech attaches its front only to hosts much larger than itself, rasps through the host’s skin with eversible jaws, and sucks the blood that oozes from the tripartite wound. Anticoagulants and anti-platelet substances are exuded into the wound to keep blood oozing from the cut during the 10– 30 min that it takes a leech to complete feeding. In fact, blood continues to ooze from the leech bite for several hours after the leech completes its meal. This ability to remove quantities of Fig. 25. Feeding behavior. (A) Measuring muscle activity in seven locations along the leech as it feeds. The drawing shows a leech attached to a membrane blood was the initial appeal of leeches medicinally (Payton, stretched over a tube containing blood. Bipolar EMG electrodes have been 1981). They are used in modern medicine to keep blood flowing inserted into the longitudinal muscles at seven locations along the leech’s body, through surgically replaced appendages until their venous in segments 9–15. (B) EMG recordings during peristaltic movements from front supply can be re-established, usually after 3–4 days (Whitaker to back. The large gray arrows indicate the direction of the peristalsis. (C) et al., 2004). Recordings in the same animal when the peristalsis initially progressed back-to- front, then reversed about midway through the recording to go front-to-back for two cycles, only to reverse again near the end of the recording (from Wilson and 3.6.1. Behavior Kleinhaus, 2000). While medicinal leeches feed, they produce slow peristaltic movements (Fig. 25), both front-to-back and back-to-front (Wilson and Kleinhaus, 2000). These movements serve to presumably because gut filling is part of the feedback propel the blood rearward into the gut and lateral gut pouches mechanism to terminate swimming (Lent and Dickinson, present in segments 7–13. Blood can be stored in these pouches 1987). for several months without decaying. Ultimately, the blood is For many weeks after feeding, leeches are sluggish and moved into the intestine, where commensal bacteria help to passive. They tend to hide and, when disturbed, withdraw rather digest the red blood cells (Braschler et al., 2003). While they than advance or locomote (Lent and Dickinson, 1984; Groome are feeding, leeches are unresponsive even to very strong et al., 1993). Hungry leeches, on the other hand, are found near mechanosensory stimulation that would normally elicit a the surface of a pond, are very active, and tend to move toward a vigorous response (Misell et al., 1998). In fact, leeches can be disturbance rather than to withdraw from it (Sawyer, 1981, cut open while they are feeding and will continue to eat (Lent, 1986; Lent et al., 1988). In the lab, hungry leeches orient toward 1985). Such half leeches, or semi-intact leeches with only the the source of surface water waves and then swim towards that head and a few anterior segments intact (Wilson et al., 1996; source (Young et al., 1981). This behavior can be elicited by the Zhang et al., 2000), will feed for longer times than normal, water waves in the dark and or visually by shadow waves 314 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 without any actual water movements (Carlton and McVean, variety of segments, and the waves can move either forward or 1993). Hence, leeches use both mechanosensory and visual backward. The nature of the pattern generator is not known. signals to orient toward the source of waves in the water. Saliva is released from the giant (greater that 200 mmin Leeches bite surfaces that are warm and/or have appropriate diameter) salivary gland cells. These cells produce salivary chemical stimuli. For instance, they will bite through warmed secretions into a tubular system that opens into the esophagus parafilm (Dickinson and Lent, 1984), and they will bite through near the mouth opening (Lent et al., 1989). These glandular a room-temperature sausage casing filled with blood or a blood cells produce very broad, Ca2+-dependent action potentials surrogate (Galun and Kindler, 1966). Thermal receptors have (Marshall and Lent, 1984). These cells are activated by adding not yet been identified in leeches, but a fair amount is known 5-HT to the bathing solution or by stimulating serotonergic about the chemoreceptors. neurons (Marshall and Lent, 1988). Hence, it is possible that the release of saliva – with its anticoagulants, anti-platelet 3.6.2. Chemosensation substances, and local anesthetic – is produced by 5-HT Galun and collaborators were the first to study chemosensa- released into the leech’s bloodstream by the same sensory cues tion, by testing the attraction of leeches to various organic that trigger the start of biting. substances (Galun and Kindler, 1966). They found that leeches The duration of a feeding episode is affected by a variety of would feed on several sugars and amino acids, particularly influences. Inflating the body with blood, saline, or even air can glucose, galactose and arginine, but only if the solution also terminate an episode prematurely, and removing the ingested contained Na+ (Galun, 1975; Elliott, 1986). The site of these blood can greatly prolong an episode (Lent and Dickinson, receptors was localized to sensory placodes on the upper lip of 1987). This suggests that stretch receptors (in the alimentary the anterior sucker (Elliott, 1987). Receptors send their axons tract or even in the body wall) help to stop feeding. There is no along particular nerves into the anterior brain (Perruccio and evidence that the chemical nature of the ingested liquid Kleinhaus, 1996). These receptors respond to the same influences the duration of feeding: leeches ingest blood and substances that leeches avidly feed upon (Li et al., 2001). arginine/NaCl solutions indistinguishably. However, there does Interestingly, both the basal level of chemosensory activity and appear to be a chemoreceptor in the gut responsive to noxious their evoked responses are decreased by adding quinidine or stimuli because leeches quickly terminated a feeding episode denatonium to the mix, substances that inhibit feeding when quinidine was added to the ingested solution (Kornreich responses (Kornreich and Kleinhaus, 1999). and Kleinhaus, 1999).

3.6.3. Motor patterns 3.6.4. Regulation and plasticity As leeches prepare to feed, they attach their anterior suckers The close association of 5-HT with feeding has suggested to the prey and elongate and stiffen the head region (Fig. 25A; that 5-HT, acting either as a hormone or as a neurotransmitter, Wilson and Kleinhaus, 2000). The rasping jaws are then everted may be the signal that regulates feeding: it builds up as the through the open mouth with the rasping teeth pressed against animal gets hungrier and triggers both food-seeking behaviors the substrate to be cut (usually the skin of the host animal), and and the onset of ingestion (Lent, 1985; Lent and Dickinson, the teeth are pulled back and forth (Dickinson and Lent, 1984). 1984, 1988). The localization of 5-HT in the nervous system As fluid flows through the cut, the pharynx starts to contract and elsewhere presents a complex picture (Lent et al., 1991; rhythmically to extract the fluid and move it into the gut. As Groome et al., 1993), and the relationship of serotonergic fluid collects in the gut, it is moved around by peristaltic neurons to feeding is complicated at best (Groome et al., 1995; movements of the body (Fig. 25). Goldburt et al., 1994; Wilson et al., 1996). A carnivorous leech, Nothing is known about the control of head positioning or Haemopis marmarota, has been classically conditioned to the rasping movements, however, dopamine may play a role avoid one of two kinds of meat (chicken versus liver) by pairing because the accessory ganglia and musculature associated with either of them with quinine and an aversive stimulus (Karrer the three jaws is richly innervated by dopaminergic fibers and Sahley, 1988). Because of the long times between meals in (Crisp et al., 2002). Movements can be elicited in the isolated sanguivorous leeches, similar experiments have not been pharynx by a variety of transmitters and modulatory substances, attempted in Hirudo. including 5-HT (Lent et al., 1989; O’Gara et al., 1999b), ACh (O’Gara et al., 1999a), and FMRFamide (Li and Calabrese, 3.6.5. Gaps in our knowledge 1987; O’Gara et al., 2000). Hence, it is possible that there is no As for crawling behavior, very little is known about the neuronal pattern generator for pharyngeal pumping; instead, CPGs for either food ingestion or peristalsis. To date, feeding this component of feeding might be activated by neuromodu- activity has not been described in an isolated nerve cord lators released onto the pharyngeal muscles directly. The preparation, making identification of the relevant neuronal peristaltic rhythm that moves blood around inside the gut is circuits nearly impossible. It might, however, be possible to expressed by muscles in the body wall that can be recorded with locate the peristalsis CPG in a semi-intact preparation EMG electrodes placed into the body wall of a feeding leech consisting of an intact front third of the animal, with much (Wilson and Kleinhaus, 2000). (The esophagus is cannulated to of the rear two-thirds dissected away to expose the nerve cord collect the blood or other fluid that is being pumped into the (Zhang et al., 2000). Because neuronal responses to food can be alimentary tract.) The peristaltic waves can originate in a detected in preparations of the dorsal lip attached to an W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 315 otherwise isolated nerve cord (P.D. Brodfuehrer, personal communication), it may even be feasible to identify INs related to feeding behavior in the brain.

3.7. Interactions among behaviors

Local bending occurs during crawling (Cacciatore et al., 2000). This bending is probably used to avoid rubbing the skin on foliage and rocks as the animal crawls along the bottom of a pond. Local bending does not appear to occur during swimming (Kristan and Stent, 1975). In part, this may be because the body wall is held too rigid during swimming to see indentations caused by light touch, or because some of the LBIs are also CPG neurons for swimming (Lockery and Kristan, 1990b). Local bending also cannot be elicited during feeding, nor can any other behavior (Misell et al., 1998). It appears that the act of Fig. 26. Interactions between the circuits underlying shortening and swimming feeding turns off all other behaviors, including any that can be behaviors. Turning on shortening, by activating T and P mechanoreceptors near elicited by even very strong somatosensory stimulation. the front of the animal turns off swimming. The connections from the pathway Shortening always predominates over swimming, whether that triggers shortening are shown. Surprisingly, most of the connections are excitatory. The letters and numbers inside circles represent the identities of the the animal is at rest (and stimuli that would produce shortening neurons (e.g. cells 204, 61, and 21 are individual segmental neurons) or cell and swimming are presented simultaneously) or is in the midst types: oscillator neurons are of three varieties – 40, 150 and 2408 – based upon of swimming (Shaw and Kristan, 1995). In part, this inhibition the phase of their oscillation, and the MNs are of four sorts (DE, VE, DI, and VI) of swimming results from a strong inhibition of cell 204, one of as explained previously. The shaded boxes enclose neurons with similar the most effective swim gating INs (Shaw and Kristan, 1997). A function. surprising result of these studies was that other swim gating INs, as well as both types of swim trigger INs tested, were actually excited by stimuli that elicited shortening (Fig. 26). stimulating somewhat away from the posterior end produces This means that several INs that elicit swimming when swimming (Kristan and Calabrese, 1976). Stimulating inter- stimulated individually are excited when shortening is elicited. mediate segments sometimes elicits swimming and other times Minimally, this implies that decision-making INs are not crawling. Recording the activity of many neurons (up to 150) dedicated to particular behaviors, but are instead multi- with voltage-sensitive dyes revealed a small number of neurons functional; i.e. they are active during two or more different whose membrane potential trajectories predicted which behaviors. This result has spawned the hypothesis that such co- behavior would be selected (Briggman et al., 2005). activation is a necessary part of decision-making, so that all Depolarizing and hyperpolarizing one neuron, cell 208, biased behavioral decisions are made by a combinatorial code of such the choice toward either swimming or crawling. Interestingly, multiplexed INs (Fig. 27). cell 208 had previously been shown to be a member of the CPG In line with this interpretation, cell R3b1, produces either swimming or crawling in a context-dependent manner, and is inhibited when shortening is elicited (Fig. 28; Esch et al., 2002; Esch and Kristan, 2002). When R3b1 is stimulated in semi- intact leeches, the expression of behavior depends on the depth of the water surrounding the intact part of the leech: in deep water, it swims, whereas in shallow water or on dry land, it crawls (Fig. 28). Thus another effect of sensory feedback on crawling is the determination of behavioral choice. Cell R3b1 appears to narrow the behavioral choice to two options – swimming or crawling – and the ﬁnal choice is made on the bases of sensory feedback related to the water level. Interneurons in the posterior brain also are probably important for crawling behavior, because some neurons in this ganglion have activity that is phase-locked to the crawl cycle (Baader and Fig. 27. Summary of the proposed combinatorial code for making behavioral Bachtold, 1997). decisions. This schematic diagram indicated which neurons are activated (black In isolated nerve cords, electrically stimulating nerves at rectangle) during four different behaviors, based upon intracellular recordings different segmental locations elicits different motor patterns: from these neurons as each of the behaviors is activated. Three cell types have been documented; for these the cell numbers of representative examples are stimulating nerves near the anterior end reliably elicits indicated below the cell types. Two others (D and E) represent other types of shortening (Shaw and Kristan, 1997), stimulating at the neurons, as yet unidentiﬁed, that could help to explain how behaviors are posterior end elicits crawling (Eisenhart et al., 2000), and chosen. 316 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

through the release of FMRFamide; Kuhlman et al., 1985a,b) implying that modulation of the heartbeat circuitry occurs in parallel with activation of the swim CPG. Other behaviors such as shortening and exploratory movements similarly show parallel modulation of the heartbeat circuitry, albeit it these effects have not be tested in the absence of sensory feedback as with swimming.

3.8. Methodologies and approaches for further research

Several lines of research and experimental techniques have been applied to leech nervous systems, but have not yet culminated in deep insights into behavioral circuits, their origins during development and functional plasticity. This section brieﬂy presents several of these, pointing out their future potential for contributing to our understanding of how leech behaviors are produced.

3.8.1. Functional indicator dyes The leech nervous system has proven useful for developing dyes sensitive to the electrical potential across the neuronal membrane (Salzberg et al., 1973; Canepari et al., 1996)and for tracking down neuronal connections (Farber and Grinvald, 1983). A class of fluorescent dyes with significantly higher sensitivity was developed recently (Gonzalez and Tsien, Fig. 28. Stimulating a single IN (R3b1) produces different behaviors, depend- 1997). These made possible the recording of many neurons at ing upon the sensory context. (A) Semi-intact preparation used to stimulate once during motor pattern generation in a segmental ganglion R3b1 in the subesophageal ganglion of the anterior brain. The brain and first (Cacciatore et al., 1999). Fluorescence changes are produced three segmental ganglia are denervated and pinned out; the rest of the animal is by fluorescent resonant energy transfer (FRET) between two free to move. Cell R3b1 was stimulated with identical stimuli three times in one dyes that are dissolved into the membranes of neuronal preparation, evoking three different behaviors (B–D). (B) When the water level was low (less than the thickness of the animal’s body at rest), the animal always somata. One of the dyes remains at the outer face of the crawled. E and C represent periods of elongation and contraction determined by membrane and is immobile; the other dye dissolves in the observing the movements in the intact part of the animal. The MN in the DP lipid bilayer of the membrane, is mobile, and carries a small nerve recording in segment 3 bursts in phase with the behavior. (C) When the negative charge (Fig. 29A). The mobile molecule moves near water level was high (greater than four times the animal’s thickness at rest), the the inside surface when the membrane is depolarized and near animal swam, as determined both by observing its movements and from the 1 Hz oscillations of the motor bursts in the nerve recording (see expanded the outer surface when the cell is hyperpolarized (Fig. 29A trace). (D) At an intermediate water depth (slightly deeper than the thickness of and B). To generate a signal that indicates the membrane the animal), the leech would sometimes produce a combined crawl and swim: potential, the cells are illuminated with a wavelength it would go through elongations and contractions typical of crawling, with absorbed by the immobile molecule. When the mobile episodes of swimming during the elongation phase of the crawl (from molecule is far from the stationary one (i.e. the cell is Esch et al., 2002). depolarized), the stationary molecule emits a photon at its characteristic wavelength. When the two types of dye for swimming, although an unusual one—it is the only unpaired molecules are close together (i.e. the cell is hyperpolarized), CPG neuron, and the only one that makes excitatory the mobile molecule absorbs some of the energy from connections to other CPG neurons (Weeks, 1982a,b,c). the immobile molecule and emits a photon at a longer Taking a more global approach, the patterns of spontaneous wavelength. A useful dye combination is coumarin and movements in intact leeches were analyzed, using an automated oxonol, when these are illuminated with violet light. If it is tracking system, for patterns of such distinguishable move- hyperpolarized, the neuron membrane emits mostly blue ments as swimming, crawling, ventilating, and searching over light; when the neuron depolarizes, the blue signal decreases the course of days (Mazzoni et al., 2005). There appeared to be and the red signal increases. This dye combination has a some distinct patterns in the sequences of movements, significantly greater signal-to-noise ratio than earlier voltage- suggesting that internal states of the animal produce choices sensitive dyes. With improved CCD cameras, voltage changes of behaviors in a somewhat ordered (non-random) way (Garcia- of 2 mV can be distinguished in small neurons (Taylor et al., Perez et al., 2005). 2003). As an example, in recording from a ganglion in an The rate of the heartbeat increases during swimming (Arbas isolated nerve cord (Fig. 29C), the intensity of fluorescence and Calabrese, 1984). This increase can be triggered in the in MNs varies significantly during swimming activity isolated nervous system by depolarization of cell 204 (perhaps (Fig. 29D). W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 317

Fig. 29. Using voltage-sensitive dyes to measure activity of neurons during the production of behaviors. (A and B) Schematic diagram of thevoltage-sensitive dyes, based upon fluorescence resonance energy transfer (FRET). Two fluorescent molecules are dissolved into the membrane. One (light gray) remains at the outer surface of the membrane, whereas the other (dark gray), which is dissolved within the membrane, moves towhichever surface is more positively charged, owing to its own small negative charge. Light of an appropriate wavelength is shown on the membrane to excite the molecule immobilized at the surface of the membrane. If the membrane is depolarized (A), the excited molecules emit photons of their characteristic fluorescence. If, instead, the membrane is hyperpolarized (B), the mobile molecule collects near the outer surface of the membrane, very close to the immobilized molecule. In this case, the mobile molecule absorbs some of the energy from the immobile molecule and emits photons at its own, longer wavelength. Therefore, the intensity of the photon emissions from either molecule – or better, the ratio of the emissions from the two – provides a measure of changes in the membrane potential. (C) Measuring the intensity of a FRET dye in MNs during the swimming motor program. The diagram shows the preparation used: a nerve cord from segment 2 through the posterior brain. Ganglion 10 (represented inside the large rectangle) was stained with both dyes and a part of its dorsal surface (inside the smaller rectangle) was imaged. In addition, extracellular records were obtained from a DP nerve. (D) The DP recording (top recording trace) shows the MN bursts characteristic of swimming. Above the recording are a series of images of the cluster of neurons at different phases of the swim cycle: the first, third, and fifth images are during dorsal motor neuronal bursts in the DP nerve recording and the second, fourth, and sixth images are from the interburst periods. Two neuronal somata are indicated, cell 3 (a dorsal longitudinal MN that fires in phase with the bursts in the DP nerve) and cell 4 (a ventral MN that fires out of phase withtheDPbursts). The bottom two recordings are the averaged intensities of the pixels representing cells 3 and 4, obtained at a rate of 8 Hz. Even at this low rate of data acquisition, the oscillations in the fluorescence signal – indicating membrane potential oscillations – can be seen without averaging.

As a test of their function, these dyes have been used to as well as to record the activity of up to 100 neurons on the locate oscillator INs that help to generate the swimming motor ventral surface of a segmental ganglion in response to stimuli pattern (Cacciatore et al., 1999). More recently, they have been that sometimes produce swimming and other times non- used to locate postsynaptic targets of Tr2 (Taylor et al., 2003), swimming responses (Briggman et al., 2005). The optical 318 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 analysis required to characterize these responses takes only 10– response (Lockery and Kristan, 1991), but it is not known 15 min, so that neurons whose activity looks interesting can be whether the changes are due to changes in neuronal properties recorded intracellularly and their electrophysiological proper- or to changes in the strengths of synapses. If synaptic strengths ties tested more directly. They can also be stimulated change, it makes a qualitative difference whether the changes individually to determine their effects on the initiation of are at the P cell-to-interneuronal synapses or at the IN-to-MN behaviors (Taylor et al., 2003). These techniques will be very synapses (see Fig. 9). If the latter synapses change their useful in tracking down potential decision-making neurons and strengths, P cell interactions adjacent to the tetanized one would the connection among them. also be strengthened; i.e. there would be generalization of the A second imaging technique employs calcium-sensitive response between sensory pathways. dyes (Ross et al., 1987), which were injected into single leech Whole-body shortening has been used extensively to study neurons, to determine the nature (i.e. spiking versus non- the neuronal bases of behavioral plasticity (Sahley, 1995)— spiking mechanisms) of transmitter release in the heartbeat mostly habituation and sensitization, although this response CPG (Ivanov and Calabrese, 2000, 2003). These dyes can also exhibits classical conditioning (Henderson and Strong, potentially be used for making finer distinctions among 1972; Sahley and Ready, 1988). The starting points for many of neuronal interactions. For instance, Ca2+-sensitive dyes might the studies attempting to find a neuronal mechanism of these be used to record from different branches of the same neuron, in behavioral plasticities are the observations that 5-HT induces order to determine whether different inputs affect particular changes in whole-body shortening resembling habituation regions of the dendritic tree differentially. Such localized (Biondi et al., 1982; Belardetti et al., 1982) and dishabituation synaptic effects could allow different regions of the dendritic (Beron et al., 1987) and that depleting the nervous system of 5- tree to act somewhat autonomously. HT severely modifies such plasticity (Ehrlich et al., 1992; Sahley et al., 1994; Modney et al., 1997). A fair amount is 3.8.2. Modeling known about the effects of 5-HT on sensitization and Several models have been produced to simulate various dishabituation (Sahley, 1995), as well as some of the second aspects of heartbeat (Nadim et al., 1995; Olsen et al., 1995; Hill messenger pathways involved in these plasticities (Burrell and et al., 2001, 2002; Jezzini et al., 2004), local bending (Lockery Sahley, 2001). In addition, blocking some 5-HT receptors et al., 1989; Lockery and Sejnowski, 1992, 1993; Lewis and eliminates sensitization (Beron et al., 1987). The site of 5-HT Kristan, 1998b), swimming (Friesen and Stent, 1977; Pearce action is thought to be the synapse from the sensory neurons to and Friesen, 1988; Taylor et al., 2000; Wolpert et al., 2000; the INs responsible for shortening and, perhaps, the pathway Wolpert and Friesen, 2000; Friesen and Cang, 2001; Zheng from the mechanoreceptors to the Retzius cells (Sahley, 1995), et al., 2004), and crawling (Cacciatore et al., 2000; which are reservoirs for much of the 5-HT contained within Skierczynski et al., 1996; Kristan et al., 2000). These models each ganglion (Glover and Lent, 1991). The interconnected incorporated simplified neuronal properties and a limited network of S cells carries the sensitizing signal throughout the number of simulated neurons. In some cases, the simplifications body; cutting Faivre’s nerve (Sahley et al., 1994) or the axon of were used because the relevant properties were not known. In a single S cell (Modney et al., 1997) prevents sensitization from other cases, however, the simplifications were required because spreading beyond the site of the lesion. In an elegant of limited computer time and memory. This limitation no confirmation that the S cell is necessary for sensitization to longer exists, so that large and complex models are now readily occur, sensitization reappeared after regeneration of the accomplished. It should now be possible to make a very connection between adjacent S cells that had been previously realistic model of intersegmental coordination in the swimming ablated (Modney et al., 1997; Burrell et al., 2003). circuit, for instance, and to determine the nature of the Bath-applied 5-HT makes the S cell itself more excitable coordination of the heart MNs as a function of the strengths of (Burrell et al., 2002), suggesting one potential site for the their connections from the heart INs. sensitization caused by endogenous release of 5-HT. An intriguing suggestion is that part of the increased excitability 3.8.3. Plasticity could be caused by reflections of action potentials in the Tand P Many leech behaviors are plastic, showing both non- mechanosensory neurons, thereby increasing the effectiveness contingent changes (e.g. habituation and sensitization) and of the sensory input (Baccus et al., 2000, 2001). Strong stimuli contingent changes (e.g. classical conditioning). For example, that induce sensitization strongly activate Retzius cells (Sahley, the local bend response shows non-contingent learning: 1994; Lockery and Kristan, 1991), and 5-HT is released from stimulating either the skin or an identified IN produces the somata of Retzius cells in a paracrine manner (Trueta et al., sensitization (increased response to a test tactile stimulus) with 2003, 2004; De-Miguel and Trueta, 2005), thereby delivering at least two different time constants, and stimulating one large quantities of 5-HT in much the same way as bath particular IN can decrease the size of the response (Lockery and application. These studies make a very good case that both the S Kristan, 1991). Because these changes are apparent in MN cell and the Retzius cells are critical for plasticity in whole- responses, the sites of such plasticity surely reside in the central body shortening, and that 5-HT is a critical neuromodulator in nervous system, however, the critical synapses that are modified the leech. have not been identified. Moreover, stimulating one P cell at a Tactile stimulation of the leech body wall sufficient to high rate also produces sensitization of the local bending activate Tand P mechanosensory neurons (Nicholls and Baylor, W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 319

1968), often elicits swimming in intact and semi-intact leeches (Fig. 16). Given this result, it is surprising that stimulation of individual T and P cells often fails to elicit swimming in isolated nerve cords. One explanation for this failure is that responses to stimulation of mechanoreceptors, in both intact (by mechanical stroking) and isolated preparations (by current injection), decrements rapidly (Debski and Friesen, 1985, 1987). Successive stimulus trials lead to briefer swim episodes until swim initiation abruptly ceases. Because leech dissection involves cutting T, P and N cell axons, strong injury-induced stimulation of these cells may have induced habituation. Habituation is most obvious for light touch, which selectively activates T cells, but also occurs during P and N cell stimulation (Debski and Friesen, 1987). The degree of habituation is correlated with a reduction of excitation in swim-gating neurons, but additional, unknown factors involving these cells and the central oscillator are also important (Debski and Friesen, 1986). Bath application of 5-HT does not reverse habituation of the swimming response to tactile stimulation (Debski and Friesen, 1987).

3.8.4. Development The nervous system of the leech, like the rest of its body, develops in a very stereotyped way (Stent et al., 1992). Fig. 30. Development of behaviors in the embryonic leech. (A) Summary plot Essentially all neurons in a segmental ganglion are born (i.e. of behaviors, both spontaneous and elicited, during embryonic time (ET). Along they have undergone their last cell division) at about the same the line are the percentages of embryonic time (100% takes about 30 days at time. [A staging scheme (Reynolds et al., 1998a) divides 20 8C.) Above the time line are the earliest times that the major behaviors, either development into percentage of embryonic development or spontaneous or elicited, can be seen. (The components of crawling, including %ED. At 20 8C, leech development takes 30 days; it is faster at ‘‘elongation’’ appear over an extended period of time.) Below the line are shown the onset times of electrogenesis, electrical synaptic connections, and chemical higher temperatures and slower at lower ones.] Between 35 and synapses in MNs. (B) Switch from circumferential indentation to local bending 40% ED, all the neurons in midbody ganglia start growing during embryogenesis. Two different behaviors – circumferential indentation processes at the same time. They grow their longest process first and local bending – can be elicited by light touch of the skin in the middle of the (i.e. sensory and MNs grow processes to the periphery, and INs animal starting just after 50% ET. Circumferential indentation begins first, then grow their interganglionic processes), which are destined to wanes in probability as local bending becomes more prominent. become axons. At about 50% ED the central processes, which will become the sites of synaptic contacts, appear. The first Hence, the progression from circumferential indentation to signs of synaptic interactions appear at about 52% ED, at the local bending results from the establishment of electrical same time as coordinated behaviors – both spontaneous and connections first, followed chemical ones, among the relevant evoked – become visible (Reynolds et al., 1998b). neurons. Swimming starts later in development, at about 62% Simple, local behaviors, such as shortening and the ED, with the appearance of alternating dorsal and ventral precursor to local bending, are seen earliest, followed by flexions (French et al., 2005). At this stage, the peaks and crawling and swimming, which appear at about 70% ED troughs do not pass through the body, i.e. there are no (Fig. 30A). Initially, the behaviors are weak and labile, undulations that propel the animal through the water. Over the becoming adult-like over the course of 10–15% ED after they course of a week, the front-to-back undulations appear and first appear. Local bending develops in two stages. At about become more pronounced, so that swimming is adult-like by 50% of ED (about two weeks after fertilized eggs are deposited the end of embryonic life. Leech neurons are sufficiently large into a cocoon), touching a mid-body segment causes it to and accessible to allow intracellular recording from the genesis contract all around, a behavior termed ‘‘circumferential of electrogenic and synaptic contacts (Marin-Burgin et al., indentation’’ (Fig. 30B). Over the course of a week, this 2005), so it should be possible to determine how these indentation gives way to adult-like local bending. The onset of behaviorally relevant neuronal circuits are formed during circumferential indentation coincides with the formation of the embryogenesis. initial contacts among the MNs, which are exclusively electrical (Marin-Burgin et al., 2005). Local bending appears 4. Conclusion as the chemical connections – particularly the inhibitory ones – are established. (Some of the electrical connections disappear, This overview of the neuronal mechanisms underlying six but many of them are retained into adulthood; they are largely distinct behaviors in the medicinal leech illuminates the utility overridden by chemical inputs during adult local bending.) of this animal for gaining insights into rhythmic movements, 320 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 sensory feedback, neuromodulatory control, interactions Arshavsky, Y.I., Deliagina, T.G., Orlovsky, G.N., Panchin, Y.V., Popova, L.B., among behaviors, and behavioral choice. The gaps in our Sadreyev, R.I., 1998. Analysis of the central pattern generator for swimming in the mollusk Clione. Ann. N.Y. Acad. Sci. 860, 51–69. understanding of the biophysical and developmental bases of Baader, A.P., Bachtold, D., 1997. Temporal correlation between neuronal tail these behaviors, in the face of so much existing information, ganglion activity and locomotion in the leech, Hirudo medicinalis. Invert. make the leech a prime contender for further study. With the Neurosci. 2, 245–251. function of about one-third of its neurons already identified, it is Baader, A.P., Kristan Jr., W.B., 1992. Monitoring neuronal activity during not unrealistic to aim for a description of the functional roles of discrete behaviors: a crawling, swimming, and shortening device for tethered leeches. J. Neurosci. Meth. 43, 215–223. all neurons in the leech nerve cord. Far-ranging ramifications of Baader, A.P., Kristan Jr., W.B., 1995. Parallel pathways coordinate crawling in such an accomplishment would include a deeper understanding the medicinal leech, Hirudo medicinalis. J. Comp. Physiol. 176, 715–726. of electrical coupling and its role in behavioral circuits, of the Baca, S.M., Kristan Jr., W.B., 2001. Influence of inhibition in the local bend functions of non-spiking versus spiking interactions in sensory– response in the medicinal leech (Hirudo medicinalis). Soc. Neurosci. Abstr. motor systems, generation of a wide range of animal behaviors 27, 518.5. Baca, S.M., Thomson, E.E., Kristan, W.B., 2005. Location and intensity from a limited set of neurons participating in overlapping discrimination in the leech local bending response quantified using optic functions, and the neuronal ontology of behaviors. The leech flow and principal components analysis. J. Neurophysiol. 93, 3560– has served an important model animal for neuroscience since 3572. the late 19th century (Mann, 1962; Sawyer, 1986). It will Baccus, S.A., Burrell, B.D., Sahley, C.L., Muller, K.J., 2000. Action potential clearly continue to play that role, for even now research papers reflection and failure at axon branch points cause stepwise changes in EPSPs in a neuron essential for learning. J. Neurophysiol. 83, 1693–1700. concerning leech biology are published at the rate of about 100 Baccus, S.A., Sahley, C.L., Muller, K.J., 2001. Multiple sites of action potential per annum. initiation increase neuronal firing rate. J. Neurophysiol. 86, 1226–1236. Bagnoli, P., Brunelli, M., Magni, F., 1975. The neuron of the fast conducting system in Hirudo medicinalis: identification and synaptic connections with Acknowledgements primary afferent neurons. Arch. Ital. Biol. 113, 21–43. Baptista, C.A., Macagno, E.R., 1988. The role of the sexual organs in the Funding was provided by NIH grants MH43396 and generation of postembryonic neurons in the leech Hirudo medicinalis.J. NS35336, and a grant from Microsoft Research Labs (to Neurobiol. 19, 707–726. WBK); NSF grant IBN-0110607 and NIH grant NIMH- Beenhakker, M.P., Blitz, D.M., Nusbaum, M.P., 2004. Long-lasting activation of rhythmic neuronal activity by a novel mechanosensory system in the MH63855 (to WOF); and NIH grant NS24072 (to RLC). We crustacean stomatogastric nervous system. J. Neurophysiol. 91, 78–91. express our deep gratitude to numerous students and colleagues Belanger, J.H., Orchard, I., 1988. Release of octopamine by Leydig cells in the who collaborated with us over a span of more than three decades. central nervous system of the leech Marcobdella decora, and its possible Without their expert and active participation in the research neurohormonal role. J. Comp. Physiol. A 162, 405–412. described here, this review would have been much shorter. We are Belardetti, P., Biondi, C., Colombaioni, L., Brunelli, M., Trevisani, A., Zavagno, C., 1982. Role of serotonin and cyclic AMP on the facilitation especially grateful to Prof. Gunther Stent, under whose tutelage of the fast conducting system activity in the leech, Hirudo medicinalis. the neuronal circuits described here saw first light. Finally, we Brain. Res. 246, 89–103. thank Courtney Milianes for expert editorial assistance. Beron, F., Brunelli, M., Catarsi, S., Garcia-Gil, M., Traina, G., Tongiorgi, E., 1987. Role of cAMP and serotonin on short- and long-term sensitization in H. medicinalis. Pflug. Archiv. 408, S52. References Biondi, C., Belardetti, F., Brunelli, M., Trevisani, A., 1982. Modulation of cAMP levels by neurotransmitters in excitable tissues of the leech, Hirudo Adrian, E.D., 1931. Potential changes in the isolated nervous system of Dytiscus medicinalis. Comp. Biochem. Physiol. 72C, 33–37. marginalis. J. Physiol. 72, 132–151. Blackshaw, S.E., 1981. Morphology and distribution of touch cell terminals in Alscher, C., Beyn, W.J., 1998. Simulating the motion of the leech: A biome- the skin of the leech. J. Physiol. Lond. 320, 219–228. chanical application of DAEs. Numer. Algorithm 19, 1–12. Blackshaw, S.E., 1993. Stretch receptors and body wall muscle in leeches. Angstadt, J.D., Calabrese, R.L., 1989. A hyperpolarization-activated inward Comp. Biochem. Physiol. 105A, 643–652. current in heart interneurons of the medicinal leech. J. Neurosci. 9, 2846– Blackshaw, S.E., Kristan Jr., W.B., 1990. Input from single stretch receptor 2857. neurons influences the centrally generated swim motor pattern in the leech. Angstadt, J.D., Calabrese, R.L., 1991. Calcium currents and graded synaptic J. Physiol. Lond. 425, 93P. transmission between heart interneurons of the leech. J. Neurosci. 11, 746– Blackshaw, S.E., Nicholls, J.G., Parnas, I., 1982. Physiological responses, 759. receptive fields and terminal arborizations of nociceptive cells in the leech. Angstadt, J.D., Friesen, W.O., 1993a. Modulation of swimming behavior in the J. Physiol. Lond. 326, 251–260. medicinal leech. I. Effects of serotonin on the electrical properties of swim- Blackshaw, S.E., Thompson, S.W.N., 1988. Hyperpolarizing responses to gating cell 204. J. Comp. Physiol. A 59, 223–234. stretch in neurones innervating leech body wall muscle. J. Physiol. Lond. Angstadt, J.D., Friesen, W.O., 1993b. Modulation of swimming behavior in the 396, 121–138. medicinal leech. II. Ionic conductances underlying serotonergic modulation Braschler, T.R., Merino, S., Tomas, J.M., Graf, J., 2003. Complement resistance of swim-gating cell 204. J. Comp. Physiol. A 59, 235–248. is essential for colonization of the digestive tract of Hirudo medicinalis by Angstadt, J.D., Moore, W.H., 1997. A circadian rhythm of swimming behavior Aeromonas strains. Appl. Environ. Microbiol. 69, 4268–4271. in a predatory leech of the family Erpobdellidae. Amer. Midland Nat. 137, Brembs, B., Lorenzetti, F.D., Reyes, F.D., Baxter, D.A., Byrne, J.H., 2002. 165–172. Operant reward learning in Aplysia: neuronal correlates and mechanisms. Arbas, E.A., Calabrese, R.L., 1984. Rate modification in the heartbeat central Science 296, 1706–1709. pattern generator of the medicinal leech. J. Comp. Physiol. 155, 783–794. Briggman, K.L., Abarbanel, H.D.I., Kristan Jr., W.B., 2005. Optical imaging of Arisi, I., Zoccolan, D., Torre, V., 2001. Distributed motor pattern underlying neuronal populations during decision-making. Science 307, 896–901. whole-body shortening in the medicinal leech. J. Neurophysiol. 86, 2475– Bristol, A.S., Marinesco, S., Carew, T.J., 2004. Neural circuit of tail-elicited 2488. siphon withdrawal in Aplysia. II. Role of gated inhibition in differential W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 321

lateralization of sensitization and dishabituation. J. Neurophysiol. 91, 678– Cang, J., Friesen, W.O., 2000. Sensory modification of leech swimming; 692. rhythmic activity of ventral stretch receptors can change intersegmental Brodfuehrer, P.D., Burns, A., 1995. Neuronal factors influencing the decision to phase relationships. J. Neurosci. 20, 7822–7829. swim in the medicinal leech. Neurobiol. Learn. Mem. 63, 192–199. Cang, J., Friesen, W.O., 2002. Model for intersegmental coordination of leech Brodfuehrer, P.D., Cohen, A.H., 1990. Initiation of swimming activity in the swimming, central and sensory mechanisms. J. Neurophysiol. 87, 2760– medicinal leech by glutamate, quisqualate, and kainate. J. Exp. Biol. 154, 2769. 567–572. Cang, J., Yu, X., Friesen, W.O., 2001. Sensory modification of leech swimming, Brodfuehrer, P.D., Cohen, A.H., 1992. Glutamate-like immunoreactivity in the interactions between ventral stretch receptors and swim-related neurons. J. leech central nervous system. Histochemistry 97, 511–516. Comp. Physiol. A 187, 569–579. Brodfuehrer, P.D., Friesen, W.O., 1984. A sensory system initiating swimming Carlton, T., McVean, A., 1993. A comparison of the performance of two sensory activity in the medicinal leech. J. Exp. Biol. 108, 341–355. systems in host detection and location in the medicinal leech Hirudo Brodfuehrer, P.D., Friesen, W.O., 1986a. Initiation of swimming activity by medicinalis. Comp. Biochem. Physiol. Comp. Physiol. 104, 273–277. trigger neurons in the leech subesophageal ganglion. I. Output pathways of Cellucci, C.J., Brodfuehrer, P.D., Acera-Pozzi, R., Dobrovolny, H., Engler, E., Tr1 and Tr2. J. Comp. Physiol. 159, 485–502. Los, J., Thompson, R., Albano, A.M., 2000. Linear and nonlinear measures Brodfuehrer, P.D., Friesen, W.O., 1986b. Initiation of swimming activity by predict swimming in the leech. Phys. Rev. 62, 4826–4834. trigger neurons in the leech subesophageal ganglion. II. Role of segmental Cline, H.T., 1986. Evidence for GABA as a neurotransmitter in the leech. J. swim-initiating interneurons. J. Comp. Physiol. A 159, 503–510. Neurosci. 6, 2848–2856. Brodfuehrer, P.D., Friesen, W.O., 1986c. Initiation of swimming activity by Cohen, A.H., Walleń, P., 1980. The neuronal correlate of locomotion in fish, trigger neurons in the leech subesophageal ganglion. III. Sensory inputs to ‘‘Fictive swimming’’ induced in an in vitro preparation of the lamprey spinal Tr1 and Tr2. J. Comp. Physiol. A 159, 511–519. cord. Exp. Brain Res. 41, 11–18. Brodfuehrer, P.D., Friesen, W.O., 1986d. Control of leech swimming activity by Coleman, M.J., Meyrand, P., Nusbaum, M.P., 1995. A switch between two the cephalic ganglia. J. Neurobiol. 17, 697–705. modes of synaptic transmission mediated by presynaptic inhibition. Nature Brodfuehrer, P.D., Friesen, W.O., 1986e. From stimulation to undulation: an 378, 502–505. identified pathway for the control of swimming activity in the leech. Science Combes, D., Merrywest, S.D., Simmers, J., Sillar, K.T., 2004. Developmental 234, 1002–1004. segregation of spinal networks driving axial and hindlimb based locomotion Brodfuehrer, P.D., Kogelnik, A.M., Friesen, W.O., Cohen, A.H., 1993. Effect of in metamorphosing Xenopus laevis. J. Physiol. 559, 17–24. the tail ganglion on swimming activity in the leech. Behav. Neural Biol. 59, Crisp, K.M., Klukas, K.A., Gilchrist, L.S., Nartey, A.J., Mesce, K.A., 2002. 162–166. Distribution and development of dopamine- and octopamine-synthesizing Brodfuehrer, P.D., Parker, H.J., Burns, A., Berg, M., 1995a. Regulation of the neurons in the medicinal leech. J. Comp. Neurol. 442, 115–129. segmental swim-generating system by a pair of identified interneurons in the Crisp, K.M., Mesce, K.A., 2003. To swim or not to swim: regional effects of leech head ganglion. J. Neurophysiol. 73, 983–992. serotonin, octopamine and amine mixtures in the medicinal leech. J. Comp. Brodfuehrer, P.D., Debski, E.A., O’Gara, B.A., Friesen, W.O., 1995b. Neuronal Physiol. A 189, 461–470. control of leech swimming. J. Neurobiol. 27, 403–418. Crisp, K.M., Mesce, K.A., 2004. A cephalic projection neuron involved in Brodfuehrer, P.D., Thorogood, M.S.E., 2001. Identified neurons and leech locomotion is dye coupled to the dopaminergic neural network in the swimming behavior. Prog. Neurobiol. 63, 371–381. medicinal leech. J. Exp. Biol. 207, 4535–4542. Brown, T.G., 1911. The intrinsic factors in the act of progression in the Cymbalyuk, G.S., Gaudry, Q., Masino, M.A., Calabrese, R.L., 2002. Bursting in mammal. Proc. Roy. Soc. Lond. 84, 308–319. leech heart interneurons, cell-autonomous and network-based mechanisms. Burrell, B.D., Sahley, C.L., Muller, K.J., 2003. Progressive recovery of learning J. Neurosci. 22, 10580–10592. during regeneration of a single synapse in the medicinal leech. J. Comp. Daly, K.C., Christensen, T.A., Lei, H., Smith, B.H., Hildebrand, J.G., 2004. Neurol. 457, 67–74. Learning modulates the ensemble representations for odors in primary Burrell, B.D., Sahley, C.L., 2001. Learning in simple systems. Curr. Opin. olfactory networks. Proc. Natl. Acad. Sci. 101, 10476–10481. Neurobiol. 11, 757–764. Debski, E.A., Friesen, W.O., 1985. Habituation of swimming activity in the Burrell, B.D., Sahley, C.L., Muller, K.J., 2002. Differential effects of serotonin medicinal leech. J. Exp. Biol. 116, 169–188. enhance activity of an electrically coupled neural network. J. Neurophysiol. Debski, E.A., Friesen, W.O., 1986. The role of central interneurons in the 87, 2889–2895. habituation of swimming activity in the medicinal leech. J. Neurophysiol. Cacciatore, T.W., Brodfuehrer, P.D., Gonzalez, J.E., Jiang, T., Adams, S.R., 55, 977–994. Tsien, R.Y., Kristan Jr., W.B., Kleinfeld, D., 1999. Identification of neural Debski, E.A., Friesen, W.O., 1987. Intracellular stimulation of sensory cells circuits by imaging coherent electrical activity with FRET-based dyes. elicits swimming activity in the medicinal leech. J. Comp. Physiol. A 160, Neuron 23, 449–459. 447–457. Cacciatore, T.W., Rozenshteyn, R., Kristan Jr., W.B., 2000. Kinematics and Delcomyn, F., 1980. Neural basis of rhythmic behavior in animals. Science 210, modeling of leech crawling: evidence for an oscillatory behavior produced 492–498. by propagating waves of excitation. J. Neurosci. 20, 1643–1655. Dembrow, N.C., Jing, J., Proekt, A., Romero, A., Vilim, F.S., Cropper, E.C., Calabrese, R.L., 1977. The neural control of alternate heartbeat coordination Weiss, K.R., 2003. A newly identified buccal interneuron initiates and modu- states in the leech. J. Comp. Physiol. 122, 111–143. lates feeding motor programs in Aplysia. J. Neurophysiol. 90, 2190–2204. Calabrese, R.L., Maranto, A.R., 1986. Cholinergic action on the heart of the De-Miguel, F.F., Trueta, C., 2005. Synaptic and extrasynaptic secretion of leech Hirudo medicinalis. J. Exp. Biol. 125, 205–224. serotonin. Cell. Mol. Neurobiol. 25, 297–312. Calabrese, R.L., Nadim, F., Olsen, O.H., 1995. Heartbeat control in the DeRosa, Y.S., Friesen, W.O., 1981. Morphology of leech sensilla, observations medicinal leech: a model system for understanding the origin, coordination, with the scanning electron microscope. Biol. Bull. 160, 383–393. and modulation of rhythmic motor patterns. J. Neurobiol. 27, 390–402. Dickinson, M.H., Lent, C.M., 1984. Feeding behavior of the medicinal leech, Calabrese, R.L., Peterson, E., 1983. Neural control of heartbeat in the leech, Hirudo medicinalis L. J. Comp. Physiol. 154, 449–455. Hirudo medicinalis. In: Roberts, A., Roberts, B. (Eds.), Proceedings of the Douglass, J.K., Strausfeld, N.J., 2000. Optic flow representation in the optic XXXVII Symposium Society for Experimental Biology on Neural Origin of lobes of Diptera: modeling innervation matrices onto collators and their Rhythmic Movements. pp. 195–221. evolutionary implications. J. Comp. Physiol. A 186, 799–811. Camhi, J.M., Levy, A., 1989. The code for stimulus direction in a cell assembly Eckert, R., 1963. Electrical interaction of paired ganglion cells in the leech. J. in the cockroach. J. Comp. Physiol. A 165, 83–97. Gen. Physiol. 46, 573–587. Canepari, M., Campani, M., Spadavecchia, L., Torre, V., 1996. CCD imaging of Ehrlich, J.S., Boulis, N.M., Karrer, T., Sahley, C.L., 1992. Differential effects of the electrical activity in the leech nervous system. Eur. Biophys. J. 24, 359– serotonin depletion on sensitization and dishabituation in the leech, Hirudo 370. medicinalis. J. Neurobiol. 23, 270–279. 322 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

Ekeberg, O¨ ., Grillner, S., 1999. Simulations of neuromuscular control in Galun, R., 1975. The role of host blood in the feeding of ectoparasites. In: lamprey swimming. Phil. Trans. Roy. Soc. Lond. B 354, 895–902. Zuckerman, A. (Ed.), Dynamic Aspects of Host-Parasite Relationships, Eisenhart, F.J., Cacciatore, T.W., Kristan Jr., W.B., 2000. A central pattern vol. II. Academic Press, New York, pp. 132–162. generator underlies crawling in the medicinal leech. J. Comp. Physiol. 186, Garcia-Perez, E., Zoccolan, D., Pinato, G., Torre, V., 2004. Dynamics and 631–643. reproducibility of a moderately complex sensory-motor response in the Elliott, E.J., 1986. Chemosensory stimuli in feeding behavior of the leech medicinal leech. J. Neurophysiol. 92, 1783–1795. Hirudo medicinalis. J. Comp. Physiol. A 159, 391–401. Garcia-Perez, E., Mazzoni, A., et al., 2005. Statistics of decision-making in the Elliott, E.J., 1987. Morphology of chemosensory organs required for feeding in leech. J. Neurosci 25, 2597–2608. the leech Hirudo medicinalis. J. Morphol. 192, 181–187. Gascoigne, L., McVean, A., 1991. Water movement sensitive cells in leech Esch, T., Kristan Jr., W.B., 2002. Decision-making in the leech nervous system. CNS. Phil. Trans. Roy. Soc. Lond. B. 332, 261–270. Integr. Comp. Biol. 42, 716–724. Georgopoulos, A.P., Kettner, R.E., Schwartz, A.B., 1988. Primate motor cortex Esch, T., Mesce, K.A., Kristan Jr., W.B., 2002. Evidence for sequen- and free arm movements to visual targets in three-dimensional space. II. tial decision-making in the medicinal leech. J. Neurosci. 22, 11045– Coding of the direction of movement by a neuronal population. J. Neurosci. 11054. 8, 2928–2937. Fan, R.-J., Marin-Burgin, A., French, K.A., Friesen, W.O., 2005. A dye mixture Gilchrist, L.S., Klukas, K.A., Jellies, J.A., Rapus, J., Eckert, M., Mesce, K.A., (Neurobiotin and Alexa 488) reveals extensive dye-coupling among swim- 1995. Distribution and developmental expression of octopamine-immunor- related neurons in leeches; physiology confirms the connections. J. Comp. eactive neurons in the central nervous system of the leech. J. Comp. Neurol. Physiol. A, doi:10.1007/s00359-005-0047-8. 353, 451–463. Fan, R.-J., Friesen, W.O., 2005. Characterization of central axon terminals of Gilchrist, L.S., Mesce, K.A., 1997. Coactivation of putative octopamine- and putative stretch receptors in leeches. J. Comp. Neurol, in press. serotonin-containing interneurons in the medicinal leech. J. Neurophysiol. Farber, I.C., Grinvald, A., 1983. Identification of presynaptic neurons by laser 78, 2108–2115. photostimulation. Science 222, 1025–1027. Glover, J.C., Kramer, A.P., 1982. Serotonin analog selectively ablates identified Frank, E., Jansen, J.K., Rinvik, E., 1975. A multisomatic axon in the central neurons in the leech embryo. Science 216, 317–319. nervous system of the leech. J. Comp. Neurol. 159, 1–13. Glover, J.C., Lent, C.M., 1991. Serotonin is released from isolated leech ganglia French, K.A., Chang, J., Reynolds, S., Gonzalez, R., Kristan III, W.B., Kristan by potassium-induced depolarization. Comp. Biochem. Physiol. C 99, 437– Jr., W.B., 2005. Development of swimming in the medicinal leech; the 443. gradual acquisition of a behavior. J. Comp. Physiol., doi: 10.1007/s00359- Goldburt, V., Sabban, B.A., Kleinhaus, A.L., 1994. Serotonin depletion 005-0003-7. inhibits feeding in carnivorous leeches (Haemopis). Behav. Neural Biol. Friesen, W.O., 1981. Physiology of water motion detection in the medicinal 61, 47–53. leech. J. Exp. Biol. 92, 255–275. Gonzalez, J.E., Tsien, R.Y., 1997. Improved indicators of cell membrane Friesen, W.O., 1985. Neuronal control of leech swimming movements: inter- potential that use fluorescence resonance energy transfer. Chem. Biol. 4, actions between cell 60 and previously described oscillator neurons. J. 269–277. Comp. Physiol. A 156, 231–242. Gramoll, S., Schmidt, J., Calabrese, R., 1994. Switching in the activity state of Friesen, W.O., 1989a. Neuronal control of leech swimming movements. I. an interneuron that controls coordination of the hearts in the medicinal leech Inhibitory interactions between motor neurons. J. Comp. Physiol. A 166, Hirudo medicinalis. J. Exp. Biol. 186, 157–171. 195–203. Granzow, B., Friesen, W.O., Kristan Jr., W.B., 1985. Physiological and mor- Friesen, W.O., 1989b. Neuronal control of leech swimming movements. II. phological analysis of synaptic transmission between leech motor neurons. Motor neuron feedback to oscillator cells 115 and 28. J. Comp. Physiol. A J. Neurosci. 5, 2035–2050. 166, 205–215. Granzow, B., Kristan Jr., W.B., 1986. Inhibitory connections between motor Friesen, W.O., 1989c. Hierarchical organization of neuronal circuits that neurons modify a centrally generated motor pattern in the leech nervous control animal locomotion. In: Erber, J., Menzel, R., Pflu¨ger, H.-J., Todt, system. Brain Res. 369, 321–325. D. (Eds.), Neural Mech. Behav. Georg Thieme Verlag, New York, pp. 105– Gray, J., 1950. The role of peripheral sense organs during locomotion in 110. vertebrates. Symp. Soc. Exp. Biol. 4, 112–126. Friesen, W.O., 1989d. Neuronal control of leech swimming movements. In: Gray, J., 1958. The movement of the spermatozoa of the bull. J. Exp. Biol. 35, Jacklet, J.W. (Ed.), Cellular and Neuronal Oscillators. Marcel Dekker, New 96–108. York, pp. 269–316. Gray, J., 1968. Animal locomotion. In: Carrington, R. (Ed.), The World Friesen, W.O., 1994. Reciprocal inhibition: a mechanism underlying oscillatory Naturalist. W.W. Norton and Co. Inc., New York, pp. 400–459. animal movements. Neurosci. Biobehav. Rev. 18, 547–553. Gray, J., Lissmann, H.W., Pumphrey, R.J., 1938. The mechanism of locomotion Friesen, W.O., Brodfuehrer, P.D., 1984. Identification of neurons in the leech in the leech Hirudo medicinalis Ray. J. Exp. Biol. 15, 408–430. through local ionic manipulations. J. Exp. Biol. 113, 455–460. Grillner, S., 2003. The motor infrastructure: from ion channels to neuronal Friesen, W.O., Cang, J., 2001. Sensory and central mechanisms control of networks. Nat. Rev. Neurosci. 4, 573–586. intersegmental coordination. Curr. Opin. Neurobiol. 11, 678–683. Grillner, S., Walleń, P., Brodin, L., Lansner, A., 1991. Neuronal network Friesen, W.O., Hocker, C.G., 2001. Functional analyses of the leech swim generating locomotor behavior in lamprey, circuitry, transmitters, mem- oscillator. J. Neurophysiol. 86, 824–835. brane properties, and simulation. Ann. Rev. Neurosci. 14, 169–199. Friesen, W.O., Pearce, R.A., 1993. Mechanisms of intersegmental coordination Groome, J.R., Clark, M., Lent, C.M., 1993. The behavioural state of satiation in in leech locomotion. Sem. Neurosci. 5, 41–47. the leech is regulated by body distension and mimicked by serotonin Friesen, W.O., Poon, M., Stent, G.S., 1976. An oscillatory neuronal circuit depletion. J. Exp. Biol. 182, 265–270. generating a locomotory rhythm. PNAS 73, 3734–3738. Groome, J.R., Vaughan, D.K., Lent, C.M., 1995. Ingestive sensory inputs Friesen, W.O., Poon, M., Stent, G.S., 1978. Neuronal control of swimming in excite serotonin effector neurones and promote serotonin depletion from the medicinal leech. IV. Identification of a network of oscillatory inter- the leech central nervous system and periphery. J. Exp. Biol. 198, 1233– neurones. J. Exp. Biol. 75, 25–43. 1242. Friesen, W.O., Stent, G.S., 1977. Generation of a locomotory rhythm by Hagiwara, S., Morita, H., 1962. Electrotonic transmission between two nerve a neural network with recurrent cyclic inhibition. Biol. Cybern. 28, 27– cells in leech ganglion. J. Neurophysiol. 25, 721–731. 40. Harris-Warrick, R.M., Baro, D.J., Cogiglio, L.M., Johnson, B.R., Levini, R.M., Friesen, W.O., Stent, G.S., 1978. Neural circuits for generating rhythmic Peck, J.H., Zhang, B., 1997. Chemical modulation of crustacean stomato- movements. Ann. Rev. Biophys. Bioeng. 7, 37–61. gastric pattern generator networks. In: Stein, P.S.G., Grillner, S., Selver- Galun, R., Kindler, S.H., 1966. Chemical specificity of the feeding response in ston, A.I., Stuart, D.G. (Eds.), Neurons Networks and Motor Behavior. MIT Hirudo medicinalis (L.) Comp. Biochem. Physiol. 17, 69–73. Press, Cambridge, MA, pp. 209–216. W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 323

Hartline, D.K., 1967. Impulse identification and axon mapping of the nine Kristan Jr., W.B., 1982. Sensory and motor neurones responsible for the local neurons in the cardiac ganglion of the lobster Homarus americanus. J. Exp. bending response in leeches. J. Exp. Biol. 96, 161–180. Biol. 47, 327–340. Kristan Jr., W.B., 1983. The neurobiology of swimming in the leech. Trends Hashemzadeh-Gargari, H., Friesen, W.O., 1989. Modulation of swimming Neurosci. 6, 84–88. activity in the medicinal leech by serotonin and octopamine. Comp. Kristan Jr., W.B., 2000. Distributed processing vs. dedicated neurons in the Biochem. Physiol. 94C, 295–302. production of simple behavioral acts. In: Cruse, H., Dean, J., Ritter, H. Henderson, T.B., Strong, P.N., 1972. Classical conditioning in the leech, (Eds.), Pre-rational Intelligence, Adaptive Behavior and Intelligent Sys- Macrobdella ditetra, as a function of CS and UCS intensity. Cond. Reflex tems without Symbols and Logic, vol. 1. Kluwer, Boston, pp. 243–266. 7, 210–215. Kristan Jr., W.B., Calabrese, R.L., 1976. Rhythmic swimming activity in Higashijima, S., Masino, M.A., Mandel, G., Fetcho, J.R., 2003. Imaging neurons of the isolated nerve cord of the leech. J. Exp. Biol. 65, 643–668. neuronal activity during zebrafish behavior with a genetically encoded Kristan Jr., W.B., Guthrie, P.B., 1977. Acquisition of swimming behavior in calcium indicator. J. Neurophysiol. 90, 3986–3997. chronically isolated single segments of the leech. Brain Res. 131, 191–195. Hildebrandt, K.-P., 1988. Circulation in the leech, Hirudo medicinalis. J. Exp. Kristan Jr., W.B., Lockery, S.R., Lewis, J.E., 1995. Using reflexive behaviors of Biol. 134, 235–246. the medicinal leech to study information processing. J. Neurobiol. 27, 380– Hill, A.A.V., Lu, J., Masino, M.A., Olsen, Ø.H., Calabrese, R.L., 2001. A model 389. of a segmental oscillator in the leech heartbeat neuronal network. J. Comp. Kristan Jr., W.B., McGirr, S.J., Simpson, G.V., 1982. Behavioural and mechan- Neurosci. 10, 281–302. osensory neurone responses to skin stimulation in leeches. J. Exp. Biol. 96, Hill, A.A.V., Masino, M.A., Calabrese, R.L., 2002. A model of intersegmental 143–160. coordination in the leech heartbeat neuronal network. J. Neurophysiol. 87, Kristan Jr., W.B., Nusbaum, M.P., 1983. The dual role of serotonin in leech 1586–1602. swimming. J. Physiol. 78, 743–747. Hill, A.A., Masino, M.A., Calabrese, R.L., 2003. Intersegmental coordination Kristan Jr., W.B., Skalak, R., Wilson, R.J.A., Skierczynski, B.A., Murray, J.A., of rhythmic motor patterns. J. Neurophysiol. 90, 531–538. Eisenhart, F.J., Cacciatore, T.W., 2000. Biomechanics of hydroskeletons, Hocker, C.G., Yu, X., Friesen, W.O., 2000. Heterogeneous circuits generate lessons learned from studies of crawling in the medicinal leech. In: Winters, swimming movements in the medicinal leech. J. Comp. Physiol. 186, 871– J., Crago, P. (Eds.), Biomechanics and Neural Control of Movement. 883. Springer-Verlag, New York, pp. 206–220. Huang, Y., Jellies, J., Johansen, K.M., Johansen, J., 1998. Development and Kristan Jr., W.B., Stent, G.S., 1975. Peripheral feedback in the leech swimming. pathway formation of peripheral neurons during leech embryogenesis. J. In: Proceedings of the CSH Symposium on Quantitative Biology: The Comp. Neurol. 397, 394–402. Synapse, vol. 40. pp. 663–674. Ikeda, K., Wiersma, C.A.G., 1964. Autogenic rhythmicity in the abdominal Kristan Jr., W.B., Stent, G.S., Ort, C.A., 1974a. Neuronal control of swimming ganglia of the crayfish, the control of the swimmeret movements. Comp. in the medicinal leech. I. Dynamics of the swimming rhythm. J. Comp. Biochem. Physiol. 12, 107–115. Physiol. 94, 97–119. Ivanov, A.I., Calabrese, R.L., 2000. Intracellular Ca2+ dynamics during spon- Kristan Jr., W.B., Stent, G.S., Ort, C.A., 1974b. Neuronal control of swimming taneous and evoked activity of leech heart interneurons. Low-threshold Ca in the medicinal leech. III. Impulse patterns of the motor neurons. J. Comp. Currents and graded synaptic transmission. J. Neurosci. 20, 4930–4943. Physiol. 94, 155–176. Ivanov, A.I., Calabrese, R.L., 2003. Modulation of spike-mediated synaptic Kristan Jr., W.B., Weeks, J.C., 1983. Neurons controlling the initiation, gen- transmission by presynaptic background Ca2+ in leech heart interneurons. J. eration, and modulation of leech swimming. In: Roberts, A., Roberts, B. Neurosci. 23, 1206–1218. (Eds.), Neural Origin of Rhythmic Movements. Cambridge University Jezzini, S.H., Hill, A.A.V., Kuzyk, P., Calabrese, R.L., 2004. Detailed model of Press, Cambridge, pp. 243–260. intersegmental coordination in the timing network of the leech heartbeat Kuffler, S.W., Potter, D.D., 1964. Glia in the central nervous system, physio- central pattern generator. J. Neurophysiol. 91, 958–977. logical properties and neuron-glial relationship. J. Neurophysiol. 27, 290– Jing, J., Gillette, R., 2003. Directional avoidance turns encoded by single 320. interneurons and sustained by multifunctional serotonergic cells. J. Neu- Kuhlman, J.R., Li, C., Calabrese, R.L., 1985a. FMRFamide-like substances in rosci. 23, 3039–3051. the leech. I. Immunocytochemical localization. J. Neurosci. 5, 2301–2309. Jordan, C.E., 1998. Scale effects in the kinematics and dynamics of swimming Kuhlman, J.R., Li, C., Calabrese, R.L., 1985b. FMRFamide-like substances in leeches. Can. J. Zool. 76, 1869–1877. the leech. II. Bioactivity on the heartbeat system. J. Neurosci. 5, 2310–2317. Karrer, T., Sahley, C.L., 1988. Discriminative conditioning alters food prefer- Lent, C.M., 1977. The Retzius cells within the central nervous system of ences in the leech, Haemopis marmorata. Behav. Neural Biol. 50, 311–324. leeches. Prog. Neurobiol. 8, 81–117. Katz, P.S., 1995. Intrinsic and extrinsic neuromodulation of motor circuits. Curr. Lent, C.M., 1985. Serotonergic modulation of the feeding behavior of the Opin. Neurobiol. 5, 799–808. medicinal leech. Brain Res. Bull. 14, 643–655. Kiehn, O., Butt, S.J., 2003. Physiological, anatomical and genetic identification Lent, C.M., Dickinson, M.H., 1984. Serotonin integrates the feeding behavior of of CPG neurons in the developing mammalian spinal cord. Prog. Neurobiol. the medicinal leech. J. Comp. Physiol. A 154, 457–471. 70, 347–361. Lent, C.M., Dickinson, M.H., 1987. On the termination of ingestive behaviour Kier, W.M., Smith, W.K., 1985. Tongues, tentacles, and trunks: the biomecha- by the medicinal leech. J. Exp. Biol. 131, 1–15. nics of movement in muscular-hydrostats. Zool. J. Linn. Soc. 83, 307–324. Lent, C.M., Dickinson, M.H., 1988. The neurobiology of feeding in leeches. Kornreich, L., Kleinhaus, A.L., 1999. Postingestive chemosensation and feed- Sci. Am. 258, 98–103. ing by leeches. Physiol. Behav. 67, 635–641. Lent, C.M., Dickinson, M.H., Marshall, C.G., 1989. Serotonin and leech Krahl, B., Zerbst-Boroffka, I., 1983. Blood pressure in the leech, Hirudo feeding behavior: obligatory neuromodulation. Am. Zool. 29, 1241– medicinalis. J. Exp. Biol. 107, 163–168. 1254. Kramer, A.P., 1981. The nervous system of the glossiphoniid leech, Haemen- Lent, C.M., Fliegner, K.H., Freedman, E., Dickinson, M.H., 1988. Ingestive teria ghiliani. II. Synaptic pathways controlling body wall shortening. J. behaviour and physiology of the medicinal leech. J. Exp. Biol. 137, 513– Comp. Physiol. A 144, 449–458. 527. Kretz, J.R., Stent, G.S., Kristan Jr., W.B., 1976. Photosensory input pathways in Lent, C.M., Zundel, D., Freedman, E., Groome, J.R., 1991. Serotonin in the the medicinal leech. J. Comp. Physiol. 106, 1–37. leech central nervous system: anatomical correlates and behavioral effects. Kristan Jr., W.B., 1974. Neural control of swimming in the leech. Am. Zool. 14, J. Comp. Physiol. A 168, 191–200. 991–1001. Lewis, J.E., 1999. Sensory processing and the network mechanisms for reading Kristan Jr., W.B., 1980. Generation of rhythmic motor patterns. In: Pinsker, neuronal population codes. J. Comp. Physiol. 185a, 373–378. H.M., Willis, W.D. (Eds.), Information Processing in the Nervous System. Lewis, J.E., Kristan Jr., W.B., 1998a. A neuronal network for computing Raven Press, New York, pp. 241–261. population vectors in the leech. Nature 391, 76–79. 324 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

Lewis, J.E., Kristan Jr., W.B., 1998b. Quantitative analysis of a directed Masino, M.A., Calabrese, R.L., 2002a. Phase relationships between segmen- behavior in the medicinal leech; implications for organizing motor output. tally organized oscillators in the leech heartbeat pattern generating network. J. Neurosci. 18, 1571–1582. J. Neurophys. 87, 1572–1585. Lewis, J.E., Kristan Jr., W.B., 1998c. Representation of touch localization by a Masino, M.A., Calabrese, R.L., 2002b. A functional asymmetry in the leech population of leech touch sensitive neurons. J. Neurophysiol. 80, 2584– heartbeat timing network is revealed by driving the network across various 2592. cycle periods. J. Neurosci. 22, 4418–4427. Li, C., Calabrese, R.L., 1987. FMRFamide-like substances in the leech. III. Masino, M.A., Calabrese, R.L., 2002c. Period differences between segmental Biochemical characterization and physiological effects. J. Neurosci. 7, 595– oscillators produce intersegmental phase differences in the leech heartbeat 603. pattern generating network. J. Neurophys. 87, 1603–1615. Li, Y., Perruccio, E., Zhang, X., Kleinhaus, A.L., 2001. Bitter substances Mason, A., Kristan Jr., W.B., 1982. Neuronal excitation, inhibition and mod- suppress afferent responses to an appetitive mixture: evidence for peripheral ulation of leech longitudinal muscle. J. Comp. Physiol. 146, 527–536. integration of chemosensory stimuli. J. Neurobiol. 49, 255–263. Mesce, K.A., Crisp, K.M., Gilchrist, L.S., 2001. Mixtures of octopamine and Lockery, S.R., Kristan Jr., W.B., 1990a. Distributed processing of sensory serotonin have nonadditive effects on the CNS of the medicinal leech. J. information in the leech. I. Input-output relations of the local bending reflex. Neurophysiol. 85, 2039–2046. J. Neurosci. 1811–1815. Mazzoni, A., Garcia-Perez, E., Zoccolan, D., Graziosi, S., Torre, V., 2005. Lockery, S.R., Kristan Jr., W.B., 1990b. Distributed processing of sensory Quantitative characterization and classification of leech behavior. J. Neu- information in the leech. II. Identification of interneurons contributing to the rophysiol. 93, 580–593. local bending reflex. J. Neurosci. 10, 1816–1829. Miller, J.B., 1975. The length-tension relationship of the dorsal longitudinal Lockery, S.R., Kristan Jr., W.B., 1991. Two forms of sensitization of the local muscle of a leech. J. Exp. Biol. 62, 43–53. bending reflex of the medicinal leech. J. Comp. Physiol. A 168, 165–177. Misell, L.M., Shaw, B.K., Kristan Jr., W.B., 1998. Behavioral hierarchy in the Lockery, S.R., Sejnowski, T.J., 1992. Distributed processing of sensory infor- medicinal leech, Hirudo medicinalis; feeding as a dominant behavior. mation in the leech. III. A dynamical neural network model of the local Behav. Brain Res. 90, 13–21. bending reflex. J. Neurosci. 12, 3877–3895. Mistick, D.C., 1978. Neurons in the leech that facilitate an avoidance behaviour Lockery, S.R., Sejnowski, T.J., 1993. A lower bound on the delectability of following nearfield water disturbances. J. Exp. Biol. 75, 1–23. nonassociative learning in the local bending reflex on the medicinal leech. Modney, B.K., Sahley, C.L., Muller, K.J., 1997. Regeneration of a central Behav. Neural Biol. 59, 208–224. synapse restores nonassociative learning. J. Neurosci. 17, 6478–6482. Lockery, S.R., Wittenberg, G., Kristan Jr., W.B., Cottrell, G.W., 1989. Function Muller, K.J., Nicholls, J.G., Stent, G.S., 1981. Neurobiology of the Leech. Cold of identified interneurons in the leech elucidated using networks trained by Spring Harbor Laboratory, Cold Spring Harbor, New York. back-propagation. Nature 340, 468–471. Mulloney, B., Murchison, D., Chrachri, A., 1993. Modular organization of Lu, J., Dalton IV, J.F., Stokes, D.R., Calabrese, R.L., 1997. Functional role of pattern-generating circuits in a segmental motor system, the swimmerets of Ca2+ currents in graded and spike-mediated synaptic transmission between crayfish. Semin. Neurosci. 5, 49–57. leech heart interneurons. J. Neurophysiol. 77, 1779–1794. Mulloney, B., Selverston, A.I., 1974. Organization of the stomatogastric gang- Lu, J., Gramoll, S., Schmidt, J., Calabrese, R.L., 1999. Motor pattern switching lion of the spiny lobster. I. Neurons driving the lateral teeth. J. Comp. in the heartbeat pattern generator of the medicinal leech; membrane proper- Physiol. 91, 1–32. ties of switch interneurons and their role in switching. J. Comp. Physiol. A Nadim, F., Olsen, 1H., DeSchutter, E., Calabrese, R.L., 1995. Modeling the 184, 311–324. leech heartbeat elemental oscillator. I. Interactions of intrinsic and synaptic Macagno, E.R., 1980. Number and distribution of neurons in leech segmental currents. J. Comput. Neurosci. 2, 215–235. ganglia. J. Comp. Neurol. 15, 283–302. Nicholls, J.G., Baylor, D.A., 1968. Specific modalities and receptive fields of Magni, F., Pellegrino, M., 1978. Neural mechanism underlying the segmental sensory neurons in CNS of the leech. J. Neurophysiol. 31, 740–756. and generalized cord shortening reflexes in the leech. J. Comp. Physiol. A Nicholls, J.G., Kuffler, S.W., 1964. Extracellular space as a pathway for 124, 339–351. exchange between blood and neurons in the central nervous system of Mangan, P.S., Cometa, A.K., Friesen, W.O., 1994a. Modulation of swimming the leech, ionic composition of glial cells and neurons. J. Neurophysiol. 27, behavior in the medicinal leech. IV. Serotonin-induced alterations of 645–671. synaptic interactions between neurons of the swim circuit. J. Comp. Physiol. Nicholls, J.G., Purves, D., 1970. Monosynaptic chemical and electrical con- 175, 723–736. nexions between sensory and motor cells in the central nervous system of Mangan, P.S., Curran, G.A., Hurney, C.A., Friesen, W.O., 1994b. Modulation the leech. J. Physiol. 209, 647–667. of swimming behavior in the medicinal leech. III. Control of cellular Nicholls, J.G., Wallace, B.G., 1978. Modulation of transmission at an inhibitory properties in motor neurons by serotonin. J. Comp. Physiol. 175, 709– synapse in the central nervous system of the leech. J. Physiol. 281, 157– 722. 170. Mann, K.M., 1962. Leeches (Hirudinea). Their structure, physiology, ecology Norris, B.J., Calabrese, R.L., 1987. Identification of motor neurons that contain and embryology. Pergamon Press, New York, pp. 79–100. a FMRFamide-like peptide and the effects of FMRFamide on longitudinal Maranto, A.R., Calabrese, R.L., 1984a. Neural control of the hearts in the leech, muscle in the medicinal leech, Hirudo medicinalis. J. Comp. Neurol. 266, Hirudo medicinalis. I. Structure, innervation and electrical coupling of the 95–111. hearts. J. Comp. Physiol. A 154, 367–380. Nusbaum, M.P., 1986. Synaptic basis of swim initiation in the leech. III. Maranto, A.R., Calabrese, R.L., 1984b. Neural control of the hearts in the leech, Synaptic effects of serotonin-containing interneurones cells 21 and 61 on Hirudo medicinalis. II. Myogenic activity and phasic control by heart motor swim CPG neurones cells 18 and 208. J. Exp. Biol. 122, 303–321. neurons. J. Comp. Physiol. A 154, 389–391. Nusbaum, M.P., El Manira, A., Gossard, J.-P., Rossignol, S., 1997. Presynaptic Marder, E., Calabrese, R.L., 1996. Principles of rhythmic motor pattern mechanisms during rhythmic activity in vertebrates and invertebrates. In: generation. Physiol. Rev. 763, 687–717. Stein, P.S.G., Grillner, S., Selverston, A.I., Stuart, D.G. (Eds.), Neurons Marin-Burgin, A., Eisenhart, J.F., Baca, S.M., Kristan Jr., W.B., French, K.A., Networks and Motor Behavior. MIT Press, Cambridge, MA, pp. 237– 2005. Sequential development of electrical and chemical synaptic connec- 254. tions generates a specific behavioral circuit in the leech. J. Neurosci. 25, Nusbaum, M.P., Friesen, W.O., Kristan Jr., W.B., Pearce, R.A., 1987. Neural 2478–2489. mechanisms generating the leech swimming rhythm; swim-initiator neu- Marshall, C.G., Lent, C.M., 1984. Calcium-dependent action potentials in leech rons excite the network of swim oscillator neurons. J. Comp. Physiol. A 161, giant salivary cells. J. Exp. Biol. 113, 367–380. 355–366. Marshall, C.G., Lent, C.M., 1988. Excitability and secretory activity in the Nusbaum, M.P., Kristan Jr., W.B., 1986. Swim initiation in the leech by salivary gland cells of jawed leeches (Hirudinea: Gnathobdellida). J. Exp. serotonin-containing interneurones, cells 21 and 61. J. Exp. Biol. 122, Biol. 137, 89–105. 277–302. W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 325

O’Gara, B.A., Abbasi, A., Kaniecki, K., Sarder, F., Liu, J., Narine, L.H., 1999a. Peterson, E.L., 1983b. Generation and coordination of heartbeat timing oscilla- Pharmacological characterization of the response of the leech pharynx to tion in the medicinal leech. II. Intersegmental coordination. J. Neurophy- acetylcholine. J. Exp. Zool. 284, 729–741. siol. 49, 627–638. O’Gara, B.A., Illuzzi, F.A., Chung, M., Portnoy, A.D., Fraga, K., Frieman, V.B., Peterson, E.L., Calabrese, R.L., 1982. Dynamic analysis of a rhythmic neural 1999b. Serotonin induces four pharmacologically separable contractile circuit in the leech Hirudo medicinalis. J. Neurophysiol. 47, 256–271. responses in the pharynx of the leech Hirudo medicinalis. Gen. Pharmacol. Phillips, C.E., Friesen, W.O., 1982. Ultrastructure of the water-movement- 32, 669–681. sensitive sensilla in the medicinal leech. J. Neurobiol. 13, 473–486. O’Gara, B.A., Brown, P.L., Dlugosch, D., Kandiel, A., Ku, J.W., Geier, J.K., Pinato, G., Battiston, S., Torre, V., 2000. Statistical independence and neural Henggeler, N.C., Abbasi, A., Kounalakis, N., 2000. Regulation of phar- computation in the leech ganglion. Biol. Cybern. 83, 119–130. yngeal motility by FMRFamide and related peptides in the medicinal leech, Poon, M., 1976. A neuronal network generating the swimming rhythm of the Hirudo medicinalis. Invert. Neurosci. 4, 41–53. leech. Ph.D. Thesis, University of California, Berkeley. O’Gara, B.A., Chae, H., Latham, L.B., Friesen, W.O., 1991. Modification of Poon, M., Friesen, W.O., Stent, G.S., 1978. Neural control of swimming in the leech behavior patterns by reserpine-induced amine depletion. J. Neurosci. medicinal leech. V. Connexions between the oscillatory interneurones and 11, 96–110. the motor neurones. J. Exp. Biol. 75, 45–63. O’Gara, B.A., Friesen, W.O., 1995. Termination of leech swimming activity by Prinz, A.A., Thirumalai, V., Marder, E., 2003. The functional consequences of a previously identified swim trigger neuron. J. Comp. Physiol. 177, 627– changes in the strength and duration of synaptic inputs to oscillatory 636. neurons. J. Neurosci. 23, 943–954. Olsen, 1.H., Calabrese, R.L., 1996. Activation of intrinsic and synaptic Reynolds, S., French, K., Baader, A., Kristan Jr., W.B., 1998a. Staging of currents in leech heart interneurons by realistic waveforms. J. Neurosci. middle and late embryonic development in the medicinal leech, Hirudo 16, 4958–4970. medicinalis. J. Comp. Neurol. 402, 155–167. Olsen, 1.H., Nadim, F., Calabrese, R.L., 1995. Modeling a neuronal oscillator. Reynolds, S., French, K., Baader, A., Kristan Jr., W.B., 1998b. Development of II. Exploring the parameter space. J. Comput. Neurosci. 2, 237–257. spontaneous and evoked behaviors in the medicinal leech. J. Comp. Neurol. Opdyke, C.A., Calabrese, R.L., 1994. A persistent sodium current contributes to 402, 168–180. oscillatory activity in heart interneurons of the medicinal leech. J. Comp. Riley, J.R., Greggers, U., Smith, A.D., Stach, S., Reynolds, D.R., Stollhoff, N., Physiol. A 175, 781–789. Brandt, R., Schaupp, F., Menzel, R., 2003. The automatic pilot of honey- Orlovsky, G.N., Deliangina, T.G., Grillner, S., 1999. Neuronal Control of bees. Proc. Roy. Soc. Lond. B 270, 2421–2424. Locomotion, From Mollusc to Man. Oxford, New York. Roberts, A., Walford, A., Soffe, S.R., Yoshida, M., 1999. Motoneurons of the Ort, C.A., Kristan Jr., W.B., Stent, G.S., 1974. Neuronal control of swimming in axial swimming muscles in hatchling Xenopus tadpoles: features, distribu- the medicinal leech. II. Identification and connections of motor neurons. J. tion, and central synapses. J. Comp. Neurol. 411, 472–486. Comp. Physiol. 94, 121–154. Ross, W.N., Arechiga, H., Nicholls, J.G., 1987. Optical recording of calcium Pastor, J., Soria, B., Belmonte, C., 1996. Properties of the nociceptive neurons and voltage transients following impulses in cell bodies and processes of of the leech segmental ganglion. J. Neurophysiol. 75, 2268–2279. identified leech neurons in culture. J. Neurosci. 7, 3877–3887. Payton, B., 1981. History of medicinal leeching and early medical references. Sahley, C.L., 1994. Serotonin depletion impairs but does not eliminate classical In: Muller, K.J., Nicholls, J.G., Stent, G.S. (Eds.), Neurobiology of the conditioning in the leech Hirudo medicinalis. Behav. Neurosci. 108, 1043– Leech. Cold Spring Harbor, New York, pp. 27–34. 1052. Pearce, R.A., Friesen, W.O., 1984. Intersegmental coordination of leech Sahley, C.L., 1995. What we have learned from the study of learning in the swimming, comparison of in situ and isolated nerve cord activity with leech. J. Neurobiol. 27, 434–445. body wall movement. Brain Res. 299, 363–366. Sahley, C.L., Modney, B.K., Boulis, N.M., Muller, K.J., 1994. The S cell: an Pearce, R.A., Friesen, W.O., 1985a. Intersegmental coordination of the leech interneuron essential for sensitization and full dishabituation of leech swimming rhythm. I. Intrinsic period gradient and coupling strength. J. shortening. J. Neurosci. 14, 6715–6721. Neurophysiol. 54, 1444–1459. Sahley, C.L., Ready, D.F., 1988. Associative learning modifies two behaviors in Pearce, R.A., Friesen, W.O., 1985b. Intersegmental coordination of the leech the leech, Hirudo medicinalis. J. Neurosci. 8, 4612–4620. swimming rhythm. II. Comparison of long and short chains of ganglia. J. Sakurai, A., Katz, P.S., 2003. Spike timing-dependent serotonergic neuromo- Neurophysiol. 54, 1460–1472. dulation of synaptic strength intrinsic to a central pattern generator circuit. J. Pearce, R.A., Friesen, W.O., 1988. A model for intersegmental coordination in Neurosci. 23, 10745–10755. the leech nerve cord. Biol. Cybern. 58, 301–311. Salinas, E., Abbott, L.F., 1995. Transfer of coded information from sensory to Pearson, K.G., 2000. Motor Systems. Curr. Opin. Neurobiol. 10, 649–654. motor networks. J. Neurosci. 15, 6461–6474. Pearson, K.G., Ramirez, J.M., 1990. Influence of input from the forewing Salzberg, B.M., Davila, H.V., Cohen, L.B., 1973. Optical recording of impulses stretch receptors on motor neurones in flying locusts. J. Exp. Biol. 151, 317– in individual neurones of an invertebrate central nervous system. Nature 340. 246, 508–509. Pearson, K.G., Ramirez, J.M., 1997. Sensory modulation of pattern-generat- Sargent, P.B., 1977. Synthesis of acetylcholine by excitatory motoneurons in ing circuits. In: Stein, P.S.G., Grillner, S., Selverston, A.I., Stuart, D.G. central nervous system of the leech. J. Neurophysiol. 40, 453–460. (Eds.), Neurons, Networks, and Motor Behavior. MIT Press, Cambridge, Sasaki, K., Burrows, M., 2003. Proprioceptors monitoring forces in a locust MA, pp. 225–235. hind leg during kicking form negative feedback loops with flexor tibiae Pearson, K.G., Reye, D.N., Robertson, R.M., 1983. Phase-dependent influences motor neurons. J. Exp. Biol. 206, 759–769. of wing stretch receptors flight rhythm in the locusts. J. Neurophysiol. 49, Satterlie, R.A., Norekian, T.P., Pirtle, T.J., 2000. Serotonin-induced spike 1168–1181. narrowing in a locomotor pattern generator permits increases in cycle Perrins, R., Weiss, K.R., 1998. Compartmentalization of information processing frequency during accelerations. J. Neurophysiol. 83, 2163–2170. in an Aplysia feeding circuit interneuron through membrane properties and Sawada, M., Wilkinson, J.M., McAdoo, D.J., Coggeshall, R.E., 1976. The synaptic interactions. J. Neurosci. 18, 3977–3989. identification of two inhibitory cells in each segmental ganglion of the leech Perruccio, L., Kleinhaus, A.L., 1996. Anatomical pathways connecting lip and studies on the ionic mechanism of the inhibitory junctional potential sensory structures and central nervous system in hirudinid leeches visua- produced by these cells. J. Neurobiol. 7, 435–445. lized by carbocyanine dyes and laser scanning confocal microscopy. Invert. Sawyer, R., 1981. Leech biology and behavior. In: Muller, K.J., Nicholls, Neurosci. 2, 183–188. J.G., Stent, G.S. (Eds.), Neurobiology of the Leech. Cold Spring Harbor, Peterson, E.L., 1983a. Generation and coordination of heartbeat timing oscilla- New York, pp. 7–26. tion in the medicinal leech. I. Oscillation in isolated ganglia. J. Neurophy- Sawyer, R., 1986. Leech Biology and Behaviour, vols. 1–3. Clarendon Press, siol. 49, 611–626. Oxford. 326 W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327

Schmidt, J., Calabrese, R.L., 1992. Evidence that acetylcholine is an inhibitory Taylor, G., 1952. Analysis of the swimming of long and narrow animals. Proc. transmitter of heart interneurons in the leech. J. Exp. Biol. 171, 329–347. Roy. Soc. Lond. A. 214, 158–183. Schu¨lter, E., 1933. Die Bedeutung des Centralnervensystems von Hirudo Teshiba, T., Shamsian, A., Yashar, B., Yeh, S.R., Edwards, D.H., Krasne, F.B., medicinalis fur Locomotion and Raumorientierung. Z. Wiss. Zool. 143, 2001. Dual and opposing modulatory effects of serotonin on crayfish lateral 538–593. giant escape command neurons. J. Neurosci. 21, 4523–4529. Sekirnjak, C., Vissel, B., Bollinger, J., Faulstich, M., du Lac, S., 2003. Purkinje Theunissen, F., Roddey, J.C., Stufflebeam, S., Clague, H., Miller, J.P., 1996. cell synapses target physiologically unique brainstem neurons. J. Neurosci. Information theoretic analysis of dynamical encoding by four identified 23, 6392–6398. primary sensory interneurons in the cricket cercal system. J. Neurophysiol. Selverston, A.I. (Ed.), 1985. Model Neural Networks and Behavior. Plenum 75, 1345–1364. Press, New York. Thompson, W.J., Stent, G.S., 1976a. Neuronal control of heartbeat in the Selverston, A.I., Rabinovich, M.I., Abarbanel, H.D., Elson, R., Szucs, A., Pinto, medicinal leech. I. Generation of the vascular constriction rhythm by heart R.D., Huerta, R., Varona, P., 2000. Reliable circuits from irregular neurons: motor neurons. J. Comp. Physiol. 111, 261–279. a dynamical approach to understanding central pattern generators. J. Thompson, W.J., Stent, G.S., 1976b. Neuronal control of heartbeat in the Physiol. Paris 94, 357–374. medicinal leech. II. Intersegmental coordination of heart motor neuron Sharp, A.A., O’Neil, M.B., Abbott, L.F., Marder, E., 1993. Dynamic clamp, activity by heart interneurons. J. Comp. Physiol. 111, 281–307. artificial conductances in biological neurons. TINS 16, 389–394. Thompson, W.J., Stent, G.S., 1976c. Neuronal control of heartbeat in the Shaw, B.K., Kristan Jr., W.B., 1995. The whole-body shortening reflex of the medicinal leech. III. Synaptic relations of heart interneurons. J. Comp. medicinal leech: motor pattern, sensory basis, and interneuronal pathways. Physiol. A 111, 309–333. J. Comp. Physiol. A 177, 667–681. Thorogood, M.S., Brodfuehrer, P.D., 1995. The role of glutamate in swim Shaw, B.K., Kristan Jr., W.B., 1997. The neuronal basis of behavioral choice initiation in the medicinal leech. Invert. Neurosci. 1, 223–233. between swimming, shortening in the leech; control is not selectively Trueta, C., Mendez, B., De-Miguel, 2003. Somatic exocytosis of serotonin exercised at higher circuit levels. J. Neurosci. 17, 786–795. mediated by L-type calcium channels in cultured leech neurones. J. Physiol. Shaw, B.K., Kristan Jr., W.B., 1999. The relative roles of the S cell network and 547, 405–416. parallel interneuronal pathways in the whole-body shortening reflex of the Trueta, C., Sanchez-Armass, S., Morales, M.A., De-Miguel, F.F., 2004. Cal- medicinal leech. J. Neurophysiol. 82, 1114–1123. cium-induced calcium release contributes to somatic secretion of serotonin Sherrington, C.S., 1906. The Integrative Action of the Nervous System. Yale in leech Retzius neurons. J. Neurobiol. 61, 309–316. University Press, New Haven. Uexku¨ll, J., 1905. Studien u¨ber den Tonus. III. Die Blutegel. Z. Biol. 46, 372– Simon, T.W., Opdyke, C.A., Calabrese, R.L., 1992. Modulatory effects of 402.

FMRF-NH2 on outward currents and oscillatory activity in heart interneur- Wadepuhl, M., Beyn, W.-J., 1989. Computer simulation of the hydrostatic ons of the medicinal leech. J. Neurosci. 12, 525–537. skeleton. The physical equivalent, mathematics and application to worm- Simon, T.W., Schmidt, J., Calabrese, R.L., 1994. Modulation of high threshold like forms. J. Theor. Biol. 136, 379–402. transmission between heart interneurons of the medicinal leech by FMRF- Walle´n, P., Williams, T.L., 1984. Fictive locomotion in the lamprey spinal cord

NH2. J. Neurophysiol. 71, 454–466. in vitro compared with swimming in the intact and spinal animal. J. Physiol. Skierczynski, B.A., Wilson, R.J.A., Kristan Jr., W.B., Skalak, R., 1996. A 347, 225–239. biomechanical model of the leech. J. Theoret. Biol. 181, 329–342. Wang, J.W., Wong, A.M., Flores, J., Vosshall, L.B., Axel, R., 2003. Two-photon Skinner, F.K., Mulloney, B., 1998a. Intersegmental coordination in inverte- calcium imaging reveals an odor-evoked map of activity in the fly brain. Cell brates and vertebrates. Curr. Opin. Neurobiol. 86, 725–732. 112, 271–282. Smith, P.H., Page, C.H., 1974. Nerve cord sheath receptors activate the large Weeks, J.C., 1981. Neuronal basis of leech swimming: separation of swim fiber system in the leech. J. Comp. Physiol. 90, 311–320. initiation, pattern generation and intersegmental coordination by selective Sparks, D.L., Kristan Jr., W.B., Shaw, B.K., 1997. The role of population coding lesions. J. Neurophysiol. 45, 698–723. in the control of movement. In: Stein, P.S.G., Grillner, S., Selverston, A.I., Weeks, J.C., 1982a. Synaptic basis of swim initiation in the leech. I. Connec- Stuart, D.G. (Eds.), Neurons, Networks, and Motor Behavior. MIT Press, tions of a swim-initiating neuron cell 204. with motor neurons and pattern- Cambridge, MA, pp. 21–32. generating ‘‘oscillator’’ neurons. J. Comp. Physiol. A 148, 253–263. Staras, K., Kemenes, G., Benjamin, P.R., 1999. Cellular traces of behavioral Weeks, J.C., 1982b. Synaptic basis of swim initiation in the leech. II. A pattern- classical conditioning can be recorded at several specific sites in a simple generating neuron cell 208. which mediates motor effects of swim-initiating nervous system. J. Neurosci. 19, 347–357. neurons. J. Comp. Physiol. 148, 265–279. Stein, P.S.G., Grillner, S., Selverston, A.I., Stuart, D.G. (Eds.), 1997. Neurons Weeks, J.C., 1982c. Segmental specialization of a leech swim-initiating inter- Networks and Motor Behavior. MIT Press, Cambridge, MA. neuron, cell 205. J. Neurosci. 2, 972–985. Stent, P.G.S., Kristan Jr., W.B., Friesen, W.O., Ort, C.A., Poon, M., Calabrese, Weeks, J.C., Kristan Jr., W.B., 1978. Initiation, maintenance and modulation of R.L., 1978. Neuronal generation of the leech swimming movement. Science swimming in the medicinal leech by the activity of a single neuron. J. Exp. 200, 1348–1356. Biol. 77, 71–88. Stent, P.G.S., Kristan Jr., W.B., Torrence, S.A., French, K.A., Weisblat, D.A., Wenning, A., Cymbalyuk, G.S., Calabrese, R.L., 2004a. Heartbeat Control in 1992. Development of the leech nervous system. Int. Rev. Neurobiol. 33, Leeches. I. Constriction pattern and neural modulation of blood pressure in 109–193. intact animals. J. Neurophysiol. 91, 382–396. Stern-Tomlinson, W., Nusbaum, M.P., Perez, L.E., Kristan Jr., W.B., 1986. A Wenning, A., Hill, A.A., Calabrese, R.L., 2004b. Heartbeat control in leeches. kinematic study of crawling behavior in the leech, Hirudo medicinalis.J. II. Fictive motor pattern. J. Neurophysiol. 91, 397–409. Comp. Physiol. 158, 593–603. Whitaker, I.S., Rao, J., Izadi, D., Butler, P.E., 2004. Historical article: Hirudo Stuart, A.E., 1970. Physiological and morphological properties of motorneur- medicinalis: ancient origins of, and trends in the use of medicinal leeches ones in the central nervous system of the leech. J. Physiol. Lond. 209, 627– throughout history. Br. J. Oral Maxillofac. Surg. 42, 133–137. 646. Willard, A.L., 1981. Effects of serotonin on the generation of the motor program Sze´kely, G., 1964. Logical network for controlling limb movement in Urodela. for swimming by the medicinal leech. J. Neurosci. 1, 936–944. Acta. Physiol. Acad. Sci. Hung. 27, 285–289. Wilson, R.J., Kleinhaus, A.L., 2000. Segmental control of midbody peristalsis Taylor, A.L., Cottrell, G.W., Kleinfeld, D., Kristan Jr., W.B., 2003. Imaging during the consummatory phase of feeding in the medicinal leech, Hirudo reveals synaptic targets of a swim-terminating neuron in the leech CNS. J. medicinalis. Behav. Neurosci. 114, 635–646. Neurosci. 23, 11402–11410. Wilson, D.M., 1961. The central nervous control of the flight in a locust. J. Exp. Taylor, A.L., Cottrell, G.W., Kristan Jr., W.B., 2000. A model of the leech Biol. 38, 471–490. segmental swim central pattern generator. Neurocomputing 32–33, 573– Wilson, R.I., Turner, G.C., Laurent, G., 2004. Transformation of olfactory 584. representations in the Drosophila antennal lobe. Science 303, 366–370. W.B. Kristan Jr. et al. / Progress in Neurobiology 76 (2005) 279–327 327

Wilson, R.J.A., Lewis, J.E., Kristan Jr., W.B., Kleinhaus, A.L., 1996. An Yu, X., 2001. Sensory Feedback Control of Swimming Activity in the Medicinal increase in activity of serotonergic Retzius neurons may not be necessary Leech. Doctoral Dissertation. University of Virginia. for the consummatory phase of feeding in the leech, Hirudo medicinalis.J. Yu, X., Friesen, W.O., 2004. Entrainment of leech swimming activity by the Exp. Biol. 29, 1405–1414. ventral stretch receptor. J. Comp. Physiol. A 190, 939–949. Wilson, R.J.A., Skierczynski, B.A., Blackwood, S., Skalak, R., Kristan Jr., Yu, X., Nguyen, B., Friesen, W.O., 1999. Sensory feedback can coordinate the W.B., 1996b. Mapping motor neurone activity to overt behaviour in the swimming activity of the leech. J. Neurosci. 19, 4634–4643. leech. Internal pressures produced during locomotion. J. Exp. Biol. 199, Yvert, B., Branchereau, P., Meyrand, P., 2004. Multiple spontaneous rhythmic 1415–1428. activity patterns generated by the embryonic mouse spinal cord occur Wilson, R.J.A., Skierczynski, B.A., Meyer, J.K., Skalak, R., Kristan Jr., W.B., within a specific developmental time window. J. Neurophysiol. 91, 1996a. Mappingmotor neuron activity to overtbehavior in the leech. I. Passive 2101–2109. biomechanical properties of the body wall. J. Comp. Physiol. 178, 637–654. Zhang, X., Wilson, R.J., Li, Y., Kleinhaus, A.L., 2000. Chemical and thermal Wittenberg, G., Kristan Jr., W.B., 1992a. Analysis and modeling of the multi- stimuli have short-lived effects on the Retzius cell in the medicinal leech. J. segmental coordination of shortening behavior in the medicinal leech. I. Neurobiol. 43, 304–311. Motor output pattern. J. Neurophysiol. 68, 1683–1692. Zheng, M., Iwasak, T., Friesen, W.O., 2004. Systems approach to modeling the Wittenberg, G., Kristan Jr., W.B., 1992b. Analysis and modeling of the multi- neuronal CPG for leech swimming. In: Proceedings of the Engineering in segmental coordination of shortening behavior in the medicinal leech. II. Medicine and Biology Society Conference, vol. 1. pp. 703–706. Role of identified interneurons. J. Neurophysiol. 68, 1693–1707. Zipser, B., 1979. Identifiable neurons controlling penile eversion in the leech. J. Wolpert, S.X., Friesen, W.O., 2000. On the parametric stability of a central Neurophysiol. 42, 455–464. pattern generator. Neurocomputing 32, 603–608. Zoccolan, D., Giachetti, A., Torre, V., 2001. The use of optical flow to Wolpert, S.X., Friesen, W.O., Laffely, A., 2000. Silicon model of the Hirudo characterize muscle contraction. J. Neurosci. Meth. 110, 65–80. swim oscillator. IEEE Eng. Med. Biol. 19, 64–75. Zoccolan, D., Pinato, G., Torre, V., 2002. Highly variable spike trains underlie Yau, K.W., 1976. Physiological properties and receptive fields of mechano- reproducible sensorimotor responses in the medicinal leech. J. Neurosci. 22, sensory neurones in the head ganglion of the leech: comparison with 10790–10800. homologous cells in segmental ganglia. J. Physiol. 263, 489–512. Zoccolan, D., Torre, V., 2002. Using optical flow to characterize sensory-motor Young, S.R., Dedwylder, R.D., Friesen, W.O., 1981. Responses of the medicinal interactions in a segment of the medicinal leech. J. Neurosci. 22, 2283– leech to water waves. J. Comp. Physiol. A 144, 111–116. 2298.