The Telson Flexor Neuromuscular System of the Crayfish: I. Homology with the Fast Flexor System

In this paper we describe a neuromuscular system in the tailfan of the crayfish and establish its homology with the neuromuscular systems that provide the flexion powerstroke for tailflip escape responses. This is intended to provide the basis for an examination of the alteration of these circuits during their evolution. More specifically, by improving our understanding of crayfish escape circuitry, and extending it to include the tailfan, we hope to be able to decide whether specific components of these circuits are the result of selection of adaptive features or are due to the influence of non-adaptive evolutionary determinants such as developmental constraints (see Gould & Lewontin, 1979).

Our study takes advantage of the considerable amount of existing data on the organization of crayfish escape behaviour (see below), and of a key feature of arthropod and annelid evolution: the differential modification of repeated body segments within and across species. These differences provide an opportunity to make intersegmental comparisons within a single species among serially homologous neural circuits, with the aim of documenting differences in the structure and function of identified, homologous neurones. Within one animal, adjacent segments in a given compartment (such as two mid-abdominal ganglia in a crayfish) are very similar, so that it is relatively easy to establish homologies and to detect subtle differences between the segments. More widely separated or specialized ganglia within the same compartment (such as the mid-abdominal and terminal ganglia of crayfish), while appearing quite dissimilar on direct comparison, may be homolo-gized by a stepwise comparison via the intervening ganglia. Ganglia in different compartments, such as thoracic and abdominal ganglia in crayfish, may be so different that homologies cannot be established without developmental data (see Bate, Goodman & Spitzer, 1981). However, even in highly dissimilar ganglia some features are usually conserved and can serve to orientate the investigator.

The fast flexor system of the crayfish abdomen is well suited for both intersegmental and interspecific comparisons. The abdomen consists of five similar segments and a highly specialized terminal segment, the tailfan. The fast flexor system, which mediates the powerstroke of the crayfish tailflip escape response, has been studied extensively (Takeda & Kennedy, 1964; Kennedy & Takeda, 1965; Mittenthal & Wine, 1973, 1978; Miller, Hagiwara & Wine, 1985). The neuromuscular system of the tailfan has also been studied previously (Larimer & Kennedy, 1969; Larimer, Eggleston, MasukaWa & Kennedy, 1971; Kramer, Krasne & Wine, 1981) but none of the studies of the tailfan include identification of the motos neurones, or more than tentative establishment of homology.

The fast flexor escape system exists in a wide range of modern species. For example, it appears to be homologous in crayfish and the syncarid crustacean, Anaspides tasmaniae (Silvey & Wilson, 1979), which diverged about 300 million years ago. The telson flexor neuromuscular system is an especially good subject for comparative study because systems that are presumably homologous have been described in other decapod Crustacea (Paul, 1981 ; Paul, Then & Magnuson, 1985).

Additional and more parochial interest in the telson system arises from a controversy about whether the system is a fusion product of two ancestral segments. Johnson (1924) proposed that the sixth abdominal ganglion (G6) is a fusion product, because of the presence of two lateral giant (LG) segments (see also Kondoh & Hisada, 1983) and two motor giant neurones (MoGs) in this hemiganglion, as opposed to one of each in other abdominal hemiganglia. However, Johnson’s identifications were based solely on morphology. Larimer & Kennedy (1969), using physiological techniques, proposed that only 10 phasic motor neurones innervate the telson flexors, including only a single peripheral inhibitor. Since their count matched earlier counts of fast flexor axons in the flexor motor nerve of ganglion 3, and since they found no evidence for either an extra MoG or LG segment, they concluded that the evidence for fusion was not convincing. The issue of fusion is important because, if it did occur, it must have been accompanied by considerable reorganization if we are to account for the doubling of some neurones with no increase in the total number of efferents.

Our task was made easier by recent research which has made it clear that the fast flexor system is not composed of a segmentally iterated series of identical ganglionic networks. Intersegmental differences have been found in the number of motor neurones (Mittenthal & Wine, 1978), the number of interneurones (Kramer et al 1981) and the presence (Mittenthal & Wine, 1973) or strength (Miller et al. 1985) of synaptic connections. Both segmental and intersegmental neurones have been found interposed between fast flexor motor neurones and the giant command neurones that had previously been thought to fire them directly (Kramer et al. 1981 ; Roberts et al 1982 ) Major intersegmental differences have been found in various fast flexor pathways (Mittenthal & Wine, 1973; Miller et al. 1985). In this respect, the sixth ganglion appears to be particularly specialized (Kramer et al. 1981), and therefore worth further examination.

In this paper we present the evidence for homology between the telson flexor neuromuscular system and the fast flexor system of anterior ganglia. In the following papers (Dumont & Wine, 1986a,b), we examine the differences between the fast flexor pathways of the anterior segments and the telson.

Crayfish, Procambarus clarkii, were purchased from a variety of commercial Suppliers and kept in communal tanks.

The telson motor neurones were backfilled with cobaltous chloride (Pitman, Tweedie & Cohen, 1972) using nerve cords isolated from the last three segments of the abdomen (see below). When a cut sixth nerve was immersed in cobaltous chloride solution (2·5 moll⁻¹) for 9h, only the large, phasic telson motor neurones were filled. If a nerve was immersed overnight, the tonic motor neurones were also filled. The cobalt was then precipitated with a few drops of ammonium sulphide. To identify the innervation of particular muscles, individual branches of the sixth nerve were backfilled. The branches innervating the dorsal and ventral portions of the posterior telson flexor (PTF) muscle were quite distinct, but the motor neurones innervating the ventral telson flexor (VTF) leave the main nerve in a range of positions as it runs alongside the VTF muscle, so they could not be backfilled selectively. To determine the innervation of this muscle we compared the innervation of the PTF with backfills of the whole nerve.

Female crayfish fertilize their eggs with stored sperm as the eggs are laid, and then attach them to the swimmerets. Females were isolated as soon as the eggs were laid, so we knew on what day the brood started, and were then kept at room temperature in aerated tap water. Part of each brood was kept until hatching to determine the time required for full development (about 30 days). The embryos (1–2mm long) were dissected away from the egg in Van Harreveld’s saline (1936), along with a small amount of the surrounding chorionic membrane. Crayfish embryos develop with their abdomens flexed, so, after the overlying membranes had been removed, the abdomen was gently extended. The embryo was then pinned out, dorsal side up, on Sylgard using the very fine, barbed spines of the prickly pear cactus. In older embryos the dorsal side of the abdomen was pulled open with sharpened forceps. If the flexor muscles were developed enough to be joined along the midline, they were separated with the tip of a glass microelectrode. The hindgut was removed and the cut edges of the dorsal abdomen were pinned back with cactus spines to expose the nerve cord.

The embryos were lightly fixed for 1h in 2% paraformaldehyde in Van Harreveld’s solution, then incubated for 1 h at 33°C in chitinase (Sigma, 0·05 mg ml⁻¹ in Millonig’s buffer). The embryos were then incubated overnight at 4°C in a monoclonal antibody that had been raised against developing grasshopper central nervous system (Chang, Ho & Goodman, 1983) in a phosphate-buffered saline solution containing 2% bovine serum albumin and 0·2% saponin. The distribution of the primary was then examined by incubating it with an HRP-conjugated antimouse IgG (Vectastain, Vector Labs Inc.). The horseradish peroxidase (HRP) was reacted with diaminobenzidine (0·5mgml⁻¹) and hydrogen peroxide (0·03–0·003%). Embryos were cleared and mounted in 100% glycerin. They were then examined at 1000 × with a Zeiss compound microscope using a Leitz 50× water immersion lens and Zeiss Nomarski optics.

These preparations were used for motor neurone identification, as the connections to the muscles could be kept intact and the preparations remained healthy for long periods. Crayfish 4–6 cm long were cooled in ice and perfused for 30–40 min with cooled, oxygenated Van Harreveld’s solution. They were then restrained ventral side up in a Sylgard-lined dish. A midline strip of ventral cuticle was removed to expose the nerve cord. The flexor motor nerves of the first five abdominal ganglia (Gl—G5) were cut to prevent the axial fast flexors from twitching. Nerves 1 and 2 (NI, N2) of G3 and G4 were tied with thread and then cut. These two ganglia and the intervening connective were turned dorsal side up and secured by the thread. This gave access to the lateral giant (LG) and medial giant (MG) axons. The rectangle of hard cuticle just anterior to the anus on which the VTFs insert was cut along the midline and freed from the surrounding tissue. The VTFs could then be pinned back to give access to N6 G6 and the anterior telson (AT) muscle. The ventral artery was ligatured anterior and posterior to G6, and the section over G6 was removed. In this way blood circulation could be maintained throughout the experiment, though its access to G6 was reduced. G6 was stabilized by pinning it to a small, wax-coated platform, which allowed us to maintain intracellular recordings from the motor neurones during muscle twitches. The ganglionic sheath was softened with pronase for 15–20 s prior to recording from the somata. Using this preparation, the crayfish remained in good condition for 6h or more, judging by the eye withdrawal reflex, spontaneous movements and heartbeat.

For improved resolution of the PSPs in the motor neurones, it was necessary to record from the processes of the motor neurones in the neuropile. This was impractical with the semi-intact preparations, so we used isolated abdominal nerve cords. Crayfish 8–10 cm long were cooled in ice. The abdomen was then separated from the thorax, cutting the nerve cord anterior to the most posterior thoracic ganglion. The ventral abdominal cuticle was removed and all the nerves cut except one N6G6. This nerve was then freed from all the muscles it innervates except the PTF. The anterior insertion of the PTF is a tendon attached to the uropod and to the anterior fast flexor muscles. This was cut to release the anterior end of the PTF. The telson cuticle on which the posterior end of the PTF inserts was then cut free, yielding the isolated nerve cord still attached to one intact PTF muscle. The cord was pinned out dorsal side up in a Sylgard-lined dish, and the sheath was removed from G6. This allowed us to examine the innervation properties of the motor neurones while recording in the neuropile. Also, the axons of the PTF motor neurones were kept intact, which greatly improved their survival time.

Intracellular recordings were made from cell bodies or neuropile using capillary Microelectrodes. The tips of these were usually filled with 3 % Lucifer Yellow, and the barrel was backfilled with 2·5 moll⁻¹ LiCl (Stewart, 1978). Electrode resistances were 80–250MΩ. At other times we used electrodes containing HRP solution (containing 0·2moll⁻¹ KC1, 0·1 moll⁻¹ Trizma, pH7-4). These had resistances of about 40MΩ. Occasionally, when large amounts of current had to be passed, we used 3 mol I⁻¹ KC1 electrodes (10–20 MΩ).

Extracellular stimulation and recordings were accomplished with suction electrodes positioned on the dorsal connective (i.e. over the LGs and MGs) or on the peripheral nerves or muscles. The LGs and MGs are large enough to be stimulated individually in the connective. To stimulate the flexor inhibitor (FI) axons, we made use of their dual innervation of the VTF and the PTF muscles and stimulated the FIs antidromically in one muscle while recording from the other. The stimulus duration was usually 0·1 ms.

At the end of the experiment, the neurones were filled. Lucifer Yellow was injected by passing 1–10nA d.c. hyperpolarizing current for 20–60min. HRP was injected by passing 500-ms depolarizing pulses of about 5 nA at 1 Hz for 20 min. The HRP preparations were incubated at 10°C for 2–6 h and then developed in HankerYates reagent (Sigma; 1 mg ml⁻¹ in Van Harreveld’s solution containing 0·1 moll⁻¹ Trizma, pH7-4, see Hanker, Yates, Metz & Rustioni, 1977). All ganglia (including cobalt backfills) were fixed overnight in 10% formalin in phosphate buffer, dehydrated in an alcohol series, cleared with methyl salicylate and viewed in whole mount. The preparations were either drawn using a camera lucida or photographed in a focus-through series and then reconstructed from the photographs. HRP fills were embedded in plastic (Durcupan) and sectioned at 5 or 10μm. These sections were sometimes counterstained with 2·5% Toluidine Blue for 45–60 s. They were then viewed on a Zeiss microscope using a dark-field condenser or Nomarski optics and a water-immersion lens.

LGs and MGs are large enough to be recognized through a dissecting microscope by their positions in the cord. They are also large enough to be stimulated individually with extracellular electrodes. In the semi-intact preparations, the activation of these neurones was verified by recording from identified muscles, principally AT (specifically activated by LG) and PTF or VTF (activated by MG; Larimer et al. 1971). The corollary discharge interneurones (12, 13) were identified according to the criteria of Kramer et al. (1981). These were: size and timing of spike when recorded extracellularly from the connective after stimulating LG or MG; ganglion of origin, determined by stimulating or recording from the cord in different segments; or initiating activity in individual I2s or I3s by antidromic activation of the corresponding segmental giant (SG) in the first nerves of G2 or G3 Usually only one or two of these criteria were used.

 
• AT

  Anterior telson (motor neurone and muscle). The AT muscle moves an appendage in the tailfan.

 
• FAC

  A fast flexor motor neurone with its soma in the anterior, contralateral (relative to its axon) cluster.

 
• FFn

  Fast flexor in ganglion n. Axial fast flexor muscles control the flexion phase of non-giant swimming. There are 5–9 fast flexor motor neurones in each abdominal hemiganglion, and they have axons in N3 in G1-G5, N6 in G6; their neuromuscular synapses facilitate. FFs include FAC, FPI and FMC cells.

 
• Fin

  Flexor inhibitor in ganglion n. Peripheral inhibitor of all fast flexor muscles in its hemisegment. It is centrally activated with a delay by the escape command neurones (LG and MG) and receives direct input from sensory cells (particularly in caudal ganglia). FI is repeated segmentally; its axon is in N3 in Gl—G5, N6 in G6.

 
• FMC

  A fast flexor motor neurone with its soma in the medial, contralateral (relative to its axon) cluster.

 
• FPI

  A fast flexor motor neurone with its soma in the posterior, ipsilateral (relative to its axon) cluster.

 
• Gn

  Abdominal ganglion numbered n (1–6, G6 is the terminal ganglion).

 
• 12

  Interneurone with cell body in G2. This corollary discharge interneurone is fired by the segmental giant in G2 and excites FF motor neurones caudal to G2.

 
• 13

  Interneurone with cell body in G3. This corollary discharge interneurone is fired by the segmental giant in G3 and excites flexor motor neurones caudal to G3; it is capable of firing the telson flexor motor giants.

 
• LG

  Lateral giant. Giant command interneurone with the abdomen as its receptive field; gives rise to the forward-pitch-type escape tailflip. It fires the motor giant neurones in G1-G3 and the segmental giant premotor cells in all abdominal ganglia. It consists of fused, segmentally repeated cells and extends the entire length of the animal.

 
• MG

  Medial giant. Giant command interneurone with the rostral animal as its receptive field; gives rise to the dart-backward-type escape tailflip. It fires the motor giant and segmental giant premotor cells in all abdominal ganglia. Its soma and input regions are in the brain, but its axon extends the length of the animal.

 
• MoGn

  Motor giant in ganglion n. This segmentally repeated motor neurone is fired directly by the LGs (in G1-G3) and MGs (in all abdominal ganglia) and activates all the axial fast flexor muscles in its hemisegment via depression-prone synapses.

 
• MoGI

  Motor giant inhibitor. These cells are responsible for depolarizing IPSPs in the MoGs. Both projecting and local MoGIs have been characterized.

 
• Nn

  Ganglionic nerve numbered n.

 
• PTF

  Posterior telson flexor. This is a homologue of axial fast flexor muscles in the sixth segment. It is innervated by efferents (homologues of axial fast flexor efferents) with axons in N6 of G6.

 
• SGn

  Segmental giant in ganglion n. This segmentally repeated cell is fired by the giant command interneurones in each abdominal ganglion and fires (in G1-G3) or gives subthreshold excitation (in G4—G6) to non-giant, fast flexor motor neurones. In addition, segmental giants fire corollary discharge interneurones. Each SG has an axon in its ipsilateral NI (N2 in G6); the efferent action of the SGs is unknown.

 
• VTF

  Ventral telson flexor. This is a homologue of axial fast flexor muscles in the sixth segment. It is innervated by efferents (homologues of axial fast flexor efferents) with axons in N6 of G6.

It has been reported that three pairs of muscles act to flex the telson (Schmidt, 1915; Larimer & Kennedy, 1969). These are the posterior and ventral telson flexor muscles and the anterior telson muscle (PTF, VTF and AT : Fig. IB). On the basis of their position, effect and physiology, Larimer & Kennedy (1969) suggested that all of these muscles are homologous to the fast flexor muscles of anterior segments (Pilgrim & Wiersma, 1963). It will be shown below that this homology can be substantiated for the PTF and VTF muscles, but not for the AT. We will use the term ‘telson flexors’ to refer to the VTF and PTF muscles; the AT, which was previously called the anterior telson flexor muscle, will be treated separately below.

All three muscles are innervated by the purely motor sixth nerve of the sixth ganglion (N6G6). Cross sections of N6G6 reveal the profiles of 11 large axons (diameters >10μm in 6-cm animals; Fig. 1C) and five smaller axons. The smaller axons are tonic motor neurones and are not considered further in this report. The 11 large axons are those of the efferents to the telson flexors and the AT. When N6 G6 is backfilled with cobaltous chloride, the cell bodies of 11 large neurones are seen in G6. The cell bodies are clustered into three groups: an ipsilateral (relative to nerve) group of two cells, a contralateral, anterior group of five cells, and a contralateral, posterior group of four cells (Fig. IF).

A simple comparison of the soma map for the N6 G6 neurones with the soma map of fast flexors in G2 or G3 does not immediately suggest any obvious homology (Fig. 1D,F). We used three strategies to analyse the system. First, we compared this ganglion with its nearest neighbour, G5 (Fig. IE), whose intermediate degree of change from more anterior ganglia formed a bridge to the last ganglion. Second, we examined embryos to confirm that the last ganglion is a fusion product of two segmental ganglia (Johnson, 1924). Finally, to establish cell-by-cell homologies between the telson flexor and fast flexor motor neurones, selected neurones wean examined individually.

Crayfish embryos were treated with a monoclonal antibody (Chang et al. 1983) that stains developing nerve cells and fibre tracts. Embryonic development normally lasts 28–30 days at room temperature. At 60% of embryonic development, G1-G5 have the normal complement of two commissures each (Fig. 2A). G6, however, contains five commissures (Fig. 2B). Examinations of embryonic cords at various stages of development make it clear that the anterior pair of commissures are derived from the ancestral sixth ganglion and the next posterior pair are from the ancestral seventh ganglion. The last commissure is formed by a cluster of cells that innervate the gut via N7 G6. A similar partial ganglion is found at the posterior end of most arthropod nervous systems (M. Bate & C. S. Goodman, personal communication). This posterior group of cells has no homologue in the anterior five abdominal ganglia, and is not considered in this paper. At a slightly earlier stage (50%), the sixth, seventh and terminal neuromeres are still separate.

Our investigation shows that the ancestral ganglia have become fused in a pattern of sequential compression, rather than superposition. Moreover, the six nerves of the terminal ganglion do not represent two sets of the three nerves of anterior ganglia. For example N1-N3 of G6 all innervate the sixth segment appendages (the uropods), whereas in anterior segments the appendages are innervated only by N1. Similarly, axons from N6 G6 can be traced back to commissures from both the ancestral sixth and seventh segments.

From these results we predicted that many elements within the telson flexor motor system would show doubling compared to their more rostral, axial fast flexor homologues.

In G1-G5 (Mittenthal & Wine, 1978) and thoracic ganglia 1–3 (Crabtree, 1981), each hemisegment contains a unique, identified motor neurone called the motor giant (MoG). On the basis of morphology alone, two potential MoG homologues (called ‘telson MoGs’) were located in G6 (Fig. 3, see also Johnson, 1924). They share with all other identified MoGs the following six morphological characteristics (see Mittenthal & Wine, 1973; references are to studies of anterior ganglia).

• (1) Distinctive soma position, in a cluster of three other flexor motor neurones (Fig. 1D-F).

• (2) Decussating axon.

• (3) Extensive peripheral branching. Each of the telson MoGs innervates all the fibres in one muscle. The same muscle also receives innervation from 2–4 other excitatory motor neurones. These cells overlap the MoG innervation but tend not to overlap with each other.

• (4) Largest cell bodies in the ganglion, measuring about 85 pm in diameter compared with 40–75 μm in diameter for the other telson flexor motor neurones (in 4-to 6-cm animals) (Fig. 3A,B).

• (5) Largest axons in the flexor motor nerve, measuring 24μm in diameter compared with 6–13 μm for other telson flexor motor neurone axons (Fig. 3C).

• (6) Restricted dendritic branching, mainly limited to short processes contacting the giant flexion command interneurones (Fig. 3D).

  In addition, six physiological features confirm the homology of these cells to anterior MoGs.

• (7) The initial excitatory junctional potential (EJP) reliably produces a twitch in the muscle, but subsequent stimuli at 1 Hz cause profound, low-frequency depression with a modest component of facilitation (11 muscle fibres in three preparations; see Fig. 4). To our knowledge, no other neuromuscular junction in the crayfish shows depression even remotely comparable to that shown by the MoG (Bruner & Kennedy, 1970).

• (8) The synapses between the MG axons and the telson MoGs are rectifying electrical synapses (Fig. 3E) of the kind first described in anterior MoGs by Furshpan & Potter (1959a). The synapses are powerful. In healthy preparations, they cause the telson MoGs to fire with a latency of > 0·5 ms following a single action potential in either MG axon; latencies in FF motor neurones are at least twice as long.

• (9) Activation of the LGs does not produce any short-latency EPSP (Fig. 5A,B; see Dumont & Wine, 1986a, for a description and analysis of the polysynaptic input). This imbalance of the input in favour of the MGs is also found in the MoGs of G4 and G5, but in no other flexor motor neurones (see Mittenthal & Wine, 1973; Fig. 5A,B).

• (10) Activation of the LG or MG axons, or antidromic activation of any fast flexor motor nerve in the abdomen, produces a long-lasting, depolarizing IPSP in the telson MoGs and in all MoGs of anterior ganglia, but not in any other flexor motor neurones (see Wine, 1977; Fig. 5A-C).

• (11) The telson MoGs and the MoG5 (and possibly more anterior ones) receive ‘spontaneous’ IPSPs (Furshpan & Potter, 1959b) from shared sources (Fig. 5D).

• (12) The appearance of MG inputs to the MoGs, when recorded from the soma, is unlike that of any other motor neurone (Fig. 5A). The action potential is attenuated, since the soma membrane is inactive and the shape of the cell leads to severe filtering of the faster components of electrical activity, and it is always followed by a depolarizing IPSP (Wine, 1977). The attenuation of the action potential is greater in the MoGs of anterior ganglia, presumably due to the greater separation of soma and axon.

Of these 12 features, all but the first three are shared by all MoGs, but by no other flexor motor neurones. We conclude that both of the telson flexor MoGs are homologues of anterior MoGs. One has its soma in the anterior contralateral cluster and its neurite crosses the midline in an anterior commissure. Hence it is called MoG6. It innervates the PTF muscle, and has been recorded and structurally identified 20 times. The other has its soma in the posterior contralateral cluster and its neurite in a posterior commissure. It is called MoG7. It innervates the VTF muscle, and has been recorded from and structurally identified 10 times. The nomenclature we use indicates our conclusions as to the ganglionic origins of the motor neurones. This subject will be dealt with more fully after we have established the homologies of the various telson flexor motor neurones with the axial fast flexor motor neurones of anterior abdominal ganglia.

The anterior five abdominal ganglia each contain a second unique, identified fast flexor efferent, the flexor inhibitor (FI). Each FI inhibits all the fast flexor muscles in its hemisegment and provides the only inhibitory innervation of these muscles (Selverston & Remler, 1972). Of all the neurones innervating the fast flexors, only FI has an extensive bilateral dendritic arborization (Fig. 6A). This unique anatomical feature allowed us to recognize two putative FI homologues innervating the telson flexors (Fig. 6B-D). Six other anatomical and physiological features are shared by these neurones and by FI neurones in more anterior abdominal ganglia; the first three of these are not found in any other fast flexor motor neurones.

• (1) The synaptic potential they produce in the muscle fibres is inhibitory, reversing at approximately –72mV (Fig. 7B).

• (2) They receive excitatory input from the LGs, which (in trains) leads to a smoothly rising EPSP and delayed action potentials (Fig. 7A).

• (3) The LG input depresses with repetition at 1 Hz (Fig. 7C).

• (4) They have the most extensive axonal branching of any telson efferent with the possible exception of the MoGs.

• (5) The somata are adjacent to those of the MoGs.

• (6) The axons are decussating. References for FI features in anterior ganglia are as follows: 1–3, 5, 6, Wine & Mistick, 1977; 4, Selverston & Render, 1972; 5, 6, Mittenthal & Wine, 1978.)

One telson FI has its soma adjacent to MoG6 in the anterior contralateral cluster and crosses the midline in an anterior commissure. The other FI has its soma adjacent to MoG7 in the posterior cluster and crosses the midline in a posterior commissure. We have termed these neurones FI6 and FI7, respectively. The FI neurones are the only telson flexor efferents which branch to innervate both the VTF and PTF muscles. In four of five preparations examined, we found no overlap in the innervation of muscle fibres by the FIs, although in all but one of these experiments only the PTF muscle was examined. In the PTF, FI6 innervates the ventral portion of the muscle and FI7 innervates the dorsal portion (Fig. 10). [In their experiments, Larimer & Kennedy (1969) also reported that no fibre received input from more than one FL] In the one exception to this arrangement, extensive, dual innervation by both FIs was noted among PTF muscle fibres. The explanation for this exception cannot be that the electrode was fortuitously sampling from a narrow region of overlap present in all animals, since widely separated fibres, all in the ventral portion of the PTF, received dual IPSPs. We do not have an explanation for this anomalous animal; apparently both FI6 and FI7 were co-innervating the dorsal PTF. FI6 was recorded from and identified seven times, FI7 nine times.

Of the seven remaining efferents, one innervates the AT muscle and six innervate the telson flexors. The AT motor neurone is not thought to be a fast flexor homologue for reasons that will be outlined below. The six motor neurones that innervate the VTF and PTF muscles appear to be homologues of the non-giant motor neurones of anterior ganglia on the basis of the following five shared characteristics.

• (1) As a group, the telson flexor motor neurones innervate the same muscle fibres as the MoGs and FIs. Thus, each VTF and PTF muscle fibre receives input from at least three efferents: one MoG, one FI, and one or more non-giant telson flexor motor neurones. This is the same pattern of innervation as in the fast flexor muscles in anterior segments.

• (2) The EJP from a telson flexor motor neurone is frequently suprathreshold. When it is subthreshold it is either stable or shows modest facilitation (Fig. 4).

• (3) The dendritic arborizations of telson flexor motor neurones are primarily ipsilateral to their axons, in the region of the neuropile directly beneath the giant intemeurones (Fig. 8D).

• (4) The size of their somata relative to those of the MoGs is the same as that of the anterior FFs. In G1-G5, the mean FF soma diameter is 69% of that of the MoGs. In the sixth ganglion, the average diameter of the telson flexor motor neurones is 66 % of the mean MoG soma diameter.

• (5) Telson flexor motor neurones receive excitatory input from identified intersegmental interneurones: LG, MG, 13 and 12 (Fig. 8E,F).

These five features were previously established for anterior FFs as follows: 1, 3, Kennedy & Takeda (1965); 2, Miller et al. (1985); 4, Roberts et al. (1982); 5, Mittenthal & Wine (1978).

These six motor neurones form three pairs within the soma map: one pair is in the anterior contralateral cluster, one is in the posterior contralateral cluster, and the third pair forms the ipsilateral cluster.

The morphology of the motor neurones and their positions within G6 enable us to establish their ganglia of origin. In each of the anterior five abdominal ganglia there is one cluster of motor neurones with posteriorly directed, decussating axons. This cluster, termed the FMC group (flexors: medial contralateral, based on soma position; Mittenthal & Wine, 1978), consists of the MoG, the FI and two non-giant fast flexor motor neurones termed FMCs (Fig. ID). The FMC group is highly conserved, being found in thoracic as well as all abdominal segments (Crabtree, 1981). Our evidence shows that it is also conserved in both the sixth and seventh Segments, since each MoG-FI pair in the terminal ganglion is adjacent to two otherion-giant telson flexor motor neurones (see Figs 3B, 6D, 8A,B). Furthermore, all the G6 motor neurones in the anterior cluster cross the midline in the same anterior commissure, whereas all the motor neurones in the posterior cluster cross the midline in a posterior commissure. Our study of the embryonic terminal ganglion indicated that the anterior posterior position of the commissures reflects their ganglion of origin. This has now been confirmed in the adult by Kondoh & Hisada (1986). We conclude that the anterior and posterior clusters are the FMC groups derived from the ancestral sixth and seventh ganglia, respectively. We have therefore called the non-giant flexor motor neurones in the anterior cluster FMC6s and those in the posterior cluster FMC7s.

The remaining two telson flexor motor neurones have cell bodies ipsilateral to their axons and are therefore considered homologous with the FPI group (flexors: posterior ipsilateral; Mittenthal & Wine, 1978) of anterior ganglia. Their position relative to the FMC groups is consistent with a derivation from the ancestral sixth ganglion. This interpretation is strengthened by the position of their neurites Although these do not cross the midline, they do approach it at the same point than the neurites of the anterior cluster cross it, well anterior to the neurites of the posterior cluster. We have therefore termed these neurones FPI6s. The FPI group of neurones contains four cells in the first four abdominal ganglia, but only three cells in G5. In G6, we find only two FPIs, suggesting this group has been further reduced in the sixth segment, and it is missing entirely from the seventh.

The pathways of the neurites described above are shown by a cobalt backfill of the neurones innervating the dorsal portion of the PTF muscle (Fig. 9). This shows two neurones from the anterior contralateral cluster crossing in the anterior commissure and the FI7 from the posterior cluster crossing in a posterior commissure. The neurite from the FPI6 approaches the midline next to the anterior commissure. Cobalt backfills of the whole nerve show that the other neurones follow the same pattern, but we have shown fills of only four neurones here for the sake of clarity.

It became apparent early in this study that the VTF and PTF muscles could not be assigned specific segmental origins. Because MoG6 innervates the PTF and MoG7 innervates the VTF, it might have been thought that those two muscles are segmentally homologous. However, when the other cells are included, it emerges that each muscle receives inhibition and excitation from neurones of both sixth and seventh segments. The pattern of innervation of the PTF and VTF muscles is shown in Fig. 10.

The hypothesis that the telson flexor muscles (i.e. PTF and VTF) are homologuen of the anterior fast flexor muscles is supported by the extensive evidence of neuronal homology between their respective motor neurones. However, the AT muscle lacks most of the characteristic features of fast flexor muscles, and we propose that it is not homologous with them.

Perhaps because the three muscles innervated by the sixth nerve all attach to the telson, they had been assumed to have the similar function of flexing the tailfan, and until now the AT muscle was called the anterior telson flexor muscle (Schmidt, 1915 ; Larimer & Kennedy, 1969). However, the AT muscle attaches to the anterior portion of the telson, close to the hinge, and runs ventrally and laterally to attach to the base of the uropod via a tendon. Its direction of action is therefore orthogonal to all other fast flexor muscles and ill-suited for telson flexion. Experiments in which the AT muscle was activated selectively, or in the context of the tailfan muscle activity produced by LG impulses (see below), show that its main effect is to rotate the uropod about its long axis and close it on the telson (pronation and abduction, Fig. 11C). The AT motor neurone is selectively fired by the LGs (Larimer et al. 1971; Fig. 1 IE), but telson flexion is not part of the LG tailflip pattern (Wine & Krasne, 1972). In addition to having a different functional role, the AT muscle shares none of the characteristic innervation of the fast flexors. It lacks innervation by either the MoGs or the FIs and is not polyinnervated: the entire muscle is innervated by a single motor neurone which fires the fibres via suprathreshold, non-depressing synaptic potentials. This absence of inhibitory input is not found in any other fast flexor muscle, but it is found in many uropod muscles (Larimer & Kennedy, 1969) and has been reported in the thoracic appendages of other crustaceans (Atwood, Parnas & Wiersma, 1967; Hoyle & Burrows, 1973).

The single AT motor neurone shares three of the five characteristics with telson and anterior FFs, but lacks the two most important ones: the AT motor neurone receives suprathreshold input from the LGs and negligible input from the MGs (Fig. 11D,E), in contrast to all other non-giant fast flexor and telson flexor motor neurones which receive equal input from the giant interneurones (Fig. 5A,B); and the muscle innervated by the AT motor neurone is separate from those innervated by the MoGs and FIs. Thus, we were unable to establish that the AT motor neurone is a homologue of any fast flexor motor neurone. As a result, we renamed the muscle the anterior telson muscle, as this reflects the position of the muscle without making any assumptions about its function, which is only known for LG tailflips. We have renamed the motor neurone accordingly. It was identified 12 times.

In anterior abdominal ganglia, the segmental giant (SG) acts as a driver neurone between the giant flexor command axons and the FF motor neurones (Roberts et al. 1982). The SGs are thought to be modified motor neurones, as they have axons in the swimmeret motor nerves (the peripheral terminations of the SG axons have not been found). A putative homologue of the SG has been found in G6. It shares the following characteristics of the SGs of more anterior segments (for anterior ganglia, see Kramer et al. 1981; Roberts et al. 1982).

• (1) It has very large neuropilar processes in close apposition with both the LG and the MG.

• (2) Its soma is small relative to the large size of its processes (Fig. 12A,B).

• (3) It has a large, rapidly conducting axon in N2. This nerve is not exactly homologous to N1 (swimmeret nerve) in anterior ganglia, but it does innervate the muscles of the uropod, which are homologues of the swimmerets in anterior segments. The conduction velocity of the SG axon is greater than that of any other efferent in the nerve, although a unit activated by the LG is almost its equal in N2G6.

• (4) It fires in response to single action potentials in either MG or LG axons, and follows one-for-one with trains of action potentials in either giant axon at frequencies up to 20 Hz.

• (5) The action potential has a characteristic shape (Fig. 12C,D).

• (6) When directly fired, the SG produces an EPSP in the non-giant telson flexor motor neurones (Fig. 12E).

We have found no evidence for a second SG in the terminal ganglion. There are no other peripheral axons in G6 which are fired reliably and at short latency by both MG and LG impulses.

We have identified 11 efferents to three phasic telson muscles and have provided evidence that two of the muscles and the 10 motor neurones that innervate them are homologues of the axial fast flexor muscles and motor neurones. The identification of telson flexor motor neurones completes the description of the efferents of the abdominal fast flexor system. This consists of 58 pairs of motor neurones in six segments, and its counterparts, although less well described, have been traced into the thorax as far anterior as the first thoracic ganglion (Table 1 ; Crabtree, 1981). In addition we have identified one homologue of the SG, a ganglionic premotor neurone with inputs to the telson FFs. The system therefore provides a good opportunity for a comparative study of its central connections and functions (Dumont & Wine, 1986a,b). Before doing this, it is necessary to evaluate the evidence for homology, and relate this study to previous work.

Homologous structures are those which have evolved from a common precursor. In terms of individual neurones, this can be most accurately approximated by considering development. Homologous neurones are those that share a common descent from an identified, homologous precursor. This pushes the problem back a step. It now becomes one of deciding which precursor cells are homologous. In some systems, the neuroblastoma cells form highly stereotyped arrays in which some precursor cells can be individually identified. Tracing adult neurones back to their neuroblast of origin for the purpose of establishing homology has been applied in two studies, both in the grasshopper (Bate et al. 1981; Pearson, Boyan, Bastiani & Goodman, 1985). In these studies, neurones with different morphologies or physiologies were identified as homologues on the basis of embryological evidence.

If embryological development has not been established, homology between two cells can be inferred if the cells share many features which serve to distinguish them from other cells. Features used in previous studies of neurones include: soma position; major neurite morphology; pattern and place of the dendritic arborization; soma membrane biophysical properties; peripheral projections (sensory field or muscle); synaptic output properties (for motor neurones EJP sign, size, facilitation or depression); common input from identified sources; and neurochemistry. We recognize that no listing of common properties, no matter how long or how idiosyncratic the individual features, can prove that two cells share a common precursor. However, when common properties at the single cell level extend to a whole population of related neurones and to their connections with other neurones or muscles, alternate explanations become highly unlikely.

In this study we have used the analysis of shared, idiosyncratic features to establish segmental homology. Five of the neurones were identified by such features: we identified nine features for the MoGs, four for the FIs, and five for the SG. The individual homologies are greatly strengthened by showing that the whole fast flexor system of G6 is homologous with the equivalent systems in anterior ganglia. Thus we can show that the entire FMC group of four cells is the same in each abdominal ganglion, and the doubling of this group in G6 is consistent with the developmental data indicating fusion of two segments. We can also determine which neurones are derived from which ancestral ganglion. Furthermore, we know that the same interneurones are presynaptic to the fast flexor motor neurones in G6 and anterior ganglia. This contrasts with the studies of interspecies homology of the metacerebral giant in gastropods (Granzow & Rowell, 1981), in which homology could only be proposed on the basis of the properties of the cell itself, since the other components of the feeding circuitry have not been well established.

Finally, where differences exist between the fast flexor system in G6 and G1-G5, they are explicable in the light of knowledge derived from other sources. The absence of FPI homologues in the sixth and seventh segmental components of G6 is consistent with the pattern of reduction of motor neurone number towards the anterior and posterior ends of the fast flexor chain, which is in turn correlated with a reduction in size and number of muscles to be innervated (Table 1; Mittenthal & Wine, 1978; Crabtree, 1981). Similarly, we might have expected two pairs of SGs in G6, whereas we have identified only one. However, the SGs have peripheral axons in the nerves that innervate the segmental appendages, and presumably have or once had an efferent function. The absence of a second pair of appendages in the sixth and seventh segments, therefore, is consistent with the absence of a second pair of SGs, particularly as the evidence from studies in palaeontology (Schram, 1981; Hessler, Marcotte, Newman & Maddocks, 1982) and comparative embryology (Manton, 1928, 1934) suggests that the appendages of the ancestral seventh segment were lost prior to segmental fusion. An alternative possibility is that the SG from the ancestral seventh segment exists, but has been modified beyond recognition.

Since we have identified fewer distinguishing features for muscles than for neurones, the evidence for the homology of the telson and flexor muscles with the anterior fast flexor muscles rests largely on their pattern of innervation. This is also the case with our claim that the AT muscle is not a homologue of the anterior fast flexor muscles. Fortunately, there are other studies which indicate that homologous motor systems can undergo fundamental changes and yet conserve the relationship between muscles and motor neurones. The metathoracic leg of the locust is specialized for jumping, yet the motor neurones innervate the homologous muscles as in the pro- and mesothoracic walking legs (Wilson & Hoyle, 1978), even though the physiological properties and functional role of two of the neurones have changed radically. Similarly, it has been shown that many of the tailfan muscles of the decapod crustacean Procambarus and the sand crabs Blepharipoda and Emerita are homologous (Paul et al. 1985); and the motor neurones innervating these muscles also appear to be homologous, even though the tailfans in these three species are used in quite different forms of locomotion and consequently are morphologically extremely different (Paul, 1981). As a third example, the motor neurones innervating the abdominal fast flexor muscles have been found to be homologous in Procambarus, in the hermit crab Pagurus (Mittenthal & Wine, 1978) and in the squat lobster Galathea (Sillar & Heitler, 1982), even though the abdominal morphology differs in all three genera, particularly in the hermit crab which uses the abdomen mainly for holding on to its shell.

There is a certain element of circularity in this argument, as highly modified muscles or motor neurones are unlikely to be recognized as homologues. Moreover, in the study of tailfan musculature, innervation patterns were one of the features used to establish motor neurone homology. Muscle innervation is certainly not immutable. In particular, the sand crabs appear to lack MoGs (see Paul, 1981), and the number of motor neurones innervating a muscle may change (Govind & Atwood, 1982; Mittenthal & Wine, 1978), but when remaining neurones are identified it is found that they are homologous with motor neurones innervating the homologous muscle. On the whole, the conservation of innervation patterns of homologous muscles does appear to be fairly reliable.

The origin of the AT muscle is not known. Since it moves the uropods, it may be homologous to anterior appendage muscles. However, Paul et al. ( 1985) suggest that it has no anterior homologue.

The PTF and VTF muscles receive innervation like that of anterior fast flexor muscles. Each fibre is innervated by one MoG, one FI and one or, more rarely, two FFs. Each muscle is innervated by neurones from both ancestral ganglia. We do not know if this is a difference from anterior segments. The only study of the innervation patterns of the abdominal fast flexor muscles that looked at several ganglia found that individual muscles can traverse three segments and receive innervation from the FFs in ganglia of all three (Otsuka, Kravitz & Potter, 1967). Unfortunately, there has been no comparable study of possible intersegmental projections of the FIs and MoGs in anterior segments. If we consider only the contralateral motor neurones, we find that both PTF and VTF muscles receive the same innervation: one MoG, two FIs, one FMC6 and one FMC7. Also it has been shown for the PTF and the VTF (Larimer & Kennedy, 1969) that no single fibre receives input from both FIs or both FMCs.

We cannot yet say which muscles are derived from which segment. Judging by their innervation they may both be derived from ancestral segments 6 and 7, but are smaller than their anterior homologues and have a more restricted innervation.

We wish to thank Dr C. S. Goodman for generously providing the monoclonal antibody. We were greatly helped by technical advice and assistance from G. Hagiwara, Dr M. Kirk, Dr C. Bishop, F. Thomas, Dr M. Bastiani, R. Ho, J. Berlot and Dr M. Plummer. Also, we are grateful to Dr F. Krasne for his comments on the manuscript, and J. Ruby for her assistance in preparing it. This study was supported by NSF grants BNS 80–15583 and BNS 81-12431 to JJW and a Stanford University Biological Sciences Department Teaching Assistantship to JPCD.