Functional organization of crayfish abdominal ganglia. III. Swimmeret motor neurons

Backfills of motor axons

Motor neurons in A2, A3, A4, and A5 were filled through their axons in N1 (Fig. 1A), the nerve that projected from these ganglia to innervate a swimmeret. The proximal ends of cut axons in branches of N1 were backfilled (Iles and Mulloney, 1971) with Co⁺⁺ using the procedure described in Leise et al. (1986). The Co⁺⁺ was precipitated as a sulfide, and the ganglia were fixed and processed (Leise et al., 1986). The same surgical procedure was used to fill axons in these branches with 5% biocytin (Horikawa and Armstrong, 1988) in distilled water. These preparations were then fixed for 2 hours in 4% formaldehyde in phosphate-buffered saline, pH 7.4 (PBS; Sigma D5773) at 4°C, rinsed in 0.1 M glycine in PBS, and dehydrated in an ethanol series to render them more permeable to proteins. To label the biocytin, the ganglia were incubated overnight in a solution that contained Streptavidin-HRP (Amersham RPN.1231) 1:300 and 0.3% Triton X-100 (Aldrich 28,210-3) in PBS. The HRP was visualized with DAB (ICN 36315), embedded, and sectioned using the procedures described in Mulloney and Hall (1991) and Acevedo et al. (1994). We prepared 32 ventral nerve cords for Co⁺⁺ backfills, and 54 nerve cords for biocytin backfills.

Cleared wholemounts were photographed with Kodak Techpan 120 film using oblique incident light reflected off a white background.

Labeling of synapses in the core of the ganglia

To discover the distribution of chemical synapses within the neuropils of these ganglia, we used an antibody directed against synapsin (Klagges et al., 1996; antibody provided by E. Buchner) to label synapses in wholemount preparations, with a method adapted from Harzsch et al. (1998). Nerve cords were fixed quickly in picric acid formaldehyde overnight at 4°C (Acevedo et al., 1994). Ganglia were then incubated in primary antisera (1:30) overnight, and in secondary antisera (Jackson Laboratories goat anti-mouse CY3, 1:200) overnight. Cleared wholemounts were photographed using 35 mm Ektachrome film.

Sections of OEG-stained ganglia and peripheral nerves

Abdominal nerve cords were fixed in situ with glutaraldehyde, then stained with OEG (Leise and Mulloney, 1986), embedded in Spurr's resin, and sectioned using glass knives. Serial sections were studied with high numerical-aperture optics, and illustrative sections were photographed using planapochromatic lenses and Kodak Techpan 120 film.

Preparation of illustrations

Both 120 negatives and 35 mm color slides were scanned using Agfa Fotolook and then assembled into plates using Adobe Photoshop. Each digitized image was first flipped to make a positive image, and its brightness and contrast adjusted to the same values as other images in the same plate.

Census of filled cell bodies

Serial sections of ganglia that contained filled neurons were studied with high numerical-aperture optics. Every section that contained parts of filled cell bodies was aligned and drawn with a camera lucida, and the outline of each filled neuron was added to the drawing. When a cell body appeared in more than one section of the series, the location of its nucleus was noted on the drawing. The filled neurons were then counted by counting the number of cells in each drawing.

Positions of cell bodies of swimmeret motor neurons within the abdominal ganglia

To reveal the innervation of a single swimmeret, we filled individual N1s and individual branches of these nerves. Every backfill of an entire N1 always revealed the same overall picture: most cell bodies of filled neurons were clustered in two groups anterior and posterior to the base of the nerve (Fig. 3). One filled cell body occurred near the midline, and one occurred contralaterally near the posterior margin of the ganglion (Fig. 3A_V). The processes that projected across to the opposite LN did not connect to any cell bodies on the contralateral side of the ganglion (Fig. 3). Some thin afferent axons went below the LN to enter the horseshoe neuropil (HN; Skinner, 1985b), and a few continued into the connective anteriorly to reach the HN of the next anterior ganglion. No filled axons projected posteriorly, and no cell bodies were filled in neighboring ganglia. Therefore, all the motor innervation of each swimmeret originates in the ganglion whose N1 innervates it, and with the exception of the one midline cell body whose axon runs in the posterior branch, the cell bodies of motor neurons that innervate each swimmeret are segregated from those that serve the contralateral swimmeret of the same segment.

The shapes of individual motor neurons were most easily seen in preparations in which only some axons filled (Fig. 3B). All had primary processes in the LN above the base of the filled N1. Their cell bodies first sent a neurite into the LN. This neurite then branched to form an axon that exited through N1 and a major integrative process that sent secondary branches throughout the LN. Branches of some of these neurons continued across the midline to the opposite LN (Fig. 3B_D). This figure also shows the shape of the anomalous neuron with its cell body at the midline.

Cell bodies of most PS motor neurons were located posterior to the base of N1

To map the motor neurons in more detail, we took advantage of the initial division of each N1 into an anterior and a posterior branch (Fig. 1A). Backfills of the posterior branch, N1p, consistently filled a compact cluster of small cell bodies at the lateral margin to the ganglion, just posterior to the base of the filled nerve (Fig. 4A). A few more cell bodies were directly ventral to the base of N1 or just anterior to it, and one cell body on the ventral midline usually filled, too (not visible in Fig. 4A; see Fig. 3B_V). Because of the physiological evidence that the posterior branch of N1 contains PS axons but not RS axons (Fig. 1D), we infer that these filled neurons include the PS motor neurons.

Cell bodies of most RS motor neurons were located anterior to the base of N1

Backfills of the anterior branch, N1a, consistently filled a looser cluster of cell bodies anterior to the base of N1, and a few small cell bodies just posterior to the base of the filled nerve (Fig. 4B). These most posterior cells were mixed in with the PS cell bodies. One more filled cell body lay contralateral to the backfilled N1a, near the ganglion's posterior margin (Figs. 3A, 4B). The neurite of this one contralateral cell followed a different course within the ganglion than did the rest of the RS motor neurons. Again, because of the physiological evidence that N1a contains RS motor axons but not PS axons (Fig. 1C), we infer that this group of filled neurons includes the RS motor neurons.

Despite these differences in positions of their cell bodies, the structures of most RS motor neurons within the ganglia were not otherwise noticeably different from those of most PS motor neurons. The cell bodies of RS motor neurons sent their primary neurite into the LN, where they branched and sent the axon on into N1a on the same path as did PS neurons. Some also sent a process to the contralateral LN. In preparations of N1a backfills with many filled axons, the branches of RS motor neurons filled the LN (Fig. 4B).

Some swimmeret motor neurons sent processes to the opposite LN through dorsal commissure 1 (DC1)

Not all swimmeret motor neurons have branches that extend medially beyond the boundaries of their home LN (e.g., Fig 3B, and Sherff and Mulloney, 1997) but some do. These branches that project to the opposite LN are potentially significant for bilateral coordination of swimmeret movements and are relevant to the definition of a swimmeret module. Serial sections of ganglia that contained filled neurons revealed that these projections are largely restricted to a particular commissure, DC1 (Fig. 5A_iv), that is located just ventral to the giant axons in each ganglion (Fig. 5B), near the anterior end of the ganglion's core (Fig. 5E). Both PS and RS motor neurons have processes in DC1. None of these processes in DC1 were connected to a cell body contralateral to the backfilled nerve.

Sensory axons projected ventrally to the horseshoe neuropil (HN) and into the anterior ventral commissure (AVC)

In frontal and transverse sections, axons filled from N1p appeared to divide into two bundles. One bundle ran ventral to the LN and dispersed into the HN (Fig. 5C). Some axons continued across the midline through the AVC and projected anteriorly into the interganglionic connective (Fig. 2). These axons that reached the AVC were larger in diameter than the many axons that terminated in HN. We did not observe axons filled from N1 projecting into the connectives posterior to the ganglion.

The other bundle rose directly into the LN (Fig. 5C). This bundle contained the axons of swimmeret motor neurons. In preparations in which many neurons were filled, their processes permeated the entire LN above the base of N1 (Fig. 5A_iii).

Census of motor neurons

To examine the size-distribution of axons innervating a swimmeret, we sectioned N1 near the ganglion, and sectioned its branches distal to their origins (Fig. 1B). N1 contains hundreds of axons that range in diameter from less than 1μm to greater than 100μm. Both branches contain many axons of intermediate size, and hundreds of thin axons that are ensheathed by connective tissue (Somers and Nunnemacher, 1970). The anterior branch, N1a, contains the two largest axons (Fig. 1B). At each point, we counted the numbers of axons with diameters equal to or larger than 2 μm in the left and right N1s of an A4 and an A5; smaller axons could not be counted in the light microscope. Near its base, N1 had 277 ± 26 (mean ± SD; n=4) axons ≥2 μm. More distally, these numbers increased; N1a had 328 ± 39 and N1p had 137 ± 43 axons (n=4). These totals are larger than the numbers of cell bodies filled from each branch, so it is probable that many sensory afferent axons have diameters larger than 2 μm and were included in these counts. Both from these numbers and from the photographs, it is apparent that these axons are tapered; their diameters decrease as they near the ganglion.

To estimate the number of swimmeret motor neurons in each segment, we counted cell bodies in serial sections of ganglia that contained cobalt or biocytin backfills of N1a or N1p. We took the maximum number of filled cell bodies counted in ganglia from each segment to be a measure of the true number of motor neurons innervating the swimmerets in that segment. For each preparation, we corrected for other non-motor neurons with axons in N1 (see Discussion). These maximum numbers varied only slightly from ganglion to ganglion (Table 1), and we observed no systematic differences between segments. The results from analysis of cobalt backfills gave 36 ± 2 cell bodies per swimmeret with axons in N1a, and 35 ± 3 cell bodies per swimmeret with axons in N1p. Therefore, there are about the same numbers of PS and RS motor neurons innervating each swimmeret.

The maximum numbers of cell bodies filled in each ganglion depended on the material used to fill the cells; biocytin preparations consistently had more filled cell bodies than did cobalt preparations (Table 1). N1a biocytin fills yielded 46 ± 4 cell bodies, and N1p fills yielded 43 ± 3 cell bodies. Like the cobalt backfills, there was no evidence of systematic differences in the number of swimmeret motor neurons in different segments. Possible causes of this methodological difference are discussed below.

Is the unit of organization one module per ganglion or two?

Movements of the two swimmerets on each segment are normally synchronized (Fig. 1D; Davis, 1969). In each ganglion, several PS and RS motor neurons send processes through DC1 to the LN on the opposite side of the ganglion (Figs. 3, 4, 5A_iv). These projections and connections might form the basis of a synaptic circuit that synchronizes motor output in both N1s and so render meaningless the notion of a separate module for each swimmeret. Evidence from microelectrode recordings contradicts this idea. Swimmeret motor neurons appear to have only one spike-initiating zone, located in N1 near the lateral edge of the ganglion (Sherff and Mulloney, 1997). Recordings from different sites in their arbors within the LN yield only attenuated impulses spreading back from the axon (Sherff and Mulloney, 1997). These contralateral projections therefore do not support action potentials. Moreover, when sodium action potentials are blocked with TTX, the circuits on opposite sides of one ganglion can oscillate independently (Murchison et al., 1993). If these projections did contribute to synchronization, they should be more effective in the presence of TTX. Under those conditions, the length-constants of these contralateral projections would increase because the total membrane conductance would be reduced, making them more electrically compact, which would increase their effectiveness in synchronizing left and right sides. That is not what happens. Despite the presence of these branches on some swimmeret motor neurons that extend through DC1 to the contralateral LN, the neurons on opposite sides of each ganglion form a separate functional module.

How many motor neurons are there in each swimmeret module?

We expected that the number of motor neurons innervating each swimmeret would be the same in every individual, except for developmental anomalies. Three sources of error would contribute to variation in the numbers of neurons observed: incomplete fills of the selected set of axons, spread of marker through gap-junctions (Paul and Mulloney, 1985a,b), and mistakes in analyzing the sections. Incomplete fills would lead to an underestimate of the true number. Either including other neurons with N1 axons or including interneurons labeled by movement of dye through gap-junctions would lead to an overestimate.

In addition to the swimmeret motor neurons, three other neurons have axons in N1 and cell bodies in the ganglion. These neurons were inevitably filled in many of our preparations, but because of their distinctive anatomy, we could recognize them. The segmental giants (Roberts et al., 1982) have gross processes, visible in Fig. 3B_D, that contact the LG and MG axons. These processes are strikingly different from those of swimmeret motor neurons. The non-spiking stretch receptors (Heitler, 1982; Paul 1989) have the two largest axons in N1a (Fig. 1B), sparse, stunted central processes, and large cell bodies located near the base of N1 (visible in Fig. 4B_V). In our census of swimmeret motor neurons, we could exclude them.

Given that the methods used to count these cell bodies in sectioned material were not prone to systematic error, and that we could correct for the these three other neurons, the question remained what measure to use to estimate the number of PS and RS neurons in each module. Backfills of any nerve do not always yield the same picture; uncontrolled differences in protocol and chance combine to give different numbers of filled neurons in different preparations. We have taken the maximum number of filled cells in each ganglion as the best measure of the true number of motor neurons. This measure says that each swimmeret has about 35 PS motor neurons and 35 RS motor neurons (Cobalt column in Table 1), and that there is little variation in the numbers of motor neurons innervating swimmerets on different segments. These results reinforce the idea that the swimmeret systems consists of a segmental series of modules that can be considered functionally identical (Skinner and Mulloney, 1998).

We were surprised that these maximum numbers were sensitive to the method used to fill the neurons, but biocytin did yield larger numbers than cobalt ion (Table 1). Both biocytin and cobalt will cross conductive gap junctions (Giaume and Korn, 1984), but in our experience the movement of biocytin between electrically-coupled neurons is detected more sensitively than the movement of cobalt. At least two factors might contribute to these differences: cobalt is toxic to neurons while biocytin is not, and the HRP method used to visualize biocytin amplifies even faint signals effectively.

One possible interpretation of the different maximum numbers of cell bodies in biocytin and cobalt preparations is that the biocytin numbers include interneurons coupled to the motor neurons. Some swimmeret motor neurons make electrical synapses with local interneurons in the LN (Paul and Mulloney, 1985a,b). The enzymatic amplification by HRP might have made visible the faint signal in local interneurons electrically-coupled to swimmeret motor neurons. If this is correct, then the difference between the biocytin and the cobalt numbers predicts there are about eight interneurons coupled to neurons with axons in the anterior and posterior branches of each N1. These coupled neurons might include members of the pattern-generating kernel of each module (Skinner and Mulloney, 1998).

Functional significance of cell body position

In this species, cell bodies of PS and RS motor neurons are partially segregated into clusters anterior to and posterior to the base of the swimmeret nerve. This conclusion is based on the separation of their axons into different branches of N1 (Sherff and Mulloney, 1996, 1997), and physiological recordings of motor activity in these branches (Fig. 1). In Homarus americanus, microelectrode recordings from individual swimmeret motor neurons led Davis (1971) to conclude that “soma position is not related to the spatio-temporal features of the normal motor output program.” It is possible that this is a species difference, but given the methods available for that pioneering study, the correlation of cell-body position and motor output was necessarily indirect. Reviewing the figures in that paper, particularly Davis's Fig. 2G, suggests to us that the same partial segregation into anterior and posterior clusters also occurs in Homarus swimmeret motor neurons.

Differences between the motor pools of the trunk and limbs

Cell bodies of motor neurons that innervate muscles of the trunk are distributed within the abdominal ganglia in a pattern significantly different from that of the swimmeret motor neurons (Fig. 6). The forces that control movements about joints between neighboring abdominal segments are produced by separate sets of slow- and fast-flexor muscles and slow- and fast-extensor muscles. In each segment, these muscles are innervated by four separate sets of identified motor neurons (Wine and Krasne, 1982; Leise et al., 1986, 1987; Drummond and Macmillan, 1998a,b). The numbers of these trunk motor neurons are smaller than the number innervating a swimmeret, but several of these trunk motor neurons have contralateral cell bodies, while others have cell bodies in the next posterior ganglion (Fig. 6). These neurons receive synaptic input bilaterally (Edwards and Mulloney, 1987; Drummond and Macmillan, 1998a,b), and the neurites of contralateral homologues are usually intertwined in the tract neuropil of each ganglion (Leise et al., 1986, 1987).

In contrast, the cell bodies of motor neurons that innervate walking legs or uropods are distributed in the same way as are swimmeret motor cell bodies. Each walking leg has its own population of motor neurons, whose cell bodies are restricted to one thoracic ganglion. Within each thoracic ganglion, the cell bodies of motor neurons for the left and right legs are well separated and occur in clusters near the base of the nerves out which their axons project (Bevengut et al., 1983; Faulkes and Paul, 1997). Their neurites also project first to the thoracic homologues of the LNs (Elson, 1996). The common inhibitor neurons are an interesting exception (Wiens and Wolf, 1993). Like the midline PS neuron (Fig. 3B), their cell bodies are at the ventral midline. Each uropod also has its own set of ipsilateral motor neurons that project to the LN (Kondoh and Hisada, 1986). Although walking legs are larger and mechanically more complex than swimmerets, their motor innervation appears to be serially homologous to the innervation of the swimmerets. It also appears to be organized into modules, and these modules have about the same number of motor neurons as do swimmeret modules (Faulkes and Paul, 1997).

The lateral neuropil is the synaptic core of each module

These anatomical features are consistent with the idea that each swimmeret is controlled by its own neural module (Murchison et al., 1993; Skinner and Mulloney, 1998). According to this idea, each module consists of the motor neurons that innervate one swimmeret, a set of unilateral nonspiking local interneurons that drive these motor neurons, and the sensory afferents that originate in the swimmeret itself. These modules produce the basic alternating bursts of impulses in PS and RS motor neurons required for effective movements of these limbs. Where in the ganglion does the synaptic organization of these units occur?

The structures of PS and RS motor neurons described here recall the structures of unilateral nonspiking local interneurons that drive subsets of these motor neurons (Paul and Mulloney, 1985ab). All the non-spiking local interneurons that are thought to form the kernel of the module's pattern-generating circuit (Skinner and Mulloney, 1998) have their processes largely restricted to the LN. Every swimmeret motor neuron has processes in its LN. Microelectrode recordings from these processes within the LN indicate that this is where motor neurons and local interneurons synapse with one another (Sherff and Mulloney, 1996). We conclude that each LN (Fig. 1C) is the site of the synaptic circuit that generates the basic motor pattern that drives one limb, and the functional core of one swimmeret module.

The behavioral significance of the LN's output is clear; PS-RS alternation is essential for effective forward swimming, and for the other behaviors to which the swimmerets contribute. From the perspective of understanding how the nervous system generates the animal's overt behavior, discovering the organization of these LNs presents a well-defined problem whose solution will be a significant step forward.