- Email: [email protected]
On the dendritic topology and activation of cockroach giant interneurons
J. Insect Physiol., 1971, Vol. 17, pp. 607 to 623. Pergamon Press. Printed in Great Britain
ON THE DENDRITIC TOPOLOGY AND ACTIVATION COCKROACH GIANT INTERNEURONS NANCY S. MILBURN and DAVID
OF
R. BENTLEY
Biology Department, Tufts University, Medford, Massachusetts, Department, University of California at Berkeley
and Zoology
(Received 12 September 1970)
Abstract-Each cercus of Periplaneta americana has 650 to 750 sensilla. The smaller cereal nerve contains about 1000 fibres; the larger contains about 2500. Many of these are under O-5 pm in diameter. The sensilla found to be most effective in driving the giant fibres were cereal thread hairs and cereal bristle hairs. The cereal nerve tract leads directly into the dendritic trees of the giant interneurons. The dendritic trees of two giant intemeurons were filled with Procion Yellow dye. Both display very extensive ipsiluted (to axon) ramification through a large volume of neuropile of the last abdominal ganglion. The location and extent of ramification of the two dendritic trees differs markedly, indicating that the two giant intemeurons probably mediate different integrative activities. One of the giant fibres is attached to a single neuron soma, demonstrating that, at least in one case, the giant intemeurons are not syncitial.
INTRODUCTION
EARLY work on synaptic transmission of nervous impulses through the last abdominal ganglion of the cockroach (PUMPHREY and RAWDON-SMITH, 1937; ROEDER, 1948) resulted in establishing the cereal nerve-giant fibre synapse preparation as one which physiologists have used widely for both research and teaching purposes. This synapse is the first in a series which forms the neuronal basis of the evasion response of the cockroach; synaptic transmission at this site has a readily observable behavioural correlate (ROEDER, 1963). Despite the long sustained interest in this system, few of the details concerning the synaptic architecture, the post-synaptic potentials, and the function of the individual giant fibres in the evasive response have been elucidated until the last few years (CALLEC and BOISTEL, 1966; DAGAN and PAFWG, 1970; PICHON and CALLEC, 1970). Recent work (HESS, 1958; PIPA et al., 1959; CALLEC and BOISTEL, 1966; FARLEY and MILBURN, 1969) has corroborated the older postulate (ROEDER, 1948) that the giant nerve fibres of cockroaches arise from cell bodies located in the sixth abdominal ganglion. Sensory fibres from the cerci synapse on dendrites of the giant fibre system in the sixth abdominal ganglion. Some of these synapses have been identified by electron microscopy (MILBURN, 1968 ; FARLEYand MILBURN, 1969). Other sensory fibres and some interneurons may also end on the dendritic branches of the giant fibres. There is evidence that interneurons exist which 607
608
NANCY
S.
MILBURN
ANDDAVIDR. BENTLEY
modify the activity at the cereal nerve-giant fibre synapses in the sixth abdominal ganglion ( CALLECand BOISTEL,1966 ; MILBURN,1968 ; HARRIS,1969 ; CONSTANTINE, 1970). The extent and exact location of the incoming pre-synaptic fibres and the dendritic trees of several giant fibres and the location of the cell bodies which support the giant fibres have remained in doubt. Although it has been proposed that each of the giant fibres might originate by the fusion of several cells and might be supported by several cell bodies (ROEDER,1948; FARLEYand MILBURN, 1969), the relation of specific cell bodies to the individual giant fibres has never been visualized. Studies of the giant nerve fibres of crayfish and of earthworms using the technique of dye injection have shown that these fibres are each connected to a single neuron soma (REMLERet al., 1968; MULLONEY,1970). The paired median giant fibres of crayfish and lobsters are neigher syncitial nor segmented. They run from the brain, where their cell bodies are located, to drive motor cells found in the abdominal ganglia. The paired lateral giant fibres of these crustaceans, however, are segmental in origin. Within each segment, septa occlude or partially occlude the axons, and each intrasegmental section is supported by a single soma located contralaterally to the axon (REMLERet al., 1968). The giant fibres of earthworms have a similar segmental arrangement and these are also connected to a single soma in each segment. The cell bodies of the earthworm lateral giant fibres are found contralateral to the axons, while the soma attached to each segment of the median giant fibre lies in a mid-ventral position beneath the axon (MULLONEY, 1970). The giant fibres of lobsters, crayfish, and earthworms stand in marked contrast to the well-known giant fibres of cephalopod molluscs which are multicellular in their structure, and to certain giant fibres of other annelids for which the present evidence still indicates a multicellular fusion within single segments of the worm (YOUNG, 1936; NICOL, 1948). About eight giant fibres are found on each side of the cockroach nerve cord. Four of these are located in a dorsal group and the other four are disposed ventrally in the neuropile. Different physiological functions have not yet been associated with different fibres. Neither the complete input to any single giant fibre nor the complete distribution of giant fibre axonal branches have been mapped. Physiological and anatomical studies indicate that the giant fibres may give off collateral branches to interneurons in the thoracic ganglia (FARLEYand MILBURN, 1969) and some or all of the giant fibres continue on through the cord to end in ganglia of the head (PIPA et al., 1959; FARLEYand MILBURN, 1969; SPIRAet al., 1969) where they may serve the function of arousing the animal to prepare for escape (DAGANand PARNAS,1970). Using the dye injection technique developed by STRETTONand KRAVITZ(1968) and adapted for insects by BENTLEY(1970), nineteen giant axons were partially filled with Procion Yellow. The complete dendritic trees of two different giant interneurons in the sixth abdominal ganglion of Periplaneta americana were visualized. In one case it was possible to see the single small contralateral soma which apparently supports the neuron. A number of additional experiments were
DENDRITIC TOPOLOGY ANDACTIVATION OFCOCKROACH GIANTINTERNEURONS 609 conducted to explore the number, type, distribution, sensory structures in driving the giant interneurons. MATERIALS
and effectiveness
of cereal
AND METHODS
Adult male Peri$zneta were immobilized with strips of Plasticine and dissected from the dorsal side to expose the nerve cord. Tracheal connexions to the sixth ganglion were opened at the surface of the perfusion fluid (modified Treherne’s) to maintain an oxygen supply. One connective was desheathed between the fifth and sixth ganglia, and fibres were stripped away until the giant axons were exposed. A glass micropipette containing 30/b Procion Yellow was inserted into the axon and dye was injected by electrophoresis or hydrostatic pressure for periods of up to 8 hr (BENTLEY, 1970). Following the injection the abdominal nerve cords were dissected, fixed in Bouin’s solution, washed, and dehydrated in ethyl alcohol and cleared in methyl benzoate. The ganglia were examined as whole mounts using a fluorescence microscope. Those ganglia which contained a giant fibre well filled with Procion Yellow were embedded in paraffin, cut into sections 15 pm thick, mounted on slides in fluoromount, and examined again by fluorescence microscopy. Photographs of the serial sections were projected onto acetate sheets and the outlines in each section of the fluorescing cell and the ganglion border were drawn. The sheetS were stacked with appropriate spacers and photographed so that a projection of the dendritic tree of each injected cell within the ganglion could be visualized. Measurements were made of the extent of arborization of the cell within the ganglion from this model. Counts of cereal sensilla were made from sections of four cerci prepared for light and electron microscopy, from twelve whole cerci examined under a dissecting scope, and from two cerci examined with a scanning electron microscope. Fibre counts of six cereal nerves were made from material prepared for electron microscopy, embedded in Epon, and examined with an RCA EMU 3G electron microscope. In twelve preparations, observations were made on the generation of giant fibre action potentials and cereal sensory nerve action potentials following stimulation of sensilla of the cerci, styli, and other areas of the caudal region. Extracellular recordings were made with tapered silver hook electrodes or with insulated steel microtip electrodes and conventional electrophysiological recording equipment. Individual hair sensilla were moved by hand with fine glass needles or with a glass stylus attached to a loudspeaker coil and vibrator to provide a Standard vibratory stimulus. The speaker coil was driven by an electronic switch which provided a gradual increase and decrease in amplitude to eliminate the production of very fast transient frequencies as the stimulus was switched on and off. RESULTS
Giant interneurons
Projections of the dendritic trees of two of the giant fibres are shown in Figs. l(A) and l(B). Th ese fibres, which were the most completely filled of the nineteen
610
NANCY S. MILBURN ANDDAVID R. BENTLEY
FIG. 1. The dendritic trees of two dye-injected left ventral giant interneurons (A and B) in the last abdominal ganglion. The figures are retouched photographs taken through acetate models of the ganglia (see text). X indicates the thick segment of fibre A, where most of the dendritic branches originate. P, is the short, thick process with few side branches which is found in most ganglia. Pa is the neurite which extends to the soma of fibre A, and probably also of fibre B. Size Cal: 100 I*
DENDRITICTOPOLOGY
AND
ACTIVATIONOF COCKROACHGIANT INTERNEURONS
611
injectedaxons, will be designated as fibre A and fibre B for purposes of discussion. Although we cannot generalize about the topology of all giant fibre dendritic trees, the structure of these two can be described in some detail. The dendritic tree of fibre A occupies a widespread area in the left half of the neuropile of the last abdominal ganglion. The giant axon originates near the branchpoint of the cell (X) close to the horizontal mid-plane of the ganglion. The axon travels a ventrally directed path anteriorly to enter the ventro-medial region of the connective. Several dendrites branch away from the axon in the anterior quarter of the ganglion, providing a dense arborization of post-synaptic processes. Most of the arborization, however, originates at (X) about 50 pm to the left of the ganglion centre. At the confluence of most of the dendritic branches is a short, thick segment of fibre. The thick segment (Figs. 1, 2) narrows anteriorly and then broadens to form the giant axon. This configuration has been found in a number of cockroach motor neurons (ROWEet al., 1969), and may be an electrically inexcitable integrating segment which precedes the spike initiating zone (SANDEMAN,1969). Four major dendritic branches as well as numerous smaller dendrites leave the axon of fibre A at X. The dendritic branches are directed anteriorly, laterally, and posteriorly. They give rise to a dense feltwork of smaller dendritic branches which must ultimately receive a large number of synaptic inputs. Examination of the very fine dendritic branches with the highest available magnification (950 x ) reveals minute fluorescent strands which twist and bend through the neuropile, occasionally assuming a helical configuration. Along such a strand at regular intervals, it is possible to see small varicosities filled with dye. If these varicosities are a post-synaptic structure then the number of synaptic knobs must be exceedingly high. The dimensions after fixation of the ganglion pictured in Fig. l(A) are 465 pm long, 411 pm wide, and 320 pm thick. The dendritic tree of this single giant fibre extends through a roughly oval portion of the neuropile 320 pm long, 154 pm wide, and 255 pm deep. More than one-third of the entire neuropile is invaded by ramifications of the dendrites of this cell. Although the dendritic tree is confined to one side of the neuropile, two processes cross the mid-line (PI, PZ). The more posterior of these is a thick fibre with few side branches which appears to end shortly beyond the mid-line. Similar processes were seen in several dendritic trees. Fibres displaying this configuration mediate electronic junctions between earthworm giant fibres (MULLONEY, 1970) so these processes may provide communication between contralateral giant fibres. The more anterior process (Ps) is the neurite of this ceil. It crosses the neuropile to the right side of the ganglion and enters a single soma which is located contralateral to the giant axon in the ganglion rind. Just at the point where the neurite crosses from the neuropile into the rind, it executes a downward loop to avoid collision with a small bundle of fibres (Fig. 3). Movement of the fluorescent dye into the soma seems unusually difficult. Apparently, the path of infusion is partially blocked by finger-like processes of the glial trophospongium which
612
NANCY S.
MILBURN
AND
DAVIDR. BENTLEY
penetrate deeply into the initial segment of the neurite, forming a valve-like configuration. However, long rays of dye do extend into the soma in regions where access to somatic cytoplasm is unobstructed (Fig. 3). The soma itself is small (14 ~1 in diameter) and is not one of the largest somata in the ganglion cortex. However, its nucleus is as large as those of cells 25 to 30 p in diameter. Since the neurite was heavily filled with dye and no side branches could be detected, this does appear to be the only neuron soma supporting the giant axon. Giant fibre B (Figs. lB, 2) has a considerably different arrangement of its dendritic ramifications. The axon extends from the ‘integrating’ segment near the mid-ventral portion of the neuropile to a ventrolateral position in the connective. No dendritic branching occurs in the anterior portion of the ganglion. This axon also has a thick initial segment and also gives off two large processes (Pi, PJ which cross the neuropile to the left-hand side of the ganglion. It is probable that P, is the neurite leading to the soma, while P, is homologous to P, of fibre A. The other processes are all of the type which branch immediately and repeatedly to give rise to a dense dendritic tree. The volume of neuropilar space which the tree occupies is roughly spherical and 150 pm in length, 115 pm in width, and 135 pm in depth. It is confined completely to the posterior left-hand quarter of the neuropile; the tract of primary sensory axons from the cereal nerve (Fig. 2) can be seen leading directly into this dendritic tree. Although the dendritic tree of fibre B is less extensive than that of fibre A, it seems to display a more dense arborization and certainly provides a surface for an enormous number of synaptic contacts. Sensory input Curiosity concerning the input to the giant fibres led us to examine the anatomy of cereal receptors and the cereal nerves and to try stimulating individual sensilla and groups of sensilla to see which ones gave rise to giant fibre activity. Cerci of adult uninjured PeripZuneta are composed of 16 to 18 segments in the male and 12 to 13 segments in the female. Three kinds of sensilla are borne on the cerci (Fig. 4). There are short, stout ‘bristle-hairs’, very long, thin ‘thread-hairs’, and small campaniform sensilla (SIHLER, 1924). All three varieties of sensilla probably contribute to the response of the animal to airborne vibration and to gross air movements; the cerci are also sensitive to substrate vibrations. The range of sensitivity of cereal sensilla is impressive-from 50 c/s to 3000-4000 c/s. This sensitivity to low pitch is unusual both in comparison to the sensitivities of the human ear and to acoustic receptors of other insects. The numbers of sensilla on each cercus have been examined by PACKARD (1871) and SIHLER (1924). Packard counted 93 to 105 campaniform sensilla on the cerci of several Periplaneta but his description of them indicates that he counted the bases of thread-hairs as campaniform sensilla. SIHLER (1924) reports that there are only about 50 campaniform sensilla and that they are most abundant on the dorsal side at the base of the cercus and at the cereal tip. Sihler also states that
613
ganglion showing part of FIG. 2. A 45 p section through the terminal abdominal B (filled with Procion yellow). Note the dense dendritic tree, giant : interneuron the ilncoming giant axon, the neurite (solid arrow), and the thick segment where the The tract of sensory fibres axon , the neurite, and most of the dendrites originate. from the cereal nerve can be seen running into the dendritic tree of giant fibre B (unfilled arrows). Size cal: 100 cc.
614
FIG. 3. The soma and part of the neurite of giant interneuron A (filled with Procion Yellow). Entry of the dye into the soma (unfilled arrow) seems to be retarded at the junction of the neurite and the soma by a trophospongium; however, rays of dye can be seen extending through the trophospongium into the region of the nucleus (black arrow). Immediately after leaving the soma, the neurite loops out of the section (15 p) to avoid a bundle of fibres (white arrow). Size cal: 20 p.
61.5
FIG. 4. Scanning electron micrograph of the ventral side of four of the middle segments of a cercus from a male adult Periplaneta americana. Three types of sensilla are present: thread-hairs (t), bristle-hairs (b), and campaniform sensilla (arrows). Size cal: 100 CL.
617
FIG. 6. Giant fibre action potentials generated by stimulation of cereal sensilla. Potentials were recorded with insulated steel microelectrodes in the connective between ganglia V and VI. Fibres could not be activated by stimulation of single sensilla. In these records, two thread-hairs and four bristle-hairs were moved by a This group of hairs was sensitive glass stylus driven by a magnetic coil vibrator. in the range of 90 to 180 c/s; other groups were sensitive in other ranges. (a) 60 c/s, (b) 90 c/s, (c) 130 c/s, (d) 140 c/s, (e) 180 c/s, (f) 200 c/s. Stimulus duration: 60 msec.
DENDRITIC TOPOLOGY AND ACTIVATION
OF COCKROACH GIANT INTERNEURONS
619
are about 100 thread-hairs on each of the cerci of adult males and up to 50 of the bristle-hairs on each of the larger segments. Our counts confirm those of Sihler rather well. We find anywhere from 90 to 110 thread-hairs and 500 to 600 bristle-hairs on each cercus. Most of the hairs are mounted on the convex ventral side of each cercus. Campaniform sensilla are rather hard to find and the sockets of the thread-hairs are easy to mistake for them. We located around 70 campaniform sensilla per cercus, situated laterally or dorsally in most cases. This means that each cercus has 650 to 750 sensilla. How many sensory neurons subserve each of these sensilla ? Both SIHLER (1924) and Hsii (1938) suggest that the majority of the cockroach cereal sensilla of all three varieties are supplied by single sense cells. However, some of the bristle-hairs and thread-hairs of crickets are multiply innervated (Hsti, 1938; J, Edwards, personal communication). In the electron microscope, we examined cross-sections of the cereal nerve taken from the abdominal portion of the nerve and from various points in the cercus (Fig. 5). The smaller cereal nerve contains around 1000 fibres while the larger contains up to 2500 fibres. Many of these fibres are less than 0.5 pm in diameter. The smaller fibres tend to occur in bundles surrounded by a glial sheath. These high numbers of fibres indicate that either the sensory fibres from cereal sense cells branch extensively, or a great many of the sensilla are served by several sense cells. Many of the sensory nerve fibres form synapses with the giant fibre dendrites in the last abdominal ganglion. Another group of cereal fibres must end on the slow fibres which are part of the network carrying impulses to the thoracic motor neurons (DAGANand PARNAS,1970). A third group of cereal fibres probably ends on interneurons which are responsible for the generation of inhibitory potentials in giant fibres as a result of stimulation of the contralateral cercus (CALLEC and BOISTEL,1966). Physiological degeneration experiments (ROEDER,1948; R. Farley and N. Milburn, unpublished) indicate that few if any cereal fibres terminate in ganglia other than the last abdominal ganglion. In order to determine which cereal receptors do end on or drive the giant interneurons, we stimulated individual sensilla with a fine glass needle, while recording from the cereal nerve and from the connective. As other workers have observed (ROEDER, 1948; WOZNIAKet al., 1967) the groups of sensory fibres of the cereal nerve show little adaptation to continued low-frequency stimulation, although individual fibres may adapt. Stimulation of single thread-hairs or of single bristle-hairs is not effective in firing the giant fibres. If the hairs are bent in groups of two to five or more, firing of the giant interneurons occurs (Fig. 6). Different groups of hairs are sensitive to different frequency ranges. Additional giant fibres are recruited as stimulus intensity is increased. Stimulation of the thread-hairs is most effective in firing the giant fibres; stimulation of the bristlehairs is somewhat less effective; and stimulation of both types of hairs together is effective. This supports the conclusion of GUTHRIE and TINDALL (1968) that the acoustic receptors which give rise to a giant interneuron response are the there
620
NANCY S.
MILBVRNANDDAVIDR. BENTJ_EY
thread-hairs. Deformation of either the cereal tip or the cereal base to excite the campaniform sensilla appears to generate many small nerve spikes in abdominal cord, but little giant fibre activity. Low frequency vibration applied directly to the cereal cuticle is very effective in stimulating giant fibres, however. It is possible that the campaniform sensilla contribute to this response. We conclude that the most effective sensory input to the giant interneurons seems to be provided by the cereal hair sensilla. DISCUSSION The giant fibres of Peripluneta are unusual in many ways among the larger class of arthropod giant fibres. Cockroach giant fibres originate in the last abdominal ganglion and dominate by their size (20 to 45 pm in diameter) the neurons of the abdominal cord. They run nearly the entire length of the animal (HESS, 1958; PIPA et al., 1959; FARLEYand MILBURN, 1969; SPIRA et al., 1969). Unlike many giant fibres they lack septa and are not segmental in origin or structure. The size of the axon, the lack of any unusually large cell bodies in the last abdominal ganglion and the apparent origin of giant cell neurites in groups of small neuron somata in the ganglion have led several workers (ROEDER, 1948 ; FARLEY and MILBURN, 1969) to propose that each giant fibre is fused outgrowth of a cluster of somata as is the case with some cephalopod giant neurons (YOUNG, 1936). We have shown that at least one giant axon is attached to only a single cell body. Therefore, the cockroach giant interneurons do not appear to be an exception to the general arthropod avoidance of longitudinally fused axons. The comparatively small size of the soma of the giant interneuron is rather surprising. DAVIS (1971) has demonstrated a strong direct relationship between soma size and axon size in some arthropod motor neurons. Unless the giant axons are a unique case, it appears that this correspondence cannot be generalized to interneurons. Since the last abdominal ganglion is really a fusion of several, probably four, primitive ganglia (GUTHRIE and TINDALL, 1968), the multiplicity of giant fibres may have arisen phylogenetically through the contribution of a pair of neurons by each primitive ganglion. However, the continued presence of several similar giant interneurons raises the question of whether they transmit redundant information, or whether each mediates a different response. Attempts to resolve this question through physiological experiments have so far been inconclusive, although Jensen (personal communication) reports some progress. Several lines of evidence indicate that the control of each giant fibre is fairly complex. Intracellular recordings of CALLEXand BOISTEL(1966) sh ow that the giant interneurons are fired by summed excitatory post-synaptic potentials, and that inhibitory potentials often occur when sensory neurons of the contralateral cercus are stimulated. CONSTANTINE(1970) reports that the giant fibre system displays habituation and dishabituation, and she attributes this to intraganglionic inhibitory interneurons. A secondary or delayed response has been found in some giant fibres by HARRIS(1969) who postulates a special neural network within the ganglion to account for it. Finally, several
DENDRITIC TOPOLOGY ANDACTIVATION OF COCKROACH GIANTINTSRNSuRONS 621 anatomical types of presynaptic endings have been observed in electron micrographs of giant fibre dendrites (FARLEY and MILBURN, 1969), which may represent some of the interneurons involved in these networks. The demonstration that the dendritic trees of fibres A and B differ markedly indicates that these two giant interneurons receive very different inputs and probably serve different functions. EDWARDS(1969, and personal communication) has recently completed a study of the distribution of sensory fibres from the cricket cereal nerve within the last abdominal ganglion. He finds that, like the giant fibre dendritic trees, the terminations of these fibres are confined to a single half of the neuropile (ipsilateral). The fibres can be further divided into two distinct groups. One of these forms a dense glomerulus in the posterior half of the ipsilateral neuropile, coinciding with the position of the dendritic tree of giant fibre B. Fig. 2 indicates that a large group of fibres also terminates around interneuron B in the cockroach. The other group of sensory axons continues through the neuropile and terminates in the anterior half of the ganglion; thus these neurons would be in a position to drive giant fibre A, but not giant fibre B. The structured distribution of sensory terminations in the neuropile, and the superimposition of these terminations on the giant fibre dendritic trees supports the conclusion that the giant interneurons are not mutually redundant, and do transmit unique information. ROEDER (1948) and subsequent workers demonstrated that some giant fibres can be driven by stimulation of either cercus, although the ipsilateral cercus is more effective. Therefore, either the giant fibre dendritic tree must extend to both sides of the neuropile, or the input pathway (sensory fibres or conceivably additional interneurons) must do so. DAVIS (1970) has shown that in parts of the crustacean motor system, synaptic contacts on both sides of the ganglion are primarily established by ramification of the post-synaptic neurons; this is also the case in some vertebrate systems (ECCLES et al., 1967). If the ipsilateral (to axon) distribution of fibres A and B is representative, it appears that in the cereal nervegiant interneuron synapse it is pre-synaptic (sensory neurons or interneurons) rather than post-synaptic cells which establish bilateral communication. Despite all the research on cockroach giant fibres, the question of why they should be so conspicuously larger than other fibres in the connective has never been satisfactorily answered. The relatively slight advantage in transmission velocity has often been cited. Another reason for the existence of large fibres may lie in the fact that Ptmphneta, like other dictyopterous and orthopterous insects, has a highly extensible abdomen. When the abdomen of a cockroach is fully extended, the nerve cord is stretched to about 150 per cent of its slack length. However, transmission time of impulses from the last abdominal ganglion to the first is affected surprisingly little by such stretching (STAFFORD, 1950). Giant fibres of earthworms actually increase in conduction velocity with moderate stretching (GOLDMAN, 1963). Large axon diameter may, therefore, be a mechanism for reliable and invariant impulse transmission through an extensible region.
622
NANCY S. MILEURN ANDDAVIDR. BENTLEY
Acknowledgements-We thank Professor HOWARDBERN of the University of California at Berkeley who provided laboratory facilities for N. MILBURN while she was on sabbatic leave from Tufts University, and Professor KENNETHROEDERfor reviewing the manuscript. The photographic work was done by JOHN UNDERHILLand E. GLENN WRIGHT and the scanning electron micrograph was taken by Dr. JOSEPH GELLER of Photometrics Inc., Lexington, Mass. Support for this research was provided by N.I.H. Grant No. 1 R01 NS09074 to Dr. BENTLEY and National Institutes of Health Grants AI 00947 and IFLO NBO 2155 to Professor ROEDERand to Dr. MILEIURN. REFERENCES BENTLEY D. R. (1970) A topological map of the locust flight system motor neurons. J. Insect Physiol. 16, 905-918. CALLECJ. J. and BOISTELJ. (1966) Etude de divers types d’activites Clectriques enregistrees par microelectrodes capillaires au niveau du dernier ganglion abdominal de la blatte, Periplaneta americana L. C. R. Sot. Biol., Paris 160, 1943-1947. CONSTANTINE S. (1970) Studies of a neural analog and concomitant to habituation in the cockroach, Periplaneta americana. Thesis submitted in partial fulfilment of doctoral degree requirements, Columbia University. DAGAND. and PARNUSI. (1970) Giant fiber and small fiber pathways involved in the evasive response of the cockroach, Periplaneta americana. J. exp. Biol. 52, 313-324. DAVISW. (1970) Motoneuron morphology and synaptic contacts: determination by intracellular dye injection. Science, Wash. 168, 1358-1360. DAVIS W. (1971) Functional significance of motoneuron size and soma position in the swimmeret system of the lobster. J. Neurophyriol. (In press.) ECCLE~J. C., ITO M., and SZENTAGOTHAI (1967) The Cerebellum as a Neuronal Machine. Springer, New York. EDWARDSJ. (1969) The composition of an insect sensory nerve, the cereal nerve of the house cricket, Acheta domestica. 27th Ann. Proc. Electron Micr. Sot. Am. FARLEY R. and MILBURN N. (1969) Structure and function of the giant fibre system in the cockroach, Periplaneta americana. J. Insect Physiol. 15,457-476. GOLDMANL. (1963) The effects of stretch on impulse propagation in the median giant fiber of Lumbricus. J. cell. camp. Physiol. 62, 105-112. GUT~~IB D. M. and TINDALLA. R. (1968) The Biology of the Cockroach. St. Martins Press, London. HARRISC. L. (1969) Existence of a neural circuit for delayed excitation of giant axons in the American cockroach. Thesis submitted in partial fulfilment of Doctoral Degree requirements, The Pennsylvania State University. Hzss A. (1958) Experimental anatomical studies of pathways in the severed central nerve cord of the cockroach. J. Morph. 103,479-502. Hsfi F. (1938) Etude cytologique et comparee sur les sensilla des insectes. Cellule, 147, l-60. MILBUFINN. (1968) Fine structure of the cereal nerve-giant fiber synapse in the sixth abdominal ganglion of Periplaneta americana. Am. Zoologist 8, 778. MULLONEY B. (1970) Structure of the giant fibers of earthworms. Science, Wash. 168, 994-996. NICOL J. A. C. (1948) The giant axons of annelids. Q. Rew. Biol. X3,291-323. PACKARD A. S. (1871) The caudal styles of insects. Sense organs, i.e. abdominal antennae. Am. Nat. 4,620-621. PICHON Y. and CALLECJ. J. (1970) Further studies on synaptic transmission in insects-I. External recording of synaptic potentials in a single giant axon of the cockroach, Periplaneta americana L. J. exp. Biol. 52, 257-265. PIPA R., COOKE. and RICHARDSA. G. (1959) Studies on the hexapod nervous system-II. The histology of the thoracic ganglia of the adult cockroach, PeripEaneta americana L. J. exp. Biol. 15, 101-113.
DENDRITIC TOPOLOGYANDACTIVATION OF COCKROACH GIANTINl’ERNEURONS
623
PUMPHREYR. J. and RAWDON-SMITH A. F. (1937) Synaptic transmission of nervous impulses through the last abdominalganglionof the cockroach. Proc. R. Sot. (B) 122, 106-118. F&MLERM. P., SELVERSTON A. I. and KENNEDY D. (1968) Lateral giant fibers of crayfish: Location of somata by dye injection. Science, Wash. 162,271-283. ROEDERK. D. (1948) Organization of the ascending fiber system in the cockroach, Periplaneta americana. J. exp. Zool. 108, 243-262. ROEDERK. D. (1963) Nerve CelZsand Insect Behavior. Harvard University Press, Cambridge, Mass. ROWE E. C., MO~ERLY B. J., HOWARDH. M., and COHEN M. J. (1969) Morphology of branches of functionally-identified motor neurons in cockroach neuropile. Am. Zoologist 9, 1107. SANDEMAN D. C. (1969) The site of synaptic activity and impulse initiation in an identified motoneuron in the crab brain. J. e@. Riol. 50,771-784. SIHLERH. (1924) Die Sinnesorgane and en Cerci der Insekten. Zool._%. (Physiol.) 45,519~580. SPIRA M., PARNA~I., and BERGMANF. (1969) Histological and electrophysiological studies on the giant axons of the cockroach Periplaneta americana. J. exp. Biol. 50, 615-627. STAFFORDN. J. (19.50) The effects of mechanical stretching on the roach nerve cord. Thesis submitted for the Master of Science degree, Tufts College. STRETTONA. and KRAVITZE. (1968) Neuronal geometry: determination with a technique of intracellular dye injection. Science, Wash. 162, 132-134. WOZNIAKA., ALVAREZR., WILSON E., and GARCIA-AUSTT E. (1967) Cereal potentials in the Periplaneta americana. Acta physiol. Latinoam. 17, 102-l 11. YOUNG J. Z. (1936) The giant nerve fibers and epistellar body of cephalopods. Quart.y. micr. Sn’. 78, 367-386.
ON THE DENDRITIC TOPOLOGY AND ACTIVATION COCKROACH GIANT INTERNEURONS NANCY S. MILBURN and DAVID
OF
R. BENTLEY
Biology Department, Tufts University, Medford, Massachusetts, Department, University of California at Berkeley
and Zoology
(Received 12 September 1970)
Abstract-Each cercus of Periplaneta americana has 650 to 750 sensilla. The smaller cereal nerve contains about 1000 fibres; the larger contains about 2500. Many of these are under O-5 pm in diameter. The sensilla found to be most effective in driving the giant fibres were cereal thread hairs and cereal bristle hairs. The cereal nerve tract leads directly into the dendritic trees of the giant interneurons. The dendritic trees of two giant intemeurons were filled with Procion Yellow dye. Both display very extensive ipsiluted (to axon) ramification through a large volume of neuropile of the last abdominal ganglion. The location and extent of ramification of the two dendritic trees differs markedly, indicating that the two giant intemeurons probably mediate different integrative activities. One of the giant fibres is attached to a single neuron soma, demonstrating that, at least in one case, the giant intemeurons are not syncitial.
INTRODUCTION
EARLY work on synaptic transmission of nervous impulses through the last abdominal ganglion of the cockroach (PUMPHREY and RAWDON-SMITH, 1937; ROEDER, 1948) resulted in establishing the cereal nerve-giant fibre synapse preparation as one which physiologists have used widely for both research and teaching purposes. This synapse is the first in a series which forms the neuronal basis of the evasion response of the cockroach; synaptic transmission at this site has a readily observable behavioural correlate (ROEDER, 1963). Despite the long sustained interest in this system, few of the details concerning the synaptic architecture, the post-synaptic potentials, and the function of the individual giant fibres in the evasive response have been elucidated until the last few years (CALLEC and BOISTEL, 1966; DAGAN and PAFWG, 1970; PICHON and CALLEC, 1970). Recent work (HESS, 1958; PIPA et al., 1959; CALLEC and BOISTEL, 1966; FARLEY and MILBURN, 1969) has corroborated the older postulate (ROEDER, 1948) that the giant nerve fibres of cockroaches arise from cell bodies located in the sixth abdominal ganglion. Sensory fibres from the cerci synapse on dendrites of the giant fibre system in the sixth abdominal ganglion. Some of these synapses have been identified by electron microscopy (MILBURN, 1968 ; FARLEYand MILBURN, 1969). Other sensory fibres and some interneurons may also end on the dendritic branches of the giant fibres. There is evidence that interneurons exist which 607
608
NANCY
S.
MILBURN
ANDDAVIDR. BENTLEY
modify the activity at the cereal nerve-giant fibre synapses in the sixth abdominal ganglion ( CALLECand BOISTEL,1966 ; MILBURN,1968 ; HARRIS,1969 ; CONSTANTINE, 1970). The extent and exact location of the incoming pre-synaptic fibres and the dendritic trees of several giant fibres and the location of the cell bodies which support the giant fibres have remained in doubt. Although it has been proposed that each of the giant fibres might originate by the fusion of several cells and might be supported by several cell bodies (ROEDER,1948; FARLEYand MILBURN, 1969), the relation of specific cell bodies to the individual giant fibres has never been visualized. Studies of the giant nerve fibres of crayfish and of earthworms using the technique of dye injection have shown that these fibres are each connected to a single neuron soma (REMLERet al., 1968; MULLONEY,1970). The paired median giant fibres of crayfish and lobsters are neigher syncitial nor segmented. They run from the brain, where their cell bodies are located, to drive motor cells found in the abdominal ganglia. The paired lateral giant fibres of these crustaceans, however, are segmental in origin. Within each segment, septa occlude or partially occlude the axons, and each intrasegmental section is supported by a single soma located contralaterally to the axon (REMLERet al., 1968). The giant fibres of earthworms have a similar segmental arrangement and these are also connected to a single soma in each segment. The cell bodies of the earthworm lateral giant fibres are found contralateral to the axons, while the soma attached to each segment of the median giant fibre lies in a mid-ventral position beneath the axon (MULLONEY, 1970). The giant fibres of lobsters, crayfish, and earthworms stand in marked contrast to the well-known giant fibres of cephalopod molluscs which are multicellular in their structure, and to certain giant fibres of other annelids for which the present evidence still indicates a multicellular fusion within single segments of the worm (YOUNG, 1936; NICOL, 1948). About eight giant fibres are found on each side of the cockroach nerve cord. Four of these are located in a dorsal group and the other four are disposed ventrally in the neuropile. Different physiological functions have not yet been associated with different fibres. Neither the complete input to any single giant fibre nor the complete distribution of giant fibre axonal branches have been mapped. Physiological and anatomical studies indicate that the giant fibres may give off collateral branches to interneurons in the thoracic ganglia (FARLEYand MILBURN, 1969) and some or all of the giant fibres continue on through the cord to end in ganglia of the head (PIPA et al., 1959; FARLEYand MILBURN, 1969; SPIRAet al., 1969) where they may serve the function of arousing the animal to prepare for escape (DAGANand PARNAS,1970). Using the dye injection technique developed by STRETTONand KRAVITZ(1968) and adapted for insects by BENTLEY(1970), nineteen giant axons were partially filled with Procion Yellow. The complete dendritic trees of two different giant interneurons in the sixth abdominal ganglion of Periplaneta americana were visualized. In one case it was possible to see the single small contralateral soma which apparently supports the neuron. A number of additional experiments were
DENDRITIC TOPOLOGY ANDACTIVATION OFCOCKROACH GIANTINTERNEURONS 609 conducted to explore the number, type, distribution, sensory structures in driving the giant interneurons. MATERIALS
and effectiveness
of cereal
AND METHODS
Adult male Peri$zneta were immobilized with strips of Plasticine and dissected from the dorsal side to expose the nerve cord. Tracheal connexions to the sixth ganglion were opened at the surface of the perfusion fluid (modified Treherne’s) to maintain an oxygen supply. One connective was desheathed between the fifth and sixth ganglia, and fibres were stripped away until the giant axons were exposed. A glass micropipette containing 30/b Procion Yellow was inserted into the axon and dye was injected by electrophoresis or hydrostatic pressure for periods of up to 8 hr (BENTLEY, 1970). Following the injection the abdominal nerve cords were dissected, fixed in Bouin’s solution, washed, and dehydrated in ethyl alcohol and cleared in methyl benzoate. The ganglia were examined as whole mounts using a fluorescence microscope. Those ganglia which contained a giant fibre well filled with Procion Yellow were embedded in paraffin, cut into sections 15 pm thick, mounted on slides in fluoromount, and examined again by fluorescence microscopy. Photographs of the serial sections were projected onto acetate sheets and the outlines in each section of the fluorescing cell and the ganglion border were drawn. The sheetS were stacked with appropriate spacers and photographed so that a projection of the dendritic tree of each injected cell within the ganglion could be visualized. Measurements were made of the extent of arborization of the cell within the ganglion from this model. Counts of cereal sensilla were made from sections of four cerci prepared for light and electron microscopy, from twelve whole cerci examined under a dissecting scope, and from two cerci examined with a scanning electron microscope. Fibre counts of six cereal nerves were made from material prepared for electron microscopy, embedded in Epon, and examined with an RCA EMU 3G electron microscope. In twelve preparations, observations were made on the generation of giant fibre action potentials and cereal sensory nerve action potentials following stimulation of sensilla of the cerci, styli, and other areas of the caudal region. Extracellular recordings were made with tapered silver hook electrodes or with insulated steel microtip electrodes and conventional electrophysiological recording equipment. Individual hair sensilla were moved by hand with fine glass needles or with a glass stylus attached to a loudspeaker coil and vibrator to provide a Standard vibratory stimulus. The speaker coil was driven by an electronic switch which provided a gradual increase and decrease in amplitude to eliminate the production of very fast transient frequencies as the stimulus was switched on and off. RESULTS
Giant interneurons
Projections of the dendritic trees of two of the giant fibres are shown in Figs. l(A) and l(B). Th ese fibres, which were the most completely filled of the nineteen
610
NANCY S. MILBURN ANDDAVID R. BENTLEY
FIG. 1. The dendritic trees of two dye-injected left ventral giant interneurons (A and B) in the last abdominal ganglion. The figures are retouched photographs taken through acetate models of the ganglia (see text). X indicates the thick segment of fibre A, where most of the dendritic branches originate. P, is the short, thick process with few side branches which is found in most ganglia. Pa is the neurite which extends to the soma of fibre A, and probably also of fibre B. Size Cal: 100 I*
DENDRITICTOPOLOGY
AND
ACTIVATIONOF COCKROACHGIANT INTERNEURONS
611
injectedaxons, will be designated as fibre A and fibre B for purposes of discussion. Although we cannot generalize about the topology of all giant fibre dendritic trees, the structure of these two can be described in some detail. The dendritic tree of fibre A occupies a widespread area in the left half of the neuropile of the last abdominal ganglion. The giant axon originates near the branchpoint of the cell (X) close to the horizontal mid-plane of the ganglion. The axon travels a ventrally directed path anteriorly to enter the ventro-medial region of the connective. Several dendrites branch away from the axon in the anterior quarter of the ganglion, providing a dense arborization of post-synaptic processes. Most of the arborization, however, originates at (X) about 50 pm to the left of the ganglion centre. At the confluence of most of the dendritic branches is a short, thick segment of fibre. The thick segment (Figs. 1, 2) narrows anteriorly and then broadens to form the giant axon. This configuration has been found in a number of cockroach motor neurons (ROWEet al., 1969), and may be an electrically inexcitable integrating segment which precedes the spike initiating zone (SANDEMAN,1969). Four major dendritic branches as well as numerous smaller dendrites leave the axon of fibre A at X. The dendritic branches are directed anteriorly, laterally, and posteriorly. They give rise to a dense feltwork of smaller dendritic branches which must ultimately receive a large number of synaptic inputs. Examination of the very fine dendritic branches with the highest available magnification (950 x ) reveals minute fluorescent strands which twist and bend through the neuropile, occasionally assuming a helical configuration. Along such a strand at regular intervals, it is possible to see small varicosities filled with dye. If these varicosities are a post-synaptic structure then the number of synaptic knobs must be exceedingly high. The dimensions after fixation of the ganglion pictured in Fig. l(A) are 465 pm long, 411 pm wide, and 320 pm thick. The dendritic tree of this single giant fibre extends through a roughly oval portion of the neuropile 320 pm long, 154 pm wide, and 255 pm deep. More than one-third of the entire neuropile is invaded by ramifications of the dendrites of this cell. Although the dendritic tree is confined to one side of the neuropile, two processes cross the mid-line (PI, PZ). The more posterior of these is a thick fibre with few side branches which appears to end shortly beyond the mid-line. Similar processes were seen in several dendritic trees. Fibres displaying this configuration mediate electronic junctions between earthworm giant fibres (MULLONEY, 1970) so these processes may provide communication between contralateral giant fibres. The more anterior process (Ps) is the neurite of this ceil. It crosses the neuropile to the right side of the ganglion and enters a single soma which is located contralateral to the giant axon in the ganglion rind. Just at the point where the neurite crosses from the neuropile into the rind, it executes a downward loop to avoid collision with a small bundle of fibres (Fig. 3). Movement of the fluorescent dye into the soma seems unusually difficult. Apparently, the path of infusion is partially blocked by finger-like processes of the glial trophospongium which
612
NANCY S.
MILBURN
AND
DAVIDR. BENTLEY
penetrate deeply into the initial segment of the neurite, forming a valve-like configuration. However, long rays of dye do extend into the soma in regions where access to somatic cytoplasm is unobstructed (Fig. 3). The soma itself is small (14 ~1 in diameter) and is not one of the largest somata in the ganglion cortex. However, its nucleus is as large as those of cells 25 to 30 p in diameter. Since the neurite was heavily filled with dye and no side branches could be detected, this does appear to be the only neuron soma supporting the giant axon. Giant fibre B (Figs. lB, 2) has a considerably different arrangement of its dendritic ramifications. The axon extends from the ‘integrating’ segment near the mid-ventral portion of the neuropile to a ventrolateral position in the connective. No dendritic branching occurs in the anterior portion of the ganglion. This axon also has a thick initial segment and also gives off two large processes (Pi, PJ which cross the neuropile to the left-hand side of the ganglion. It is probable that P, is the neurite leading to the soma, while P, is homologous to P, of fibre A. The other processes are all of the type which branch immediately and repeatedly to give rise to a dense dendritic tree. The volume of neuropilar space which the tree occupies is roughly spherical and 150 pm in length, 115 pm in width, and 135 pm in depth. It is confined completely to the posterior left-hand quarter of the neuropile; the tract of primary sensory axons from the cereal nerve (Fig. 2) can be seen leading directly into this dendritic tree. Although the dendritic tree of fibre B is less extensive than that of fibre A, it seems to display a more dense arborization and certainly provides a surface for an enormous number of synaptic contacts. Sensory input Curiosity concerning the input to the giant fibres led us to examine the anatomy of cereal receptors and the cereal nerves and to try stimulating individual sensilla and groups of sensilla to see which ones gave rise to giant fibre activity. Cerci of adult uninjured PeripZuneta are composed of 16 to 18 segments in the male and 12 to 13 segments in the female. Three kinds of sensilla are borne on the cerci (Fig. 4). There are short, stout ‘bristle-hairs’, very long, thin ‘thread-hairs’, and small campaniform sensilla (SIHLER, 1924). All three varieties of sensilla probably contribute to the response of the animal to airborne vibration and to gross air movements; the cerci are also sensitive to substrate vibrations. The range of sensitivity of cereal sensilla is impressive-from 50 c/s to 3000-4000 c/s. This sensitivity to low pitch is unusual both in comparison to the sensitivities of the human ear and to acoustic receptors of other insects. The numbers of sensilla on each cercus have been examined by PACKARD (1871) and SIHLER (1924). Packard counted 93 to 105 campaniform sensilla on the cerci of several Periplaneta but his description of them indicates that he counted the bases of thread-hairs as campaniform sensilla. SIHLER (1924) reports that there are only about 50 campaniform sensilla and that they are most abundant on the dorsal side at the base of the cercus and at the cereal tip. Sihler also states that
613
ganglion showing part of FIG. 2. A 45 p section through the terminal abdominal B (filled with Procion yellow). Note the dense dendritic tree, giant : interneuron the ilncoming giant axon, the neurite (solid arrow), and the thick segment where the The tract of sensory fibres axon , the neurite, and most of the dendrites originate. from the cereal nerve can be seen running into the dendritic tree of giant fibre B (unfilled arrows). Size cal: 100 cc.
614
FIG. 3. The soma and part of the neurite of giant interneuron A (filled with Procion Yellow). Entry of the dye into the soma (unfilled arrow) seems to be retarded at the junction of the neurite and the soma by a trophospongium; however, rays of dye can be seen extending through the trophospongium into the region of the nucleus (black arrow). Immediately after leaving the soma, the neurite loops out of the section (15 p) to avoid a bundle of fibres (white arrow). Size cal: 20 p.
61.5
FIG. 4. Scanning electron micrograph of the ventral side of four of the middle segments of a cercus from a male adult Periplaneta americana. Three types of sensilla are present: thread-hairs (t), bristle-hairs (b), and campaniform sensilla (arrows). Size cal: 100 CL.
617
FIG. 6. Giant fibre action potentials generated by stimulation of cereal sensilla. Potentials were recorded with insulated steel microelectrodes in the connective between ganglia V and VI. Fibres could not be activated by stimulation of single sensilla. In these records, two thread-hairs and four bristle-hairs were moved by a This group of hairs was sensitive glass stylus driven by a magnetic coil vibrator. in the range of 90 to 180 c/s; other groups were sensitive in other ranges. (a) 60 c/s, (b) 90 c/s, (c) 130 c/s, (d) 140 c/s, (e) 180 c/s, (f) 200 c/s. Stimulus duration: 60 msec.
DENDRITIC TOPOLOGY AND ACTIVATION
OF COCKROACH GIANT INTERNEURONS
619
are about 100 thread-hairs on each of the cerci of adult males and up to 50 of the bristle-hairs on each of the larger segments. Our counts confirm those of Sihler rather well. We find anywhere from 90 to 110 thread-hairs and 500 to 600 bristle-hairs on each cercus. Most of the hairs are mounted on the convex ventral side of each cercus. Campaniform sensilla are rather hard to find and the sockets of the thread-hairs are easy to mistake for them. We located around 70 campaniform sensilla per cercus, situated laterally or dorsally in most cases. This means that each cercus has 650 to 750 sensilla. How many sensory neurons subserve each of these sensilla ? Both SIHLER (1924) and Hsii (1938) suggest that the majority of the cockroach cereal sensilla of all three varieties are supplied by single sense cells. However, some of the bristle-hairs and thread-hairs of crickets are multiply innervated (Hsti, 1938; J, Edwards, personal communication). In the electron microscope, we examined cross-sections of the cereal nerve taken from the abdominal portion of the nerve and from various points in the cercus (Fig. 5). The smaller cereal nerve contains around 1000 fibres while the larger contains up to 2500 fibres. Many of these fibres are less than 0.5 pm in diameter. The smaller fibres tend to occur in bundles surrounded by a glial sheath. These high numbers of fibres indicate that either the sensory fibres from cereal sense cells branch extensively, or a great many of the sensilla are served by several sense cells. Many of the sensory nerve fibres form synapses with the giant fibre dendrites in the last abdominal ganglion. Another group of cereal fibres must end on the slow fibres which are part of the network carrying impulses to the thoracic motor neurons (DAGANand PARNAS,1970). A third group of cereal fibres probably ends on interneurons which are responsible for the generation of inhibitory potentials in giant fibres as a result of stimulation of the contralateral cercus (CALLEC and BOISTEL,1966). Physiological degeneration experiments (ROEDER,1948; R. Farley and N. Milburn, unpublished) indicate that few if any cereal fibres terminate in ganglia other than the last abdominal ganglion. In order to determine which cereal receptors do end on or drive the giant interneurons, we stimulated individual sensilla with a fine glass needle, while recording from the cereal nerve and from the connective. As other workers have observed (ROEDER, 1948; WOZNIAKet al., 1967) the groups of sensory fibres of the cereal nerve show little adaptation to continued low-frequency stimulation, although individual fibres may adapt. Stimulation of single thread-hairs or of single bristle-hairs is not effective in firing the giant fibres. If the hairs are bent in groups of two to five or more, firing of the giant interneurons occurs (Fig. 6). Different groups of hairs are sensitive to different frequency ranges. Additional giant fibres are recruited as stimulus intensity is increased. Stimulation of the thread-hairs is most effective in firing the giant fibres; stimulation of the bristlehairs is somewhat less effective; and stimulation of both types of hairs together is effective. This supports the conclusion of GUTHRIE and TINDALL (1968) that the acoustic receptors which give rise to a giant interneuron response are the there
620
NANCY S.
MILBVRNANDDAVIDR. BENTJ_EY
thread-hairs. Deformation of either the cereal tip or the cereal base to excite the campaniform sensilla appears to generate many small nerve spikes in abdominal cord, but little giant fibre activity. Low frequency vibration applied directly to the cereal cuticle is very effective in stimulating giant fibres, however. It is possible that the campaniform sensilla contribute to this response. We conclude that the most effective sensory input to the giant interneurons seems to be provided by the cereal hair sensilla. DISCUSSION The giant fibres of Peripluneta are unusual in many ways among the larger class of arthropod giant fibres. Cockroach giant fibres originate in the last abdominal ganglion and dominate by their size (20 to 45 pm in diameter) the neurons of the abdominal cord. They run nearly the entire length of the animal (HESS, 1958; PIPA et al., 1959; FARLEYand MILBURN, 1969; SPIRA et al., 1969). Unlike many giant fibres they lack septa and are not segmental in origin or structure. The size of the axon, the lack of any unusually large cell bodies in the last abdominal ganglion and the apparent origin of giant cell neurites in groups of small neuron somata in the ganglion have led several workers (ROEDER, 1948 ; FARLEY and MILBURN, 1969) to propose that each giant fibre is fused outgrowth of a cluster of somata as is the case with some cephalopod giant neurons (YOUNG, 1936). We have shown that at least one giant axon is attached to only a single cell body. Therefore, the cockroach giant interneurons do not appear to be an exception to the general arthropod avoidance of longitudinally fused axons. The comparatively small size of the soma of the giant interneuron is rather surprising. DAVIS (1971) has demonstrated a strong direct relationship between soma size and axon size in some arthropod motor neurons. Unless the giant axons are a unique case, it appears that this correspondence cannot be generalized to interneurons. Since the last abdominal ganglion is really a fusion of several, probably four, primitive ganglia (GUTHRIE and TINDALL, 1968), the multiplicity of giant fibres may have arisen phylogenetically through the contribution of a pair of neurons by each primitive ganglion. However, the continued presence of several similar giant interneurons raises the question of whether they transmit redundant information, or whether each mediates a different response. Attempts to resolve this question through physiological experiments have so far been inconclusive, although Jensen (personal communication) reports some progress. Several lines of evidence indicate that the control of each giant fibre is fairly complex. Intracellular recordings of CALLEXand BOISTEL(1966) sh ow that the giant interneurons are fired by summed excitatory post-synaptic potentials, and that inhibitory potentials often occur when sensory neurons of the contralateral cercus are stimulated. CONSTANTINE(1970) reports that the giant fibre system displays habituation and dishabituation, and she attributes this to intraganglionic inhibitory interneurons. A secondary or delayed response has been found in some giant fibres by HARRIS(1969) who postulates a special neural network within the ganglion to account for it. Finally, several
DENDRITIC TOPOLOGY ANDACTIVATION OF COCKROACH GIANTINTSRNSuRONS 621 anatomical types of presynaptic endings have been observed in electron micrographs of giant fibre dendrites (FARLEY and MILBURN, 1969), which may represent some of the interneurons involved in these networks. The demonstration that the dendritic trees of fibres A and B differ markedly indicates that these two giant interneurons receive very different inputs and probably serve different functions. EDWARDS(1969, and personal communication) has recently completed a study of the distribution of sensory fibres from the cricket cereal nerve within the last abdominal ganglion. He finds that, like the giant fibre dendritic trees, the terminations of these fibres are confined to a single half of the neuropile (ipsilateral). The fibres can be further divided into two distinct groups. One of these forms a dense glomerulus in the posterior half of the ipsilateral neuropile, coinciding with the position of the dendritic tree of giant fibre B. Fig. 2 indicates that a large group of fibres also terminates around interneuron B in the cockroach. The other group of sensory axons continues through the neuropile and terminates in the anterior half of the ganglion; thus these neurons would be in a position to drive giant fibre A, but not giant fibre B. The structured distribution of sensory terminations in the neuropile, and the superimposition of these terminations on the giant fibre dendritic trees supports the conclusion that the giant interneurons are not mutually redundant, and do transmit unique information. ROEDER (1948) and subsequent workers demonstrated that some giant fibres can be driven by stimulation of either cercus, although the ipsilateral cercus is more effective. Therefore, either the giant fibre dendritic tree must extend to both sides of the neuropile, or the input pathway (sensory fibres or conceivably additional interneurons) must do so. DAVIS (1970) has shown that in parts of the crustacean motor system, synaptic contacts on both sides of the ganglion are primarily established by ramification of the post-synaptic neurons; this is also the case in some vertebrate systems (ECCLES et al., 1967). If the ipsilateral (to axon) distribution of fibres A and B is representative, it appears that in the cereal nervegiant interneuron synapse it is pre-synaptic (sensory neurons or interneurons) rather than post-synaptic cells which establish bilateral communication. Despite all the research on cockroach giant fibres, the question of why they should be so conspicuously larger than other fibres in the connective has never been satisfactorily answered. The relatively slight advantage in transmission velocity has often been cited. Another reason for the existence of large fibres may lie in the fact that Ptmphneta, like other dictyopterous and orthopterous insects, has a highly extensible abdomen. When the abdomen of a cockroach is fully extended, the nerve cord is stretched to about 150 per cent of its slack length. However, transmission time of impulses from the last abdominal ganglion to the first is affected surprisingly little by such stretching (STAFFORD, 1950). Giant fibres of earthworms actually increase in conduction velocity with moderate stretching (GOLDMAN, 1963). Large axon diameter may, therefore, be a mechanism for reliable and invariant impulse transmission through an extensible region.
622
NANCY S. MILEURN ANDDAVIDR. BENTLEY
Acknowledgements-We thank Professor HOWARDBERN of the University of California at Berkeley who provided laboratory facilities for N. MILBURN while she was on sabbatic leave from Tufts University, and Professor KENNETHROEDERfor reviewing the manuscript. The photographic work was done by JOHN UNDERHILLand E. GLENN WRIGHT and the scanning electron micrograph was taken by Dr. JOSEPH GELLER of Photometrics Inc., Lexington, Mass. Support for this research was provided by N.I.H. Grant No. 1 R01 NS09074 to Dr. BENTLEY and National Institutes of Health Grants AI 00947 and IFLO NBO 2155 to Professor ROEDERand to Dr. MILEIURN. REFERENCES BENTLEY D. R. (1970) A topological map of the locust flight system motor neurons. J. Insect Physiol. 16, 905-918. CALLECJ. J. and BOISTELJ. (1966) Etude de divers types d’activites Clectriques enregistrees par microelectrodes capillaires au niveau du dernier ganglion abdominal de la blatte, Periplaneta americana L. C. R. Sot. Biol., Paris 160, 1943-1947. CONSTANTINE S. (1970) Studies of a neural analog and concomitant to habituation in the cockroach, Periplaneta americana. Thesis submitted in partial fulfilment of doctoral degree requirements, Columbia University. DAGAND. and PARNUSI. (1970) Giant fiber and small fiber pathways involved in the evasive response of the cockroach, Periplaneta americana. J. exp. Biol. 52, 313-324. DAVISW. (1970) Motoneuron morphology and synaptic contacts: determination by intracellular dye injection. Science, Wash. 168, 1358-1360. DAVIS W. (1971) Functional significance of motoneuron size and soma position in the swimmeret system of the lobster. J. Neurophyriol. (In press.) ECCLE~J. C., ITO M., and SZENTAGOTHAI (1967) The Cerebellum as a Neuronal Machine. Springer, New York. EDWARDSJ. (1969) The composition of an insect sensory nerve, the cereal nerve of the house cricket, Acheta domestica. 27th Ann. Proc. Electron Micr. Sot. Am. FARLEY R. and MILBURN N. (1969) Structure and function of the giant fibre system in the cockroach, Periplaneta americana. J. Insect Physiol. 15,457-476. GOLDMANL. (1963) The effects of stretch on impulse propagation in the median giant fiber of Lumbricus. J. cell. camp. Physiol. 62, 105-112. GUT~~IB D. M. and TINDALLA. R. (1968) The Biology of the Cockroach. St. Martins Press, London. HARRISC. L. (1969) Existence of a neural circuit for delayed excitation of giant axons in the American cockroach. Thesis submitted in partial fulfilment of Doctoral Degree requirements, The Pennsylvania State University. Hzss A. (1958) Experimental anatomical studies of pathways in the severed central nerve cord of the cockroach. J. Morph. 103,479-502. Hsfi F. (1938) Etude cytologique et comparee sur les sensilla des insectes. Cellule, 147, l-60. MILBUFINN. (1968) Fine structure of the cereal nerve-giant fiber synapse in the sixth abdominal ganglion of Periplaneta americana. Am. Zoologist 8, 778. MULLONEY B. (1970) Structure of the giant fibers of earthworms. Science, Wash. 168, 994-996. NICOL J. A. C. (1948) The giant axons of annelids. Q. Rew. Biol. X3,291-323. PACKARD A. S. (1871) The caudal styles of insects. Sense organs, i.e. abdominal antennae. Am. Nat. 4,620-621. PICHON Y. and CALLECJ. J. (1970) Further studies on synaptic transmission in insects-I. External recording of synaptic potentials in a single giant axon of the cockroach, Periplaneta americana L. J. exp. Biol. 52, 257-265. PIPA R., COOKE. and RICHARDSA. G. (1959) Studies on the hexapod nervous system-II. The histology of the thoracic ganglia of the adult cockroach, PeripEaneta americana L. J. exp. Biol. 15, 101-113.
DENDRITIC TOPOLOGYANDACTIVATION OF COCKROACH GIANTINl’ERNEURONS
623
PUMPHREYR. J. and RAWDON-SMITH A. F. (1937) Synaptic transmission of nervous impulses through the last abdominalganglionof the cockroach. Proc. R. Sot. (B) 122, 106-118. F&MLERM. P., SELVERSTON A. I. and KENNEDY D. (1968) Lateral giant fibers of crayfish: Location of somata by dye injection. Science, Wash. 162,271-283. ROEDERK. D. (1948) Organization of the ascending fiber system in the cockroach, Periplaneta americana. J. exp. Zool. 108, 243-262. ROEDERK. D. (1963) Nerve CelZsand Insect Behavior. Harvard University Press, Cambridge, Mass. ROWE E. C., MO~ERLY B. J., HOWARDH. M., and COHEN M. J. (1969) Morphology of branches of functionally-identified motor neurons in cockroach neuropile. Am. Zoologist 9, 1107. SANDEMAN D. C. (1969) The site of synaptic activity and impulse initiation in an identified motoneuron in the crab brain. J. e@. Riol. 50,771-784. SIHLERH. (1924) Die Sinnesorgane and en Cerci der Insekten. Zool._%. (Physiol.) 45,519~580. SPIRA M., PARNA~I., and BERGMANF. (1969) Histological and electrophysiological studies on the giant axons of the cockroach Periplaneta americana. J. exp. Biol. 50, 615-627. STAFFORDN. J. (19.50) The effects of mechanical stretching on the roach nerve cord. Thesis submitted for the Master of Science degree, Tufts College. STRETTONA. and KRAVITZE. (1968) Neuronal geometry: determination with a technique of intracellular dye injection. Science, Wash. 162, 132-134. WOZNIAKA., ALVAREZR., WILSON E., and GARCIA-AUSTT E. (1967) Cereal potentials in the Periplaneta americana. Acta physiol. Latinoam. 17, 102-l 11. YOUNG J. Z. (1936) The giant nerve fibers and epistellar body of cephalopods. Quart.y. micr. Sn’. 78, 367-386.