JEZ 780 THE JOURNAL OF EXPERIMENTAL ZOOLOGY 278:119–132 (1997) Estimation of the Size and Directional Output of Functional Groups of Interneurons Underlying Abdominal Positioning Behaviors in Crayfish LAWRENCE D. BREWER AND JAMES L. LARIMER* Department of Zoology, University of Texas at Austin, Austin, Texas ABSTRACT Quantitative studies were made of a large population of interneurons that controls postural flexion and extension of the crayfish abdomen. The number of interneurons needed to produce a motor program was estimated by stimulating a single abdominal positioning interneuron and recording interneuronal activity that was evoked from rostral and caudal connectives in an isolated abdominal nerve cord. We also examined the role that these functional groups have in producing a stronger motor output in either a rostral or caudal direction and thus specifying various abdominal geometrics. The average number of interneurons responding to stimulation of a single abdominal positioning interneuron was 32 (range: 3–50; n = 27). The average number of interneurons that decreased activity was 10 (range: 2–32). Of 653 activated interneurons from 20 preparations, approximately 43% fired between 2 and 5 Hz, 33% fired between 6 and 15 Hz, and 25% fired >15 Hz. The size of a recruited group was usually but not always correlated with the strength of its motor response or with the direction of motor bias. Therefore, the contribution of a group may depend upon the number of active elements as well as synaptic efficacy. J. Exp. Zool. 278:119–132, 1997. © 1997 Wiley-Liss, Inc. Vertebrate as well as invertebrate movements are encoded within the central nervous system (CNS) by populations of cells (Hensler, ’88; Lee et al., ’88; Zecevic et al., ’89; Churchland and Sejnowski, ’92; Georgopoulos et al., ’92; Tsau et al., ’94). For example, in crayfish and lobsters groups of synaptically interacting premotor interneurons control postural movements of the segmented abdomen (Miall and Larimer, ’82a; Jellies and Larimer, ’85, ’86; Jones and Page, ’86; Larimer ’88; Murphy et al., ’89). Abdominal positioning is a relatively simple behavior involving extension and flexion, however, the number of abdominal positions ranging from full extension to full flexion is quite large. Many of these movements are produced by different intensities of motor output in different segments or directions of the abdominal nerve cord, sometimes referred to as a motor bias (Kennedy et al., ’67; Jones and Page,’86; Larimer and Pease, ’88). A biased motor output is considered a major mechanism to produce the almost infinite number of abdominal geometries routinely seen in animals such as crayfish and lobsters. In crayfish hundreds of premotor abdominal positioning interneurons (APIs) have been described both physiologically and morphologically in abdominal ganglia one through six and many are © 1997 WILEY-LISS, INC. considered identified cells (Miall and Larimer, ’82a, b; Larimer and Jellies, ’83; Larimer and Moore, ’84; Jellies and Larimer, ’85, ’86; Larimer and Pease, ’88; Murphy et al., ’89). An API may originate in any of the six abdominal ganglia with one or more axons projecting in the rostral and/or caudal directions. All axons that project caudally terminate in the sixth abdominal ganglion (A6), and axons that project rostrally extend through A1 into the thoracic ganglia and some terminate in the brain (see Larimer and Jellies, ’83: Fig. 7; Miall and Larimer, ’82b; Larimer and Moore, ’84). Furthermore, each API extends dendrites into most if not all ganglia that it projects through. This type of organization has the potential for extensive recruitment and other synaptic interactions among large groups of cells throughout the abdominal nerve cord. Numerous experiments have indicated that APIs operate as members or elements of a group to produce abdominal positioning movements (Miall and Larimer, ’82a; Jellies and Larimer, ’85, ’86; Murphy et al., ’89). Using population statistics (Lincoln index and *Correspondence to: James L. Larimer, Department of Zoology, University of Texas at Austin, Austin, TX 78712. E-mail: neuroserf @mail.utexas.edu Received 7 November 1996; Revision accepted 23 December 1996 120 L.D. BREWER AND J.L. LARIMER a maximum likelihood method) Larimer and Pease (’88) estimated that there are about 360 APIs in abdominal ganglia one through six. However, the number of APIs that are active during any given behavior is not known. We have designed a method to estimate the number of interneurons or APIs that are synaptically recruited when current is injected into a single API in an isolated abdominal nerve cord. The number of recruited interneurons was estimated by recording from the connectives both rostral and caudal to the ganglion where the impaled API was located. The number of interneurons or other APIs that are excited can be determined since the axon(s) of each API projects through the abdominal nerve cord into the thoracic nerve cord and/or terminates in the sixth and last abdominal ganglion. This study also addressed the question of whether interneuronal activity can be used to predict the direction of motor bias. Motor bias is defined as having a stronger motor output in either the rostral or caudal ganglia. The direction of motor bias was compared to the size and firing frequency of the recruited interneuronal group. The use of isolated nerve cords offers the advantage of eliminating the confusion of sensory feedback in the system. For example, in an intact behaving animal it would be almost impossible to distinguish APIs from sensory and other cells. Our method, however, does not provide a direct measure of the number of interneurons activated in a freely behaving animal; rather, these results represent an estimate of the number of interneurons required to produce a fictive behavior. The motor outputs produced by this method are very similar if not identical to motor outputs observed in restrained, semi-intact preparations (Jellies and Larimer, ’86; Murphy et al., ’89) (Larimer, personal observation). This type of motor output has also been shown to produce abdominal movements in behaving animals (Kennedy et al., ’67; Larimer and Eggleston, ’71). The results reported here support the model proposed by Larimer (’88) that APIs operate within distributed circuits and that functional groups arise from synaptic recruitment. Evidence indicated that in addition to the size of an interneuronal group, firing frequency, synaptic efficacy, and local circuits also influence abdominal positioning motor outputs including motor bias. MATERIALS AND METHODS Both male and female crayfish, Procambarus clarkii, were used in these experiments. All indi- viduals had a rostral-telson length of approximately 8–15 cm. The animals were obtained from Waubun Laboratories (Schriever, LA), maintained in dechlorinated tap water at 18°C, and fed dry commercial cat food. Animals were anesthetized on ice and then dissected in van Harreveld’s saline (van Harreveld, ’36). Isolated abdominal nerve cords with motor roots attached were pinned ventral side up in a Sylgard-lined glass Petri dish filled with fresh saline. The first and sixth abdominal ganglia were removed and the 1–2 (rostral) and 5–6 (caudal) connectives were desheathed and teased into bundles for extracellular recording (Fig. 1). The third abdominal ganglion was desheathed with fine forceps for intracellular impalement with microelectrodes. All APIs were impaled in the third abdominal ganglion (A3). This ganglion was selected because a large number of abdominal positioning elements were identified in A3 in earlier studies. A WPI 767 intracellular probe with bridge balance (up to 1,000 MΩ) was coupled to microelectrodes for stimulating and recording from impaled APIs. Microelectrodes were sometimes filled with 3 M KCl (20–40 MΩ). Alternatively, microelectrode tips were filled with an aqueous 3% Lucifer Yellow CH solution (Sigma, St. Louis, MO), and the shanks were filled with 1 M lithium chloride (50– 150 MΩ) (Stewart, ’78). In most experiments, following physiological examination, the fluorescent dye Lucifer Yellow CH (Sigma) was injected into impaled APIs by passing hyperpolarizing pulses between 3 and 5 nA for 500–750 ms every second or 1.5 s for 5–30 min. Nerve cords were then fixed in 4% paraformaldehyde for 12–24 h, rinsed in Sorenson’s buffer (3 × 30 min washes), dehydrated in an ethanol series from 50 to 100%, and cleared in methyl salicylate for 30 min. Dye-filled APIs were viewed with a compound microscope under near ultraviolet (UV) light (430 nm) and drawn using a camera lucida. Suction electrodes were used to record spontaneous and evoked interneuronal activity from axon bundles in the rostral and caudal connectives (Fig. 1). Fictive tonic abdominal flexion motor activity was usually recorded from the third superficial motor roots of the second abdominal ganglion (A2F) and from the fourth abdominal ganglion (A4F). Extensor activity was recorded from the second motor root of the fourth abdominal ganglion (A4E) (Fig. 1). All extracellular electrical activity was amplified using differential AC amplifiers (AM Systems, model 1700; Everett, WA). SIZE AND OUTPUT OF GROUPS OF INTERNEURONS Fig. 1. Diagram showing an isolated abdominal nerve cord preparation with the first and sixth ganglia removed. A microelectrode (IN) was used to stimulate, record, and dye-fill APIs in the third abdominal ganglion (A3). Suction electrodes were used to monitor flexor motor activity from the superficial third roots (R3) of the second (A2F) and fourth (A4F) abdominal ganglia. Extensor motor activity was monitored from the second root (R2) of the fourth abdominal ganglion (A4E). Suction electrodes were also used to record interneuronal activity from the rostral (RC) and caudal connectives (CC), which were teased into bundles. 121 Any changes in interneuronal impulse activity in the ventral nerve cord in response to stimulation of an impaled API were recorded from the rostral and caudal connectives (Fig. 1). This allowed us to determine the number of interneurons recruited during a fictive motor program (see below). Furthermore, stimulation of some APIs produces a stronger motor output (bias) in the rostral or caudal direction. Hence, we were also able to correlate this directional motor bias with interneuronal activity in the nerve cord. Motor activity was displayed and photographed on a Tektronix 5111A storage oscilloscope (Beaverton, OR) or recorded directly with an Astro-Med Dash IV chart recorder (West Warrick, RI). Interneuronal activity was stored on tape using a Vetter FM VHS recorder (Rebersburg, PA) and later displayed with the Dash IV using a pretriggered data capture module. Waveforms (action potentials) were sampled at 10 kHz per channel with 64 kilosamples of total memory per channel. This digital, high frequency record was used to identify individual waveforms. Specific interneuronal records from each animal taken from the Dash IV were copied as transparencies. Each unique waveform (impulse) on the transparency served as a template to identify similar waveshapes from the complete chart recording (Fig. 2A–C). Data were analyzed 1 s before, 1 s during, and 1 s after API stimulation. Most stimulations were of 1 s duration. Only enough current was injected to produce a ficitive motor output typical of a naturally produced and observable abdominal positioning movement. The following considerations led us to this stimulus paradigm. First, stimulating different APIs with a standard current cannot be used because many APIs require different amounts of current to give a motor output. Also, different APIs were impaled at different locations. Second, one cannot simply use a standard firing frequency for each API because some APIs give a normal-looking motor output at 50 Hz while others require 90 Hz. Third, it is generally known that with crayfish and other motor systems, an increase in stimulus intensity results in an increase in motor neuronal recruitment and firing frequency (Atwood and Wiersma, ’67; Evoy and Kennedy, ’67; Davis and Kennedy, ’72). At extremely low stimulus intensities little or no motor output occurs, and at extremely high stimulus intensities a very strong motor output is produced that is probably not characteristic of behaving animals. Therefore, the best criterion was to set the stimulus strength to give a normal-looking motor program that 122 L.D. BREWER AND J.L. LARIMER would have produced an observable movement in an intact animal. Occasionally coincidental impulses that occurred during high neuronal activity caused distorted waveforms; hence, these data could not be analyzed. While the visual overlay method is time consuming, it was found to be much more reliable than one computer software package that we tested which was designed for this kind of analysis. Interpretation of interneuronal data from the connectives Fig. 2. Chart record used to identify action potentials produced by specific interneurons from the abdominal nerve cord. A: Electrophysiological records of interneuronal activity from two rostral connective bundles (RC) and one caudal connective bundle (CC) before, during, and after stimulation of an API (IN = intracellular electrode). The open arrow refers to a portion of the record shown in B (prestimulus) and the solid arrow refers to a portion of the record shown in C (stimulus = 6 nA). B, C: Chart record of interneuronal activity sampled at 10 kHz. B: The open arrowhead refers to a waveform that was used as a template. A transparency of the template is used to identify other waveforms which match the template as shown in C. C: An example of comparing the template Interneuronal activity was characterized by the number of interneurons participating in a motor program as well as by their firing frequencies. These data are presented in histogram format. The contribution of the impaled cell is not included in these figures even though they were participating in these behaviors. Criteria for determining if a neuron was participating in a behavior using this type of experimental paradigm are not well established so we were forced to design our own criteria. These criteria and their derivations are discussed below. Stimulation of each impaled API evoked activity that can readily be divided into two groups: 1) recruited interneurons that were silent before API stimulation and 2) interneurons that were spontaneously active before API stimulation and increased their activity during the stimulus. Interneurons were further pooled into categories according to their firing frequency: 2–5, 6–10, 11–15, 16–20, and >20 Hz. Spontaneously active interneurons were considered as participating in a fictive positioning behavior if they meet the following criteria. 1) Interneurons that increased activity to above 10 Hz must have increased their firing rate by at least 50% (e.g., prestimulus rate = 10 Hz and firing rate during stimulus = 15 Hz); or interneurons that increased their firing rate by 40–49% were included if their poststimulus firing rate decreased to their prestimulus rate. 2) Interneurons that fired below 11 Hz were only included if their activity increased by 100% (i.e., doubled), or if their firing rate increased by at least 60% and their poststimulus firing rate returned to their with other waveforms. The open arrowhead refers to a waveform that matches the template and is therefore considered to be an action potential from the same neuron. The solid arrowheads refer to waveforms which do not match the template and are therefore considered to be action potentials from different neurons. SIZE AND OUTPUT OF GROUPS OF INTERNEURONS prestimulus rate. 3) Low activity cells that only increased their rates from 1 to 2 Hz were ignored. 4) Interneurons which were inhibited were also analyzed. Interneurons that had a prestimulus rate below 11 Hz were counted as inhibited if their rates decreased by at least 50%. Cells that had a prestimulus rate above 10 Hz were considered to be inhibited if their firing rates decreased by at least 40% during the stimulus. These criteria were determined in part by analyzing data from a single experiment in which the same API was stimulated three times (in preparation). By repetition we could observe how interneurons recorded from the connectives responded or failed to respond to repeated stimulation. This also allowed us to estimate the “noise” that was present in the abdominal nerve cord. While these criteria are not perfect (i.e., some cells that are included may not be part of the abdominal positioning system and vice versa), we feel that they are reasonable based on these observations. In order to accurately determine the number of APIs participating during a fictive movement, activity from neurons that have bidirectional axons should only be counted once. However, this task was not always possible and we may have underestimated the number of bidirectional neurons due to several reasons: 1) branch point failure (Grossman et al., ’79a, b; Nicholls et al., ’92) (see Discussion for further explanation); 2) axonal inhibition in other parts of the nerve cord; 3) waveform distortion; and 4) axonal damage during dissection. In some experiments it was only possible to record from one rostral and caudal hemiconnective. Since APIs and nerve cord activity for flexion and extension are almost entirely bilaterally symmetrical, counts of interneurons from these experiments were doubled. two groups: those that were recruited and those that showed increased activity. Interneurons classed as recruited were silent before API stimulation. Other interneurons were spontaneously active before stimulation but increased their firing rate during the stimulus. The average number of interneurons recruited in response to stimulation of an API was 18 (range: 3–50; n = 27 preparations). The average number of interneurons that increased activity was 14 (range: 3– 32) and the average number of interneurons that decreased activity was 10 (range: 2–32). The number of interneurons activated and inhibited during each of these API stimulations is summarized in Table 1. Stimulation of most APIs produced a flexor motor output, one API produced an extension motor program, and activation of two other APIs excited both flexor and extensor motor neurons. This type of mixed motor output has been TABLE 1. Number of recruited interneurons, number of interneurons that increased (+) activity, and number of interneurons that decreased (–) activity in response to intracellular stimulation of 27 APIs1 Comments a a a a, b a a a a c RESULTS Single APIs were impaled and depolarized to give a motor output typical of those underlying postural movements. At the same time recordings were made of evoked axonal activity from the rostral and caudal connectives. Several features regarding the interneuronal circuitry controlling abdominal positioning were determined. First, we estimated the size of each interneuronal group and the firing frequency of each unit. Second, we studied the relationship between interneuronal group activity and the production of different motor outputs in different segments of the nerve cord. Evoked interneuronal activity was divided into 123 a a a, d c Average (SD) 1 Recruited 30 22 8 8 6 50 26 10 38 8 8 13 11 7 12 3 18 6 10 26 28 34 28 22 17 20 23 18 (11.5) +Activity 8 10 3 4 6 18 32 8 20 11 10 16 11 16 17 11 13 20 16 16 22 20 14 15 20 10 10 14 (6.3) –Activity 8 4 3 3 6 8 18 2 14 7 9 8 12 10 9 11 9 8 9 10 10 10 32 10 11 7 14 10 (5.7) All APIs produced fictive abdominal flexion unless otherwise indicated. a: Numbers were doubled since this represents a recording from a hemiconnective; b: flexion-producing ingibitor interneuron (this interneuron decreased motor activity); c: mixed motor output (excitatory flexor and extensor motor neurons were activated); d: extension-producing interneuron. 124 L.D. BREWER AND J.L. LARIMER observed previously (Murphy et al., ’89). Stimulation of another type of API activated a group of interneurons that decreased or inhibited flexor motor activity. The number of interneurons recruited and the number that only increased activity were categorized according to their firing frequency (2–5, 6–10, 11–15, 16–20, and >20 Hz). Figure 3 summarizes how interneurons responded during stimulation from 20 experiments in which 653 interneurons were considered to be affected by API stimulation. Forty-three percent of all interneurons that were recruited or that increased activity fired between 2 and 5 Hz. Twenty-four percent of affected interneurons fired between 6–10 Hz. The total percentage of affected interneurons that fired above 10 Hz was about 33%. Of this percentage about half of these fired above 20 Hz. Only very weak motor outputs resulted unless at least one or two interneurons fired at about 20 Hz or more. Interneurons that increased their firing frequencies usually did not exceed 35 Hz, but a few fired as high as 70 Hz. The ranges of firing frequencies of recruited interneurons observed in these experiments were similar to those observed in semi-intact animals (Jellies and Larimer, ’86; Murphy et al., ’89). The latency from stimulus to firing of most recruited interneurons was between 20 and 75 ms. Those interneurons that showed longer latencies tended to fire at low frequencies, generally between 2 and 5 Hz. Following API stimulation, most of the evoked interneuronal activity returned to prestimulus firing rates within 20 ms; however, occasionally some interneurons continued to fire higher than their prestimulus rate. How group size and firing frequency affected motor output Representative examples of the morphology and motor output of some impaled APIs are shown below (Figs. 4 and 5). As indicated in the model proposed by Larimer (’88), stimulation of each API activated a set of interneurons. As might be expected, larger interneuronal groups produced stronger motor outputs than smaller interneuronal groups, as defined by more motor neurons firing at high frequency. However, in a few instances (n = 4 of 27), the size and firing frequency of the interneuronal population were not always correlated with the strength of motor output; some large groups produced a weaker motor output than some smaller groups. Flexion-producing interneuronal population Fig. 3. Percentage of interneurons that were recruited (lined bars) or increased (+) activity (open bars) according to firing frequency (Hz) in response to stimulation of an API. A total of 653 interneurons from 20 preparations were considered to be affected by stimulation of an API. Approximately 43% of all interneurons fired at very low frequencies (2–5 Hz) and 15% fired above 20 Hz. Usually only very weak motor outputs were produced unless a group contained at least one or two cells that fired at about 20 Hz. Stimulation of the API shown in Figure 4A produced a strong flexor motor output (Fig. 4B). This API is one of the most commonly encountered flexion-producing APIs (Jellies and Larimer, ’85; Larimer and Jellies, ’83; Larimer and Moore, ’84; Larimer and Pease, ’88). Analysis of nerve cord activity identified 64 interneurons that were affected by API stimulation. Of these interneurons 34 were recruited (i.e., neurons that were previously silent), 20 increased activity, and 10 were inhibited. The distribution according to firing frequency of interneurons that were recruited and increased activity is presented in Figure 4C. Most interneurons (n = 27) fired at low frequencies (2–5 Hz); however, a relatively large population of APIs (n = 16) fired at higher frequencies (>15 Hz). This group appeared to have a high level of synaptic interactions that produced a strong motor response. This API may be important in evoking natural flexion due to its extensive ability to recruit other neurons. Stimulation of other APIs recruited very few if any interneurons at higher frequencies (i.e., Fig. 5C, column A). SIZE AND OUTPUT OF GROUPS OF INTERNEURONS 125 Different flexion-producing interneuronal groups provide different and unexpected motor outputs While the strength of motor output was usually correlated with the size and firing frequency of the interneuronal group, the examples presented in Figure 5 show that these criteria did not always indicate the strength of motor activity. The interneuronal groups and the motor outputs produced by stimulating two different flexion APIs were compared. Each API had two axons, one that projected in the rostral direction and one that projected in the caudal direction (Fig. 5A 1, B 1). During current injection the firing frequencies of the impaled APIs shown in Figures 5A 1, B 1 were 50 and 60 Hz, respectively. Stimulation of the API shown in Figure 5A1 produced a strong flexor motor output in the fourth abdominal ganglion (A4F) and an even more robust motor output in the second abdominal ganglion (A2F) (Fig. 5A2). Stimulation of the API shown in Figure 5B 1 produced a weaker motor response, especially in the second abdominal ganglion (A2F) (Fig. 5B2). Stimulation of the API (Fig. 5B1) that produced the weaker motor output activated more than six times as many interneurons at frequencies above 10 Hz (Fig. 5C, column B) than stimulation of the API that produced the stronger motor output (Fig. 5C, column A). Thus the size of an interneuronal group did not always indicate the strength of a motor response. Figure 4. Fig. 4. Interneuronal group activated in response to stimulation of an identified flexion-producing API. A: Morphology of a flexion-producing API. This API was located entirely on one side of the midline and had an axon that projected in both the rostral and caudal directions. Dashed line represents the midline of the ganglion. Rostral is toward the top. B: Flexor motor output (A4F and A2F) produced during stimulation of this API. The motor neuron with the largest action potential firing in trace A4E is the extensor peripheral inhibitor. I = current; IN = intracellular electrode; A4F = flexor output from fourth abdominal ganglion; A4E = extensor output from fourth abdominal ganglion; A2F = flexor output from second abdominal ganglion. C: Frequency histogram of the number of interneurons, according to firing frequency, that were recruited or increased (+) activity in response to stimulation of the API. Each column represents the combined number of interneurons that were recruited (lined bar) and increased activity (open bar) at a particular frequency. Hence, the total number of affected interneurons as shown in column A at 2–5 Hz was 27. Twenty of these interneurons were recruited and seven showed increased activity. A relatively large number of interneurons (n = 16) were fired above 15 Hz. 126 L.D. BREWER AND J.L. LARIMER Comparison between the strength and direction of interneuronal activity and motor output Interneuronal group activity in the rostral and caudal connectives was compared with the direction and strength of motor output in an effort to explain motor bias. A motor output that is stronger in either the more rostral or caudal motor roots is defined as biased. Rostral motor output refers to motor activity anterior to the site of intracellular impalement (A3) and caudal motor output refers to motor activity posterior to A3 (see Fig. 1). The activity (firing frequency) and the axonal direction of the impaled API were considered when assessing the strength of group activity; however, the contribution of the impaled API itself was not included in the following frequency histograms. Only the number of recruited interneurons and the number of interneurons that increased activity are included in these figures. In 8 of 12 experiments the direction and strength of interneuronal activity were closely correlated with the direction and strength of motor output (see Table 2). Interneuronal activity resulting from stimulation of the remaining four APIs did not reflect the strength of motor output (see Table 2). However, in 10 of 12 experiments the strength and direction of motor output were correlated with the direction of the stimulated APIs axon (Table 2). Fig. 5. Comparison of two interneuronal groups activated by different flexion-producing interneurons. A1, B1: Morphology of two flexion-producing APIs. Each API had an axon that projected in both the rostral and caudal directions. Dashed line represents the midline of the ganglion. Rostral is toward the top. A2B2: Flexor motor outputs produced during stimulation of the command elements in A 1 and B1, respectively. Stimulation of the API shown in A1 produced a slightly stronger flexor motor output (A2) than stimulation of the API shown in B1 (B2). The motor output in root A2F was much stronger in A2 than in B2. During current injection the firing frequencies of the impaled APIs shown in A1 and B1 were 50 and 60 Hz, respectively. The intracellular trace (IN) was not bridge balanced in A 2. See Figure 4 for explanation of the abbreviations. C: Frequency histogram of the number of interneurons, according to spiking frequency, that were recruited or increased (+) activity in response to API stimulation. Data from columns A and B are counts of interneurons resulting from stimulation of the APIs shown in A 1 and B 1, respectively. Very few interneurons were active above 10 Hz during stimulation of the API shown in A1 (column A), but a large population of interneurons was activated above 10 Hz during stimulation of B1 (column B). The population that contained a larger number of interneurons that fired at frequencies above 10 Hz (column B) produced a weaker motor output than the smaller population (column A). This figure follows the same format as explained in Figure 4. TABLE 2. Direction of the stimulated APIs axon, and direction and strength of both the interneuronal and motor activity1 Axonal direction Nerve cord activity Motor activity R and C2 R and C2 R and C2 C2 C2 C2 R2 R2 R and C3 R and C3 R and C3 R3 R=C R=C R<C R<C R<C R<C R=C R>C R<C R<C R>C R=C R=C R =C R<C C only R<C R<C R=C R>C R=C R=C R=C R>C 1 All 3 variables are presented in relation to each other from 12 experiments. All impaled APIs were flexion-producing. R = rostral; c = caudal; > = stronger than; < = weaker than; = indicates no bias observed. 2 Direction and strength of activity in nerve cord were reflected in motor output. 3 Direction and strength of activity in verve cord did not reflect motor output. SIZE AND OUTPUT OF GROUPS OF INTERNEURONS Specific example of interneuronal activity and motor output that was correlated Stimulation of an API with a single axon that projected in the caudal direction (Fig. 6A) resulted 127 in a stronger output from the caudal flexor motor roots (Fig. 6B, trace A4F), while rostral motor output was considerably weaker (Fig. 6B, trace A2F). The number of interneurons responding to API stimulation from the rostral and caudal connectives was about equal, 12 and 14, respectively, but the number of interneurons firing above 10 Hz was four times greater in the caudal direction (Fig. 6C). Eight axons from the caudal connectives fired above 10 Hz, but only two axons from the rostral connectives fired at higher frequencies. Hence, both interneuronal and motor activity were biased (stronger) in the caudal direction. Specific example of interneuronal activity and motor output that was not correlated Stimulation of an API with axons that projected in both the rostral and caudal directions (Fig. 7A) resulted in an equally robust output from the rostral and caudal flexor motor roots (Fig. 7B). However, interneuronal activity in the caudal connectives was stronger than that in the rostral connectives (Fig. 7C). Six axons fired above 20 Hz in the caudal connectives, while only one axon fired above 20 Hz in the rostral connectives. Hence, the stronger interneuronal activity in the caudal connectives might be expected to produce a stronger caudal motor output. However, this was not observed; instead, motor output was not biased in either direction. Inconsistencies of this kind suggest that factors such as synaptic efficacy and the type of interneurons recruited are of importance in explaining bias as well as the impulse traffic traveling in a particular direction. Fig. 6. Stimulation of a flexion-producing interneuron that resulted in both interneuronal and motor output activities that were biased in the caudal direction. A: Morphology of a flexion-producing API that projected an axon in the caudal direction. Dashed line represents the midline of the ganglion. Rostral is toward the top. B: Stimulation of this API resulted in a stronger caudal (A4F) than rostral (A2F) flexor motor output. The intracellular trace (IN) was not bridge balanced. See Figure 4 for explanation of the abbreviations. C: Frequency histogram of the number of interneurons, according to spiking frequency, that were recruited or increased (+) activity in the rostral (column A) and caudal (column B) connectives as a result of intracellular stimulation of the API. Each column represents the combined number of interneurons that were recruited and increased activity at a particular frequency. Hence, the total number of affected interneurons as shown in column A at 2–5 Hz was six. Two of these interneurons were recruited and four showed increased activity. 128 L.D. BREWER AND J.L. LARIMER DISCUSSION The number of interneurons that comprised a functional group controlling the fictive abdominal positioning movements examined here ranged from 11 to 68. A total of 360 APIs have been estimated in abdominal ganglia one through six by Larimer and Pease (’88); therefore, only a subset of this total is active during a fictive behavior. Only APIs in ganglia two through five were examined in this study, and from these ganglia Larimer and Pease (’88) estimated that 214 APIs are present. Since the APIs from ganglia one and six were not included, the number of APIs participating in an abdominal positioning behavior is greater than that reported here. However, we feel our sample gives us some indication of the organization and functional activity of a group of interneurons participating in a specific behavior. While we cannot be sure that every interneuron counted participated in a fictive behavior, previous evidence suggests that most of these interneurons are APIs that are synaptically recruited into functional groups (Jellies and Larimer, ’86; Murphy et al., ’89). Usually interneuronal groups that contained relatively large numbers of interneurons firing at high frequencies (>15 Hz) produced strong motor outputs. Thus coding for abdominal positioning movements appears to be determined by the total spike frequency emanating from a recruited group of APIs. This type of coding also controls the direction of cockroach turning during escape (Liebenthal et al., ’94). Little is known about how APIs are recruited, but it is probable that the strength of sensory or descending inputs influences the number and firing frequencies of the APIs that form a functional group. However, the total number of spikes did not al- Figure 7. Fig. 7. Stimulation of a flexion-producing interneuron that resulted in an interneuronal activity that was stronger in the caudal connectives, but the motor output was equally strong in the rostral and caudal directions. A: Morphology of a flexion-producing API that projected an axon in the rostral direction. Dashed line represents the midline of the ganglion. Rostral is toward the top. B: Stimulation of this API resulted in a flexor motor output that was equally strong in the caudal (A4F) and rostral (A2F) motor roots. See Figure 4 for explanation of the abbreviations. C: Frequency histogram of the number of interneurons, according to spiking frequency, that were recruited or increased (+) activity in the rostral (column A) and caudal (column B) connectives as a result of intracellular stimulation of the API. Interneuronal activity was stronger in the caudal connectives, but the motor outputs were equally strong in the rostral and caudal directions. This figure follows the same format as explained in Figure 6. SIZE AND OUTPUT OF GROUPS OF INTERNEURONS ways indicate the strength of motor output. In a few instances strong motor outputs occurred even though relatively few interneurons fired at high frequencies and vice versa (Fig. 5). Several explanations can be offered to account for these results. Strong synaptic inputs from only a few interneurons could have produced a robust motor output. The large number of interneurons firing at low frequencies could also have contributed toward a strong motor output through temporal and/or spatial summation. The relatively weak motor outputs may have occurred even though interneuronal group activity was strong because very few of these interneurons may have had synaptic inputs onto the motor neurons, or the connections were comparatively weak (Fig. 5). There is also the possibility that inhibitory inputs onto the interneurons or the motor neurons resulted in a weaker motor output. Thus other factors in addition to the size of the recruited group may influence motor neuron firing. These and other possible explanations are further discussed in the following section. Interneuronal activity and motor bias Kennedy et al. (’67) found that stimulation of certain APIs, then called command neurons, produced a stronger motor output in either the rostral or caudal direction (motor bias). The production of motor bias programs is believed to be an important mechanism of achieving a wide variety of abdominal movements. Larimer and Pease (’88) found a strong correlation between the direction of motor bias and the axonal projection of a stimulated API; however, there were exceptions to this rule. The results reported here confirm and extend the findings of Larimer and Pease (’88). We have added a new dimension to the analysis of bias by comparing the direction and magnitude of bias with the direction, size, and firing frequencies of the interneuronal groups that underlie fictive behaviors. The strength and direction of interneuronal activity were compared with the strength and the direction of motor output from interneuronal groups recruited by stimulation of 12 different APIs (Table 2). Six of these 12 groups exhibited a motor bias, and in 5 of these 6 the strength and direction of interneuronal activity reflected the direction of motor bias. The most likely mechanisms that could produce a motor bias would include a larger population of APIs firing in one direction than another, stronger synaptic inputs in one direction, or inhibitory inputs onto the APIs or onto motor neurons, in particular ganglia. For example, 129 in one experiment there was a rostral motor bias but the nerve cord activity was equally robust in the rostral and caudal directions (Table 2). In this case we would predict that there were stronger synaptic inputs onto the motor neurons in the more rostral ganglia, or inhibitory inputs onto the interneurons or motor neurons in the more caudal ganglia resulted in a weaker caudal motor output. However, we cannot rule out the possibility that some of the interneuronal activity in the caudal direction may have been primarily concerned with other motor systems (Burdohan and Larimer, ’95). Some of the data from Table 2 can be interpreted in another manner. Six of the 12 APIs stimulated had bidirectional axons. Stimulation of three of these APIs resulted in a correlation between interneuronal and motor activity. Stimulation of the remaining three APIs produced a stronger interneuronal activity in the rostral or caudal direction, but the motor output was not biased (Table 2). Therefore, we must conclude that the direction of interneuronal activity does not always indicate the direction of a motor bias. Several explanations can be offered to account for these discrepancies. First, APIs are involved in coordinating with several motor systems (Murchison and Larimer, ’90, ’92; Chrachri et al., ’94; Burdohan and Larimer, ’95); therefore, some of these interneurons may have been involved primarily with motor systems other than abdominal positioning. Second, the synaptic efficacy of APIs could greatly determine the strength of motor output. Third, flexor inhibitory interneurons are frequently encountered and some APIs frequently receive inhibitory postsynaptic potentials (personal observation). Therefore, an API providing excitatory inputs onto motor neurons in one ganglion may be inhibited in another ganglion. One inhibitory interneuron was analyzed (Table 1), and stimulation of this flexor inhibitor evoked activity in 12 interneurons that fired at frequencies that ranged from 3 to 33 Hz. Little is known about how these inhibitory interneurons interact within the abdominal positioning system. Fourth, neuromodulators have been shown to affect swimmeret and abdominal positioning motor systems (Livingstone et al., ’80; Harris-Warrick and Kravitz, ’84; Kravitz, ’88; Ma et al., ’92; Barthe et al., ’93; Chrachri and Neil, ’93). It is conceivable that a modulator could be affecting some ganglia more than others. Fifth, local interneurons may be important modulators within the abdominal positioning system. At least two iden- 130 L.D. BREWER AND J.L. LARIMER tified local nonspiking interneurons have been shown to inhibit local excitatory flexor motor activity in P. clarkii (Jellies and Larimer, ’85; Toga et al., ’90). Sixth, interneurons with bidirectional axons could also be important in producing a motor bias. Based on previous morphological data, about one third of all APIs have bidirectional axons, thus about two thirds of the remaining APIs have axons that course only in the rostral or caudal direction (Miall and Larimer, ’82a, b; Larimer and Jellies, ’83; Larimer and Moore, ’84; Jellies and Larimer, ’85, ’86). Only occasionally was it possible to show that a given recruited API has a bidirectional axon based on analyzing firing frequency in the rostral and caudal connectives. Therefore, the number of recruited APIs with bidirectional axons was probably underestimated. We can speculate that the inability to detect these recruited APIs may have been due to several factors: 1) axonal inhibition in other parts of the nerve cord, 2) axonal damage during dissection, 3) waveform distortion, or 4) branch point failure (Grossman et al., ’79a, b; Nicholls et al., ’92). Conduction block or branch point failure could produce a motor bias since one axon firing at a high frequency could produce a vigorous motor response in one direction while the other axon from the same interneuron firing at a lower frequency would produce a weaker response in the other direction. In this manner the parts of the one neuron could function as two separate units. Organization of the abdominal positioning system Why are so many interneurons required for this behavior? At least two explanations can be offered. First, APIs are known to participate in several behaviors, including swimmeret (Murchison and Larimer, ’90, ’92) and uropod movements (Takahata and Hisada, ’85, ’86a, b; Burdohan and Larimer, ’95). Since many APIs have axonal projections that course through the thoracic ganglia (Larimer and Jellies, ’83; Larimer and Moore, ’84) they probably interact with other neural centers as well (Barthe et al., ’93; Chrachri and Neil, ’93; Chrachri et al., ’94). This type of distributed organization has been observed in a wide variety of animals (Cleary and Byrne, ’93; Otto and Hennig, ’94; Tsau et al., ’94; Wu et al., ’94). A second reason for the large number of APIs may be that no single API codes for an entire behavior. Rather, positioning behaviors are encoded by ensembles of interneurons (APIs) organized into functional groups that act in concert (Larimer, ’88). During any one movement, a subset (group) of the total population of APIs is activated. Within a group each API makes only a fractional contribution toward a behavior (Larimer, ’88); however, some APIs probably make larger contributions than others based on firing frequency and their synaptic efficacy with motor neurons (discussed below). Each API may belong to more than one group, and as a result the contribution of any one API during different behaviors may be quite variable. For example, an API may fire at a high frequency while acting as a member of one group, but as part of another group the same API may fire at a lower frequency. The combination of many interneurons or APIs into a seemingly infinite number of groups provides the variety required to account for the large repertoire of abdominal positioning behaviors. While the number of groups formed are extremely large, we have recently completed studies that indicate that at least some of these groups may have some degree of cohesion (Brewer and Larimer, ’94). For example, repetitively stimulating the same API with a microelectrode in the same animal consistently activates the same group of interneurons provided synaptic fatigue does not occur. Furthermore, stimulating the same API in different animals activates a similar but not identical number of interneurons (Brewer and Larimer, ’94). Comparison of abdominal positioning system with other motor systems Several large populations of neurons that control behavior in invertebrates as well as vertebrates are now being studied. As with the abdominal positioning system, other systems appear to use subsets from the larger interneuronal population to perform particular tasks. During cockroach escape behavior a large population of thoracic interneurons (over 100) controls leg movements. These thoracic interneurons project through several ganglia to affect leg motor neurons and are organized into parallel and serial groups (Ritzmann and Pollack, ’86, ’90; Casagrand and Ritzmann, ’91). Furthermore, some interneurons or groups affect motor neuronal output more strongly than other groups (Ritzmann and Pollack, ’90). Several other systems are also organized similarly, including the interneurons that control the head movements of crickets during eye cleaning (Hensler, ’88), the local bending reflex of leeches (Lockery and Kristan, ’90), and the descending propriospinal neurons that control SIZE AND OUTPUT OF GROUPS OF INTERNEURONS fictive scratching in turtles (Berkowitz and Stein, ’94a, b). Hence, the interneuronal organization that controls these behaviors is very similar to the abdominal positioning system. In summary, abdominal positioning movements are encoded by a synaptically recruited function group of interneurons that represents a subset formed from a larger population of APIs. Our observations show that some APIs recruited large groups of interneurons while others recruited much smaller groups. Usually, large groups with interneurons firing at high frequencies produced strong motor outputs while smaller groups with interneurons firing at lower frequencies produced weaker motor outputs. However, on occasion some large groups with interneurons firing at high frequencies produced relatively weak motor outputs, and some smaller groups produced relatively strong motor outputs. The contribution of a group therefore may depend on such factors as group size, firing frequency, synaptic efficacy, the presence of inhibitory inputs, neuromodulation, and the role of local circuits. ACKNOWLEDGMENTS Figures were prepared by Gwen Gage, Janet Young, and Kristina Schlegel. We thank Drs. Wesly Thompson and John Burdohan for reading an early version of the manuscript. This research was supported by NIH grant NS05423, a Jacob Javits award to J.L.L. LITERATURE CITED Atwood, H.L., and C.A. Wiersma (1967) Command interneurons in the crayfish nervous system. J. Exp. Biol., 46:249–261. Barthe, J.-Y., M. Bevengut, and F. Clarac (1993) In vitro, proctolin and serotonin induced modulations of the abdominal motor system activities in crayfish. Brain Res., 623:101–109. Berkowitz, A., and P.S.G. Stein (1994a) Activity of descending propriospinal axons in the turtle hindlimb enlargement during two forms of fictive scratching: Broad tuning to regions of the body surface. J. Neurosci., 14:5089–5104. Berkowitz, A., and P.S.G. Stein (1994b) Activity of descending propriospinal axons in the turtle hindlimb enlargement during two forms of fictive scratching: Phase analyses. J. Neurosci., 14:5105–5119. Brewer, L.D., and J.L. Larimer (1994) Estimation of identified interneurons forming a functional group controlling abdominal positioning in the crayfish. Soc. Neurosci. Abstr., 20:1407. Burdohan, J.A., and J.L. Larimer (1995) Interneurons involved in the control of multiple motor centers in crayfish. J. Exp. Zool., 273:204–215. Casagrand, J.L., and R.E. Ritzmann (1991) Localization of ventral giant interneuron connections to the ventral me- 131 dian branch of thoracic interneurons in the cockroach. J. Neurobiol., 22:643–658. Chrachri, A., and D.M. Neil (1993) Interaction and synchronization between two abdominal motor systems in crayfish. J. Neurophysiol., 69:1373–1383. Chrachri, A., D.M. Neil, and B. Mulloney (1994) Statedependent responses of two motor systems in the crayfish, Pacifastacus leniusculus. J. Comp. Physiol. A., 175:371–380. Churchland, P.S., and T.J. Sejnowski (1992) The Computational Brain. MIT Press, Cambridge, MA. Cleary, L.J., and J.H. Byrne (1993) Identification and characterization of a multifunction neuron contributing to defensive arousal in Aplysia. J. Neurophysiol., 70:1767–1776. Davis, W.J., and D. Kennedy (1972) Command interneurons controlling swimmeret movements in the lobster. II. Interactions of effects on motoneurons. J. Neurophysiol., 35:13–19. Evoy, W.H., and D. Kennedy (1967) Central nervous organization underlying control of antagonistic muscles in the crayfish. I. Types of command fibers. J. Exp. Zool., 165:223–238. Georgopoulos, A.P., J. Ashe, N. Smyrnis, and M. Taira (1992) The motor cortex and the coding of force. Science, 256:1692–1695. Grossman, Y., I. Parnas, and M.E. Spira (1979a) Differential conduction block in branches of a bifurcating axon. J. Physiol., 295:283–305. Grossman, Y., I. Parnas, and M.E. Spira (1979b) Ionic mechanisms involved in differential conduction of action potentials at high frequency in a branching axon. J. Physiol., 295:307–322. Harris-Warrick, R.M., and E.A. Kravitz (1984) Cellular mechanisms for modulation of posture by octopamine and serotonin in the lobster. J. Neurosci., 4:1976–1993. Hensler, K. (1988) Intersegmental interneurons involved in the control of head movements in crickets. J. Comp. Physiol. A., 162:111–126. Jellies, J., and J.L. Larimer (1985) Synaptic interactions between neurons involved in the production of abdominal posture in the crayfish. J. Comp. Physiol. A, 156:861–873. Jellies, J., and J.L. Larimer (1986) Activity of crayfish abdominal positioning interneurons during spontaneous and sensory-evoked movements. J. Exp. Biol., 120:173–188. Jones, K.A., and C.H. Page (1986) Postural interneurons in the abdominal nervous system of lobster. II. Evidence for neurons having both command and driver roles. J. Comp. Physiol. A, 158:273–280. Kennedy, D., W.H. Evoy, B. Dan, and J.T. Hanawalt (1967) The central nervous organization underlying control of antagonistic muscles in the crayfish. II. Coding of position by command fibers. J. Exp. Zool., 165:239–248. Kravitz, E.A. (1988) Hormonal control of behavior: Amines and the biasing of behavioral output in lobsters. Science, 241:1775–1781. Larimer, J.L. (1988) The command hypothesis: A new view using an old example. Trends Neurosci., 11:506–510. Larimer, J.L., and A. Eggleston (1971) Motor programs for abdominal positioning in crayfish. Z. Vgl. Physiol., 74:388–402. Larimer, J.L., and J. Jellies (1983) The organization of flexion-evoking interneurons in the abdominal nerve cord of the crayfish, Procambarus clarkii. J. Exp. Zool., 226:341–351. Larimer, J.L., and D. Moore (1984) Abdominal positioning in- 132 L.D. BREWER AND J.L. LARIMER terneurons in crayfish: Projections to and synaptic activation by higher CNS centers. J. Exp. Zool., 230:1–10. Larimer, J.L., and C.M. Pease (1988) A quantitative study of command elements for abdominal positioning behavior in the crayfish, Procambarus clarkii. J. Exp. Zool., 247:45–55. Lee, C., W.H. Rohrer, and D.L. Sparks (1988) Population coding of saccadic eye movements by neurons in the superior colliculus. Nature, 332:357–360. Liebenthal, E., O. Uhlmann, and J.M. Camhi (1994) Critical parameters of the spike trains in a cell assembly: Coding of turn direction by the giant interneurons of the cockroach. J. Comp. Physiol. A, 174:281–296. Livingstone, M., R.M. Harris-Warrick, and E.A. Kravitz (1980) Serotonin and octopamine produce opposite postures in lobsters. Science, 208:76–79. Lockery, S.R., and W.B. Kristan (1990) Distributed processing of sensory information in the leech. II. Identification of interneurons contributing to the local bending reflex. J. Neurosci., 10:1816–1829. Ma, P.M., B.S. Beltz, and E.A. Kravitz (1992) Serotonin-containing neurons in lobsters: Their role as “gain-setters” in postural control mechanisms. J. Neurophysiol., 68:36–54. Miall, R.C., and J.L. Larimer (1982a) Interneurons involved in abdominal posture in crayfish: Structure, function and command fiber responses. J. Comp. Physiol. A, 148:159–173. Miall, R.C., and J.L. Larimer (1982b) Central organization of crustacean abdominal posture motoneurons: Connectivity and command fiber inputs. J. Exp. Zool., 224:45–56. Murchison, D., and J.L. Larimer (1990) Dual motor output interneurons in the abdominal ganglia of the crayfish Procambrus clarkii: Synaptic activation of motor outputs in both the swimmeret and abdominal positioning systems by single interneurons. J. Exp. Biol., 150:269–293. Murchison, D., and J.L. Larimer (1992) Synaptic interactions among neurons that coordinate swimmeret and abdominal movements in the crayfish. J. Comp. Physiol. A, 170:739–747. Murphy, B.F., M.L. McAnelly, and J.L. Larimer (1989) Abdominal positioning interneurons in crayfish: Participation in behavioral acts. J. Comp. Physiol. A, 165:461–470. Nicholls, J.N., A.R. Martin, and B.G. Wallace (1992) From Neuron to Brain, Ed. 3. Sinaurer Associates, Inc., Sunderland, MA, pp. 140, 450. Otto, D., and R.M. Hennig (1993) Interneurons descending from the cricket subesophageal ganglion control stridulation and ventilation. Naturwissenschaften, 80:36–38. Ritzmann, R.E., and A.L. Pollack (1986) Identification of thoracic interneurons that mediate giant interneuron-to-motor pathways in the cockroach. J. Comp. Physiol. A, 159:639–654. Ritzmann, R.E., and A.L. Pollack (1990) Parallel motor pathways from thoracic interneurons of the ventral giant interneuron system of the cockroach, Periplaneta americana. J Neurobiol., 21:1219–1235. Stewart, W.W. (1978) Functional connections between cells as revealed by dye-coupling with a highly fluorescent napthalamide tracer. Cell, 14:741–759. Takahata, M., and M. Hisada (1985) Interactions between the motor systems controlling uropod steering and abdominal posture in crayfish. J. Comp. Physiol. A, 157:547–554. Takahata, M., and M. Hisada (1986a) Local nonspiking interneurons involved in gating of the descending motor pathway in crayfish. J. Neurophysiol., 56:718–731. Takahata, M., and M. Hisada (1986b) Sustained membrane potential change of uropod motor neurons during the fictive abdominal posture movement in crayfish. J. Neurophysiol., 56:702–717. Toga, T., M. Takahata, and M. Hisada (1990) An identified set of local nonspiking interneurons which control the activity of abdominal postural motoneurones in crayfish. J. Exp. Biol., 148:477–482. Tsau, Y., J.-Y. Wu, H.P. Hopp, L.B. Cohen, D. Schiminovich, and C.X. Falk (1994) Distributed aspects of the response to siphon touch in Aplysia: Spread of stimulus information and cross correlation analysis. J. Neurosci., 14:4167–4184. van Harreveld, A. (1936) A physiological solution for freshwater crustaceans. Proc. Soc. Exp. Biol., 34:428–432. Wu, J.-Y., L.B. Cohen, and C.X. Falk (1994) Neuronal activity during different behaviors in Aplysia: A distributed organization? Science, 263:820–823. Zecevic, D., J.-Y. Wu, L.B. Cohen, J.A. London, H.P. Hopp, and C.X. Falk (1989) Hundreds of neurons in the Aplysia abdominal ganglion are active during the gill-withdrawal reflex. J. Neurosci., 9:3681–3689.