critical parameters of the spike trains in a cell assembly ...raghav/pdfs/animalbehavior/... ·...

J Comp Physiol A (1994) 174:281-296 dleunn ef

~

�9 Springer-Verlag 1994

Critical parameters of the spike trains in a cell assembly: coding of turn direction by the giant interneurons of the cockroach E. Liebenthal, O. Uhlmann, J. M. Camhi

Department of Cell and Animal Biology, Life Sciences Institute, Hebrew University, Jerusalem 91904, Israel

Accepted: 11 October 1993

Abstract. Cockroaches (Periplaneta americana) respond to air displacement produced by an approaching predator by turning and running away. A set of 4 bilateral pairs of ventral giant interneurons is important in determining turn direction. Wind from a given side is known to produce more spikes, an earlier onset of the spike trains, and different fine temporal patterning, in the ipsilateral vs the contralateral set of these interneurons. Here we investigate which of these spike train parameters the cockroach actually uses to determine the direction it will turn.

We delivered controlled wind puffs from the right front, together with intracellular injection of spike trains in a left ventral giant interneuron, under conditions where the animal could make normally directed turning movements of the legs and body. In trials where our stimuli caused the left side to give both the first spike and more total spikes than the right, but where our injected spike train included none of the normal fine temporal patterning, 92% of the evoked turns were to the right- opposite of normal (Figs. 4-6). In trials where the left side gave the first spike, but the right side gave more spikes, 100% of the turns were to the left-the normal direction (Figs. 8, 9). Comparable results were obtained when each of the left giant interneurons 1, 2 or 3 were electrically stimulated, and when either weak or stronger wind puffs were used. Stimulating a left giant interneuron electrically in the absence of a wind puff evoked an escape-like turn on 9% of the trials, and these were all to the right (Fig. 9).

These results indicate that fine temporal patterning in the spike trains is not necessary, and information about which side gives the first spike is not sufficient, to determine turn direction. Rather, the key parameter appears to be relative numbers of action potentials in the left vs the right group of cells. These conclusions were support-

Abbreviations: GI, giant interneuron; vGI, ventral giant interneuron; dGI, dorsal giant interneuron; LY, Lucifer yellow; CF, car- boxyfluorescein

Correspondence to: J.M. Camhi

ed by similar experiments in which extracellular stimulation of several left giant interneurons was paired with right wind (Figs. 11, 12).

Key words: Escape behavior - Neural assemblies Spike trains - Giant interneurons - Cockroach

Introduction

The brain encodes particular features of a sensory stimulus by means of groups, or assemblies, of neurons. Among the vertebrates, the coding of particular sensory features can, in some cases, be attributed to assemblies located in specific brain regions. For instance, with re- gard to directional localization, the subject of the present paper, the mid-temporal area of the monkey cortex plays a role analyzing the direction of moving visual stimuli, and can account for the consequent directed behavior (Salzman et al. 1990, 1992; Britten et al. 1992). Other defined regions or nuclei are responsible for other directional behaviors (Riquimaroux et al. 1992; Glimcher and Sparks 1992; Georgopoulos et al. 1993; Konishi 1993; Metzner 1993).

Beyond identifying responsible brain regions, there is a need to determine the mechanisms by which a cell assembly codes stimulus features. For instance, several regions responsible for stimulus localization have maps of sensory space. At a given moment, then, the region of the map that is most strongly excited corresponds to the current stimulus position (Salzman et al. 1990; Georgopou- los et al. 1993). This implies that the difference in one or more parameters of the cells' activities, across different regions of the map, accounts for the directional specification. It is often assumed that the key parameter used is the number or frequency of action potentials (e.g. Hubel and Wiesel 1968; Britten et al. 1992). However, there is little direct evidence on this question. In fact, different neural parameters tend to co-vary. For instance, at stimulus onset, a cell that is more strongly excited generally both gives more action potentials and gives its first action

282 E. Liebenthal et al.: Spike train parameters for directional behavior

potential earlier than a less strongly excited cell. Relative time is, in fact, a key parameter for directional discrimination in the auditory system (e.g., Moiseff and Konishi 1981). Also, two such cells may fire in different temporal patterns (Optican and Richmond 1987; Geisler et al. 1991; Miller et al. 1991; Theunissen and Miller 1991). Any of these parameters, as well as others, are potential candidates for information-bearing signals. However, not all of them are necessarily used by the read-out system to execute directional behavior.

We describe here a set of experiments in which we have been able partially to discriminate among different spike train parameters used in producing directional behavior. We have accomplished this by modifying experimentally specific parameters of spike trains in a behaving animal, and observing the resulting effects on the animal 's directional behavior. We have carried out this work on a particularly tractable assembly of neurons, the giant interneurons (GIs) that code for left vs right escape turns in the cockroach Periplaneta americana (Camhi 1980; Camhi and Levy 1989). The major advantages of this system are: 1) Most or all of the neurons comprising the assembly are individually identified; 2) The overall number of GIs participating in the directional discrimination is small perhaps just four bilaterally homologous pairs of cells (Comer and Dowd 1993); 3) The animal displays the directional behavior even when restrained in a man- ner that permits dissection and impalement of individual GIs.

Cockroaches respond to the normal approach of a predator by turning and then running away (Camhi and Tom 1978; Camhi et al. 1978). They detect the predator 's approach by sensing the resulting air displacement, or wind gust. Wind-receptive hairs on a pair of abdominal antenna-like organs, called cerci, are displaced by the wind, thereby activating the underlying sensory cells (Nicklaus 1965; Dagan and Camhi 1979; Westin 1979). These sensory cells excite the GIs in the last abdominal ganglion (Westin et al. 1977; Blagburn et al. 1985; Ha- m o n e t al. 1990).

The GIs are subdivided into a ventral and a dorsal group (vGIs and dGIs). The ventral group contains 4 bilateral pairs of identified cells, GIs 1, 2, 3 and 4. At least the first 3 of these are known to be involved in specifying the direction of the initial escape turn (Comer and Dowd 1993; Camhi and Levy 1989). In contrast, the dGIs are not specifically known to be involved in specifying the direction of the initial turn. The vGI axons proceed through the thoracic ganglia, where they excite thoracic interneurons that, in turn, excite leg motor neurons responsible for the escape behavior (Ritzmann and Pollack 1986, 1988; Ritzmann 1993).

Our studies have focussed on the vGI assembly, and its mechanisms of discriminating left vs right wind stimuli. 1 It is known that when a wind stimulus arrives from

1 The cockroach's directional discrimination is more refined than just left vs right. At a minimum, the 4 quadrants of horizontal space are discriminated from one another. We selected the left/right discrimination for simplicity, as the opposite behaviors that result are readily distinguishable experimentally, even in tethered cockroaches.

the front right, each of the right GIs of largest axonal diameter (GIs 1, 2 and 3) gives more spikes, begins firing earlier, and gives a different temporal pattern of spikes (generally more "bursty") than its contralateral homolog (Camhi and Levy 1989; Camhi, unpublished observations; Comer and Dowd 1993; Kolton and Camhi 1994 and unpublished).

Which of these 3 parameters are actually used by the cockroach to evoke a properly directed turn? Two prior studies have shown that the relative numbers of spikes in the left vs the right group of GIs is well correlated with the direction of the escape turn under a number of experimental conditions (Dowd and Comer 1988; Camhi and Levy 1989). However, a correlation does not prove a causative connection. Rather, to do so requires an inter- ventive experiment.

In the work described here, by injecting intracellular current pulses into a left GI, we were able to overcome the directional effect of a wind puff from the right and thereby evoke a turn to the right-into the wind. Then, by varying the parameters of our electrical stimulation, we could determine which spike train parameters are most important in specifying the turn's direction. Our results indicate that fine temporal patterning is not a necessary cue, and the relative time of onset is not a sufficient cue, for left/right discrimination. Rather, our data suggest that relative numbers of spikes in the left vs the right vGIs is a dominant cue determining whether the animal turns to the left or the right. Experiments in which we stimulated extracellularly small groups of GIs support this finding.

Methods

All experiments were carried out on adult male cockroaches Periplaneta americana. The cockroaches were raised in our labora- tory culture in plastic barrels, and were kept at 22-28 ~ C, on a 12:12 L: D cycle. They were fed on rat chow and water ad lib.

In order to assess the escape behavior of the cockroach while we manipulated experimentally the responses of the vGIs, it was necessary to restrain the animal in a way that permitted normal escape movements of the legs and body. For this, we employed a method developed previously (Camhi and Nolen 1981), in which the animal was fixed by 4~6 tiny "minuten" pins, inserted from above through the posterior lateral margins of the abdomen, into a small amount of wax situated under the abdomen (Fig. 1). This wax adhered to a transparent Lucite disk, on which the cockroach's six legs stood. The disk was oiled to reduce friction, permitting the cockroach to walk normally in place. In cockroaches restrained in this way, the initial escape movements of the legs relative to the body are virtual- ly identical to those of free-ranging animals. Also, the front end of the body turns initially away from the stimulus, though the angle of turn is limited by the restraining pins (Camhi and Levy 1988).

Prior to restraining the animal, we ablated its antennae, to pre- vent antennally mediated wind responses from influencing the evoked escape behavior (Burdohan and Comer 1990). We also ablated the wings. After pinning the animal to the wax, we dissected the abdomen from the dorsal aspect, supporting the A5 6 connec- tives on a wax-covered platform for intracellular impalement. We applied drops of saline as needed to maintain the fluid level above the supported connective (Callec and Sattelle 1973).

Microelectrodes were filled with either Lucifer yellow LY (tips with 3% LY in 0.1 M KCI, shanks with 0.1 M KC1) or 5(6)-car- boxyfluorescein-CF (tips with 6% CF in 0.44 M KCI, shanks with

E. Liebenthal et al. : Spike train parameters for directional behavior 283

intracellular extracellular stim. recording

pulses/sec

camera 250 frames/sec

mirror : j

Fig. 1. Experimental setup. A cockroach is pinned on a slick surface, permitting normal leg movements relative to the body. Electrodes are shown for intracellular electrical stimulation and extracellular recording of a single left vGI. A wind puff is presented from the front right. See text for details

1 M KCI). We then dissected out the last abdominal ganglion (ganglion A6), and for LY preparations, dehydrated in an alcohol series and cleared in methyl salicylate, or for CF preparations, mechani- cally cleaned the ganglion. We then examined the ganglion in whole mount in an epifluorescence microscope. We identified the filled GI on the basis of its soma position and dendritic tree shape (Daley et al. 1981). In some experiments, where the fluorescent image was weak, a SIT camera (Hamamatsu C2400) aided in GI identification. We categorized as putatively identified GIs those cells for which the anatomy, though suggestive, did not provide a definitive identity, but was further corroborated by physiological data (see below for physiological methods)]

After impaling the axon of a suspected vGI, (we usually targeted the left GI1), we delivered trains of depolarizing current pulses to evoke corresponding trains of action potentials in the impaled axon. The frequencies of the evoked spikes ranged, in different preparations, from 350-500/s. This is well below the maximal spike frequency seen in the vGIs at the onset of their responses to wind puffs of nearly saturating intensity, and in fact is close to the mean frequency of the vGI responses to such puffs (Westin et al. 1977). Train durations varied from 15 to 200 ms.

We recorded the intracellularly evoked spikes by means of hook electrodes wrapped around the A3m connective, and insulated with a vaseline/paraffin oil mixture. At the beginning of each experiment, we applied brief, subthreshold and suprathreshold trains of stimulus pulses to the impaled axon, to allow identification of the spike

2 The vGIs are all excited by wind stimuli (Westin et al. 1977). They have little or no ongoing activity, and are inhibited when the animal walks (Daley and Delcomyn 1980), unlike the dorsal GIs (dGIs), which are excited during walking. Their spikes are the largest in extracellular recordings, since their axons have the largest diameters. GI 3 is excited primarily by wind from the ipsilateral front, GI 2 primarily from the ipsilateral rear, and GI 1 both from ipsilateral front and rear (Kolton and Camhi 1994, and unpublished).

/1/1 Hooks ~

10 ms

Subthreshold Evoked stimulation GI spikes

Fig. 2. Subthreshold and suprathreshold intracellular stimulation of the impaled axon of a left GI 1. Top trace: a monitor that recorded only the time, and not the current strength, of each stimulus pulse. Bottom trace: extracellular recording

in the extracellular record (Fig 2). We repeated such brief trains at the end of an experiment, to verify the axon's continued responsiveness. In each such test trial, and in experimental trials, we determined from the extracellular records whether the impaled axon followed the train of stimulus pulses. We also checked for the pres- ence of additional spikes, from other axons that could have been excited as artifacts by our stimulus pulses, or that could be evoked postsynaptically by the stimulated vGI's orthodromically ascending or antidromically descending spikes. In the great majority of trials we saw no such additional spikes. As the GI axons have diameters of 25-60 mm, and as the much smaller spikes from axons of roughly 8 mm are still clearly visible in hook recordings of this type (Liber- sat et al. 1989), it is clear that no other GI axons, nor any axons of diameters down to a few pm, were active in the abdominal connec- tives as a result of our GI stimulation. In a few trials, however, we did see additional spikes during the evoked spike train. We did not analyze the data from these trials.


In most experiments, we delivered both a wind stimulus and a train of electrical stimulus pulses intracellularly. In most of these same preparations, we also gave wind without elecrical stimulation, as a control. In one set of experiments, we replaced the intracellular electrode with a pair of extracellular stimulating hooks wrapped around the left As~ connective. We recorded the injected spike train and the wind-evoked spikes with two separate pairs of hook electrodes wrapped around the left and right A~m connective. As in the intracellular experiments, we gave either wind stimuli or electrical trains of pulses, or both, with relative timing as described in the Results section.

The wind always came from 15 ~ right of head-on, at different times relative to the train of pulses. We generally used wind stimuli having a peak velocity of 1.3 m/s, which we will call "low wind". In some experiments we used puffs of 2 m/s, which we will call "high wind".

We recorded all neurobiological data on a video recorder (Neu- rocorder DR-886), having a frequency response of DC to 22 kHz, and we analyzed the data off line using the program Computer- scope (R. C. Electronics). We recorded the cockroach's behavioral responses to the electrical and/or wind stimuli with either a Low- cam camera (Redlake Corp, Santa Clara CA), or a high speed video (NAC, Tokyo), each operated at 250 frames/s. The camera's view was through a mirror positioned under the lubricated petri dish, and angled at 45 ~ upward. This permitted visualization of the legs, even when the animal retracted them underneath its body (Nye and Ritzmann 1992). The Locam camera simultaneously filmed an oscil- loscope screen displaying the onset of electrical stimulation, whereas the NAC TV screen included a video display of the recorded onset of the wind, interfaced through a wave inserter. These recordings allowed us to correlate the stimulus time with the behavioral events. The animal could not see these stimulus monitors.

We analyzed the Locam film frame by frame, by projecting onto a digitizing pad (Hipad, Houston Instruments) and entering the locations of the 6 tarsi (feet), head, rear of the body and position of the wind stimulator into an IBM AT computer. A computer program (Camhi and Levy 1988) then calculated the initial movement of each leg relative to the body in the anterior-posterior, and the medial-lateral axes, as well as the change in angle of the body's long axis. (This turn of the body resulted exclusively from a swing of the front end, as the rear end was fixed in place by the pins.) All movements were normalized to body size, permitting comparison of movements by different cockroaches. The NAC video data were similarly analyzed using a dedicated NAC computergraphic program.

We evaluated the cockroach's turn direction by analyzing the movements of the front and middle legs, 3 and the turn of the body. For each leg, we plotted as a vector the direction and magnitude of

3 The rear legs were not analyzed, for the following reasons: 1) The rear leg on the outside of the turn (right rear leg for a left turn) remains on the ground (in the stance phase) for its initial movement in only 48% of the trials in free-ranging animals (Camhi and Levy 1988), as compared to 100% for the other 5 legs. In restrained animals, it was impossible for us to discriminate on which of the trials this leg was on the ground as, owing to the slick surface, a leg moves relative to the ground even during the stance phase. Thus, we could obtain no useful data from this leg. 2) The opposite rear leg (on the inside of the turn) gave a mean movement of 0.005 body lengths in low wind control tests, and 0.02 in high wind control tests, as compared to 0.18 and 0.2 in free and restrained, undissected animals (Camhi and Levy 1988). This reduction of movement may have resulted from our need to insert six, rather than just 4 pins as we had in prior studies, and to pin more anteriorly in the abdomen, closer to the rear legs, in order to stabliize the preparation for intracellular recording. 3) As the rear leg on the inside of the turn does not move substantially in the medio-lateral plane even in unre- strained cockroaches (Camhi and Levy 1988), it presumably gener- ates little of the body torque, which rather is primarily the product of the front and middle legs.

the initial stance phase movement relative to the body (Camhi and Levy 1988). This, together with the direction of body movement, permitted us to develop a scoring system for each trial (Appendix). The range of possible scores for a given trial was between - 1 and + 1. The closer to + 1, the more the trial resembled the normal left-ward turns of the control animals away from our wind stimulus (which was on the right); and the closer to -1 , the more the trial resembled the mirror-symmetric, opposite turn. We also used these scores to carry out statistical comparisons among the sets of trials under different control or experimental conditions.

To simplify the analysis of the behavioral responses, we presented the electrical or wind stimuli only while the animal was station- ary. However, the cockroach's responsiveness to wind is much greater immediately after a bout of running than when standing for long periods (Camhi and Nolen 1981). Therefore, we typically touched the body to initiate running and then, immediately after the running ceased, we delivered the stimulus. Thus, the question arises whether the immediately ensuing behavior always resulted from our test stimulus, rather than occurring as an unrelated resumption of running behavior. We are confident that these were in fact stimulus evoked responses since, in each experiment, the great majority of trials showed a consistent direction relative to the stimuli we presented.

We considered only trials on which a criterion amount of movement occurred. This criterion could be fulfilled in one of two ways, either if: (a) the tarsi of at least 2 of the 4 observed legs (prothoracic and mesothoracic pairs) moved by at least 0.05 body lengths in the medio-lateral axis (and thus could contribute significantly to a turn), and the body angle had also changed by at least 1.5 ~ or (b) the tarsi of at least 3 of the 4 observed legs moved medio-laterally by at least 0.05 body lengths. These criteria were set by examining the smallest movements made in control trials of animals responding to wind puffs only.

Results

Control observations

It is k n o w n that dur ing the initial movemen t of the legs in the escape of a free-ranging cockroach, 5 or all 6 legs remain fixed on the g round (in stance phase) and push the body away from the source of wind s t imula t ion (Camhi and Levy 1988). Tha t is, the legs push toward the st imulus, so that the body moves away. In cockroaches re- s trained by the methods we employed, because of the slick surface, the legs slip against the g round in stance phase, rather than remain ing fixed at a given poin t on the ground. The body also turns slightly, perhaps part ly because of some remain ing friction of the legs against the ground, and perhaps because of con t rac t ion of inter-seg- menta l body muscles.

To provide control da ta for our experiments, it was necessary to determine the direct ional responses to high and low wind puffs, for cockroaches that were dissected and prepared as if for an experiment. For this purpose, we carried out the identical dissection, suppor ted the nerve cord on the platform, inserted a microelectrode into the cord ( though not into an axon) and placed the hook electrodes and vaseline as usual. We did not give any electrical st imulus, bu t ra ther jus t the wind puff, as always from 15 ~ right of head-on. For low wind puffs, in 58% of the trials (14 out of 24) there was a detectable m o v e m e n t response. In 93 % of these, the response satisfied our criteria for analysis (Methods section). All of these 13 respons-

Left Legs ~ ~ Right Legs o~ 02 "~ / , t

-L_p ' f \ / ' 1\ \ ( _ 1 1

o ' % ,

, , , _ . . ~ _ _ ~ . . . . _ . , I I I t

- free ranging cockroaches ; \ . . . . . pinned animals ~ \

�9 p inned d issected a n i m e l s - L W t / . . . . p inned d issected a n i m a l s - H W

ELect. Stim.

150ms slim (~_

Fig. 3. Comparison of leg movements in response to right front wind puffs, in cockroaches under 4 different conditions: Free ranging, or pinned, or pinned and dissected and presented with either low, or high wind puffs (LW vs HW-see Methods section). The mean movements, relative to the body, are plotted as vectors, scaled to fractions of body length. The scale drawn for the left front leg applies to all legs. Data for free-ranging and pinned animals are from Camhi and Levy (1988) where the peak wind velocities ranged from 0.75 to 2.25 m/s. Data for "pinned dissected animals, L W " are from 13 trials in 4 cockroaches, and for "pinned dissected animals, H W " are from 18 trials in 3 cockroaches. Mean body turns are shown in the upper arrow for LW and HW trials

Hooks

Wind

Response of all the

F 'GI to wind s

L Electrically evoked spikes in GI1

Behavior j Starts

J

Fig. 4. Recording of an experiment in which a train of electrical stimulus pulses was delivered to the left GI 1 at a rate of 350/s for 150 ms, together with a low wind puff from the right front. The wind trace indicates the moment the electrical signal was delivered to the wind-generating motor. Shortly thereafter, the response of all the GIs to the wind is seen in the middle trace. The moment that the behavior began was determined by filming the animal

7_o ms


es were directed to the left, away from the wind source (score +0.8 + 0.07; mean + SD). The mean latency to the first detectable movement for the criterion responses was 43 ms, and the range was 16 to 140 ms. Similarly, for high puffs, in 73% of the trials there was detectable movement, 95% of these surpassed criterion, and of these 94% (17 out of 18) were to the left (score +0.68 _ 0.27). The mean latency of the criterion responses was 36 ms, and the range was 20 to 90 ms. Figure 3 emphasizes the simi- larity of the initial leg movements in these dissected cockroaches vs free ranging and pinned, undissected individu- als. All the arrows show a push of the legs toward the wind source, so as to turn the body away.

Does injection of a spike train in left GI 1 reverse the turn direction of escape evoked by a right wind stimulus?

We wished initially to examine the importance of the bilateral difference in the number of action potentials given by the left versus right group of vGIs. How many action potentials would we need to inject in a left vGI in

150ms

>

Left Legs Right Legs

Fig. 5. Mean leg movements in response to low wind puffs from the right front, presented together with a prolonged (150 ms) electrical stimulus train to the left GI1 (solid arrows, 10 trials). These responses are compared against the mean responses to low wind puffs from the right front, mirror reversed as though presented from the left front (dotted arrows). Scale and conventions as in Fig. 3

order to achieve a greater number of left than right vGI spikes, even though the wind stimulus would come from the right? A prior study had suggested that the left-right difference is roughly 13 spikes (Camhi and Levy 1989). Thus, we needed to inject roughly 26 spikes-13 to bring the two sides into balance, plus up to a further 13 to reverse the balance. We wished to inject this spike train between the onset of wind-evoked vGI spikes and the onset of the behavior. However, as the mean latency on the control trials (i.e. wind only) was only 43 ms (low wind) or 36 ms (high wind), on the average it would not have been possible to inject more than 12-20 spikes at 350-500 Hz. Thus, it was necessary to initiate the train of stimulus pulses before the onset of the wind-evoked vGI spikes. We began these pulses 3 5 4 0 ms prior to the wind, and used a train length of 150 ms- long enough that the spike train continued at least until the onset of the behavior on all trials. Our having to begin the stimulus train early resulted in an undesired situation in which the left side gave both the larger number of spikes and the first spikes. Our rationale was to attempt, under these conditions, to reverse the turn direction and if successful, next to a t tempt to determine which parameters of the stimulus train were actually responsible for this reversal.

We targeted left vGI 1 for impalement and electrical stimulation in most experiments, as this cell shows espe- cially clear left/right discrimination of wind stimuli

12

150ms slim

O~

c- O

0 rn

8

4

/ /

/ / . . , . - - j . . - - -

ii ,

~ GH O

-4 t ............................. "'"'"" - ..... putative GI1

- 8 I I

0 50 I00 Time (ms)

Fig. 6. Body turns evoked by low wind puffs from the right front, presented together with electrical stimulation of the left GI 1. Of the 10 turns plotted, all but one (which came from an experiment with a putatively identified GI 1) were to the right. All turns are normalized to zero time representing the moment of onset on each trial. The graph for each trial terminates two frames (8 ms) after the first step had been completed in all four of the front and middle legs, unless the body movement had already ceased or reversed by that time. Each line represents one trial. The average (av) change in body angle for the 10 trials shown was 4.4 ~ to the right

E. Liebenthal et al.: Spike train parameters for directional behavior 287

�9 "r" , ~

i

t,LI

ta . _ .

" ~ 0

<

J

I I I

E , , -

.1(-

i I I i -0.5

),

I I 0

M o v e m e n t

J ,i,,il I

I I I I I I i i i

S c o r e s 0.5

Fig. 7. Movement scores for the behavioral responses of each experimental situation of this study. The scoring method is described in the Appendix. For each experimental situation, the box is centered above the score, from - l to + 1, on the scale along the bot tom of the figure. The double-headed arrows on each box indicated + one standard deviation. Positive scores indicate turns to the left, and negative scores turns to the right. The wind was always delivered from the right, except in the "Elect. Stim." (second row from the bottom), in which only an electrical stimulus train, and no wind, was presented. It can be seen that the control experiments with wind only (lowest of the 4 rows), and the experiments in which only a brief electrical stimulus train was applied (15 or 33 ms-up to 8 pulses), all received positive scores, between +0.54 and +0.8. In contrast, the experiments in which a 150 ms stimulus was applied (so that the left/right difference in the number of spikes was reversed in favor of the left vGIs-Camhi and Levy 1989) all received negative scores, between -0 .46 and -0 .7 . This correlation between the

number of spikes added to the left vGI and the location on the scale was independent of the identity of the stimulated vGI axon (GI 1, 2 or 3) and the wind puff intensity (low or high wind). Statistically, we compared the control low wind data against all experimental low wind data, and also against the "Elect. Stim." data. We also compared the control high wind data against all experimental high wind data. There were no significant differences between the control data and all experimental data appearing on the right side of this figure (P = 0.3, Mann Whitney). In contrast, the controls were highly significantly different from all data appearing on the left side of the figure ( P < 0.01, Mann Whitney). * On a third trial, it was not possible to determine a score owing to an imperfection in the film. �9 * 33 ms refers to the trials in which a brief stimulation was applied to the left GI 1, after the onset of the wind-evoked response in the vGIs. *** Data include 1 trial with a GI 3, 2 with GIs 2, and 1 with a putative GI 2. **** Data include 2 trials with GI 3, and 2 with putative GIs 2

288 E. Liebenthal et al. : Spike train parameters for directional behavior

(Westin et al. 1977; Kol ton and Camhi 1993 and unpublished). An example of an exper iment is shown in Fig. 4. The trial begins with roughly 13 electrically evoked spikes in the left G I 1, followed by a bar rage of wind- evoked spikes in all the vGIs and other cells, dur ing which the train of electrical s t imulus pulses continues. At 122 ms after the first electrically evoked spike, the behavior began, as de te rmined f rom the high speed film. 4

On 66% of the trials (25 out of 38 in 22 animals) using a low wind puff f rom the right front together with left G I 1 s t imulat ion, there was a detectable m o v e m e n t response. Of the 25 responses, 12 satisfied our criteria for analysis (6 definitively identified G I 1, and 6 putat ive G I 1). In 92% of these (11 of the 12), the turn was directed to the right, away f rom the s t imulated axon and into the wind. Figure 5 shows the mean m o v e m e n t vectors of the front and middle legs for 10 of the 12 trials. (Two of the trials could not be scored because the head was obscured on the film. However , they were bo th evaluated visually as clear right turns). As the figure shows, all the legs push to the left, thus turning the body to the right. Changes in b o d y angle in response to the s t imulat ion are shown in Fig. 6 where 9 of the 10 turns are toward the right. The mean angle of turn was 4.4 ~ right. The m o v e m e n t score for this exper iment was - 0.49 + 0.27 (Fig. 7, box labeled "150 ms" in the row labeled " L o w wind + Elect. stim."), as c o m p a r e d to + 0.8 for the control animals responding to low wind only (Fig. 7, far b o t t o m right box). Tha t is, the mean turn was to the right (negative score) ra ther than the left (positive score). The difference between the exper imentals and controls was highly significant (P <0.001, Mann-Whi tney) .

In this set of experiments , the left g roup of v G I s as a whole was made to give a higher n u m b e r of act ion potentials than the right v G I group. But the left g roup also

gave the first spike, since the electrical s t imulat ion of the left G I 1 precedes the wind-evoked spikes. In contrast , the right g roup of vGIs p r e sumab ly gave their no rma l t empora l pa t te rn of spikes, including doublets or o ther sub-groupings , initially very high spike frequencies, and a characteris t ic decrease in spike frequency dur ing the train (Westin et al. 1977; Kol ton and Camhi 1993 and unpublished). The p redominance of right turns, then, suggests that this t empora l pa t te rn ing is not essential for the cockroach ' s de termining whether to turn left or right. However , bo th the relative n u m b e r of left vs right v G I spikes, and /o r the relative t ime of onset of left vs right v G I spikes, remain as candidates for the impor t an t parameters . Moreover , this result verifies that G I 1 is pa r t of the assembly of cells control l ing the direction of the initial escape turn, since the exper imenta l manipu la - t ions in G I 1 affected the turn direction.

Does injection of a brief, early spike train in left GI I, reverse the turn direction evoked by a right wind stimulus?

We next a t t empted to discr iminate between the pa rame- ters of relative number s of spikes vs relative t ime of onset of spike trains in the left vs the right vGIs . To do so, we repeated the exper iment of Fig. 4, but this t ime delivered just 5 to 7 st imulus pulses to the G I (15 ms st imulus train at 350-500/s). The st imulus train began roughly 10 ms pr ior to the onset of the wind-evoked spikes (Fig. 8). Adding just 5 7 spikes to the sum of left v G I spikes is insufficient to reverse the left/right balance (Camhi and Levy 1989). Thus, in this experiment , the right vGIs gave more spikes, whereas the first v G I spike was on the left.

4 In these experiments, it is difficult to determine precisely the number of injected action potentials that are available to the cockroach for computation of its turn direction. First, as it is unknown over what duration a given GI spike can exert an effect, one is unsure of the degree of effectiveness of the GI 1 spikes early in the train. Second, once the wind response began, it was not possible to verify how consistently the impaled cell continued to respond to the stimulus pulses with spikes. (We assume, though, that the spikes continued uninterrupted with each pulse.) Third, it is unknown at what instant prior to the onset of the behavior the motor system makes its final commitment to a given turn direction, and thus which electrically evoked spike is the last that helps determine the initial turn's direction. (As a prior study suggested that roughly 5 ms are required for this decision (Camhi and Nolen 1981) we counted spikes up to 5 ms before the first visible movements.) And finally, the wind stimulus itself would evoke roughly 5 spikes in the impaled left GI 1 (Camhi and Levy 1988). Some of the spikes would occur during the brief intervals between electrically evoked spikes, and thus may pass through to the thorax, though these may leave the site of the electrical stimulus refractory to the next stimulus pulse. Others would be blocked by antidromically descending electrically evoked spikes. We estimate that, given the 36 stimulus pulses injected prior to the onset of the behavior in Fig. 4, the total number of spikes that our stimulus train added, from the onset of the train till 5 ms before the behavior, above those that would be evoked in the left GI 1 by the right wind stimulus alone, was roughly 32. In some of the trials, the number of added spikes was lower (but never lower than 20), due to a shorter behavioral latency and/or the failure of the axon to follow some of the stimulus pulses.

Elect. Stim.

Hooks il 1 Wind - - IBehav i~

Starts 20 ms

Fig. 8. Recording of an experiment in which a train of just 7 electrical stimulus pulses was delivered to the left GI 1, together with a low wind puff from the right front. Pulse frequency was 430/s. For- mat same as in Fig. 4


15 ms stirn


Fig. 9. Mean leg movements in response to low wind puffs from the right front, presented together with a brief (15 ms) electrical stimulus train to the left GI 1 (solid arrows, 7 trials) from two cockroaches. These responses are compared against the mean responses to low wind puffs from the right front (dotted arrows). Scale and conventions as in Fig. 3. The average change in body angle (av) for the 7 trials was 4.3 ~ to the left

is three times greater than the largest left/right time difference recorded in response to winds from the front right (Camhi and Levy 1989). Thus, if left/right time differences were a significant cue in this system, these 10 ms should have been adequate to specify a turn to the right.

Second, the stimulus in Fig. 4 contains sufficient spikes so that the left vGI sum exceeded the right sum, whereas with the stimulus of Fig. 8 the greater sum was to the right (Camhi and Levy 1989). This hints at the left/right number of spikes as the crucial parameter. It remains possible, however, that reversing the relative number of spikes, together with the relative time, in Fig. 4 were both necessary for reversing the turn direction, neither of these parameters alone being adequate to produce this turn reversal.

To investigate this question would require injecting into the left GI 1 a spike train that starts after the first right GI spikes, and that delivers, before the onset of the behavior, sufficient spikes to reverse the left/right balance. We attempted such an experiment, using intracellular stimulation at the highest possible frequency that the GI 1 axon would follow, roughly 500/s. However, since the behavioral latency was brief, we were unable to inject more than 7 or 8 spikes. Of the 5 behavioral responses of criterion level (all definitively identified GI1), all were to the left, with a mean score of +0.57 + 0.26 (Fig. 7-row labeled "Low Wind + Elect. Stim.", box labeled "33 ms"). This reinforces the results of the experiment of Figs. 8-9, where a similar number of spikes was injected, though primarily prior to the onset of the wind-evoked GI spikes, and produced similar left turns. Thus, it appears that no matter when, relative to the wind-evoked spikes, such a short spike train is injected, it fails to reverse the turn direction.

The effect of high-wind vs low-wind puffs

In 64% of the trials (9 out of 14 in 3 animals) there was a response, and 7 of these (3 definitively identified GI1, and 4 putative GI1) satisfied the criterion for analysis. All 7 of these responses were turns to the left, away from the wind, as determined by their positive movement scores. The mean vectors were all very close to those of control trials with right, front wind stimulation alone, as was the mean change in body angle, 4.3 ~ left (Fig. 9). The mean movement score for this experiment (Fig. 7-row labeled "Low Wind + Elect. Stim.", far right box) was +0.77 _+ 0.15. This was not significantly different from the control data with low wind stimulation only (P = 0.78, Mann-Whitney), but was highly significantly different from the experiment in which we delivered a long stimulus train, as graphed in Fig. 5 (P <0.01, Mann-Whitney).

What is the critical difference between the long injected spike trains, such as that in Fig. 4, which caused a reversal of turn direction, vs that of Fig. 8, which did not ? First, though in both experiments the left GI 1 fires the first spike, the time difference between the first left and the first right spike is much greater in Fig. 4 than in Fig. 8. However, the left/right difference in Fig. 8, 10 ms,

It is known that for a given stimulus direction, the greater the wind stimulus (in terms of wind speed and accelera- tion) the greater the number of spikes evoked, up to some saturation value of wind, which varies among the different vGIs (Westin et al. 1977; Camhi and Levy 1989). However, for a given stimulus direction, the difference between the number of left vs right GI spikes is highly conserved, both at the level of particular homologous pairs of vGIs, and thus also at the level of the two bilateral groups. For instance, for left vs right GIs 1, the difference in numbers of spikes in response to wind from a given direction is almost identical for puffs of 0.1 m/s as compared to 2 m/s - differing by less than 5%. Nearly as small a variation is seen for GIs 2 and 3 (Camhi and Levy 1989). This constancy in the face of a 20-fold change of stimulus strength provides a further hint that the left/ right difference in numbers of spikes is an important parameter for directional specification. In contrast, left/ right differences in time of first spike vary by a factor of 850% for GI 1 (0.2 vs 1.7 ms), and by a factor of roughly 350% for GIs 2 and 3 (Camhi and Levy 1989).

Given this, one would expect that, if one were to add the same number of intracellularly evoked spikes in left


GI 1 as we did in Fig. 4, but now give a stronger wind stimulus, this should produce a very similar reversal of turn direction. Indeed, with our high wind puff, in each of 3 trials with positively identified GIs 1 for which the behavioral response surpassed criterion level, the turns were to the right. Moreover, the score, - 0.46 _ 0.27, was nearly identical to that for low wind (Fig. 7-compare on the left the boxes "150 ms" in the two top rows). In addition, we tested high wind puffs together with 15 ms stimulus trains (5-7 spikes) injected into 3 positively identified GIs 1, in a total of 5 trials. We obtained a score of 0.71 _+ 0.17, nearly identical to those with low winds (Fig. 7- compare on the far right the boxes "15 ms" in the top two rows). These results with high winds, then, reinforce the results of the experiments with low wind, and further support the contention that the left/right difference in numbers of spikes is a key parameter for left/right behavioral discrimination.

Other vGIs

In a small number of experiments, we identified the impaled cell as left GI 2 or 3, rather than GI 1. We tested these with 150 and with 15 ms stimulus trains, using high wind puffs. The mean movement scores were -0 .43 __ 0.5 and +0.54 _+ 0.29 respectively, very similar to those of GI 1 with these long and short stimulus trains, respectively (Fig. 7-top row, boxes labeled "GIs 2 + 3"). This suggests that, like GI 1, GIs 2 and 3 are also involved in the control of the escape behavior, and specifically in determining its direction. More pointedly, it suggests that, for a left/right discrimination, it may matter little which of the vGIs of a given side carries most of the action potentials; rather, the total number of spikes on a given side appears to be the important parameter.

Does injecting a spike train in left GI 1, without wind stimulation, evoke escape to the right?

The above experiments would suggest that stimulating electrically just one vGI, without wind, may produce an escape turn in the direction away from the stimulating electrode, since there would be a clear left/right difference in number of spikes. We anticipated, however, that just one cell of the assembly would produce escape behavior- operationally defined as behaviors that satisfied our criteria for analysis-on relatively few trials.

We injected into the left GI 1 of 13 cockroaches, trains of from 50 to 200 electrical pulses. In only 36% of the trials (12 out of 33) was there any detectable movement, and in 9 of these, only one or two legs made movements that were generally small and directionally disorganized. On only 3 of the 12 trials did the responses satisfy our criteria for analysis, each of these 3 from a different animal. Thus, as expected, escape behavior was evoked in- frequently-on 9% of the trials. (Two of the cells impaled in these experiments were definitively recognized as GI 1, and the third was a putative GI 1.)

In all 3 of these responses, the turn was directed to the right, away from the electrically stimulated left axon. The


Fig. 10. Leg movements in response to electrical stimulation of left GI 1 in a single trial (solid arrows), compared to the mean control leg movements in response to a low wind puff from the left (dotted arrows). We computed these mean control leg movements in response to left wind by simply mirror-reversing the mean movements evoked by the control low wind puffs from the right. Scale and conventions as in Fig. 3

leg movements of one of these responses is shown in Fig. 10 (solid arrows), compared to the mean leg movements evoked by control low wind puffs from the left (dotted arrows). It can be seen that all legs moved to the left, and the body turned to the right; thus the overall turn was to the right. The movement scores for two trials 5 were -0 .83 and -0 .58. These scores were significantly different from those of the control experiment (P<0.01, Mann-Whitney). This is the first time that any one neu- ron in the cockroach has been shown capable of evoking a properly organized escape behavior, as opposed to evoking action potentials in motor neurons of the legs (Ritzmann and Camhi 1978; Ritzmann 1981). The latencies from the onset of the injected spikes to the onset of the behavior were 90 to 200 ms. That these latencies were longer than most of those evoked by wind stimuli (means 43 and 36 ms for control low wind and high wind trials,

5 The third trial could not be included in the mean movement score because the quality of the film did not allow a computer-graphical analyis. However, in this trial three of the four front legs made large movements to the left, as in the other two trials, so we are confident it was a right turn.


A

A6 As

I �9 I I

I ms I i Fig. IIA, B. Extracellular stimulation of the left connective. A Demonstration on a pinned-out cockroach preparation, showing that the GIs are stimulated by low voltage pulses, like those used in experiments on the behaving animal. Upper panel: superimposed sweeps in response to stimulus pulses of gradually increased voltages, from subthreshold for all spikes to just sub-threshold for the impaled left GI 2. Just three large jumps were seen on the trace, which may reflect the action potentials of just three large axons (seen most clearly as the three upward-going peaks). Lower panel: repeated gradual increase of voltage, to just supra-threshold for the intracellularly recorded left GI 2 (lower trace). This increase evoked, in addition to the three same peaks seen in the top panel, an earlier spike (lower arrow), associated with the intracellularly recorded GI spike (lower trace, 80 mV amplitude spikes), as well as an additional jump (upper arrow) not associated with the intracellular spike. Slightly increased voltage produced a sixth jump in the trace (not shown). These could well represent the 6 largest GIs, 1-3 and 5 7.

St ~ R R

ELECT. ONLY 1 / / / / / J /

.

WIND + ELECT.

A tO ms I I

Further voltage increases produced only small, later spikes. In the experiments on behaving preparations, though we had no intracellular recording, we adjusted the stimulus so that only large jumps in the trace, and few if any small, late spikes were evoked; thus we presume that primarily GIs were stimulated. B Recordings from an extracellular experiment. The arrow below each panel indicates the moment that the evoked behavior began. Top panel: wind stimulus only. Middle panel: electrical stimulus only. (In all such trials, between 4 and 10 pulses evoked the turning behavior.) Bottom panel: wind plus electrical stimulus. (In all such trials, the electrically evoked spikes began after the wind-evoked spikes, and either 4 or 5 pulses were injected before the behavior began.) The tops of the upward stimulus artifacts were erased. In the middle and bottom panels, every second stimulus pulse appears to have evoked a large initial downward spike. The turn directions evoked on these trials were: wind only-left; electric only-right; wind plus electric right

respect ively) is expected, as on ly one G I was exci ted in the presen t exper iment . The fact tha t s t imula t ing a single v G I can evoke a p r o p e r tu rn conf i rms tha t the normal , w i n d - e v o k e d t e m p o r a l pa t t e rn ing of the spike t rain, ab- sent f rom this s t imulus, is no t essential for specifying a left vs r ight turn.

Extracellular tests

As m e n t i o n e d above , there was insufficient t ime, f rom the onse t of the w i n d - e v o k e d spikes to the onset of the be-

havior , to reverse the left /r ight ba lance of n u m b e r of spikes by means of the in t race l lu la r s t imula t ion of a single vGI . However , by s t imula t ing ex t race l lu la r ly a g r o u p of left vGIs it shou ld be poss ib le to accompl i sh this, or near ly so. If, for instance, the in terva l f rom w i n d - e v o k e d spikes to behav io ra l onset was 25 ms, and if one sub t rac t s 5 ms f rom each end of this interval , in these 15 ms, wi th 500 pulses/s, s t imula t ing GIs 1, 2 and 3, one can inject 21 spikes. If rough ly half of these b lock w i n d - e v o k e d spikes, then the ac tua l n u m b e r of spikes a d d e d to those tha t would have occur red in response to the wind only, is roughly 11. This would a p p r o x i m a t e l y equal ize the left vs



ElecL ~ i,,1

Wind '

Elect. Wind

J

J

T

2 f f ,

L i

-8

Wind +

Wind Elect. Elect. I I ~ I I

i, , ~ p i i

I I I I I I I I R -6 -4 -2 0 2 4 6 8

Body Turn [o] Fig. 12. Behavioral responses to stimuli like those in Fig. 11B. Dot -

ted arrows: leg movements and body turn in response to wind only (the right mesothoracic leg's mean movement was inexplicably to the left); dashed arrows: same for electric pulses only; solid arrows:

same for both stimuli together. The differences in the medial- lateral component of the leg movements (x values) were significant (P < 0.05, T-test) for all 4 legs analyzed except the left mesothoracic (Left 2). Data from 4 animals, with 14 wind trials, 8 electric, and 10 wind plus electric. Body turn mean _+ 1 standard error of the mean

right numbers of vGI spikes. One would then predict that, if the relative number of spikes is an important cue, this electrical stimulus should bias the behavioral response away from the left turns evoked by wind alone. Even fewer pulses might accomplish this biasing, if additional axons such as those of the dGIs, some of which were surely excited by the stimulus, also contribute to the directional code.

Thus, we carried out experiments like those of the type already described, but using extracellular stimulation.

Though we could not be certain exactly which axons were excited, we presumed that the GI axons, having the largest diameters in the cord, would be stimulated by lower voltage pulses than most other axons. In fact, in preliminary tests, while recording intracellularly from a vGI while selecting the voltage level for the extracellular stimulus pulses, the vGI was indeed excited by such a low voltage stimulus (Fig. l lA). Figure l l B shows sample physiological recordings from one animal in the 3 stimulus situations used-wind only, electrical pulses only, or wind plus electrical pulses. In both situations with wind stimulation, as expected, the first large spike was on the right, the side of the wind stimulator.

Wind stimuli evoked turns to the left as expected (Fig. 12, dotted arrows). Trains of stimulus pulses delivered to the left connective evoked turns to the right (Fig. 12, dashed arrows). This corroborates the results of Fig. 10, in which stimulating just a single left GI 1 was able to evoke a right turn, though rarely. This also con- firms that trains of stimulus pulses to one side, not having the temporal structure of wind-evoked spike trains, can evoke an oppositely directed turn. Trains of stimulus pulses plus wind, where the pulses began after the onset of wind-evoked spikes (Fig. 11B, bot tom panel) caused a shift in the angle of the body's turn away from that of the wind-evoked turns, and toward that of the electrically- evoked turns (Fig. 12, solid arrows). This appears to reflect an increased percent of turns to the right (e.g., even though the mean body turn was close to 0 ~ as shown in the bot tom panel of Fig. 12, the standard error was largest for the wind plus electrical trials (solid arrow), reflecting some clear left and some clear right turns, rather than just forward runs. The fact that this stimulus, though clearly not as well defined as intracellular stimulation, was able to shift the turn direction as shown sup- ports the suggestion that the relative numbers of spikes on the two sides, by itself, can determine toward which side the cockroach will turn.

Discussion

The general question addressed in the paper is, what is the nature of the neural code? It has traditionally been thought that most information in neural systems is coded by the number or frequency of action potentials, higher spike rates being associated with stronger sensory stimuli and with central or motor signals commanding stronger muscular movement. More recently, a number of studies have pointed out that neurons can display additional, sometimes complex, coding parameters involving temporal variation (Optican and Richmond 1987; Abeles 1988; Geisler et al. 1991; Miller et al. 1991; Theunissen and Miller 1991). However, showing that neurons represent information by a given coding parameter does not prove that this parameter is used to convey the information onward in the circuit. To determine this, one needs to consider whether changes in this parameter affect subse- quent processing stages, and ultimately the output of the circuit. To date, there is little information on which of several alternative coding parameters function in a given


system to specify an appropriate behavioral output (M6rchen et al. 1987; Rheinlander and M6rchen 1979). This was the goal of the present experiments.

The key design aspect of the experiments described in this paper is the combination of manipulating the neural code of an individually identified cell, and recording on- line the resulting change in the behavior of the whole animal. This had not previously been accomplished in the cockroach escape system. Rather, in some prior studies on this, as well as other systems, a more extreme method of altering a neural assembly's code has been used, namely, the killing of selected neurons by pronase injection or by photoablation, followed by behavioral testing (Comer 1985; Comer and Dowd 1993; Libersat et al. 1989; Atkins et al. 1984). Killing a cell is, of course, an all-or-nothing procedure. Thus, it does not permit the type of analysis carried out here, where we determined the behavioral effect of altering the number and timing of action potentials in one cell relative to others. It is this fine degree of control that permitted us to discriminate among the different coding parameters in the cell assembly.

Prior experiments have implicated GIs 1, 2 and 3 as important members of the assembly that codes the direction of escape behavior in the cockroach. For instance, electrical stimulation of GI 1 with trains of spikes can evoke action potentials in leg motor neurons involved in escape (Ritzmann and Camhi 1978), and these outputs are enhanced if GI 2 or 3 is simultaneously stimulated (Ritzmann 1981). Cell killing experiments showed that eliminating a GI 3 alone, or any pair of the ipsilateral GIs 1, 2 and 3, or all three of these, resulted in a significant elevation of the percent wrong turns in response to ipsilateral wind stimuli (Comer and Dowd 1993). In fact, killing all three GIs and delivering wind from the ipsilateral front evoked 59% turns into the wind (C. Comer, personal communication). Thus, these 3 vGIs alone can account for the majority of the 92% turns into the wind in response to the same wind stimuli, seen after cutting the entire ipsilateral connective (Comer and Dowd 1987). This cut severs all the GI axons and other axons on the side of the wind stimulus. It is possible that GI 4, a particularly active cell that completes the vGI quartet (Westin et al. 1977), could likewise complete the directional code.

The role this suggests for all the vGIs in jointly discriminating between left vs right wind stimuli must be considered in relation to the directionality of the wind responses of these cells. It was first thought that, of the vGIs, only GI 1 responds differentially to wind from the two sides (Westin et al. 1977). More recently, though, each of the four vGIs was shown to respond preferential- ly to ipsilateral wind (Camhi and Levy 1989; Kolton and Camhi 1993 and unpublished). The present experiments confirm the role of GIs 1, 2 and 3 in specifying turn direction, as electrically stimulating any of these cells can strongly influence turn direction (Fig. 7). In fact, GIs 1, 2 and 3 on either side are known to excite an ipsilateral group of postsynaptic neurons in the thoracic ganglia, which themselves are known to affect the leg motor neurons (Ritzmann and Pollack 1986, 1988; Ritzmann 1993). Thus, these left and right sub-sets of thoracic interneu-

rons appear to be involved in integrating the directional output of the left vs the right groups of vGIs.

Perhaps the most striking single result of the present experiments is that injecting a train of spikes into a single left vGI caused the turn evoked by a right wind puff to be of reversed direction. It should be.noted that stimulating the GI alone, without the wind, only rarely evoked a turn. Thus, the effect of the GI stimulation together with wind was not a mere overpowering of one response by another, but rather a computation by the system that took account of both the wind-evoked and the electrically evoked spikes.

A preferred experiment would have been to inject intracellularly into a left vGI, between the onset of wind- evoked spikes and the onset of the behavior, a sufficient number of stimulus pulses that the left vGIs would give more spikes than the right vGIs. As the short behavioral latencies precluded this, we adopted two different strate- gies that gave complementary results. The first was to inject a stimulus train that was long enough to reverse the left/right balance, though this meant starting the train before the wind-evoked spikes. Thus, the left side gave both the first spike and the greater number of spikes. Finding that this stimulus train did indeed reverse the turn direction, we regarded this as baseline information, and next attempted to determine which parameters of this stimulus were responsible for the reversal of turn direction. The result of this analysis was that the normal, wind-evoked temporal pattern of spikes was not necessary, and information on the side giving the first spike was not sufficient, to determine the direction of the turn. This hinted at the importance of the left/right balance in numbers of vGI spikes.

The second strategy was to inject more stimulus pulses only in the desired time interval, but to do so using less controlled extracellular stimulation of several axons. The result was a shift of the turn direction away from the wind-evoked, and toward the electrically evoked turns.

The electrical stimulation experiment of Figs. 4-5 is subject to the criticism that an unnaturally large number of spikes was evoked in the impaled cell. The total number of spikes evoked by both the wind and the electrical pulses was in the range of 25-35, roughly 2-3 times the normal number in response to a wind puff of saturating intensity. In fact, we kept the number of excess spikes close to the minimum needed to reverse the left/right numbers of spikes, rather than injecting an arbitrarily large number of stimulus pulses. The resulting behavior did not appear at all abnormal, but rather was just a normal turn in the opposite direction. Moreover, in the extracellular stimulation experiment, whose results corroborated those of Figs. 4-5, the increase in numbers of spikes per vGI was much less, all remaining within the range of normal wind responses. Thus, we are confident that the change of turn direction reported here is not an artifact caused by working outside the system's natural range.

Our intracellularly injected spike train also resulted in an increased net number of spikes in the system. Howev- er, this net number was still well within the natural range for the low wind experiments, and just at the limit of the


natural range for the high wind experiments (Westin et al. 1977; Kolton and Camhi 1994 and unpublished).

Although temporal patterning and relative time of wind appear unimportant for left/right discrimination, we cannot eliminate the possibility that such parameters might have some minor effect, or that there could be used for other forms of discrimination, such as wind from front right vs rear right. It should be recalled, however, that in its most aroused behavioral state, a cockroach can respond with escape behavior to winds so weak that they probably evoke no more than just two spikes in any one vGI (Camhi and Nolen 1981). With just two spikes, a train in a given cell contains little temporal information, further reducing the likelihood that fine temporal patterning is used.

GI killing experiments likewise argue against the use of information on relative time of onset of left vs right spike trains (Camhi and Levy 1989). This also parallels what is known about the outputs of the vGIs. Separate left and right groups of thoracic interneurons receive ex- citatory inputs from the ipsilateral sets of vGIs (Ritz- mann and Pollack 1988; Ritzmann 1993), and thus could

well serve as separate left and right spike counters. In contrast, there is no indication that the thoracic circuitry is specifically designed to discriminate fine temporal information or relative times of onset of spike trains, though it should be added that this question has not been specifically addressed experimentally.

The fact that neither fine temporal patterning, nor relative time of onset of spike trains, appear to be important cues for left/right discrimination, even though they are clearly encoded in the vGI assembly, indicates that there is strong filtering of code parameters in this system. This emphasizes the importance of examining the output end of a system- preferably the behavior itself-using a "top-down" analysis, to determine the important parameters coded in the nervous system.

Acknowledgements. We thank Prof. Uzi Motro for advice on devel- oping the behavioral scoring method, Drs. Lihu Kolton and Roni Rado for assisting in the extracellular stimulation experiments, and Dr. Lihu Kolton plus two anonymous reviewers for advice on the manuscript. This work was supported by grant J91-03 from the Whitehall Foundation.

0"31 A _ _

e - <

0.1 AZ ,,..A

_J >,

o n~

0

-0.1

0 )

~o -0.3 n

-0.5 -0.3 -0.1 0 0.1 0.3 4:

Left Right Body Length

Fig. I3A, B. Evaluation of movement responses, relative to the body, of the left middle leg. A Control data, low wind. The 13 points represent the end positions, relative to the starting position (center point of graph), of the leg's initial stance phase movement in 13 trials. The arrowhead indicates the mean movement for these trials. Based on the position of the arrowhead, we established a scoring system as follows: Points that fell within a sector of + 30 ~ from the arrow received the highest positive scores of 5 (if the point was closer to the graph's center than the arrowhead) or 6 (if further away). Two additional sectors of up to 30 ~ were demarcated above and below this highest scoring sector. If the Y axes fell within these 30 ~ , the sector ended there. Within these two sectors, slightly lower grades of 4 or 4.5 were awarded. On either side of these, if the Y axis

3 Lett Right

Body Length had not yet been encountered, was a region of lower grades, 3 or 3.5. The scoring boundaries of the left side of the graph are a mirror image of the right, but scores on the left are negative. (A point falling on a boundary was awarded the higher of the two scores.) This scoring system rewarded with the highest grades those points clustered on the right side. This same set of scoring boundaries was used to determine scores for all experimental trials with low wind stimuli. For high wind stimuli, a similar system of boundaries was established, based on the high wind control data. B Experimental data. On the same set of scoring boundaries as in A are represented the 7 points for 7 trials, from two cockroaches, in which low wind was coupled with a 15 ms stimulus train to the left GI 1. See text for further details


Appendix: Scoring system

The method of scoring turn direction is explained in Fig. 13 which shows, by way of example, the movements of the left middle leg. The 13 points on the graph of Fig. 13A represent the movements of this leg from their starting position (normalized for all trials as point (0, 0) on the graph), on 13 responses to low wind stimuli without vGI stimulation (that is, low wind control trials). The arrow represents the mean vector of all these points (wind from the right evoked leg movements to the right, which pushed the body to the left).

The numbers shown on the graph indicate the values we awarded to the data points, depending upon the sectors into which they fell. Basically, the closer a point was to the tip of the arrow the larger its value (maximum value = +6), rendering this scoring method reflective of the degree of clustering of the points. In addition, the sectors on the graph are mirror symmetric, with negative values awarded to points that fell on the opposite side of the graph. Thus, any points that might cluster around a mirror symmetric arrow, and would thus produce an opposite turn, would receive a large negative value. This rendered the scoring method highly reflective of the turn's laterality. For each trial, each of the front and middle legs received such a value. In addition, for each trial, a body turn of greater than 1.5 ~ received a value, which was +6 if it was directed away from the wind, and - 6 if toward the wind. We then added up all values for a given trial. The maximal possible total value of a trial was 30, and the minimal possible was - 3 0 (four legs plus body turn, each with a maximum/minimum of +/-6). To obtain the final score, we simply divided by 30, thus normalizing to a range from - 1 to +1. The mean score for these 13 low wind controls was +0.8 which, being close to + 1, indicates turns clearly directed away from the wind.

A particular advantage of this scoring system was that it reflect- ed not only whether the turns were to the left or right, but also how similar the leg and body movements were in different experimental groups. We compared statistically the scores of the low wind control data against low wind experimental results (that is, low wind from the right front plus electrical stimulation of a left vGI, or electrical stimulation alone). To score these experimental trials, we used the identical scoring system, and the identical boundaries as for the low wind control data. For instance, Figure 13B shows the scoring method, again for the left middle leg, applied to a sample of 7 trials in response to low wind from the right with a 15 ms stimulus train to the left GI 1. The data are superimposed on the same set of boundaries as in Fig. 13A. The points cluster around the arrow, hinting that this electrical stimulus was largely without directional effect. The score for each trial, again based on the values of each leg's movement and the body's movement, was obtained in the same way as described above for the control trials, and was likewise normalized to the range - 1 to +1. We similarly used the boundaries from high wind control data for scoring the high wind experimental trials.

References

Abeles M (1988) Neural codes for higher brain functions. In: Markivitsch H (ed) Information processing by the brain: a physiological-cognitive perspective. Hans Huber Press, Bern, pp 223-232

Atkins GS, Ligman F, Burghardt T, Stout FJ (1984) Changes in phonotaxis by the female cricket Acheta domesticus L. after killing identified acoustic interneurons. J Comp Physiol A 154: 795-804

Blagburn JM, Beadle DJ, Sattelle DB (1984) Synapses between an identified giant interneurone and a filiform hair sensory neurone in the terminal ganglion of first instar cockroaches, Periplaneta americana. J Exp Biol 113:477-484

Britten KH, Shalden MN, Newsome WT, Movshon JA (1992) The analysis of visual motion: A comparison of neuronal and psy- chophysical performance. J Neurosci 12:4745-4765

Burdohan JA, Comer CM (1990) An antennal-derived mechanosen- sory pathway in the cockroach: Descending interneurons as a substrate for evasive behavior. Brain Res 535:347-352

Callec JJ, Satelle DB (1973) A simple technique for monitoring the synaptic action of pharmacological agents. J Exp Biol 59:725- 738

Camhi, J (1980) The escape system of the cockroach. Sci Am 243(6): 158-172

Camhi JM, Levy A (1988) Organization of a complex movement: Fixed and variable components of the cockroach escape behavior. J Comp Physiol A 163:317-328

Camhi JM, Levy A (1989) The code for stimulus direction in a cell assembly in the cockroach. J Comp Physiol A 165:83-97

Camhi JM, Nolen TG (1981) Properties of the escape system of cockroaches during walking. J Comp Physiol 142:339-346

Camhi JM, Tom W (1978) The escape system of the cockroach Periplaneta americana: I. The turning response to wind puffs. J Comp Physiol 128:193-201

Camhi JM, Tom W, Volman S (1978) The escape behavior of the cockroach Periplaneta americana: II. Detection of natural predators by air displacement. J Comp Physiol 128:203 212

Comer CM (1985) Analyzing cockroach escape behavior with lesions of individual giant interneurons. Brain Res 335:342-346

Comer CM, Dowd JP (1987) Escape turning behavior in the cockroach: Changes in directionality induced by unilateral lesions of the abdominal nervous system. J Comp Physiol A 160:571-583

Comer CM, Dowd JP (1993) Multisensory processing for movement: Antennal and cercal mediation of escape turning in the cockroach. In: Beer RD, Ritzmann RE, McKenna T (eds). Bio- logical neural networks in invertebrate neuroethology and robotics. New York, Academic Press, pp 89 112

Dagan D, Camhi JM (1979) Responses to wind recorded from the cercal nerve of the cockroach Periplaneta americana: II. Direc- tional selectivity of the sensory nerves innervating single columns of filiform hairs. J Comp Physiol 133 : 103-110

Daley DL, Vardi N, Appignani B, Camhi JM (1981) Morphology of the giant interneurons and cercal nerve projections of the Amer- ican cockroach. J Comp Neurol 196:41-52

Dowd JP, Comer C (1988) The neural basis of orienting behavior: A computational approach to the escape turn of the cockroach. Biol Cybern 60:37-48

Georgopoulos AP, Taira M, Lukashin A (1993) Cognitive neuro- physiology of the motor cortex. Science 260:42 52

Glimcher PW, Sparks DL (1992) Movement selection in advance of action in the superior colliculus. Nature 355:142-145

Hamon A, Guillet JC, Callec JJ (1990) A gradient of synaptic effica- cy and its presynaptic basis in the cercal system of the cockroach J Comp Physiol A 167:363-376

Hubel DH, Wiesel TN (1968) Receptive fields and functional archi- tecture of monkey striate cortex. J Physiol (Lond) 195:215-243

Kolton L, Camhi JM (1973) Re-evaluating the directional sensitivity of identified giant interneurons of the cockroach. Soc Neu- rosci Abstr 19: 702

Konishi M (1986) Centrally synthesized maps of sensory space. Trends Neurosci 9:163-168

Konishi M (1993) Neuroethology of sound localization in the owl. King Solomon Lectures in Neuroethology. J Comp Physiol A 173:3-7

Libersat F, Levy A, Camhi JM (1989) Multiple feedback loop in the flying cockroach: Excitation of the dorsal and inhibition of the ventral giant interneurons. J Comp Physiol A 165:651-668

Metzner W (1993) The jamming avoidance response in Eigenmanni- ais controlled by two separate motor pathways. J Neurosci 13:1862-1878

Miller JP, Jacobs GA, Theunissen FE (1991) Representation of sensory information in cricket cercal sensory systems I. Re- sponse properties of the primary interneurons. J Neurophysiol 66:1680-1703

Moiseff A, Konishi M (1981) Neural and behavioral sensitivity to binaural time differences in the owl. J Neurosci 1 :4048


M6rchen A, Rheinlaender J, Schwarzkopff J (1978) Latency shift in insect auditory nerve fibers. Naturwissenschaften 65:656-657

Nicklaus R (1965) Die Erregung einzelner Fadenhaare von Periplaneta americana in Abhfingigkeit yon der Gr6sse und Richtung der Auslenkung. Z Vergl Physiol 50:331 362

Nye SW, Ritzmann RE (1992) Motion analysis of leg joints associated with escape turns of the cockroach Periplaneta americana. J Comp Physiol A 171 : 183-194

Optican LM, Richmond BJ (1992) Temporal encoding of two-di- mensional patterns by single units in primary inferior temporal cortex III. Information theoretic analysis. J Neurophysiol 57:162-178

Rheinlaender J, M6rchen A (1979) Time-intensity trading in locust auditory interneurones. Nature 281:672-674

Riquimaroux H, Gaioni SJ, Suga N (1992) Inactivation of the DSCF area of the auditory cortex disrupts frequency discrimination in the mustached bat. J Neurophysiol 68:1613-1623

Ritzmann RE (1981) Motor responses to paired stimulation of giant interneurons in the cockroach Periplaneta americana: II. The ventral giant interneurons. J Comp Physiol 143:71 80

Ritzmann RE (1993) The neural organization of cockroach escape and its role in context-dependent orientation. In: Beer RD, Ritz- mann RE, McKenna T (eds). Biological neural networks in invertebrate neuroethology and robotics. New York, Academic Press, pp 113-137

Ritzmann RE, Camhi JM (1978) Excitation of leg motor neurons by giant interneurons in the cockroach Periplaneta americana. J Comp Physiol 125:305-316

Ritzmann RE, Pollack AJ (1986) Identification of thoracic interneurons that mediate giant interneuron-to-motor pathways in the cockroach. J Comp Physiol A 159:639~54

Ritzmann RE, Pollack AJ (1988) Wind-activated thoracic interneurons of the cockroach: II. Patterns of connection from ventral giant interneurons. J Neurobiol 19:589-564

Salzman CD, Britten KH, Newsome WT (1990) Cortical microstim- ulation influences perceptual judgements of motion direction. Nature 346:174 177

Theunissen EE, Miller JP (1991) Representations of sensory information in the cricket cercal sensory system II. Information theoretic calculation of system accuracy and optimal tuning-curve widths of four primary interneurons. J Neurophysiol 66:1690- 1702

Westin J (1979) Responses to wind recorded from the cercal nerve of the cockroach, Periplaneta americana. I. Response properties of single sensory neurons. J Comp Physiol 133:97-102

Westin J, Langberg JJ, Camhi JM (1977) Responses of the giant interneurons of the cockroach Periplaneta americana to wind puffs of different directions and velocities. J Comp Physiol 121:307-324

critical parameters of the spike trains in a cell assembly ...raghav/pdfs/animalbehavior/... ·...

Documents