of impulses per response, their means and variation; the frequency distributions of impulse numbers; the time distribution of activity during a

Transcription

1 J. Physiol. (1969), 2, With 4 text-ftgure8 Printed in Great Britain A QUANTTATVE ANALYSS OF THE RESPONSES OF CERTAN DORSAL HORN NEURONES TO MECHANCAL STMULATON OF THE LARGE FOOT PAD N CATS BY D. R. G. FULLER* AND J. A. B. GRAY+ From the Department of Physiology, University College London, Gower Street, London, W.C. 1 (Received 5 June 1968) SUMMARY 1. Responses of single units in the dorsal horns of the spinal cords of cats to mechanical stimulation of the large foot pad have been recorded. 2. The observations made with different single stimuli belonging to the set under different conditions of excitability have included: the numbers of impulses per response, their means and variation; the frequency distributions of impulse numbers; the time distribution of activity during a response and in particular the probability of an impulse occurring in each successive time interval. 3. The observations with two stimuli were designed to measure the,portion of the whole response contributed by the second stimulus and o relate the size of this contribution to the interval between the stimuli and their distance apart. 4. The results are discussed and it is concluded that the observed transformations allow representations of successive events to exist at the same time and that the purpose of the transformation may therefore be to allow particular types of interaction between successive events. A model of mechanism which is consistent with many of the observations is also discussed. NTRODUCTON The analysis of complex stimuli, i.e. numbers of events patterned in space and/or time, by a nervous system involves interaction between the representations of events within the stimulus (e.g. Hubel & Weisel, 1959; Whitfield & Evans, 1965). Each event can be defined as having a certain * Present address: Department of Physiology, New York Medical College, New York 129. t Present address: Medical Research Council, 2 Park Crescent, London, W Phy. 2

2 576 D. R. C. FULLER AND J. A. B. GRAY magnitude of a particular quantity at a defined instant and at a defined point in space. The relations between events are analysed by processes involving additions and subtractions between the representations of these events. An external event may be represented in the primary nerve fibre population either by a temporal distribution of impulses or by a spatial distribution of activity amongst different units; the latter is of advantage in situations requiring good temporal resolution between events (Gray, 1962, 1968). Thus interactions between representations are not necessarily the same as interactions between units; the latter may be either interactions within a representation or between representations. There has been little interest in the interactions between representations, as opposed to units, in systems in which the events are represented in the primary population by a spatial distribution of activity. t is likely that the first transformations in such systems, and the mechanisms involved, will be found to have features common to many situations in the animal kingdom. Basically the tasks to be performed must, in all cases, have strong similarities. These tasks must involve interactions within each primary representation of such a kind as will produce new representations which can interact by being present at the same place at the same time. Since such systems have high temporal but poor simultaneous spatial resolution, it is not unlikely that interactions between successive events are important. The studies described here are concerned with investigating such interactions and mechanisms as a general problem concerned with handling spatially represented information. The work is only incidentally concerned with the role of a particular group of cells in a particular part of the nervous system of a particular species. As a system of study we have used a group of cells in the dorsal horn of the cat spinal cord because we have useful approximations to a set of inputs; these are the primary impulse representations of a set of events which are mechanical displacements of varying amplitude applied at points on the pads of cats (Fuller & Gray, 1966; Gray, 1966). Dorsal horn cells of similar characteristics have been noted by many authors since Hunt & Kuno (1959) and the particular group responsive to these particular inputs has been studied previously (Armett, Gray & Palmer, 1961; Armett, Gray, Hunsperger & Lal, 1962) and shown to possess monosynaptic connexions to this particular primary population and to behave in a homogeneous manner. METHODS The methods used were those described by Armett et al. (1962). n summary, cats were decarebrated under temporary halothane anaesthesia and had low spinal sections. Their pads were stimulated with one or two mechanical displacements which were applied from Rochelle salt crystals at accurately defined points and each of which had a rectangular time

3 RESPONSES OF DORSAL HORN NEURONES 577 course with a rate of rise at least twice critical and a maximum displacement of about 2 t. Single units in the dorsal horn whose responses to these stimuli included an early response (Armett et al. 1962) were found and recorded from extracellularly with micropipettes having resistances between 5 and 1 MM. RESULTS The experiments can be divided into two categories, those with single and those with double stimuli. The basic plan of the single-stimulus experiments was a simple three-dimensional (stimulus amplitude, distance along one axis, and a factor related to the excitability of the junctional zone) matrix with five levels in each dimension. Because of the variability of the responses a minimum of thirty trials with each condition was thought to be necessary. Repeat observations were included in the planned experimental sequence. The total number of trials desired was never achieved on a single unit and despite many experiments the number of units giving useful results was only thirteen. t was originally intended to control excitability by stimulation of descending inhibitory tracts in the medulla (Carpenter, Lundberg & Norrsell, 1963). One successful experiment of this kind was done, but since it was very difficult to find units in such preparations, the excitability of the other units studied was altered by altering the frequency of mechanical stimulation. Both methods gave qualitatively similar results; quantitative comparisons are not possible. The experiments with two stimuli were similar in design but the only variables used were stimulus interval and distance between stimuli. Three units were used for these experiments, which can be regarded only as exploratory. Three types of simple analysis of the results were made: (a) total numbers of impulses per response, their means and their variation, (b) the probability distributions of impulse numbers, and (c) the timing of the impulses as shown by the time range occupied by the nth impulse (at all values of n), and the probabilities of an impulse falling into each millisecond interval. mpulse numbers. When the stimulus amplitude was increased, other conditions remaining constant, the mean number of impulses per trial discharged by a single second-order cell increased, the relationship to stimulus strength or to the number of impulses in the volley leaving the pad (Fuller & Gray, 1966) on a linear plot being S-shaped. With the larger stimuli the mean number of impulses per trial reached a maximum (2-9) and in some experiments decreased with the largest stimuli (Fig. la). Changing the other variables also revealed a similar maximum mean number of impulses. These facts suggest that there is some mechanism 37-2

4 578 D. R. G. FULLER AND J. A. B. GRAY which cannot only limit the maximum response of these cells but can also actively reduce it. Similar results using mechanical stimuli have been described for cells in the dorsal column nuclei (Lal, 1966; Wilson, 1967). When the position of the stimulus was changed, other conditions remaining constant, it was possible to define some sort of centre of the receptive field, there being, in general, a decline in activity as the stimulus was removed away from this region (Fig. 1c and d). The mean number of impulses discharged from the cell per trial was not necessarily greatest at this centre of the receptive field. When the cell was strongly excited, the a 6 or. 4 2 //, X,,--,>.d Stimulus strength sec 2 d C Ad-A / b \ "' V r~~~~~~o - P-ts 1 a_r \ \\ a mm mm Fig. 1. The mean numbers of impulses discharged per stimulus to the pad (ordinate) under different conditions of stimulus strength, position and interval between stimuli. (a) Abscissa, stimulus amplitude as a proportion of that of the largest used; positions *-* mm, O --O mm, Q- -O mm; interval 1 sec. (b) Stimulus amplitude -71; positions -* mm, mm; abscissa interval in sec. (c) Stimulus amplitude *-* 1, , O-- -Q -5, -O -36; abscissa, position along a straight line along the length of the large foot pad in mm on either side of the point giving the maximum response: interval 1 sec. (d) Stimulus amplitude -71; abscissa, position as (c); intervals - 3 sec, U--- 1 sec, O-- -O 5 sec, E - -2 sec.

5 RESPONSES OF DORSAL HORN NEURONES.579 graph (Fig. 1 c and d, top curves) relating position to number tended to be flat or even to be lower in the centre; the latter presumably due to the reduction in response at high levels of excitation (see above). When the frequency of mechanical stimulation of the pad was increased (Fig. 1 b and d), or on one occasion when pad stimulation was accompanied by repetitive electrical stimulation of the medullary pyramid and the strength of such stimulation increased, the response was reduced. The fluctuations in the responses were large. The standard deviations increased as the means increased and the coefficients of variation had higher values at lower values of mean. All units had coefficients of variation around 5 % when their mean number of impulses per trial was at a maximum. t was found that most points fell on a linear relationship when the logarithm of the mean number of impulses per trial was plotted against the logarithm of its variance (see below). A group of points did not fit this relationship and these appeared to be associated with the decline in impulse number with maximum excitation. To eliminate points at which a limiting mechanism might be operating, only points at which not only the slope of the relation between mean number and stimulus strength was positive but also the second differential was positive were used. The logarithm of the mean was linearly related to the logarithm of the variance for the fortyfive points from five units meeting these criteria and a correlation coefficient of -93 was obtained. Transforming the regression equation of log. variance on log. mean, the relation of standard deviation to mean was C- = 1.17 mo51, i.e. the variance was proportional to the mean. There were fifty-seven points eliminated in this analysis and of these all but five had standard deviations appreciably below that which would be predicted in this way. n most cases the difference was between a third and a half of the predicted value. The five positive deviations were all small. There appears, therefore, to be a systematic reduction in error associated with the limitation of the number of impulses discharged; there was no systematic association between the amount of reduction in error and the absolute number of impulses. Probability distribution of impulse number. The representation by the second-order population of information about a single event must involve the activity of many cells; most cells in the population can be excited from large parts of the pad. Any estimate of the behaviour of the whole population therefore requires information about the probability distribution of impulse number per trial under different conditions and in particular at different distances from the centre of the receptive field. The general characteristics of those distributions are easily summarized. Since both ends of the distribution are fixed, i.e. at zero and at the maximum number of impulses per trial for that cell, the distributions range from severely

6 58 D. R. G. FULLER AND J. A. B. CRAY skewed with tail to the right through symmetrical to skewed with the tail to the left as the excitation is increased. Timing of response. The times at which the nth impulse (taken at all values of n) occurred in the response to a stimulus were measured and the a 5 - r mm 5 - :r -, FL -1 ijfili * - PN 5 - l l 1 *5 - O 1 s msee Fig. 2a. For legend see opposite page. 1 33mm 267 mm mean times and their standard deviations calculated. Under conditions of sufficient excitation one or even two impulses occurred early (1-12 msec atter stimulus; see Figs. 2 and 3), had a small standard deviation of latency (-9-1P9 msec) compared to later impulses ( msec) and

7 RESPONSES OF DORSAL HORN NEURONES 581 were followed by a gap which might be longer than subsequent intervals in the discharge. These impulses had the characteristics of the early discharge in this system (Armett et al. 1961; Armett et al. 1962). A clearer way of displaying the temporal characteristics of the response was to plot the probability of an impulse occurring in each of a succession of time intervals after the stimulus. One-millisecond intervals have been used. nformation of this kind is required to see how well the stimulus is b r - 48 PN -5 - _ 4.5 ] o, n,n msec Fig. 2. Plots of probability of an impulse occurring in a one-millisecond interval (ordinate), against time after the beginning of the mechanical displacement in msec (abscissa). (a) Each histogram is from results obtained at a different position along the heel-to-toe axis of the pad. The distance on either side of the position of maximum response is given to the right of the figure in mm. All were obtained with stimulus -71 of maximum used and stimulus interval 1 sec: from the same unit as Fig. 1. (b) Each histogram is for a different stimulus amplitude; approximate number of prirnary units excited indicated at right (method of Armett et al. 1962). Near position of maximum response; interval 3 sec; different experiment. This and Fig. 3 blocked by integral parts of abscissa and marked by that integral part; i.e. block marked 1 includes represented at various times after the arrival of the primary volley with a view to investigating interactions between events separated in time. t was also needed since these probabilities can be predicted from the model (see Discussion). Two sets of histograms of this kind from different units are shown in Fig. 2. The most striking feature of these, particularly Fig.

8 582 D. R. G. FULLER AND J. A. B. GRAY 2a, is that there is an early peak followed by a longer period of activity. This is to be expected from what has previously been found of the mass responses (Armett et al. 1962). The figure is characteristic of the units studied in that two phases have always been found, but the balance between them has varied between the extremes of a predominant early phase (Fig. 3) and a late discharge in front of which occasional impulses characteristic of the early phase were found. The late part of the discharge declined rather abruptly at about 2-25 msec in most experiments. The *5 - O msec h J87 _ 3 2 msech 65 3 msec - l l 5 - msec Fig. 3. mpulse probabilities in msec intervals (ordinate) against time in msec (abscissa) after the onset of the first of two mechanical displacements 6-67 mm apart (in the top two, the responses to conditioning and test are shown alone). Probabilities recorded over the periods shown for different times between stimuli as indicated at right of figure; the expected contribution to the total made by the conditioning stimulus at each interval may be seen from the top histogram. Contribution of test as proportion of test alone indicated by number beside each histogram,

9 RESPONSES OF DORSAL HORN NEURONES 583 unit illustrated in Fig. 2a discharged -rather longer, though there is a decline at about 2 msec. Another regular feature of such probability histograms was the presence of oscillations with a period of about 2 msec (in particular see Fig. 2b). t seemed possible that different parts of the discharge might behave differently in respect to position and size of stimulus. Families of curves similar to those of Fig. 1 were therefore constructed for each of a number of 3 msec time periods. Such results did not show enough consistency for any conclusions to be drawn. Double stimuli Results were obtained from only three units in experiments using pairs of stimuli. The results were displayed as probability against time histograms (Fig. 3) and as graphs of the contribution due to the test response presented as a proportion of the response to the test stimulus alone (Fig. 4). These values were obtained by subtracting the mean number ofimpulses expected from the conditioning stimulus in the appropriate period of time from the mean number with both stimuli. Such calculations were made for the whole response and for different parts of the response separately. Figure 4 refers only to the whole response. The basic pattern observed is perhaps best seen by looking first at the graph for a separation of 5*33 mm (Fig. 4). At short intervals there is little or no response to the test, but at an interval of 5 msec there is a peak which is greater than unity. This is followed by a trough which is at its minimum value at 4 msec. All three units showed almost complete absence of a test response at and 1 msec. This is partly due to the properties of the primary population (Armett & Hunsperger, 1961) and partly to previous excitation of the same cell (Gray & Lal, 1965). These can be avoided by correcting for changes in the input and working with subthreshold conditioning stimuli; the over-all effect at times less than 1 msec is then found to be facilitation (Armett et al. 1962). The peaks for the three units were at 5 msec (Fig. 4), 5-1 msec (Fig. 3) and 1-2 msec; the last was associated with a large and prolonged late phase. The maxima expressed as the proportion of the conditioned response to the test response alone were 1-3 (Fig. 4), 1 1 (Fig. 3) and 1-2, and they were found respectively at stimulus distances of 5-3 mm, 6-7 mm and 4 mm. A detailed analysis at all distances was not made but the results are consistent with the idea that the optimum separation was 5-6 mm. The troughs occurred at 4 msec (Fig. 4), 3-4 msec (Fig. 3) and 3 msec and these reached values of 3, 7 and 9 respectively at the same distances as the peak optima. All units showed reductions at all times at distances less than those just mentioned (Fig. 4, 4 mm); another

10 584 D. R. G. FULLER AND J. A. B. GRAY unit (same as Fig. 3) was very similar at 4 mm but points lay between *6 and -8. The third unit was tested with both stimuli through the same crystal. The test stimulus gave rise to very much reduced (< 2 %) responses at all times; this is consistent with results with mass responses of the population (Gray & Lal, 1965). The longest distance, 8- mm, was 4 mm., mm 4 41 *5 667 mm J.5 8 mm msec Fig. 4. The contribution (ordinate) made to the total response by that to the second (test) stimulus given as a proportion of the response to the test stimulus alone plotted against the time between the two stimuli in msec (abscissa). Each plot is for a different distance between the stimuli as indicated at the right of the figure. Different unit from Fig. 3.

11 RESPONSES OF DORSAL HORN NEURONES 585 tested only on one unit (Fig. 4); such distances were nearly as great as the length of the pad and present added difficulties. n this unit it appears that at the longer distances there was little interaction except at very short time intervals. DSCUSSON The cell population studied is of homogeneous behaviour, is excited by a particular set of inputs, has monosynaptic as well as poiysynpatic connexions to these inputs and lies in a particular position in the dorsal horn (Armett et al. 1961; Armett et al. 1962; Gray & Lal, 1965). t is probable that most of the axons of these cells end in the lower part of the spinal cord, because a study by R. M. Eccles (personal communication) on cells of similar characteristics in the sural distribution showed that only about 25 % of the axons could be detected at the level of the lowest thoracic segments by the technique of antidromic excitation. Some cells used in our experiments may have been tract cells, but if they were, they had properties similar to the remainder and thus have not introduced any serious distortion of the picture. As suggested in the introduction the role of the first junctional region may be largely determined by the characteristics of the input and may be commonly found where such inputs occur. The output of a single junctional zone could be used for several purposes, as are primary signals. For these reasons it is the potential information in the output in which we are interested here rather than that which is extracted by any particular further stage. t may be noted, however, that the precision with which cats move over awkward surfaces and edges suggests the rapid transmission of sufficient information from the pads and its rapid utilization in spinal systems, but the system described in this paper may not be involved. The primary transformation converts a very brief volley into something an order of magnitude longer. The later parts of such a response can coincide with subsequent activity, representing a new event in the primary population. The interactions, both excitatory and inhibitory, which are thus possible are considered below. The precision and mechanism of the transformation are considered first. The variability of the responses to a constant stimulus of individual cells is great (see also Armett et al and Gray & Lal, 1965). This could be due to instabilities in the system giving rise to noise. Alternatively, the apparent random changes could be due to fluctuations in the excitabilities of the cells in the network resulting from inputs from other parts of the nervous system and these could be of functional significance rather than noise. The precision of the whole will depend on the number of units and this is unknown. Nonetheless, it is possible to give an idea of the order of

12 586 D. R. G. FULLER AND J. A. B. GRAY magnitude involved and a calculation of what might be expected for a population having about the same number of units as the primary population has been carried out. This has been done by making reconstructions of population activity from the results obtained from a single unit on which an especially large number of observations were made and which had a balance between early and late activity which was approximately the mean of the sample of units observed. Calculations have been made for sixty-one units in a regular array. This number is approximately the number of units in the primary population and is the figure which has been used in calculations on that population (Fuller & Gray, 1966). The calculations made were of the same kind as those described previously (Gray, 1966) and assume all the error to be uncorrelated noise. The final figures which will be quoted are confidence limits at a probability of -68. That is to say, given an impulse pattern, these are the expected characteristics of the stimuli corresponding to the impulse patterns at the stated limits. Two reconstructions were made. n one, only the early response was considered, and in the other, the total discharge was used. A comparison between these two, based as they are on the same data and the same assumptions, does give some idea of the possible importance of the late response in achieving accuracy. Confidence limits vary in a complex manner with stimulus strength and only the smallest values are quoted; the conditions at which these occurred were not all the same. The smallest limits found using the early response alone, ± 9 % for strength and + 5 mm for position, are almost the same as those found using the whole response, ± 8 % and + 6 mm. Clearly such a system need not be imprecise. The results described in the paper have been used to test a model which is an attempt to explain the mechanism of the transformation. The basis of the model, which includes recurrent excitation, was described by Armett et al. (1962). Predictions made from the developed model, which did not include inhibition, were consistent with single-stimulus experimental values over the range of conditions in which inhibition could be ignored. There is therefore a case for the further consideration of the hypothesis of Armett et al. as a way of explaining some of the mechanisms involved. The results were also used to consider what factors within the primary representation are in fact utilized by the second-order cells (Gray, 1968). The excitability changes which occur after a conditioning stimulus are the resultant of several factors. There are early changes mentioned in the results section and in earlier papers (Armett & Hunsperger, 1961; Gray & Lal, 1965). There is summation with the excitability underlying the late response and there is presynaptic inhibition (Eccles, Kostyuk & Schmidt, 1962; Schmidt, Senges & Zimmermann, 1967; see also Gray & Lal, 1965). The introduction of the facilitation due to the late response has the effect of delaying the net reduction of excitability caused by the inhibition. The input has a high temporal resolution but a poor spatial one and hence interactions between successive events, e.g. movement across the pad, may well be important. The fact that particular intervals at particular distances

13 RESPONSES OF DORSAL HORN NEURONES 587 give greater responses than others suggests that there may be some organization of this junctional zone which produces an output which is in some way related to a temporal pattern of movement. n conclusion, a mechanism which may include something functionallv equivalent to recurrent excitation transforms the primary representation into one an order of magnitude longer. The new representation, unless there is a great reduction in the number of units, potentially contains enough information for precision. The prolongation of the representation allows the interaction of the representations of successive events. n this particular situation the interactions of facilitatory and inhibitory processes might relate output to movements across the pad, and accurate and rapidly transmitted information about such movements may well be needed at a segmental level in the fine control of certain actions. We wish to express our thanks to the Medical Research Council for a Grant to one of us (J. A. B. G.) for the scientific assistance and equipment which have made it possible to carry out this work. REFERENCES ARMETT, C. J., GRAY, J. A. B., HUNSPERGER, R. W. & LAL, S. (1962). The transmission of information in primary receptor neurones and second-order neurones of a phasic system. J. Phy8iol. 164, ARMETT, C. J., GRAY, J. A. B. & PALMER, J. F. (1961). A group of neurones in the dorsal horn associated with cutaneous mechanoreceptors. J. Physiol. 156, ARMETT, C. J. & HUNSPERGER, R. W. (1961). Excitation of receptors in the pad of the cat by single and double mechanical pulses. J. Physiol. 158, CARPENTER, D., LUNDBERG, A. & NORRSELL, U. (1963). Primary afferent depolarization evolved by the sensorimotor cortex. Acta phy8iol. scand. 59, ECCLES, J. C., KoSTYUK, P. G. & SCHMDT, R. F. (1962). Presynaptic inhibition of the central actions of flexor reflex afferents. J. Physiol. 161, FULLER, D. R. G. & GRAY, J. A. B. (1966). The relation between mechanical displacements applied to a cat's pad and the resultant impulse patterns. J. Physiol. 182, GRAY, J. A. B. (1962). Coding in systems of primary receptor neurons. Symp. Soc. exp. Biol. 16, GRAY, J. A. B. (1966). The representation of information about rapid changes in a population of receptor units signalling mechanical events. Ciba Fdn Symp. Touch, Heat and Pain, pp London: Churchill. GRAY, J. A. B. (1968). Responses of certain dorsal horn cells to mechanical stimulation of the cat's pad. n The Skin Senses, pp Springfield: Thomas. GRAY, J. A. B. & LAL, S. (1965). Effects of mechanical and thermal stimulation of cats' pads on the excitability of dorsal horn neurones. J. Physiol. 179, HUBEL, D. H. & WESEL, T. N. (1959). Receptor fields of single neurones in the cat's striate cortex. J. Physiol. 148, HUNT, C. & KUNO, M. (1959). Background discharge and evoked responses of spinal interneurones. J. Physiol. 147, LAL, S. (1966). The functional properties of three systems of second-order neurones and their relationship to the coding of information. Ph.D. thesis, Univ. of London. SCHMDT, R. F., SENaES, J. & ZMMERMANN, M. (1967). Presynaptic depolarization of cutaneous mechanoreceptor afferents after mechanical skin stimulation. Expl Brain Res. 3, WHTFELD,. C. & EVANS, E. F. (1965). Response of auditory cortical neurones to stimuli of changing frequency. J. Neurophysiol. 28, WLSON, P. (1967). Functional relationship between the cuneate nucleus and the somatosensory cortex in the rat. Ph.D. thesis, Univ. of Newcastle upon Tyne, pp