change capable of exciting nerve terminals; (c) the initiation of impulses at the

Transcription

1 248 J. Physiol. (I 950) III, I I2.8I3 ACTION POTENTIALS FROM A SENSORY NERVE ENDING By BERNHARD KATZ From the Biophysics Research Unit, University College, London (Received 10 November 1949) The initiation of sensory impulses presents a series of problems: (a) the transmission of the stimulus from its external source to the receptor cell; (b) the mechanism whereby the stimulus is transformed into an electrical membrane change capable of exciting nerve terminals; (c) the initiation of impulses at the nerve ending and their propagation into the afferent axon trunk. An attempt has been made here to throw further light on these problems by studying the electrical changes in a sensory nerve fibre at a point close to its peripheral terminals. It is clear that only the last aspect of the problem (c) is directly amenable to electrical investigation, but such a study may also provide important clues to the preceding links of the process (a and b). A suitable preparation was obtained from the muscle spindle of the frog. The electric response in the sensory axon close to the spindle differs in certain respects from a simple nerve spike, and some of these features can be related to the local events which take place in the sense organ. In the present paper the action potential in the terminal portions of the fibre will be described. It will also be shown that the initiation of an impulse at a sensory nerve ending does not invariably lead to a propagated message in the main axon. In the following paper, stretching of the muscle will be shown to depolarize the sensory nerve endings and thereby give rise to a repetitive discharge of impulses. METHOD Preparation. The experiments in this and the following paper were made on stretch receptors of the M. extensor longus dig. IV of the frog, the results being obtained from seventy-nine preparations at temperatures of 15-24o C. The muscle is supplied by a branch of N. peroneus. Before entering the muscle the nerve divides into a few twigs of which the proximal one often contains a motor and two sensory axons. Occasionally, one or two sensory fibres leave the N. peron. lat. several millimetres above the point of departure of the main branch to the muscle and, surrounded by a separate connective tissue sheath, run to the proximal part of the muscle. After the muscle with its nerve had been isolated, one of the proximal axons was selected, and all other nerve fibres were cut. The remaining axon could often be seen to end within a millimetre of its point of entry into the muscle, and many of these preparations were mounted without further dissection. Some muscles were divided until the intramuscular part of the sensory fibre had been cleared and the

2 ACTION POTENTIALS FROM NERVE ENDING 249 spindle with a portion of the intrafusal muscle bundle freed from adjacent tissue (cf. Fig. 1). In such cases the recording electrode could be placed a little closer to the nerve endings, but the results were substantially the same with either type of preparation. The intramuscular course of the nerve and the position of the sensory endings was checked after many experiments by staining the preparation with osmic acid. The histological pattern of the sensory terminations varies a good deal. In some cases the axon is connected to a single end organ and 4~~~~~~~~~~~ -~~~~~~~~~ 0 A- 4 ~ ~ ~ ~ ~ ~ ~ ~~~~L~ ~~~~1O0t Fig. 1. Photomicrographs of frog muscle spindles. A, isolated living spindle, immersed in Ringer solution, showing capsule, intrafusal muscle fibres and nerve supply. The nerve contained a sensory and a motor axon which was cut. B, stained preparation (osmic acid). Muscle with two sensory axons one of which was cut. The muscle was flattened between slides before fixation. does not divide into branches until it has reached, or entered, the spindle capsule. In other cases, the sensory nerve fibre enters the muscle and then divides into two branches which supply separate spindles (cf. Young, 1950). These end organs are located either on the same or on different intrafusal muscle bundles. Finally, the axon may divide before entering the muscle. The results described in this paper were obtained in preparations in which the fibre divided distal to, and not between, the recording leads. This is important because complications arise if a bifurcation of the axon occurs between the recording electrodes; in this case an afferent impulse starting from one of the spindles passes the distal lead and returns to it by 'axon reflex', and this alters the shape of the spike potential. 17-2

3 250 BERNHARD KATZ Apparatus. The recording system has been described elsewhere (Katz, 1949). The recording electrodes consisted of chlorided silver wires connected to th'e preparation through capillary tubes filled with Agar-Ringer. The amplifier was either direct coupled or had coupling condensers causing its response to decline to one-half in about 0-25 sec. In several experiments a condenser and grid leak were placed into the input lead to check whether the grid current of the first valve (about A.) affected the nerve fibre. For antidromic stimulation, platinum electrodes were applied to the sciatic nerve, and brief thyratron shocks were used, isolated from earth by means of an aircored transformer. The tendons of the muscle were clamped by ruling pens and the whole preparation lifted into oil. Two recording electrodes were applied, usuall one to nerve and one to muscle (Fig. 2). There are obvious To amplifier advantages in placing the electrodes on relatively robust and immobile parts of the preparation, and therefore, when the whole muscle was used, the recording leads were applied preferably to the end of the muscle and to the peroneal nerve, ratherthan to theisolated portion of the axon. The muscle and nerve tissues on either side of the isolated stretch merely acted as low-resistance connectors without N1 contributing any noticeable electric response themselves. Ns S This was verified by short-circuiting the sensory axon (letting it lie along the muscle) which made the afferent response invisible, and by moving the recording electrodes along muscle or nerve trunk, there being no noticeable change in the shape M of the afferent spike until the electrodes were moved on to the isolated axon itself. However, the present method could not be used for recording antidromic impulses in the sensory fibre: in this case the central electrode had to be applied to the isolated axon itself, to avoid recording responses of cut Fig. 2. Electrode arrangement showfibres of the peroneal peroneal nerve. fibres of the nerve. ~ing stimulating (S) and recording The isolated axon, together with its sheath, was usually leads. imt, M. extensor longus f. thick and a few millimetres long. The resistance of dig. IV N, iolatensor fibre this stretch was very high, and in some experiments con- t siderably magnified the time constant of the input circuit. Hence, the spike potential was sometimes appreciably slowed and attenuated. Square pulse calibrations indicated that the time constant of the delay of the recording system varied in different preparations between less than 40 and 250 ltsec. In the former case, the distortion of the action potential was slight, but in the latter case the peak of the spike may have been reduced by as much as 50 %. The size of the spike (usually 2-3 mv.) was, however, of little interest, and the interpretation of the records was not seriously affected by this amplifier lag. To summarize the present technique, electric responses are led off from a sensory axon at two points, a few millimetres apart. The distal lead is close to the sensory terminals of the spindle, but not in direct contact with them. Even when the spindle had been isolated (Fig. 1) the terminal structures remained enclosed inside the spindle capsule, and any localized potential changes could only be recorded at a distance determined by the length of the intra-capsular nerve branches. The electrotonic spread along the fine terminal branches is bound to involve an appreciable decrement, and the residual signal recorded on the outside of the capsule must be a reduced and distorted image of the local events at the sensory endings. It is important to keep this reservation in mind when interpreting the potential changes recorded with the present method.

4 ACTION POTENTIALS FROM NERVE ENDING 251 RESULTS When the tension of the resting muscle is low, afferent impulses are discharged occasionally, at apparently irregular intervals. If these impulses are recorded in the neighbourhood of the spindle, two characteristic phenomena are observed which differ from the properties of the ordinary conducted axon spike: IT.,.:..1~ 4 R X ; A100p.j Fig. 3. (Left): Afferent impulse from a muscle spindle. Note: All records, in this and the subsequent paper, read from left to right. Downward deflexion means 'spindle negative' (i.e. a positive potential difference between the proximal and distal recording electrode). The spikes were recorded on a fast repetitive time-base, without synchronization. (Right): Photomicrograph of same muscle with single sensory axon and spindle. The preparation was treated as in Fig. 1 B. (i) instead of a diphasic wave, a triphasic potential change is seen the third phase of which will be shown to be a residual depolarization of the nerve endings; (ii) the afferent impulses are initiated by discrete local action potentials which either remain abortive or lead to a propagated spike after a small but visible delay. The first observation suggests that the terminal portions of the nerve fibre differ in certain respects from the axon trunk; the second phenomenon indicates that there are discrete regions of low safety margin, or 'partial block' in the initial path of the afferent impulse.

5 252 BERNHARD KATZ A localized 'negative after-potential' Figs. 3-5 illustrate the shape of typical afferent spikes recorded with the distal lead close to, i.e. within less than 1 mm. of, the spindle capsule. The triphasic potential change was observed in every fresh preparation, provided the sensory axon was not injured nearby or partly depolarized by preceding stimuli (cf. Katz, 1950). The third phase lasts several milliseconds at 200 C. and may reach 40-50% of the initial spike deflexion. 0 1 msec Fig. 4. Afferent (first deflexion downward) and antidromic (first deflexion upward) impulses in single sensory axons. In experiment A, the second record from below shows a superposition, by chance, of an antidromic and afferent impulse. S, stimulus artifact preceding the antidromic impulse. When the proximal lead was applied to an injured region, the afferent impulse became monophasic and showed a prolonged tail portion which corresponded to the third phase previously observed. It appears that the electric response at the terminal portions of the axon is more prolonged than along the rest of the fibre and thereby gives rise to a third phase. Such an effect may be described as a 'terminal negative after-potential'.

6 ACTION POTENTIALS FROM NERVE ENDING 253 A corollary to this phenomenon was observed when afferent and antidromic spikes were compared (Figs. 4, 5). This was done by stimulating the sciatic nerve with a short shock, without moving the recording electrodes. There are N Antidromic B, I I I I I msec. Afferent \I,-, Fig. 5. A, diagram illustrating the assumed composite nature of the afferent and antidromic spike record. In this figure 'spindle negativity' (first deflexion of afferent spike) is shown upward. M, muscle; N, nerve. B, tracings of afferent and antidromic spikes. Same experiment as in Fig. 4, B and C. obvious differences in the shapes of the two action potentials, the afferent spike being triphasic, with a brief and small second phase, while the antidromic impulse is diphasic with a large and prolonged second phase. When the two action potentials are superimposed, as in Figs. 4 and 5, their final phases are seen to be identical in time course and electrical sign, although the initial spikes

7 254 BERNHARD KATZ have opposite polarity. These observations can be explained in the following way. If there is a prolonged depolarization at the terminal portions of the fibre, this would add its effect to the second phase of the descending antidromic impulse, while it would oppose the second, and give rise to the third phase of the afferent impulse (see diagram in Fig. 5). The mechanism of this localized after-potential is not clear, but it may be pointed out that an effect of this kind would arise if the terminal parts of the nerve fibre were more strongly polarized than the rest of the axon (see Discussion). In the course of a prolonged experiment, the third phase of the afferent action potential tended to diminish and eventually to vanish before the main spike was noticeably affected, a feature which seems to be chara,cteristic of negative after-potentials generally (Lorente de No, 1947). Sensory 'pre-potentials' When the muscle was under little tension, the frequency of the afferent impulses was low and irregular, and in many preparations propagated impulses alternated at random with small monophasic action potentials which did not spread to the proximal recording electrode (Figs. 6-8). When the propagated impulses were examined on a fast time base, it was found that there were discrete 'steps' in the foot and rising phase of the action potentials (Figs. 3-5 and 7). These initial steps were clearly preliminary potential changes ('prepotentials', Arvanitaki, 1939) which, only after a perceptible delay, led to a conducted sensory impulse. The step-like rise was shown by the afferent, but not by the antidromic, spike. If a series of antidromic impulses was produced by repetitive stimulation of the sciatic nerve, the successive action potentials were all of identical shape; on the other hand, the shape of afferent spikes from the spindle was not constant, but exhibited a definite 'play' in the step-like composition of their wave front (Figs. 3-5 and 7). It was clear that successive impulses in the same axon conformed to two or three discrete patterns characterized by a different shape of foot and rising phase. Similarly, in those preparations where abortive impulses were seen, they appeared in two or three recurrent sizes. In the experiment of Fig. 6B, for instance, random responses from an almost slack spindle were recorded during a period of approximately 5 sec. Out of 74 detectable action potentials, 30 were propagated (relative size: 100), while the rest were abortive impulses of two, or possibly three discrete amplitudes (relative sizes: 25, 12 and 8-10; frequencies: 14, 1 and 29, respectively). In another experiment, during a period of 8-5 sec., 112 action potentials were recorded of which 70 were propagated (size: 100), while the others were local responses (sizes: 32, 10 and 5; frequencies: 7, 18 and 17, respectively).

8 ACTION POTENTIALS FROM NERVE ENDING 255 C) 0d OD ID 0 OD- Ca 0 0 OD 0 2-0gO Q O 4' (D co o b0 *S O0

9 256 BERNHARD KATZ It should be noted that the propagated spikes are likely to have been attenuated more severely by the present recording technique than the local action potentials (see Method), and that the relative size of the local responses given by the above figures may be somewhat too high. Another significant fact was found when the intervals between successive afferent potential changes were measured. In the almost slack preparation, these intervals varied over a wide range, but it was clear that after a propagated impulse the preparation became silent for at least 1 sec., that is no spike, fullsize or abortive, could be observed during this time. It was equally clear that after an abortive impulse no such silent period occurred. For example, in the second experiment quoted above, the least observed interval between a propagated spike and the next, local or propagated, action potential was 50 msec. (the maximum being 290 msec.), but the shortest interval between an abortive impulse and the next potential change, local or propagated, was about 1 msec. Fig msec. Tracings of propagated and abortive spikes in a sensory axon. 'Spindle negativity' upward. Same experiment as in Fig. 6B. Abortive impulses were observed when the muscle was almost slack and its afferent discharge infrequent and irregular, but they disappeared on stretching and gave way to a regular series of propagated spikes (Fig. 8). They also disappeared, together with all afferent response, when the spindle was subjected to a bombardment by antidromic impulses at a rate of about 100 per sec. The interval between the antidromic spikes at this frequency was longer than the refractory period of the nerve, but shorter than the repetition period of the afferent discharges in the almost slack muscle. The stoppage of all afferent potentials indicates that the antidromic impulses must have reached the active terminals of the spindle and made them temporarily irresponsive (see also Matthews, 1931). Thus, while the afferent potentials were not all propagated, the antidromic impulses never seemed to fail in reaching the terminal structures. The properties of the 'pre-potentials' here described differ in important respects from the local responses found in nerve during subthreshold stimulation, or at the end-plate region during neuromuscular block. The sensory pre-potentials occur in discrete quantal sizes and are not subject to continuous gradation which is quite different from the end-plate potential or the local cathodic response in nerve. This observation provides an interesting clue,

10 ACTION POTENTIALS FROM NERVE ENDING 257 and it suggests that we are dealing with discrete sensory spikes which have started in one of the terminal branches, but for some reason encounter an obstacle before propagating into the axon trunk. It appears that the spindle contains a number of terminal units which-at least under a condition of low tension-are capable of firing separately and producing miniature spikes. Such 1.PM".r t', w Fig. 8. Effect of moderate stretching on propagated and abortive spikes. In records 2,4,5 and 8-10, the muscle was under little tension, mm. long. Note occasional abortive spikes. In records 1, 3, 6 and 7 the muscle was stretched to a 1-2 mm. greater length. Note burst of propagated discharges, but no sign of abortive spikes. a miniature spike might raise the excitability of other near-by terminals by means of its electrotonic spread, and it apparently can sum with other 'steps' and build up to a propagated impulse (see Fig. 7). Once a full-size spike, whether afferent or antidromic, has arisen, it apparently invades all the terminal branches and so produces a period of complete silence (p. 256). A miniature spike which fails to propagate cannot render the other terminal units refractory, and this probably explains the absence of a silent period in these instances.

11 258 BERNHARD KATZ DISCUSSION The 'negative after-potential' at the nerve endings It has been pointed out that this effect might be caused by a gradient of membrane polarization along the terminal branches, the endings being at a higher resting potential than the trunk of the axon. It is well known (Schaefer, 1934) that the action potential becomes longer and is followed by a 'negative afterpotential' when it travels through an anelectrotonic, i.e. more strongly polarized, region. The characteristic difference between afferent and antidromic spike could, therefore, be explained if the peripheral end of the fibre were more strongly polarized and thus normally in an 'anelectrotonic' state. Unfortunately, the present preparation does not lend itself to a direct test of this matter, for with a spike amplitude of 2-3 mv., a resting potential difference of less than 1 mv. would have to be looked for,.and a steady potential difference of this size is impossible to measure in the presence of cut nerve branches. It might be argued that a steady potential difference and, therefore, the triphasic action potential may arise, not because of some special property of the nerve endings, but perhaps because the region of the fibre near the central recording electrode might have been depolarized by injury currents from the cut peroneal nerve (cf. Fig. 2). This suggestion, however, can be dismissed because the presence of the third phase depended upon the proximity of a spindle to the peripheral lead. The local after-potential was not seen in fibres which did not terminate near-by, although they must have been equally affected by any injury currents from the cut peroneal branches. The propagation of 'new-born' impulses from the spindle It might be suggested that an anelectrotonic gradient along terminal branches would also account for the apparent obstacle which bars the propagation of newborn afferent impulses and for the 'notches' and 'steps' in Figs. 3-5 and 8, which resemble those described by Erlanger & Blair (1934) during partial anodal block. But there is another and perhaps simpler way of explaining these effects. Before the sensory axon terminates it divides into several small branches (cf. Fig. 1 B). It is conceivable that the transmission of an impulse from a fine terminal, or pre-terminal, branch into the main axon occurs with a lower safety margin than in the reverse direction. A sudden increase of diameter, at some points of bifurcation, would cause the afferent action currents to be reduced in density, and the geometrical disparity between a fine and coarse adjoining portion of the axon cylinder may account for different facility of propagation in the two directions, and, in extreme cases, even for a one-way block. This is speculative, but of some general interest because similar events would be likely to occur at the dendrites of central neurones. Whatever the correct explanation, the quantal size and the brief time course of the observed ' step-potentials' distinguish them sharply from any graded local potential such as that described in the following paper. There is little doubt that the two or three

12 ACTION POTENTIALS FROM NERVE ENDING 259 local 'steps' shown above are true spikes which have started independently at terminal points of the spindle, and have yet to overcome some physiological obstacle before a sensory message is originated. The 'random discharge' at low tension An interesting question is posed by the large and apparently random variations in the frequency of the afferent discharges when the tension of the muscle is very low. There is no evidence of a regular rhythm in either propagated or local action potentials (Figs. 6, 8). Apparently, there are considerable fluctuations in the excitatory state of the stretch receptor, but their cause is at present unknown. It is pertinent that the local abortive spikes form part of this 'random discharge' at low tension, and are replaced by a regular pattern of propagated impulses when the tension is raised. It has been argued that abortive impulses arise from fractional activity within the spindle, that is from a separate excitation of individual groups of receptor terminals. If that argument is accepted, it would follow that the irregular disturbances which become so evident at low tension are due to highly localized events which'take place at individual endings within the spindle. It is not inconceivable that molecular agitation in the mechanical receptor substance, or ionic noise in the terminal nerve membrane is ultimately responsible for the observed fluctuations in the local excitatory level. SUMMARY 1. Sensory impulses from frog muscle were recorded at a point close to the spindle. The shape of these impulses differs in two respects from the ordinary axon spike: (a) there is a localized 'negative after-potential' at the nerve endings; (b) afferent spikes are initiated by brief step-like 'pre-potentials' which at times remain abortive. 2. The negative after-potential at the nerve endings follows the afferent as well as the antidromic spike and persists for several milliseconds, at 200 C. 3. At low levels of tension, successive spikes in the same axon show characteristic differences in shape. Two or three discrete patterns of 'pre-potentials' can be distinguished which recur consistently throughout the experiment. In many preparations, 'pre-potentials' are observed which occasionally fail to propagate and then form small monophasic spikes of discrete 'quantal' sizes. Such abortive spikes are wiped out by antidromic bombardment of the spindle at a rate of about 100 per sec., and by moderate stretch which gives rise to a regular series of full-size afferent impulses. 4. The presence of step-like pre-potentials and of abortive spikes indicates that there are local obstacles which delay and sometimes block the propagation of 'new-born' impulses after they have arisen at the receptor terminals. It is suggested that transmission from a fine terminal branch into the axon trunk

13 260 BERNHARD KATZ has a lower safety margin than transmission in the reverse direction, and that a delay or partial block might occur at some points of bifurcation of the sensory nerve fibre. I wish to thank Prof. A. V. Hill for the facilities provided in his laboratory and Mr J. L. Parkinson for his invaluable assistance. I am indebted to Prof. J. Z. Young and the staff of his department, in particular to Mr F. J. Pittock and Mr J. Armstrong, for frequent help. REFERENCES Arvanitaki, A. (1939). Arch. int. Physiol. 49, 209. Erlanger, J. & Blair, E. A. (1934). Amer. J. Physiol. 110, 287. Katz, B. (1949). J. exp. Biol. 26, 201. Katz, B. (1950). J. Physiol. 111, 261. Lorente de N6, R. (1947). A Study of Nerve Physiology, I and 2. In Stud. Rockefeller Inst. med. Re and 132. New York. Matthews, B. H. C. (1931). J. Physiol. 72, 153. Schaefer, H. (1934). Ergebn. Physiol. 36, 151. Young, J. Z. (1950). In preparation.