Differential atrophy of sensory and motor fibers following section of cat peripheral nerves

Brain Research, 178 (1979) 347-361 © Elsevier/North-Holland Biomedical Press

347

D I F F E R E N T I A L A T R O P H Y OF S E N S O R Y A N D M O T O R FIBERS F O L L O W I N G S E C T I O N OF C A T P E R I P H E R A L N E R V E S

J. A. HOFFER*, R. B. STEIN and T. GORDON Department of Physiology, University of Alberta, Edmonton T6G 2H7, Alberta (Canada) (Accepted April 12th, 1979)

Key words: atrophy - - axotomy - - compound action potentials - - motor fibers - sensory fibers

SUMMARY Differential effects of peripheral nerve section on myelinated sensory and motor fiber populations were investigated in 5 hindlimb nerves of cats. U p o n electrical stimulation of each nerve, monophasic compound action potentials were recorded from the L6, L7 and SI dorsal and ventral roots, and the impedance of each root was measured. The decline in the electrical charge computed from potentials 43 to 252 days after nerve section gave a measure of the effect of axotomy on the diameters of sensory and of motor fibers in each nerve. N o significant difference in the rate of atrophy of sensory and motor fibers was observed after about 45 days following nerve section. After about 145 and 245 days, however, dorsal root charge contributions had decreased significantly more than ventral root values. Exponential decay curves were fitted separately to charge data for sensory and for motor fibers. The calculated value for the endpoint of the decay was about 35 ~o of the control value for motor fibers, and not significantly different from zero for sensory fibers. These results suggest that in response to axotomy, myelinated motor fiber diameters decline at first but later stabilize, while myelinated sensory fibers continue to decline and may atrophy completely if regeneration is prevented. Possible roles of electrical activity and o f ' t r o p h i c ' interactions with the periphery in the maintenance of cell properties are discussed.

INTRODUCTION I f a mammalian peripheral nerve is severed, its fibers will initially undergo a * Present address: Laboratory of Neural Control, NINCDS, Building 36, Room 5A29, National Institutes of Health, Bethesda, Md. 20014, U.S.A.

348 retrograde degenerative processS,2L Histological studies21, 43 have demonstrated that all fibers undergo considerable atrophy, which is not reversed unless the fibers regenerate and form new functional connections. Thereafter, diameters increase again, and may approach their original sizes 2,~,9. The survival of fibers in severed peripheral nerves which are unable to form new connections has been less clear ag. Compound action potentials recorded from myelinated fibers in ligated nerves decline following single exponential decay curves with time constants of about 45 days, to about 25 ~ of preoperative control levels 11. Thus, many fibers may continue to conduct electrical activity indefinitely after ligation. Whether some fibers are more affected than others could not be determined from neural compound potentials, since these only monitor the average effects in a peripheral nerve. In recordings from peripheral nerves in cats during controlled locomotion, motor activity recovered substantially more than sensory activity after reinnervationlV, 2a,a8. One contributing factor could be that sensory fibers atrophy more severely following axotomy. However, the effects of axotomy on sensory and motor fibers in mammalian peripheral nerves have not been compared directly, either histologically or electrophysiologically. To test whether sensory fibers atrophy differentially, we have measured the charge generated in dorsal (sensory) and ventral (motor) root compound action potentials. The total charge is related to the average diameter in a population of fibers. In a preliminary study 22, acute recordings in two cats with implanted peripheral nerve electrodes suggested that several months after nerve section the electrical charge in dorsal root compound action potentials was more reduced than the charge in ventral root potentials. This paper substantiates our preliminary observations, gives the time-course of these effects, and discusses possible mechanisms for the differential response of sensory and motor nerve fibers to axotomy. Part of this work has appeared in abstract form 18. METHODS Experiments were performed on 18 adult cats of both sexes weighing 2.5-5.7 kg. Under Nembutal anesthesia and using aseptic methods, the following nerves were sectioned. Left hindlimb: lateral gastrocnemius-soleus (LGS), at the point of entry to the lateral gastrocnemius muscle; tibial, 1 cm proximal to the ankle. Right hindlimb: medial gastrocnemius (MG), at the point of entry to its muscle; common peroneal (CP), just above where it divides into its deep and superficial branches; sural, about 5 cm proximal to the ankle. The proximal stump of each sectioned nerve was ligated with a size 4-0 Mersilene suture and attached to a 1 cm square Silastic sheet sewn onto the muscle surface. The Silastic squares served as anchors to retain the original geometry of the nerve, and to impede reinnervation by fibers that might grow through the ligature. A few nerve preparations, which still showed partial motor reinnervation, were excluded from these results (see also Kuno et al.26). Cats suffered relatively minor motor impairment due to surgery and were soon able to walk in nearly normal fashion.

349 Approximately 45, 145 or 245 days later, the effects of axotomy were investigated in acute experiments under deep Nembutal anesthesia. A wide laminectomy was performed, exposing the lumbosacral cord, and the nerves of each hindlimb were revealed through a long dorsal incision. The cat was mounted in a stereotaxic frame, both hindlimbs were rigidly clamped, and paraffin oil pools were formed using skin flaps. Rectal and oil pool temperatures were thermostatically maintained at 35 4- 2 °C by a heating pad and radiant heat. The ligated nerves and their contralateral controls were mounted for stimulation on individual pairs of silver electrodes, spaced 2-3 mm apart and placed at least 8 mm proximal to the ligature, to ensure that the parent fibers in each nerve were stimulated rather than only the finer sprouts which reached into the neuroma40. Negative rectangular pulses of 0.01 msec duration and supramaximal amplitude for myelinated fibers (8-15 × threshold), were delivered once per second. A recording cuff 3.4 mm in diameter and 26 mm long was installed around each sciatic nerve (see Davis et a1.11). The dorsal and ventral roots from L6, L7 and S1 were prepared bilaterally for recording on platinum hook electrodes.

Impedance Using electrodes spaced every 2.5 mm, the electrical impedance of each root was determined just prior to recording potentials. The impedance at each electrode was measured with respect to the cut end of the root, placed on the proximal electrode, using a 10 kHz signal 36. Consecutive impedance values were plotted and the net tissue impedance was determined from the regression line, as indicated in Fig. 1B and the accompanying legend.

Compound potentials Thereafter, each ipsilateral nerve was stimulated and the following parameters of the compound potentials recorded from that root were measured: the peak amplitude (in mV), the base, the half-width, and the latency to the main peak (in msec). Each root was inspected in turn until all the nerves were mapped. To reduce variability due to progressive fiber deterioration over a recording session lasting several hours, each root was cut just before recording from it. To control for any decay in the condition of the nerves, compound action potentials were recorded simultaneously at the sciatic nerve cuff. Nerves whose potentials declined more than 15 ~ by the end of the experiment were discarded. In a few nerves (less than 10 ~), a small decline was compensated by appropriate scaling of later values.

Determination of electrical charge Comparison of potential amplitudes recorded from different roots is not meaningful unless the root impedances are taken into account. Since the S1 roots are considerably thinner than the L7, and the L6 roots are usually shorter, the electrical impedance of the tissue over which a potential develops can be markedly different (Fig. 1C, second column). Each potential was divided by the impedance to yield units of current (middle diagrams). Since the peak amplitude of a monophasic compound potential can vary markedly with conduction distance due to dispersion in conduction

350

-F

L7 Sl

Kn

SCfATI

h-'i

=

~ ~, ,o ,,.,,,,n

ROOT ROOT POTENTIAL IMPEDANCE

C

V(mV]

R(KII)

~

I:'~{nA)

B

AJ~.A CHA,gGE SUM COMPUTATION Q=Jldt ~.Q (hA-ms)

{pC)

{pC)

50 ,25mm

0

i I ms

0

0

~ / ~

0

2

Fig. 1. A: stimulation and recording procedure. Each of 5 hindlimb nerves was dissected free, cut, and prepared for stimulation on bipolar hooks (both hindlimbs were prepared; shown left). Ligated nerves (LGS and tibial in the left leg) exhibited characteristic neuromas and electrodes were placed at least 8 m m proximal to the ligature. Sciatic cuffs were installed during the acute experiment. The dorsal and ventral roots were sequentially cut and placed on a 6-hook array (left dorsal root S1 shown in the diagram). Short L6 roots sometimes spanned only 5 or 4 hooks. Compound action potentials were recorded between the two extreme electrodes ( + a n d - - ) . B: determination of root impedance. Immediately before stimulating the hindlimb nerves, the root impedance was determined using a 10 kHz test signal. Impedances measured at consecutive electrodes with respect to the recording.electrode at the cut end ( + ) are plotted as a function of interelectrode distance for a typical SI dorsal root. The y-intercept of the regression line indicates the contact impedance (Zc) of the electrode-tissue interface. The net tissue impedance of the root was computed by subtracting the contact impedance from the total value between the recording electrodes. C: schematic representation of determination of charge. The diagrams represent compound potentials recorded from the L7 and SI ventral roots in one preparation, obtained when the control M G nerve was stimulated. Recording configurations are shown at the left. Note that different roots had markedly different impedance values, which were used to scale the 190tentials accordingly. The calculations of the total charge delivered by each nerve are further described in the text.

351 velocities of different fibers, the area under the curve, which remains relatively unchanged with changes in conduction distance, is a better indicator of fiber population response 14. Computing area was equivalent to integrating current over time, giving units of charge. This procedure gave a measure in picocoulomb (pC) of the discharge of the myelinated fibers projecting from each nerve onto each root. In practice, charge values were computed by the following approximation (see Fig. 1C): peak voltage (mV) charge (pC) =

× (1/4 base (msec) -k 1/2 halfwidth (msec)) × 1000. (1)

root impedance (kf~)

The resolution by this method was 0.1 pC. Irregularly shaped potentials and dispersed potentials showing several peaks were subdivided into several trapezoidal components whose areas were computed separately, and added. Reproducibility of independent measurements of the same potential by different methods, including integration by analog circuitry, was within 4- 3 ~o. Values of charge depend on the number of fibers contributing to the potentials, and on the cross-sectional area of the fibers 37. Measurements based on compound action potential areas therefore carry a bias weighted towards large myelinated fibers, but are still preferable to conduction velocity measurements, which only measure the largest fibers. Furthermore, in contrast to peak voltages, values of charge from similar waveforms in different roots are additive (last column in Fig. 1C). Thus, the total contribution from fibers belonging to each nerve, projecting onto all dorsal roots and all ventral roots was measured. In several preparations compound action potentials were also recorded distally from each of the nerves, upon root stimulation. However, the distal portions of freshly cut nerves tended to deteriorate with time, which altered the electrical recording conditions by the time we investigated all the roots. Furthermore, in some ligated nerves (e.g. LGS, MG) the neuromas limited the length available for dissection, which hindered comparison with control nerves. These problems were not encountered when the nerves were stimulated distally and potentials were recorded in turn from freshly cut roots. Therefore, results presented in this paper are based on potentials recorded from roots, at locations remote from the sites of the original lesions. Data analysis

In order to control for variability in the contribution of each nerve to each of the roots, we compared the experimental nerve's total dorsal charge (DRCexp) with the control nerve's dorsal contribution DRCctrl and likewise for the ventral contributions of each nerve (VRCexp and VRCetrl). Thus, values were normalized within each cat. Relative changes were shown by a root ratio (RR) computed for each nerve: Y' DRCexp/y, VRCexp RR

=

.

(2)

Y' DRCctrl/y' VRCetrl In successful preparations every one of the 10 nerves and 12 roots held up well

352 for the period of time needed to determine root ratios. Frequently, however, one or more nerves blocked during the acute experiment, or individual roots were damaged during dissection. Consequently, in some cats charge values were obtained for either the control or the experimental nerve but not both. As an alternative, the pooled contributions of all nerves of each type were compared with pooled contributions of control nerves. RESULTS i

Fig. 2 shows compound action potentials recorded in one cat from S1 dorsal and ventral roots in response to stimulation of the LGS and M G nerves. Experimental nerves, ligated 139 days earlier, are shown by asterisks (LGS in the left leg, and M G in the right leg). Potentials from ligated nerves are broader and of much smaller amplitude, and the latencies to onset and to peak values are much longer. This agrees well with histological data on the decline in diameter observed for each fiber 21,43, present also in the dorsal roots where our recordings were done, proximal to the ganglia 1°. Due to this reduction in fiber diameter, extracellular potentials are smaller, and conduction velocities are slower. Irregularly shaped waveforms (e.g. the double-peaked ventral root component of LGS) usually maintained their general shape, indicating that fiber subpopulations of different diameters had shrunken to about the same relative extent. Often one or more later waves were also resolved, at latencies two to four times that of the first wave. Later waves were attributed to y fibers in ventral roots or to A6 fibers in dorsal roots. The area under these smaller peaks, which never exceeded 5 ~ of the main contribution (see also Boyd and Davey4), were included in our charge computation.

left LOS*- DR

right LGS*-VR

LOS-DR

LGS-VR

I ~ ~~ MG -DR

MG-VR

MG*-DR

MG*VR

Fig. 2. Compound action potentials recorded from dorsal and ventralroots following nerve stimulation' Potentials recorded in one preparation from the S1 roots, which received large contributions from the LGS and MG nerves. Several superimposed traces were triggered by the stimulus onset. Vertical calibration bars indicate 1 mV, horizontal bars indicate 1 msec in each case. The wider noise band in dorsal root records at higher gains was due to spontaneous discharge in intact afferent fibers projecting from gluteal and tail regions. Asterisks indicate the experimental nerves, ligated 139 days earlier. Ligated nerves had potentials of considerably smaller amplitude and longer latency and duration than control nerves.

353 TABLE 1 Average values of total root charge, distribution of charge among roots, and dorsal to ventral root charge ratio, in control nerves

Values of total charge are arithmetic means ± S.E. for samples ranging between 8 and 15 nerves each. Charge ratios are geometric means d: S.E. of values obtained for individual control nerves of each type; samples ranged between 8 and 12 nerves each. Nerve

Dorsal roots Total charge (pC)

LGS MG Sural CP Tibial

86 ± 24 83 ± 31 68 4- 34 770 ± 180 700 4- 160

Ventral roots Average charge distribution (%) L6

L7

S1

0 1 0 37 25

76 32 37 60 67

24 67 63 3 8

Total charge (pC)

219 4- 54 180 4- 55 -570 ± 130 645 4- 260

Average charge distribution (%)

Charge ratio DRC/VRC

L6

L7

S1

4 0

79 50

17 50

0.38 i 0.03 0.40 4- 0.04

29 1

69 9

2 90

1.2 4- 0.2 1.1 4- 0.3

Average charge values obtained from control nerves are given in Table I. Dorsal to ventral fiber charge ratios (last column) are in reasonable agreement with fiber counts from histological studies. Boyd and Davey 4 counted an average of 145 myelinated type I afferent fibers and 280 a motor fibers in the M G nerve (a ratio of 0.52); we obtained an average charge ratio of 0.40. The discrepancy probably reflects the slower conduction velocity found in the central processes of afferent fibers, when compared to their peripheral processes 10. However, charge is a non-linear function of fiber diameter. Therefore, mixed musculocutaneous nerves like the c o m m o n peroneal and tibial, which have many more (but, on the average, smaller diameter) sensory than m o t o r fibers, have sensory to motor charge ratios not much greater than unity (last column of Table I). Fig. 3 shows charge values computed for an intact c o m m o n peroneal nerve and for its contralateral nerve, ligated 251 days earlier. Comparison of the bar diagrams in Fig. 3 confirms the impressions from Fig. 2: (1) the charge delivered to all roots by ligated nerves was markedly reduced; (2) ligation caused a greater charge reduction in dorsal roots, expressed as a per cent of control, than in ventral roots. The root ratio of 0.3 (from Eqn. 2), quantifies the relatively greater effect of ligation on dorsal root fibers. The values of charge delivered by all the nerves in the cat of Fig. 3 were shifted towards the lumbar roots, compared to the averages in Table I. The cord in this cat was presumably 'prefixed'5, 83 with large lumbar roots and uncommonly small sacral roots. Thus, single root charge values could not be simply compared to those of other cats. Summing the contributions on all roots and using the contralateral nerves as controls reduced this variability. R o o t charge ratios obtained from 30 nerves and their controls are shown in Fig. 4A. A ratio of 1 indicates that, on the average, dorsal root fibers belonging to the experimental nerve decreased in size by the same amount as ventral root fibers in the

354

~

8 0 0f

~[

DORSAL ROOTS

~

mEX~E~,.~NTAL ~ ] CONTROL NERVE

400 f 0~

"--] ROOT RATIO = ~

ce

n," 800 f

-r"

VENTRAL ROOTS

% DRC

17 = -- = 0.3 % VRC 57

0 L

L6 L7

L6 L7

SUM

Fig. 3. Differential decline in charge from sensory and motor fibers. Values of charge generated by CP nerves on various roots are shown for one cat. Compared to the control nerve, the experimental nerve, cut 251 days earlier, generated relatively less charge on the dorsal roots (17 % of contro]) than on the ventral roots (57 %). The ratio of these two, 0.3, (the root ratio from Eqn. 2) provides a measure of the differential decline in the charge produced by sensory and motor fibers following axotomy.

2

0

A

B

+.

+

10 o

°1o

c(

+

c) O % d

+

o Cr:

~"

4-

I 0

(3

,!

+

L~ __.~

± 50

I

--

I SO

----i

0 250 DAYS

[ 50

AFTER

I

1

T5 C

l

+

J 25£]

AXOTOMY

Fig. 4. A: time course of changes in root ratios. The ratios of dorsal-to-ventral root electrical charge (see Fig. 3) are plotted here for 30 nerves (9 LGS, 10 MG, 5 CP, and. 6 tibial). Root ratios were not significantly different from I during the first 50 days following axotomy. A Student's t-test indicated that by about 145 or 245 days, root ratios were significantly less than 1 (P < 0,005 a n d P < 0.001). Geometric mean values for each cluster are shown by connected squares. B: decline in conduction velocity ratios. Semi-logarithmic plot of data obtained from 20 LGS and M G nerves. The best-fitting straight line had a slope corresponding to a time constant of 532 dz 99 days (mean i S.E.). The conduction velocity in the fibers comprising the main peak in the dorsal compound action potential declined significantly more rapidly than in the fibers contributing to the peak in the ventral root potential.

3001 .

A)

VENTRAL

+

355

B)

ROOTS

DORSAL

ROOTS

+

+

100

O

~

II 0

+

I SO

I

I 1 SO

I

I 0

250

DAYS

AFTER

SO

I SO

250

AXOTOMY

Fig. 5. A: decline in total charge recorded from ventral roots following stimulation of LGS and MG nerves. An exponential decay curve (solid line) was fitted to the data obtained from 42 nerves. Decay parameters were: time constant 75 4- 41 days; final value 75 :k 28 pC, where the indicated standard deviations are estimates based on linear theory at the solution point. A linear approximation indicated that the final value was significantly different from zero (P < 0.01). B: decline in total charge recorded from dorsal roots following stimulation of LGS and MG nerves. Parameters of the fitted exponential curve were: time constant 121 i 70 days; final value--1.4 :k 25 pC. The final value was not significantly different from zero. same nerve. N o t e that cats with no previous ligation (0 days), were bilaterally symlnetric z3 (root ratios were reasonably close to l). R o o t ratios at 43-48 days following nerve ligation were also not statistically different from 1. However, at 139-150 and 232-252 days, the average root ratios, 0.62 4- 0.18 and 0.39 4- 0.15, respectively, were significantly smaller than 1 (P < 0.005 and P < 0.001). Thus, sensory fibers were progressively more affected than m o t o r fibers following axotomy. C o n c o m i t a n t changes in conduction velocity were also monitored. C o n d u c t i o n velocity ratios (CVR) were c o m p u t e d according to the expression DRVexp/VRVexp CVR =

(3) DRVetrl/VRVetr!

where D R V and V R V are the conduction velocities of fibers in the largest peak o f the main potentials recorded f r o m dorsal and ventral roots respectively. After ligation conduction velocity ratios were close to unity at 45 days but declined at 145 and at 245 days. The best-fitting straight line (Fig. 4B) has a slope significantly less than zero (P < 0.001), indicating greater slowing o f sensory than o f m o t o r fibers. To study the time dependence o f this process, ventral and dorsal charge values obtained f r o m 42 L G S and M G nerves were plotted separately (Fig. 5). These two nerves had comparable mean charge contributions and time courses. Inherent variability a m o n g preparations caused some scatter, but the trends are clearly seen for

356 both ventral and dorsal root charge. Non-linear curve fitting techniques 25 were used to compute the parameters of the decay process. Exponential curves fit the decay in charge better than linear or quadratic functions, as was also found for neural compound action potential amplitudes It. In the general expression for a single exponential decay curve y = Ae -t/~ -< C,

(4)

A + C is the initial value of charge before ligation, B is is the time constant of the decay process, and C is the final value of charge that is approached at the end of the decline. The asymptote C for ventral root charge from the LGS and MG nerves (Fig. 5A) is 75 4- 28 pC (mean ~ S.D.), while the corresponding value for the sensory fibers (Fig. 5B) is --1.4 ~ 25 pC. The time constants B are respectively, 75 ± 41 and 121 -5= 70 days, and the values of A, 123 ~ 29 and 86 ~ 25 pC. The standard deviations are estimates at the solution point, based on linear theory 25. Assuming that charges distribute normally and a linear approximation is valid, these results indicate that the decay process in motor fiber charge approaches a final value significantly greater than zero (P < 0.01), whereas the predicted final value for sensory fiber charge is not significantly different from zero. DISCUSSION The present study shows that following axotomy, large myetinated sensory fibers are substantially more affected than motor fibers in the same peripheral nerves. Although no significant difference was observed in the first 45 days, the decline in the total electrical charge measured from dorsal root fibers several months after axotomy was significantly greater than the decline observed for ventral root fibers. Since the total charge measured from a population of fibers is a function of fiber diameter, this finding indicates that on the average, sensory fiber diameters were more reduced than motor fiber diameters. This conclusion is supported by the greater concomitant decline observed in the conduction velocity of fibers contributing to the main peak of dorsal root compound action potentials. The decay process for motor and for sensory fibers in the LGS and M G nerves could be fitted separately with single exponential curves, allowing a prediction of the final value of charge for each fiber population. Motor fiber charge, which was down to 40 O//oo f control by 245 days, was estimated to stabilize at about 35 % of control, a value significantly different from zero. Sensory charge, on the other hand, was down to 1l ~ of control by 245 days, and the endpoint of the sensory fiber decay process could not be resolved from zero with our method. Thus, sensory fibers may atrophy continuously in response to axotomy, while motor fibers eventually stabilize at a steady level.

Validity of the findings Before discussing the implications of these findings, possible methodological errors must be ruled out. Some of these concerns have been mentioned in earlier sections but will be summarized here.

357 (1) Considerable conduction delays are introduced by distal stimulation of the very fine processes that constitute a neuroma, and this slowing is more pronounced among dorsal root than ventral root fibers40, In our experiments we placed the stimulating electrodes at least 8 mm proximal to the ligature (about 12 mm proximal on the average) in order to stimulate the parent axons in each ligated nerve. Axons rarely die back more than this distance 6. (2) Root ratios lower than 1 might conceivably have been caused by systematically greater damage inflicted on dorsal roots than ventral roots during their preparation for recording. This possibility was excluded by ligating some nerves in each hindlimb, thus having some intact nerves and some ligated nerves represented in each root. (3) Sensory fibers might be more sensitive to damage during nerve dissection. Larger fibers will block first if a nerve is rapidly stretched, or compressed ag, and some of the largest axons in cat peripheral nerves are afferent4. However, control nerve fibers should have been as susceptible to this kind of trauma as experimental nerves, so any damage should not have affected the ratios. Furthermore, although lesioned peripheral nerve fibers are more sensitive to ischaemia than normal fibers, this abnormality has been demonstrated in both sensory and motor fibersZL (4) Some fibers in ligated nerves might have reinnervated end organs, which soon produces an increase in diameter11, zl. All nerves were checked, and any nerve that showed signs of motor reinnervation was excluded from the results. No attempt was made to identify sensory reinnervation which might have occurred in the absence of visible signs of motor reinnervation. However, if sensory reinnervation did sometimes occur, a greater subsequent recovery in sensory fibers would have caused root ratios to be larger (closer to 1) than with no sensory reinnervation. Thus, the validity of the results reported here still stands. (5) Recording compound action potentials introduces a bias in favor of the largest fibers, due to their larger extracellular currents and longer wavelengths 15, 29,a0, 37. However, we used the same interelectrode distance (generally 12.5 ram) for dorsal and ventral roots. Occasionally, when a shorter distance (10 mm or, rarely, 7.5 mm) was used for short L6 roots, such change was made bilaterally. Thus little or no differential bias was introduced. (6) The severity of the chromatolytic reaction depends in part on the level at which injury occurs 39. It is unlikely that in this study dorsal root fibers were more severely affected because their cell bodies are located more peripherally than motoneuron cell bodies. The distance from the tie to the dorsal root ganglia was on average only 10-15 ~ shorter than to the corresponding ventral horn regions. Furthermore, comparing L6 dorsal root fibers with S1 ventral root fibers, whose cell bodies were about equidistant from the tie, gave the same result. (7) Following denervation of synergists, 'control' muscles may have hypertrophied, and conduction velocity in the axons supplying them may have increased above normal values 13. An increase in a-motoneuron conduction velocity can indeed occur, although Walsh et al. 42 found significant changes in axonal conduction velocity of M G motoneurons only after removal of synergists, and not after simply denervating

358 some synergists. A slight hypertrophy of contralateral motor axons would have made root ratios larger (closer to l), again not threatening the validity of our finding. Whether such increases in fiber diameter also occur in 'control' sensory axons is unknown. Thus, our results must be interpreted as relative effects between ligated and contralateral fibers. (8) A small fraction of cat lumbar ventral root fibers have been identified as afferents7,24, 28. The great majority of these fibers, however, are unmyelinated, and therefore out of the scope of the present study. Myelinated fibers account for less than 4 ~ of the total studied 2a and their effect would again be in the wrong direction. (9) Finally, differential effects did not only take place proximal to the dorsal root ganglia, where our recordings were done. Dorsal root fiber conduction velocities decline similarly in the central and peripheral segments following axotomy 1°. Furthermore, compound potentials recorded distally from several nerves upon root stimulation confirmed the finding. On the basis of the considerations just mentioned, we conclude that large myelinated sensory and motor fibers do respond differently to axotomy.

Possible mechanismsfor differential atrophy The nature of the interaction between nerve fibers and the periphery is not known. It is known, however, that axons isolated from the periphery undergo atrophic changes which may be reversed when new connections are made. Cell regulation may be mediated by exchanges of chemical substances between peripheral organs and nerve cell endings ('trophic' interactions). In our experiments, motor and sensory cells were equally disconnected from their peripheral target organs, and associated 'trophic' factors. Clearly, cell bodies of both types must receive some sort of message to signal axotomy, and reinnervation in the case of regenerating fibers. Purves 3~ showed that nerve growth factor could prevent the responses to axotomy in sympathetic ganglia, while anti-nerve growth factor produced responses characteristic of axotomy in axons that were still intact. Sensory fibers, like sympathetic fibers but unlike motor fibers, are sensitive to nerve growth factor during normal development 34. Alternatively, electrical activity may itself be the putative 'message '~° in the present circumstances. By severing their connection with the periphery, sensory fibers are largely silenced electrically38, while motor fibers continue to conduct impulses from the spinal cord. We have previously demonstrated 17,z3,3s that substantial efferent activity in motor axons can be recorded for indefinite periods of time following axotomy (see also Acheson et al.1). Electrical activity arising from a neuroma ~9 may not be sufficient to prevent atrophy of large afferents which typically exhibit high firing rates, and low thresholds. During walking, cross-correlation studies of the neural traffic in large myelinated fibers of intact common peroneal nerves have shown that sensory spikes are at least three times more numerous than motor spikes. After axotomy these relative numbers are reversed 38. Interestingly, spikes are generated spontaneously in the region of a neuroma some time after axotomy4L Up to 30 ~ of all fibers terminating in a neuroma in a severed rat sciatic nerve fire with some regularity following section. This activity,

359 present mostly in dorsal root fibers, peaks about 12 days after axotomy and declines substantially by 30 days 19. Perhaps this neural traffic is related to the absence of a differential response to axotomy in our study for the first 45 days following nerve section. In summary, two possible explanations are compatible with our results: either (1) electrical activity plays a role in the maintenance of large myelinated fibers, and motoneurons are relatively spared because they continue to draw spike traffic from the cord, or (2) sensory fibers are more critically dependent than motor fibers on a trophic chemical transported from the periphery, such as nerve growth factor. Fiber shrinkage vs cell death For either sensory or motor fibers, a decline in electrical charge could imply a more or less uniform shrinkage of all fibers, or a selective death of some fibers and sparing of others. Our data do not distinguish between these possibilities. However, later peaks from 7' and c~fibers are affected to about the same relative extent as earlier peaks, suggesting that a generalized shrinkage may take place, at least in myelinated fibers. Secondary changes affecting the size or shape of the fiber potential may also occur; e.g. changes in the geometry of myelination accompany fiber shrinkage s. Considerable evidence from previous work indicates that some cell death also occurs, depending on the severity of the lesiong0, 89. Data available from human amputees at the turn of the century suggested that cell death was more severe in the dorsal root ganglion than in the ventral quadrant of the spinal cordaL Our data (Fig. 5) suggest that motor fibers in ligated nerves eventually stabilize, while sensory fibers as a whole continue to decay, and many or all may eventually die. Whether some types of sensory fibers may endure axotomy better than others, cannot be determined from these data. In addition to axotomy, some human genetic and virally mediated disorders can selectively affect motoneurons or spinal ganglion cells 12. Clinically, amputees may experience phantom limb sensation for many years. We recently were able to stimulate sensory fibers with a cuff implanted surgically around the ulnar nerve of a human amputee 3a whose arm had been amputated below the elbow over 30 years earlier. During the recent operation, the nerve was seen to end clearly in a neuroma. Yet, when the nerve was stimulated post-operatively, the amputee reported sensations on the dorsolateral surface of his phantom hand, and along the fourth and fifth fingers. This is the usual innervation of the ulnar nerve, and suggests that at least some sensory fibers had survived for over 30 years following axotomy. A clinical implication of the work presented here is that repair of nerve injuries should take place as soon as possible to maximize the chances of good sensory recovery after nerve lesion. Motor recovery may proceed successfully after much longer interruptions, although the extent of motor reinnervation may be limited by severe muscle atrophy 2,3~, rather than by the ability of severed motor axons to regenerate.

ACKNOWLEDGEMENTS The authors are grateful to Drs. G. E. Loeb, W. B. Marks, M. J. O'Donovan, Z.

360 Seltzer, J. T o o p a n d H. DeF. Webster for c o m m e n t s on the manuscript. This work was supported by the Medical Research C o u n c i l a n d the M u s c u l a r Dystrophy Association o f C a n a d a . J. A. Hoffer was a M D A C Post-Doctoral Fellow. REFERENCES 1 Acheson, G. H., Lee, E. S. and Morison, R. S., A deficiency in the phrenic respiratory discharges parallel to retrograde degeneration, J. NeurophysioL, 5 (1942) 269-273. 2 Aitken, J. T., Sharman, M. and Young, J. Z., Maturation of regenerating nerve fibres with various peripheral connexions, J. Anat. (Lond.), 81 (1947) 1-22. 3 Aitken, J. T. and Thomas, P. K., Retrograde changes in fibre size following nerve section, J. Anat. (Lond.), 96 (1962) 121-129. 4 Boyd, 1. A. and Davey, M. R., Composition of Peripheral Nerves, Livingstone, Edinburgh, 1968, 57 pp. 5 Burke, R. E., Strick, P. L., Kanda, K., Kim, C. C. and Walmsley, B., Anatomy of medial gastrocnemius and soleus motor nuclei in cat spinal cord, J. NeurophysioL, 40 (1977) 667-680. 6 Cavanaugh, M. W., Quantitative effects of the peripheral innervation area on nerves and spinal ganglion ceils, J. comp. NeuroL, 94 (1951) 181-219. 7 Clifton, G. L., Vance, W. H., Applebaum, M. L., Coggeshall, R. E. and Willis, W. D., Responses of unmyelinatedafferents in the mammalian ventral root, Brain Research, 82 (1974) t63-167. 8 Cragg, B. G., What is the signal for chromatolysis? Brain Research, 23 (1970) 1-21. 9 Cragg, B. G. and Thomas, P. K., Changes in conduction velocity and fibre size proximal to peJ ipheral nerve lesions, J. Physiol. (Lond.), 157 (1961) 315-327. 10 Czeh, G., Kudo, N. and Kuno, M., Membrane properties and conduction velocity in sensory neurones following central or peripheral axotomy, J. Physiol. (Lond.), 270 (1977) 165-180. 11 Davis, L. A., Gordon, T., Holier, J. A., Jhamandas, J. and Stein, R. B., Compound action potentials recorded from mammalian peripheral nerves following ligation or resuturing, J. PhysioL (Lond.), 285 (1978) 543-559. 12 Dyck, P. J., Pathologic alterations of the peripheral nervous system of man. In P. 3. Dyck, P. K. Thomas and E. H. Lambert (Eds.), Peripheral Neuropathy, VoL I, Saunders, Philadelphia, 1975, pp. 296-336. 13 Edds, M. V., Hypertrophy of nerve fibers to functionally overloaded muscles, J. eomp. Neurol., 93 (1950) 259-275. 14 Erlanger, J. and Gasser, H. S., Electrical Signs of Nervous Activity, University of Pennsylvania Press, Philadelphia, 1937, 225 pp. 15 Gasser, H. S., Effect of the method of leading on the recording of the nerve fiber spectrum, J. gen. PhysioL, 43 (1960) 927-940. 16 Gehuchten, A. van, L'anatomie fine de la cellule nerveuse, Cellule, 13 (1897) 315-390 (quoted by CraggS). 17 Gordon, T., Hoffer, J. A. and Stein, R. B., Continued generation of action potentials in nerves without peripheral connexions, J. Physiol. (Lond.), 272 (1977) 39P-40P. 18 Gordon, T., Hoffer, J. A. and Stein, R. B., Differential response of motor and sensory fibers to ligation of cat hindlimb peripheral nerves, NeuroscL Abstr., 4 (1978) 1504. 19 Govrin-Lippmann, R. and Devor, M., Ongoing activity in severed nerves: source and variation with time, Brain Research, 159 (1978) 406-410. 20 Grant, G. and Arvidsson, J., Transganglionicdegeneration in trigeminal primary sensory neurons, Brain Research, 95 (1975) 265-279. 21 Gutmann, E. and Sanders, F. K., Recovery of fibre numbers and diameters in the regeneration of peripheral nerves, J. Physiol. (Lond.), 101 (1943) 489-518. 22 Hoffer, J. A., Gordon, T. and Stein, R. B,, Differential atrophy of sensory and motor fibers following ligation or resuture of cat hindlimb nerves. In: D. L. Jewett and H. R. McCarroll (Eds.), Nerve Repair: lts Clinical aud Experimental Basis, C. V. Mosby, St. Louis, 1979, in press. 23 Hoffer, J. A., Stein, R. B. and Gordon, T., Effects of axotomy on neural activity during locomotion, NeuroscL Abstr., 3 (1977) 863. 24 Kato, M. and Hirata, Y., Sensory neurons in the spinal ventral roots of the cat, Brain Research, 7 (1968) 479--482. 25 Knott, G. D. and Shrager, R. I., On-line modeling by curve-fitting. In Computer Graphics: Proceedings of the SIGGRAPH Computers in Medicine Symposium, VoL 6, 1972, pp. 138-151.

361 26 Kuno, M., Miyata, Y. and Munoz-Martinez, E. J., Differential reaction of fast and slow a-motoneurones to axotomy, J. PhysioL (Lond.), 240 (1974) 725-739. 27 Lieberman, A. R., The axon reaction: a review of the principal features of perikaryal responses to axon injury, Int. Rev. Neurobiol., 14 (1971) 49-124. 28 Loeb, G. E., Ventral root projections of myelinated dorsal root ganglion cells in the cat, Brain Research, 106 (1976) 159-165. 29 Marks, W.B. and Loeb, G. E., Action currents, internodal potentials, and extracellular records of myelinated mammalian nerve fibers derived from node equations, Biophys. J., 16 (1976) 655-668. 30 Paintal, A. S., The influence of diameter of medullated nerve fibres of cats on the rising and falling phases of the spike an d its recovery. J. Physiol. (Lond.), 184 (1966) 791-811. 31 Purves, D., Long-term regulation in the vertebrate peripheral nervous system, Int. Rev. Physiol., 10 (1976) 125-178. 32 Ranson, S. W., Retrograde degeneration in the spinal nerves, J. comp. NeuroL, 16 (1906) 265-293. 33 Romanes, G. J., The motor cell columns of the lumbo-sacral spinal cord of the cat, J. comp. NeuroL, 94 (1951) 313-363. 34 Scott, D., Gutmann, E., and Horsky P., Regenelation in spinal neurons: proteosynthesis following nerve growth factor administration, Science, 152 (1966) 787-788. 35 Seddon, H., SurgicalDisorders of the Peripheral Nerves, Churchill, Livingstone 1972, pp. 299-321. 36 Stein, R. B., Charles, D., Gordon, T., Hoffer, J. A. and Jhamandas, J., Impedance properties of metal electrodes for chronic recording from mammalian nerves, 1EEE-BME, 25 (1978) 532-537. 37 Stein, R. B., Nichols, T. R., Jhamandas, J., Davis, L. and Charles, D., Stable long-term recordings from cat peripheral nerves, Brain Research, 128 (1977) 21-38. 38 Stein, R. B., I-Ioffe~,J. A., Gordon, T., Davis, L. and Charles, D., Long-term recordings from cat peripheral nerves during degeneration and regeneration: implications for human nerve repair and prosthetics. In D. L. Jewett and H. R. McCarroll (Eds.), Nerve Repair: Its ClinicalandExperimental Basis, C. V. Mosby, St. Louis, 1979, in press. 39 Sunderland, S., Nerves and Nerve Injuries, Livingstone, Edinburgh, 1968, pp. 72-77. 40 Wall, P. D. and Devor, M., Physiology of sensation after peripheral nerve injury, regeneration, and neuroma formation. In S. G. Waxman (Ed.), Physiology and Pathobiology of Axons, Raven Press New York, 1978, pp. 377-388. 41 Wall, P. D. and Gutnick, M., Ongoing activity in peripheral nerves: the physiology and pharmacology of impulses originating from a neuroma, Exp. NeuroL, 43 (1974) 580-593. 42 Walsh, J. V., Burke, R. E., Rymer, W. Z. and Tsairis, P., Effect of compensatory hypertrophy studied in individual motor units in medial gastrocnemius muscle of the cat, J. Neurophysiol., 41 (1978) 496-508. 43 Weiss, P., Edds, M. V. and Cavanaugh, M., The effect of terminal connections on the caliber of nerve fibers, Anat. Rec., 92 (1945) 215-233.