Motor projection patterns to the hind limb of normal and paralysed chick embryos

A number of recent studies have given evidence that the motor projections to the limb change during embryonic development. Lamb (1976, 1977) using retrograde horseradish-peroxidase (HRP) labelling has shown that the pattern of motor innervation of the Xenopus hind limb changes during development and that at least some of this change comes about through motoneuron death. He found that as the larvae developed the rostrocaudal extent of the spinal cord innervating some limb regions became smaller: some connexions present in early embryos were lost during the period of normal motoneuron death. Pettigrew, Lindeman & Bennett (1979) have produced both HRP and electro-physiological evidence for a similar change in innervation in the chick embryo forelimb. Changing innervation has also been seen electrophysiologically in the axolotl limb (McGrath & Bennett, 1979) and rat intercostal muscles (Harris & Dennis, 1977). On the other hand, it has been claimed, on the basis of both HRP and electrophysiological studies that no such change occurs in the chick embryo hind limb (Landmesser & Morris, 1975; Landmesser, 1978 b) and that here motoneuron death is not concerned with the removal of such inappropriate connexions.

Paralysis of embryos greatly reduces the amount of motoneuron death (Creazzo & Sohal, 1978, 1979; Laing & Prestige, 1978; Olek, 1980; Olek & Edwards, 1978; Pittman & Oppenheim, 1978, 1979). Therefore, in paralysed embryos it should be possible to see the projection pattern present prior to motoneuron death. Retrograde transport of HRP should be an efficient method of revealing the projection patterns but has a serious drawback in that the whole motoneuron population is not labelled (Lamb, 1979; Landmesser, 1978b), and it is difficult to rule out leakage of HRP to neighbouring areas. An alternative is to examine the distribution of motoneurons in the spinal cord after partial amputation. This will reveal the position of the entire population innervating the remaining part of the limb. Amputation at various proximodistal levels will ultimately reveal the pattern of innervation of the whole limb.

After partial removal of the optic tectum in the goldfish, the whole innervating retina can form a complete topographic map on the remaining tectum: the innervation can ‘compress’ into the remaining target (Gaze & Sharma, 1970; Yoon, 1971, 1976). It was therefore possible that a similar ‘compression’ could occur in the partially amputated limbs: that axons could make connexions adequate for motoneuron survival by innervating parts of the limb they do not normally innervate.

The aims of the present work were to see if the patterns of innervation of normal and of paralysed limbs were the same and to see if a ‘compression’ could occur in the limb.

White Leghorn eggs were used in all experiments. Before incubation they were prepared for windowing by boring a hole into the airsac and drilling an approximately 1 × 1 cm window in the shell with a carborundum disc attached to a dental drill. The square of shell was not removed but covered with Scotch tape and the eggs incubated at 38 °C in a forced draught incubator. The predrilled windows were opened on the day of the first manipulation. α-Bungarotoxin (α-Bgt) (Boehringer or Miami Serpentarium) was used to produce paralysis because the irreversible nature of its action (Chang & Lee, 1963) made it relatively easy to obtain complete lack of movement up to day 10 of incubation. The α-Bgt was dissolved at a concentration of 1 mg/ml in Hank’s Balanced Salts Solution (pH 7·4) containing 10 mM HEPES buffer, 50i.u./ml penicillin and 50/ μg/ml streptomycin (all Flow Laboratories). The solution for control injections was the same, except for the omission of the toxin.

All solutions were passed through a 0·45 μm Millipore filter prior to use. All operations were carried out using instruments sterilised in alcohol.

Amputations were always of the right hind limb since the rotation of the embryo makes the right limbs more accessible.

Radical amputations (virtually all limb tissue removed) were performed on day 3 of incubation (stages 17 and 18 of Hamburger & Hamilton, 1951). Evan’s blue dissolved in Hank’s Balanced Salts solution was injected in the yolk sac beneath the embryo in order to increase the definition of the embryonic structures. The extraembryonic membranes were cut and the exposed limb bud was removed by cutting around it with electrolytically sharpened tungsten needles and pulling it off the embryo.

Graded amputations were produced in embryos at stages 20–24, when the limb buds are more developed, using small pieces of broken razor blade attached by Araldite (Ciba-Geigy) to fine glass rods. One was positioned deep to the limb bud as a support while the other was used to cut through the limb bud. The cuts were made at different proximodistal levels in order to produce graded amputations such as those described by Hampé (1959). Axons first enter the hind limb but at stage 24 and reach the knee at stage 25 (Oppenheim & Heaton. 1975). Therefore, in order not to axotomise neurons, only distal cuts were made in stage-24 limb buds.

On days 3 and 4 injections were carried out by tearing the extraembryonic membranes and displacing the amniotic fluid with a 100 μI loading dose of the toxin or sham-injection solution.

On days 6 and 8, injections of 50 μI were made intraperitoneally with a 100 μI Hamilton syringe fitted with a 33-gauge needle.

Embryos amputated on day 3 were injected on days 3, 6 and 8. Embryos amputated on day 4 were injected on days 4, 6 and 8.

Embryos were killed by decapitation on day 6 and day 10 of incubation and the lumbar vertebral column dissected out. In 10-day embryos the hind limbs were left attached to the vertebral column in order to allow determination of the muscle volume. Tissues were fixed in Carnoy’s solution and processed for paraffin histology. Serial sections were cut at 8/μm and stained with haematoxylin and eosin.

The numbers of lateral motor column motoneurons in every tenth section of the lumbar lateral motor column were counted blind. Only cells containing one or more nucleoli were included in the counts. The counts thus obtained were corrected for double counting using an Abercrombie (1946) correction factor.

The distribution of motoneurons within the lumbar lateral motor column was examined by dividing the rostrocaudal extent of the column into ten ‘bins’ of equal length.

Determination of muscle volume

The volume of muscle in the hind limbs of the 10-day-old embryos was calculated from camera-lucida drawings, at×14, in the standard manner (Bueker, 1945; Konigsmark, 1970) except that photocopies were cut up and weighed. In this way the original drawings were preserved for further reference.

The muscle volume in the hind limb was determined for 17 out of 20 paralysed embryos fixed at day 10 and for 16 out of 22 sham-injected controls. Motoneurons were counted in all embryos (Tables 1 and 2). The total number of motoneurons was greater on the intact sides of paralysed embryos (16586 ± 1458, mean±s.D.) than in the controls (10445 ± 1036) (P > 0 ·001, Mann-Whitney U-test). The number of motoneurons in the control embryos was not significantly different (P > 0·4, Mann-Whitney U-test) from that of totally untreated embryos (10229 ± 1429, n = 14).

The total number of motoneurons present in 10-day paralysed embryos was not significantly different (P > 0 ·1) from that in 6-day control embryos (17241 ± 1243; n = 5). The number of motoneurons peaks at day 6 (Hamburger, 1975) and thus the paralysis can be said to have prevented normal cell death. Although the chronic paralysis increased motoneuron number, it reduced muscle volume in intact limbs from 17·69 ±4·42 mm³ to 7 ·94± 1·59 mm³ (P < 0 · 001, Mann-Whitney U-test) (Fig. 1). Degenerating muscle fibres were seen in the paralysed embryos.

There were interesting relationships between motoneuron number and muscle volume in the various limbs (Fig. 1). There was a significant correlation between muscle volume and innervating motoneuron number in the intact limbs of control embryos (r = 0·73). Thus, some of the large variation in motoneuron number in control embryos seemed to be related to between-embryo variation in muscle volume. A similar correlation between limb size and motoneuron number was noted for Rana pipiens larvae by Decker & Kollros (1969). There was no significant correlation (r = 0·37) for the intact sides of paralysed embryos.

In both paralysed and sham-injected embryos as the degree of amputation increased both muscle volume and motoneuron number became reduced towards zero. In the sham-injected embryos the regression lines for the intact and amputated sides differed (Fig. 1). This indicated a different relationship between muscle volume and motoneurons in amputated and intact limbs : for the same increase in muscle volume there was a greater increase in motoneuron number on the amputated sides.

The relationship between the muscle volume and motoneuron survival after amputation could be seen by comparing the amputated side with the contralateral intact side. The volume of muscle in the amputated limb was expressed as a percentage of the volume of muscle in the intact contralateral limb and the number of motoneurons on the amputated side as a percentage of the number on the intact side. The result of plotting the values thus obtained against each other is shown in Figure 2. There was a remarkably linear relationship between the degree of motoneuron and muscle survival after amputation in both sham-injected (r = 0 ·97) and paralysed (r = 0·96) embryos. The linear relationship indicates that there is a uniform ‘innervation density’ (number of innervating motoneurons per unit of muscle volume) throughout the proximodistal extent of the limb. This finding relates to the linear relationship between the wet weight of a muscle and the number of motoneurons projecting to it noted by Landmesser (1978a).

There was no sign of any major compression of the motoneuron innervation into the limb in either the sham-injected or paralysed embryos. In neither group was there 100% survival of motoneurons with significantly less than 100% of normal muscle volume present. This suggested that, with or without paralysis, there was very little possibility for motoneurons normally innervating distal muscles to contact the remaining proximal muscles.

The P = 0 ·1 confidence limits for the two regression lines in Fig. 2 overlapped continuously indicating that the two regressions were not significantly different. This in turn would suggest that there was a constant proportion of the innervating motoneurons dying throughout the limb. There was not a greater proportion of cell death amongst motoneurons innervating the thigh than amongst motoneurons innervating the shank.

Since motoneuron death was reduced in the paralysed embryos it was of interest to see where the excess motoneurons were within the lateral motor column. This was investigated for the rostrocaudal axis by diving the full length of the lateral motor column into ten equal ‘bins’. Segments were not used because the division of the spinal cord into segments during development is arbitrary, with the result that the lateral motor column is not always present in the same segments. Using segments adds another variable which may obscure any effect of treatment.

There were greater numbers of motoneurons in all bins on the intact sides of paralysed embryos (Fig. 3). The difference between the curves in Fig. 3 represents the number of motoneurons prevented from dying or, because prevention of motoneuron death was so good, the position of motoneurons that normally die. To allow direct comparison of the three groups of embryos, the number of motoneurons in each bin in each embryo was expressed as a percentage of the total number of motoneurons in that particular embryo. The values for all the embryos in each group were then pooled (Table 3). The distribution of motoneurons innervating intact limbs appeared to be the same in control and untreated embryos (Fig. 4A); no significant difference could be detected in any bin. However, significant differences were found in seven out of the ten bins when the treated group was compared with the combined control and untreated groups (Fig. 4B). The probability values were of the same order of magnitude whether the test used was the non-parametric Mann-Whitney U-test or the parametric t-test. The probability values were also similar whether the comparison was with the combined control and untreated groups or with the control group alone except that in bin 10 the level was changed from P < 0 · 001 to 0 · 01 > P > 0 ·001.

The fact that the two distributions in Fig. 4B are different means that motoneuron death does not affect a constant fraction of the starting population of cells throughout the rostrocaudal extent of the lateral motor column. Wherever the distribution for treated embryos lies below that for the combined control group a smaller than average fraction of motoneurons die. Wherever the treated embryo distribution lies above the control a greater than average fraction of motoneurons die. Therefore, a greater than average loss of motoneurons occurs in bins 3,4 and 5. For some reason the population of limbinnervating motoneurons originally created in these bins is subject to a greater neurothansasia than the rest of the population.

One measure of the shape of a curve is the coefficient of skewness (Kendall & Stuart, 1969). The coefficient was calculated from the motoneuron counts for each embryo prior to binning. It depends mostly on the main body of the data and is therefore largely independent of the estimation of the ends of the lateral motor column whereas binning depends on this. The skewness test is, then, even more accurate than using bins. The values for the control group ranged from – 0 · 038 to 0 · 297 (mean = 0 ·133, n = 22), for the untreated group from – 0 · 018 to 0 · 243 (mean = 0 ·115, n = 14) and for the treated group from 0 ·52 to 0 ·534 (mean = 0 ·249, n = 20). Comparing the values with the Mann-Whitney U-test, the control and untreated groups were not significantly different (P > 0 ·6) whereas the treated group was significantly different from the combined control groups (P < 0 ·001). Thus, the treated group showed greater skewness than the control group, i.e. the shapes of the curves were different.

Insight into why motoneuron death was not uniform along the length of the lateral motor column could be obtained by examining the distribution of motoneurons innervating the partially amputated limbs. The distribution of motoneurons surviving various degrees of amputation may be seen in Fig. 5. In sham-injected embryos (Fig. 5 A), as the degree of amputation became less, more motoneurons survived the operation. A peak in the distribution arose regularly in the rostral tenths, with a relative plateau in the caudal tenths. The 63 % distribution approximated to a knee amputation and thus to the distribution of motoneurons innervating the thigh musculature. A similar peak and plateau for thigh musculature could be inferred from the horseradish peroxidase data of Landmesser (1978 b). The distribution in the paralysed embryos although similar, did not follow this pattern exactly (Fig. 5 B). For example, the 64 % distribution, approximating to knee amputation, did not correspond to the intact distribution in bin 3, whereas the equivalent distribution (63 %) in the control embryos did.

To emphasize this point the data for nine embryos with approximately 60 % motoneuron survival is shown in expanded form in Fig. 6 and Table 4. Using the Mann-Whitney U-test, the distribution of motoneurons on the amputated sides was found to be different from that of the intact sides in bins 5 –9 in controls (Fig. 6 A), but different in bins 3 –9 in treated embryos (Fig. 6B). In controls the probability value for bin 4 was 0 ·058, which is bordering on significance. A (-test, which unlike the U-test, takes account of how far apart the values in the two groups are, gave a probability value of 0·05 > P > 0·01. There was therefore, considering the small number of embryos involved, probably a significant difference at bin 4 in control embryos. Now the difference between the distribution of motoneurons on the amputated and intact sides reveals the position of the motoneurons innervating the part of the limb which had been amputated. In Fig. 6 the differences represent the innervation of the shank since the 63 % and 64% survival occurred with amputation approximately at the knee. In controls the two sides were different in bins 4 to 9 (Fig. 6 A) so the shank innervation in controls was from these bins. For treated embryos there was also a difference at bin 3. Thus the shank received innervation from bin 3 in embryos in which motoneuron death was prevented but not in control embryos where motoneuron death had occurred normally. Motoneuron death thus removed this innervation not found in the adult.

The aim of the present study was to investigate the innervation of intact and amputated limbs in control and paralysed embryos. By this approach, the question of whether or not the pattern of innervation changes during development could be answered. Also, information could be expected on the plasticity of the motoneuron/limb system, answering the question of whether ‘compression’ can occur in the limb.

The data showed that there is an increase in motoneuron number in the a-Bgt injected embryos and a reduction in muscle volume. Muscle atrophy in paralysed limbs has been reported many times but usually at later stages of development (e.g. Drachman, 1964; Giacobini et al. 1973). Pittman & Oppenheim (1979) note that shank weight is reduced in 10-day curare and α-cobratoxin paralysed embryos. The reduction is to 75 % of control in the α-cobratoxin paralysed embryos which is not as great as the present reduction in muscle volume (to 45 % of control). However, as well as muscle, shank weight will include bones, tendons, etc. which may not be so seveiely affected by paralysis.

Paralysed embryos have greater numbers of motoneurons innervating a smaller volume of muscle. It is probable that a trophic feedback from muscle to nerve normally keeps embryonic motoneurons alive and this feedback is presumably greater in the paralysed embryos. Possibly related cases are the interesting finding of increased nerve growth factor (NGF) levels in denervated irides (Ebendal, Olson, Seiger & Hedlund, 1980), and the more recent finding of increased concentrations of trophic factor in denervated muscle (Hill & Bennett, 1982). The basic finding does not address the problem of whether motoneuron death comes about normally through redundancy or rejection (Prestige, 1976). The paralysis may be increasing the number of available contact sites, reducing redundancy, by keeping muscle more myoblastic (Giacobini et al. 1973) or by allowing more contacts per muscle fibre (Pittman & Oppenheim, 1979). Alternatively paralysis could block the mechanism by which axons are rejected by muscle fibres just as it slows loss of multiple innervation (Brown, Jansen & & Van Essen, 1976). Of course, both sorts of process could be occurring together; with contact site number being increased and rejection being decreased.

The different relationships between muscle volume and motoneuron number on the intact and amputated sides of sham-injected embryos may relate to the effect of supernumerary limbs on motoneuron survival. There is a puzzlingly small increase in motoneuron number when a supernumerary limb is grafted onto an embryo, 15 % average in Xenopus (Hollyday & Mendell, 1976) and 17 % average in chick (Hollyday & Hamburger, 1976). In the present experiments, for a doubling in intact limb size through natural variation there is an increase in motoneuron number of 21 % (Fig. 1, control intact sides 10 mm³ to 20 mm³). In amputated limbs a doubling of muscle volume produces a much greater increase in motoneuron number (83 % for 9 to 18 mm³: control amputated sides, Fig. 1). Thus, the relationship between muscle volume and motoneuron number after supernumerary grafting is similar to that for intact limbs and dissimilar to that for amputated limbs.

If it is supposed that in a normal, whole limb there is a complete set of post-synaptic targets (cf. Prestige & Willshaw, 1975), then the difference between grafting and amputation may be stated thus: an amputated limb contains an incomplete set of targets whereas a limb ta which a supernumerary has been added has a full set of targets but extra sites at each target position. It would seem that the number of targets is more important than the number of sites at a target in determining the number of motoneurons supported by a limb. When there is a small amount of any one post-synaptic target present it will maintain a relatively large number of motor units, a result possibly related to those of bilaterally innervated limbs (Lamb, 1980).

From the comparison of the degrees of muscle and motoneuron survival on the amputated sides (Fig. 2), it is clear that no major ‘compression’ of the motoneuron innervation into the chick hind limb takes place either with, or without, paralysis. A similar result showing lack of compression of motoneuron innervation into non-paralysed chick embryo limb has recently been reported (Whitelaw & Hollyday, 1980). α-Bgt allows greater numbers of motoneurons to make connexions with each muscle, but does not apparently allow the formation of connexions by motoneurons which normally innervate other parts of the limb. Motoneuron number in the paralysed embryos is increased by 60 %, and Lamb (1980) has shown that Xenopus limbs are capable of maintaining at least twice their normal number of motoneurons. These two observations, taken together, suggest that the supportive capacity of the limb is not fully utilised in the paralysed embryos.

Axonal guidance (Lance-Jones & Landmesser, 1980) may prevent the formation of connexions between motoneurons destined for the deleted parts and the surviving muscles. Alternatively the motoneurons destined for deleted parts may connect with the surviving muscle but die because of incompatibility (Lamb, 1981). The lack of compression observed in paralysed embryos would then show that paralysis cannot prevent the motoneuron death resulting from such an incompatibility.

There are several explanations for the regression lines in Fig. 2 not going through the origin. Firstly, waiting only until day 10 may not be allowing the effect of amputation to reach completion as in the study on the chick wing by Hamburger (1934) where at days 8 and 9 40 % of motoneurons remained when all muscle had been removed. However the effect of early (day 3) amputation of the hind limb is complete at day 10 (Laing, 1982, unpublished observations), and a similar relationship was seen in embryos fixed at days 16 and 20 after amputation at day 10 (Laing, unpublished observations). The population of cells which survives amputation is real, but it may not be as large as suggested by the data. The discrimination of motoneurons from large interneurons situated laterally in the grey matter can be difficult in the absence of other motoneurons. For this reason the number of ‘motoneurons’ counted in the radically amputated embryos may be inflated by the inclusion of interneurons. It is also difficult to recognize scattered muscle fibres and take them into account in the camera-lucida drawings. Therefore, the volume of muscle present following radical amputation may be larger than indicated. Finally, some of the surviving motoneurons may be innervating the contralateral limb (Lamb, 1980). An alternative interpretation is that the survival of some motoneurons is independent of the presence of their target tissue. This is an uncomfortable hypothesis. Lamb (1981) has recently found that bilateral amputation in Xenopus gives 100% loss of motoneurons. This shows that at least in that species the survival of all motoneurons is dependent on the limb.

The data show differences in the distributions of motoneurons between control and paralysed embryos. Since the paralysis prevents motoneuron death, it is concluded that the projection pattern present prior to motoneuron death is retained in the paralysed embryos and is different from the post-death pattern. This conclusion is at variance with that of a study using paralysis combined with retrograde HRP labelling (Oppenheim, 1981). The discrepancy may arise from the different degrees of prevention of motoneuron death in the two studies.

Unlike the present study (Fig. 3), previous data on the prevention of motoneuron death with curare and botulinum toxin (Pittman & Oppenheim, 1978, 1979) showed no effect on the rostral-most part of lateral motor column. But, Oppenheim (personal communication) has recently found that using cobratoxin and starting injections at day 4 gives a hypothanasia in the rostral part of the lateral motor column. Some of the effect may be due to the use of different neuromuscular blockers, but the disparity is probably due to the onset time of the paralysis since the botulinum toxin and curare were first applied on day 6 (Pittman & Oppenheim, 1978, 1979). If this latter is true, it suggests that the rostral motoneurons are amongst the first to die. Certainly the rostral motoneurons are amongst the first to become ‘limb-dependent’ (Prestige, 1967) as shown by radical amputation (Oppenheim, Chu-Wang & Maderdrut, 1978).

The fact that the distributions of motoneurons within the lumbar lateral motor column in paralysed and control embryos are different (Fig. 4B) means that the proportion of motoneurons dying is not constant throughout the limbinnervating pool. In an earlier study (Landmesser, 1978 b) it was concluded that motoneuron death did not alter the rostrocaudal distribution of motoneurons supplying various muscles and muscle groups. The difference in results may arise from the use of segments rather than bins to denote the position of the motoneuron pools. As mentioned in the results section the demarcation of spinal cord segments is arbitrary and variable. The segments vary in length between animals and the segmental position of the motor column also varies. Using bins nullifies any variation from these sources.

The conclusions from the present study are that neurothanasia eliminates a constant proportion of the motoneurons innervating each part of the limb (Fig. 2), but not a constant proportion within the spinal cord (Fig. 4). This could only happen through non-uniformity of death in sub-populations such as has been shown for the shank (Fig. 6) where the innervation from bin 3 entirely disappears after motoneuron death. Motoneuron death therefore plays a part in producing the normal mature pattern of innervation of the limb.

This work was carried out while I was research associate to the late M. C. Prestige and funded by the Medical Research Council of Great Britain. I should like to thank Dr R. K. Milne of the University of Western Australia Mathematics Department for statistical advice and Dr A. H. Lamb for his patient criticism during the preparation of the manuscript.