Homing behaviour of regenerating axons in the amphibian limb

Amphibian limbs become functionally reinnervated following either denervation or limb amputation and regeneration (Stephens & Holder, 1987; Wigston & Kennedy, 1987). After denervation, axons re-establish connections with their appropriate muscles, a process that may involve competitive interactions between matched and unmatched axons at the target site (Bennett & Raftos, 1977; Dennis & Yip, 1978; Wigston, 1980, 1986). However, after limb amputation and regeneration, no such target interactions occur as initial contacts between motor axons and differentiating muscle groups are correct (Wilson et al. 1988). In this case, therefore, guidance cues must exist in the regenerating limb tissue which set the growing axons on course to reach their appropriate targets. The presence of guidance cues in the axolotl limb has also been suggested following analysis of axon trajectories in the mature limb (Wilson & Holder, 1988). Axon sorting occurs at ‘decision regions’ comparable to those previously described in the developing chick limb by Landmesser and her colleagues (Lance-Jones & Landmesser, 1981; Tosney & Landmesser, 1985).

The existence of local guidance cues in limb tissue in the mature urodele amphibian has also been implied following experiments in which limb nerves are cut and misrouted (Grimm, 1971; Cass & Mark, 1975). In these cases, axons grow from the deviated nerve and locate their correct muscle targets by seeking out their original pathway (Holder et al. 1982, 1984). Using cobalt chloride axon tracking and electron microscopy, we have established previously that axons within the deviated nerve stump can break through the perineurium at or close to the site of deviation of the nerve from its normal path (Holder et al. 1984). The present study was undertaken to compare the events occurring within the deviated nerve to control nerves that had been cut but not deviated and to assess, using horseradish peroxidase filling, the behaviour of individual regrowing axons. HRP-filled regrowing axons have been examined at various times after nerve deviation and we demonstrate that regrowing axons take one of several routes to exit the deviated nerve stump. No matter which route is taken, however, once axons have left the confines of the peripheral nerve they invariably grow distally, towards the original target muscle. Much of this directed regrowth occurs along the original nerve pathway but some axons grow parallel but slightly displaced from this route. Electrophysiological and electron microscopic analysis confirm and expand on these observations and demonstrate that the deviated region of the nerve undergoes extensive degeneration as compared with the undeviated control.

Axolotls (Ambystoma mexicanum) and Italian crested newts (Trituras cristatus) were used in this study. Axolotls were spawned in the colony at King’s College and newts purchased from Xenopus Ltd. They were fed once or twice a week on raw ox heart or live tubifex. During initial experiments, in line with our previous studies (Holder et al. 1984), crested newts were used. However, axolotls were found to be preferable for HRP analysis of axon pathways. The responses of axons to deviation of peripheral nerves appear to be very similar in the two species.

Two protocols were used. In all cases, animals were anaesthetized in MS222 during surgery. In the forelimb of the axolotl, an incision was made in the skin on the dorsal surface, from the elbow to the shoulder. The skin was deflected back to expose the extensor cranialis nerve (ecn) which was cut at the level of the elbow (Fig. 1A-B). The nerve was then carefully deflected on top of biceps, and the skin flap was replaced and sutured with 10/0 ethilon sutures. This procedure has been described previously (Holder et al. 1982). The operated animals were kept in a refrigerator at 5 °C for 24 h, to prevent vigorous activity, which might have displaced the nerve, after which they were kept at room temperature. In the forelimb of the newt, the forearm flexor nerve (ffn) was deviated from the midhumerus to the shoulder region (Fig. 1C-D). To reach the nerve, a skin incision was made in the ventral midupper arm and the two fascicles of the ffn were cut. A length of nerve was teased proximally to the pectoral muscle and the ffn was pushed onto this muscle in a tunnel beneath the skin. This is essentially the method described previously (Holder et al. 1984). Control operations involved cutting, but not deviating, the ffn.

The animals were maintained for various periods following the operations. Experimental and control newts with the deviated or undeviated ffn were analysed at times from 6 days to 4 months after surgery whereas axolotls with the deviated ecn were analysed between 3 weeks to beyond a year later.

After varying periods of time animals were anaesthetized, transcardially perfused with axolotl Ringer and decapitated (see Wilson & Holder, 1988). The limb and shoulder girdle were dissected free and ecn was exposed where it entered the limb. The nerve was either crushed with HRP-encrusted forceps or jabbed with a HRP-coated micropin. In some animals, the original ecn pathway was exposed and HRP applied to one or several regenerated fascicles of axons. After HRP application, the limb and shoulder were transferred to aerated Ringer maintained at 0-5 °C for up to 24 h to allow transport of HRP.

Following incubation, the limbs were immersion fixed in 2 % glutaraldehyde during which time the entire length of the ecn was exposed from the shoulder to below the elbow. Care was taken when removing the skin from the deviated portion of the nerve to avoid tearing any axons.

Subsequent to fixation, the limbs were washed in 0·1 m-phosphate buffer (pH7·2) and HRP was demonstrated as described previously (Wilson & Holder, 1988). Incubation times were 30 min in DAB plus a further 20 min in DAB plus hydrogen peroxide. Specimens were then transferred to phosphate buffer (pH7·2).

The gross pattern of nerves and axon bundles was drawn with the aid of a camera lucida either following fixation or after reaction with DAB. The tissue was then further dissected to free the relevant nerve from the rest of the limb: both heads of biceps were cut and freed from the humerus; this muscle plus the deflected ecn, along with the connective tissue and blood vessels lying in the original ecn pathway as far as the elbow, were removed and pinned out in a Sylgardcoated dish. The tissue was dehydrated through alcohols and cleared in either methyl salicylate or cedarwood oil.

Animals were reanaesthetized in MS222 and the shoulder and limb dissected free, skinned and placed in a Sylgard-coated dish containing axolotl Ringer’s solution. In experimental animals, the recording electrode was placed proximal to the limb plexus and the stimulating electrode at one of four positions, at the cut end of the deviated nerve, midway to the point of deviation, at the point of deviation or, in some cases, in the original pathway of ffn distal to the point of deviation (see Fig. 7A). Recordings were made following stimulation from the first three positions in all cases and in the fourth position in examples surviving for longer periods of time. In controls where ffn was cut but not deviated, the stimulating electrode was placed in positions equivalent in distance from the cut as they were in the deviated nerve.

Following electrophysiological analysis, the ffn was removed from the limb, fixed in half-strength Karnovsky solution and processed for transmission electron microscopy.

The results will be described in three sections, beginning with the HRP analysis of the rerouted ecn in the axolotl and continuing with the electrophysiological and electron microscopic analysis of the rerouted newt ffn. This order is chosen because the HRP study presents clearly the basic features of the behaviour of regenerating axons that make it easier to describe the electro-physiological data.

17 reroutes were performed and animals were analysed in short-term (21 days to 14 weeks - 11 cases) and longterm (greater than a year - 6 cases) groups. 9 of the 11 short-term experiments gave good HRP fills. In contrast, the fills in the older animals were of limited value in describing individual axon behaviour as the HRP was not transported over long enough distances in these larger limbs. These cases were instructive, however, in that they allowed detailed assessment of the final pattern of regrowth and fasciculation in the reformed peripheral nerve.

The first case that gave clear HRP filling in the short-term group was examined 27 days after rerouting. Evidence from our previous studies (Holder et al. 1982, 1984) suggested that no axon outgrowth from the deviated nerve occurred before 3-4 weeks. A feature common to all preparations was retraction of the deviated nerve stump. The amount of retraction seen was variable between different animals, and was not dependent upon the time between deviation and fixation (Fig. 2), indicating that all the retraction occurs within the first 3 to 4 weeks. In some cases, the amount of retraction was almost back to the position of nerve deviation from its original pathway (Fig. 2B). The retraction left behind debris, which was usually free of axons. The degree to which individual axons retracted in each specimen before they commenced regrowth was also variable, with, occasionally, axons retracting back as far as the point of deviation. This amount of retraction was uncommon, however, and most axons retracted about the same distance, giving a distinct lateral edge to the deviated stump (Fig. 2A).

In the animals analysed between 27 and 38 days postdeviation, many of the axons had several or many sprouts from their cut and retracted ends, which were extremely fine and often very long (30-40 μm) (Fig. 3). Some of the sprouts terminated in growth-cone-like structures. Because of their fineness, it was usually not possible to trace all the sprouts of individual axons. There were also some outgrowths along the length of some axons; these took the form of filopodial- or lamellipodial-like structures (Fig. 3). Most outgrowth, however, was from the cut end of the axon: significantly, axons never systematically formed collateral sprouts at the site of deviation from their normal pathway. Sprouting and early axon regrowth within the peripheral nerve was not always directed, thereby creating a tangle of axons in the deviated stump. It is likely that most of the sprouts were retracted, as they were not as frequently apparent once axons had broken out of the deviated stump.

Unlike initial axon growth inside the nerve stump, once at the peripheral margins of the deviated nerve the great majority of axons showed directed regrowth in a distal direction. A number of characteristic axon trajectories were seen; some of these are shown in the summary drawing in Fig. 4, and Fig. 5 shows cameralucida tracings of the trajectories of axons at the margins of the deviated nerve. In Fig. 5A axon trajectories at the lateral edge of the stump are shown: one axon still possesses three terminal sprouts; two axons have made U-turns and grown back down into the deviated stump; and two axons have emerged from the end of the stump and are showing directed axon growth distally.

Fig. 5B shows directed axon regrowth at the lateral edge of the stump where a few exceptional axons can be seen directed proximally; this is the region of the degenerating, retracting piece of ecn. It is most likely, therefore, that these axons have not retracted as much as other axons although we cannot rule out the possibility that they are growing in a proximal direction.

Fig. 5E-H shows different regions of the trajectories taken by axons that leave the stump at the point of deviation from their original pathway. Although a few axons either retracted back to the deviation point or formed sprouts that broke out at this point, most axons exiting at the site of deviation had done U-tums within, or at the edge of, the deviated stump. When axons did turn around, they usually grew back along the superficial edge of the nerve rather than through the centre of the deviated stump. Some of these axons grew along the proximal edge of the deviated nerve before exiting and growing distally.

Once away from the confines of the deviated nerve stump all axonal regrowth was directed distally. Leading axons had variable terminations; most common were fine axonal extensions, sometimes ending in growth cones. There was always much evidence of axons fasciculating with each other and of fascicles uniting. Within the fascicles, there were often clusters of morphologically simple growth cones.

Not all fascicles grew along the original ecn pathway; many grew parallel, either on or close to biceps muscle. The fascicles were, however, always directed towards the elbow, where most fascicles united to enter distal limb territory (Fig. 6).

Nerves in three normal limbs and control cut nerves were analysed in both limbs of four animals 2,4,5 and 6 weeks after surgery. Recordings were made proximal to the limb plexus with the stimulating electrode in one of three positions, immediately proximal to the cut (position 3) or at more proximal positions (positions 2 and 1) corresponding to the locations at which stimulation was carried out in the deviated nerve (refer to Fig. 7A). Two weeks after simple nerve cut, responses within the proximal part of the nerve, were normal; increasing amplitude and decreased latency were evident with more proximal stimulation (cf. Fig. 7B and C). This picture was seen 4, 5 and 6 weeks after nerve cut indicating the healthy nature of the axons in the proximal nerve stump.

In sharp contrast to the cut controls, recordings from deviated ffns showed considerable degeneration within the rerouted part, which was first seen after 21 days.

Recordings following stimulation of the cut end of the rerouted nerve at 6 and 14 days appeared normal. After 3 weeks the latencies increase and the amplitude decreases considerably for stimulation at positions 3 and 2 (Fig. 7D); by 29 days, the amplitude has reduced to almost zero following stimulation at these points (Fig. 7E). Stimulation of the distal tip of the rerouted ffn at this time using a higher gain shows a multiphasic response of axons with markedly different conduction velocities indicating that only a few viable axons of variable condition remain. The same results are seen following stimulation at positions 3 and 2 at 36 and 43 days. In all cases examined, from 6 to 43 days, stimulation at point 1, which is proximal to the bend in the deviated nerve, produces essentially normal responses with consistently short latencies and large amplitude signals (Fig. 7D and E) which is in contrast to the responses evoked following stimulation at positions 2 and 3. This indicates that degeneration of axons evident in the rerouted part of ffn occurs only as far as the point of deviation. Stimulation of neurites in the original pathway of the ffn was first possible at 43 days and by 57 days such stimulation evoked action potentials in the spinal roots (Fig. 7F). By 57 days, subthreshold endplate potentials could be recorded in the biceps muscle (normally innervated by ffn and therefore denervated by rerouting) during stimulation of the rerouted ffn showing that axon sprouts had reached one of their previous targets and innervated muscle fibres. In preparations examined 64 days after surgery, action potentials could be recorded in biceps during nerve stimulation indicating maturation of the regenerating neuromuscular junctions. Action potentials could be evoked by stimulation of either the rerouted part of ffn or at the brachial plexus. In some muscle fibres, the latency of the response to stimulation of the brachial plexus was shorter than that to stimulation of the rerouted section of nerve, whereas in others the pattern was reversed. These results indicate that some axons reinnervating biceps are collaterals of axons within the ffn and others had grown from the end of the cut nerve. These observations are consistent with the variations in pattern of axon growth seen in deviated axolotl ecn with HRP.

Electron microscopic analysis of cut, but undeviated, ffn at all times between 2 and 6 weeks showed essentially normal myelinated and unmyelinated axon profiles. A large number of neurites were seen within the nerve proximal to the cut as regeneration proceeded. In sharp contrast, parts of the deviated region of the ffn showed considerable structural degeneration after only a week. Degeneration became more extensive over the next 2-3 weeks reflecting the breakdown of axonal function revealed electrophysiologically (Fig. 8A). The first few neurites were also seen in the deviated nerve at 6 days, although more appeared at 14 days. By 36 days, large groups of neurites were present within the rerouted ffn, the great majority of which being in association with a Schwann cell (Fig. 8B). In some cases, individual Schwann cells appeared associated with both degenerating axons and groups of growing neurites, an example of this appears in Fig. 8A. However, there was no evidence of remyelination at this stage.

The results presented in this paper demonstrate two central points; first, the proximal nerve undergoes a different response to transection if it is deviated away from its original pathway and distal nerve stump and second, axons, once they have grown away from the confines of the rerouted peripheral nerve, invariably grow in a distal direction showing clear homing behaviour.

The pattern of neurite outgrowth from the rerouted nerve is revealing. Inside the nerve stump itself neurites are oriented in many directions, often looping and growing back up the nerve (Figs 2-3). Neurites can exit the deviated nerve anywhere along its length from the cut tip to the point of deviation. Once without the confines of the nerve, however, growth is invariably in a distal direction. This directed growth indicates that axons are reacting to an environmental signal which may be a long-range ‘beacon’ or a localized grid network of positional values which guide axons to their targets. The existence of both mechanisms has previously been suggested following experiments performed in either the central or peripheral vertebrate nervous system. For example, retinal ganglion cells transplanted to ectopic sites in the head of frog embryos put out axons which locate their appropriate target, the optic tectum, by abnormal pathways (see for example Harris, 1986). In contrast, in other experiments involving placement of retinal ganglion cells in the head or caudal body positions (Constantine-Paton & Capranica, 1976; Katz & Lasek, 1979) axons grew in particular tracts, albeit abnormal ones, indicating a matching of axons to particular pathways (reviewed by Katz et al. 1980). Similar axonal pathfinding has been demonstrated following transplantation of Mauthner neurones (Katz & Lasek, 1981) and positional mismatch of dorsal funiculus with centrally projecting dorsal root ganglion axons (Holder et al. 1987). Strong evidence for the existence of diffusable molecules in the CNS is limited to the recent demonstration of a long-range attracting influence produced by the ventral floor plate of the embryonic rat spinal cord that is specific for a subset of spinal intemeurones (Tessier-Lavigne et al. 1988). Using a similar in vitro approach Lumsden & Davies (1986) have demonstrated the existence of a chemotropic factor produced specifically by a peripheral target, the epithelium of the trigeminal innervation field, to attract axons from an appropriate set of neurones, those of the trigeminal ganglion. This factor, although not nerve growth factor (Lumsden & Davies, 1983; Davies et al. 1987), remains unidentified.

Distal to the deviated nerve are the isolated distal nerve stump and a set of denervated muscles, either of which could be a source of a long-range chemotrophic or chemotropic molecule. Evidence for such molecules produced by distal nerve stump tissue has been produced by experiments in which such pieces of nerve are sutured into the end of silicone chambers into which neurites from the proximal nerve stump grow (Longo et al. 1983; Williams et al. 1984); and by stimulating neurite outgrowth by placing cut pieces of nerve near to a source of axons in vivo (Kuffler, 1986) or in vitro (Richardson & Ebendal, 1982). In this latter case, part of the neurite-stimulating activity was inhibited by antibodies to nerve growth factor (NGF). This molecule is now known to be produced by distal stumps following peripheral nerve section and Schwann cells in this situation express NGF receptors on their surface (Heumann et al. 1987). Although the function of NGF and its receptor in this case are not yet clear it would appear that they act to maintain axons of regrowing NGF-sensitive neurones until they re-establish contact with their targets (Johnson et al. 1988). NGF, presumably, would not support motor neurones in the same way, but nerve sheath cells from a distal stump may produce a separate neurotrophic factor which, if released from a localized source, could act neurotropically (Kuffler, 1986; Heumann et al. 1987; Johnson et al. 1988). In either case, it is not clear whether such a factor can have an effect at a distance of several millimetres, which would need to occur in the present experiments to achieve directed axon regeneration. It is also possible that the distal attraction of neurones results from contact with Schwann cells that have migrated proximally from the isolated distal nerve stumps.

An alternative mechanism would be that axons grow distally as a result of positional cues present in the surrounding limb tissue. Such positional cues are present in the mature amphibian limb because it is able to regenerate following amputation (see review by Tank & Holder, 1981). Furthermore, the connective tissue cells of the limb play a crucial role in this process (see, for example. Holder, 1989) and it is these cells that neurites will be growing over on the surface of muscles and in the connective tissue spaces within the limb. One observation that suggests that a defined pathway for a nerve is not marked out by a positional information network is that the final tracts of axons regenerated from a rerouted nerve do not follow the original path (Fig. 6). Nonetheless, as suggested by Harris (1986), axons could follow local cues, homing in on a particular point by the shortest route.

Finally, the differential axon survival seen between deviated and control nerves is worthy of comment, although its true significance remains unclear. It is certain that axon degeneration in the rerouted nerve is not due to a general signal within that length of nerve because growing neurites are plentiful, as seen using HRP (Figs 4, 5) and electron microscopy (Fig. 8). A further interesting difference is the differential response of Schwann cells within the deviated length of nerve. In one instance, they are associated with axon breakdown and clearance of degenerating cellular debris and, in the other, they surround growing neurites.

It is a pleasure to thank all members of our laboratory for their support during the course of this work, but particularly Jon Clarke for his input on the work and the manuscript. Financial support was from project grants from the MRC (N.H.), Action Research (D.T. and N.H.) and an MRC Studentship (S.W.).