Compartmental neurodegeneration and synaptic plasticity in the Wld^s mutant mouse

Axons

The primary event in WD is axonal fragmentation and degeneration (Vial, 1958; Allt, 1976; Hallpike, 1976; Nicholls et al. 1992). Subsequent breakdown and removal of the myelin sheath occurs by phagocytosis involving the invasion of myelomonocytic cells after the onset of axonal degeneration (Beuche & Friede, 1984). Once the lesioned axon has begun to fragment, the myelin sheath retracts from the nodes of Ranvier creating enlarged nodal spaces (Fig. 1). These then segregate the nerve into ‘digestive chambers’ or ‘ellipsoids’ (Allt, 1976). Within each of these compartments, the axon fragments proceed to a state of complete degradation. Electron microscopy has demonstrated that the major early axonal changes include fragmentation of endoplasmic reticulum and dissolution of neurofilaments and microtubules within 48 h (Vial, 1958; Honjin et al. 1959; Ballin & Thomas, 1969; Donat & Wisniewski, 1973). These changes have been attributed by Schlaepfer (1974) to the influx of calcium ions at the lesion site. Soon after the onset of these events it is also possible to detect a more distinctive degenerative marker, the swelling and lysis of axonal mitochondria.

As degradation of the axon continues, the ellipsoids are removed by phagocytosing Schwann cells and invading macrophages. At the same time, Schwann cells proliferate: in lesioned rabbit sciatic nerve after 25 days there are up to 13 times the number present before injury (Abercrombie & Johnson, 1946). These Schwann cells join together, tip to tip, forming longitudinal bands known as ‘bands of Büngner’. The Schwann cell bands may play a role in guiding the regenerating proximal nerve stump axons back to the denervated site. Schwann cells secrete many different growth and adhesive factors such as nerve growth factor (NGF; Heumann et al. 1987), the neural cell adhesion molecule (N-CAM; Nieke & Schachner, 1985) and cytokines including members of the interleukin (IL-x) family (Rutkowski et al. 1999; for review see Fu & Gordon, 1997).

Cell body reaction

In adults the proximal portion of an axotomised motoneurone does not normally degenerate alongside the distal stump following axotomy, at least in the short term (Romanes, 1941; Johnson & Duberley, 1998). However, marked changes occur to both the cell body and its nucleus (Nicholls et al. 1992). The cell body swells, the nucleus translocates and the Nissl substance (endoplasmic reticulum) becomes dispersed.

In contrast, survival of neonatal motoneurones is strongly dependent upon their maintaining synaptic contact with their target muscles. Thus neonatal motoneurones normally die by apoptosis within a few days of nerve section. In this case, motoneurone death is mitigated by neurotrophic factors originating in the target muscles signalling through transmembrane receptors to the Bax-Bcl-2-Bcl-X system, activating calcium-dependent proteases (Martinou et al. 1994; Knudson et al. 1995; Arce et al. 1998; Villa et al. 1998; reviewed by Pettmann & Henderson, 1998).

Immediate early genes (IEGs; see Morgan & Curran, 1991), which control the cell body response following axotomy of adult motoneurones, are activated by transcription factors including c-Jun and JunD, which are selectively expressed in axotomised peripheral and central neurones within 10-15 h of nerve lesion (for review see Herdegen & Leah, 1998). By contrast, c-Fos, FosB, JunB and Fras expression do not occur at all (Leah et al. 1991; Jenkins & Hunt, 1991; Herdegen & Zimmermann, 1994). Induction of c-Jun expression precedes the activation of several regeneration-associated genes such as GAP-43, tubulins and other cytoskeletal proteins required for axon regeneration (Mikucki & Oblinger, 1991; Tetzlaff et al. 1991; Herdegen & Zimmermann, 1994).

Neuromuscular junction

The progress of WD at the mammalian neuromuscular junction (NMJ) is well documented (Miledi & Slater, 1968, 1970; Manalov, 1974; Winlow & Usherwood, 1975). Axotomy induces degeneration of nerve terminals before the degeneration of their motor axons (Birks et al. 1960): within 24-26 h after axotomy in rodents, depending upon the length of the remaining distal nerve stump. Initial degenerative changes in the nerve terminal include (after a lag period of 3-8 h): (i) swelling, disruption of the cristae and destruction of mitochondrial membranes; (ii) a reduction in the number of and clustering of synaptic vesicles; and (iii) active invasion of terminal Schwann cell processes into the synaptic cleft. More advanced degenerative changes involve extensive fragmentation of nerve terminal membranes, accompanied by axoplasmic autolysis and phagocytic engulfment of the nerve terminal by the terminal Schwann cell. These sometimes remain apposed to the postsynaptic folds for up to 3 weeks after phagocytosis is complete. Retention of Schwann cells in the period following denervation is significant, because they promote reinnervation by extending processes that guide sprouts and regenerating axons back to denervated endplates (Son & Thompson, 1995a,b).

In summary, axotomy induces distinct cellular responses on either side of an axonal lesion. Distal axons and synaptic terminals degenerate. In neonates, axotomised neurones also die, by apoptosis. Adult neurones normally survive axotomy, however. Thus these distinct reactions of normal neurones to interruption of their axons suggest that different cellular mechanisms regulate cell survival; and these mechanisms may, to a certain extent, be compartmentalised.

Discovery of the Wld^s mouse

During their investigations into the role of recruited myelomonocytic cells in WD of mouse peripheral nerve, M. C. Brown, V. H. Perry and their colleagues discovered, quite serendipitously, a spontaneous mutation in the C57Bl/6 line of mice supplied originally by Harlan-Olac (Lunn et al. 1989). The mutant mice show no readily discernible phenotype and they breed easily. What distinguishes these mice is that WD is significantly delayed and protracted after axotomy. Thus, the distal portion of cut axons and their motor nerve terminals remain morphologically intact for as long as 2 weeks. Remarkably, the isolated distal axons are still capable of conducting action potentials, and neuromuscular synapses continue to release neurotransmitter and recycle synaptic vesicles for several days, despite being disconnected from their cell bodies (Tsao et al. 1994; Ribchester et al. 1995). Both sensory and motor axons are delayed in their degeneration, and the mutation also delays WD in the central nervous system, for example following section of the optic nerve (Perry et al. 1990a; Ludwin & Bisby, 1992). The Wld^s phenotype appears to be an age-dependent phenomenon, however; mice over 4 months of age appear to revert to the WD pattern of a wild-type (Perry et al. 1992; Tsao et al. 1994; Ribchester et al. 1995; Gillingwater et al. 2000; T. H. Gillingwater, D. Thompson, M. P. Coleman & R. R. Ribchester, manuscript in preparation; but see Crawford et al. 1995).

Molecular genetics of the Wld^s mouse

Genomic analysis of the mutant subsequently led to its redesignation as Wld^s (Lyon et al. 1993). The nature of the Wld^s mutation has now also been unequivocally demonstrated.

Perry et al. (1990c) produced genetic outcrosses and backcrosses with the Wld^s mouse and as a result showed that the mutation is controlled by a single autosomal dominant gene. Further studies mapped the Wld locus to the distal end of chromosome 4, the region homologous with the human chromosomal region 1p34-1p36 (Lyon et al. 1993). The Wld locus was subsequently identified to be an 85 kb tandem triplication within the candidate region, although in some mice there is a tandem duplication indicating that there may be some instability in the repeat copy number (Coleman et al. 1998).

Recent studies have identified exons of three genes located within the 85 kb repeat sequence that could account for the Wld^s phenotype (Conforti et al. 2000). Two genes, ubiquitin fusion degradation protein 2 (Ufd2) and a novel gene, D4Cole1e, were reported to span the proximal and distal boundaries of the repeat unit, forming a chimeric gene with an open reading frame coding for a 43 kDa fusion protein (Fig. 2). The Wld^s gene and its protein product are strongly expressed in the nervous system of Wld^s mice. The third exon, coding for Rbp7, is not expressed in the nervous system (Conforti et al. 2000).

It has since been shown that the novel sequence D4Cole1e incorporates the complete sequence encoding nicotinamide mononucleotide adenylyl transferase (Nmnat), the enzyme responsible for synthesising NAD (Emanuelli et al. 2001), but incorporating an N-terminal sequence of 18 amino acids that is not normally translated in Nmnat. Transgenic mice expressing the complete chimeric protein sequence under the control of a β-actin promoter also show the Wld^s phenotype (M. P. Coleman, personal communication).

Slow WD in the Wld^s mouse is an intrinsic property of the nerve

Despite the recent demonstration of the genetic characteristics of the Wld^s mutant, the exact function of the gene and the mechanisms by which it confers protection from the effects of axonal injury are not known. In their initial paper describing the Wld^s mouse, Lunn et al. (1989) postulated that the reason for the slow progression of WD was a problem with the recruitment of myelomonocytic cells to the distal portion of the axon. They subsequently showed in transplant experiments that the phenotype was not due to a defect in the circulating monocyte population (Perry et al. 1990b). They concluded that the mutation therefore alters intrinsic properties of axons. These findings were supported by Glass et al. (1993) who showed that Wld^s axons still degenerate slowly within grafted peripheral nerve sheaths containing wild-type Schwann cells, while axotomised wild-type axons degenerate normally within an environment containing Wld^s Schwann cells. Likewise, axotomy of Wld^s axons demyelinated by intraneural injection of lysophosphatidyl choline resulted in slow degeneration of the distal stump (Hall, 1993). Axotomised Wld^s neurites also persist in tissue culture. For example, Deckwerth & Johnson (1994) studied Wld^s sympathetic neurones from the superior cervical ganglion. In wild-type neurones the cell body and axons degenerate concurrently following the removal of NGF (Edwards & Tolkovsky, 1994). However, in cultures of Wld^s neurones where NGF was absent, the axons survived whilst cell bodies underwent apoptosis. Distal neurites also persist following physical axotomy in culture (Glass et al. 1993), although the survival of axons depends on how long the neurones are cultured before their axons are cut (Buckmaster et al. 1995). From these experiments it was suggested that the cell body of a neurone may provide a maintenance factor which is transported to the extremities of the cell via axoplasmic transport and inhibits the initiation of WD. The Wld^s mutation may either cause a stabilisation of such a factor, or impair the axonal machinery responsible for its breakdown, thereby prolonging the initiation of WD following separation from the cell body.

The hypothesis that the Wld^s phenotype may be due to the impairment or failure of axonal transport was examined by Glass and colleagues (Glass & Griffin, 1991, 1994; Watson et al. 1993), but refuted with the demonstration that bi-directional transport of neurofilaments continues at a normal rate for up to 14 days post-axotomy. Tsao et al. (1994) also found no abnormalities in neurofilament phosphorylation and stability. A number of studies have addressed the possibility that an altered regulation of Ca²⁺ ions within the nerve causes the Wld^s phenotype. For instance, both Glass et al. (1994) and Buckmaster et al. (1995) demonstrated that degeneration in Wld^s axons, as in wild-type axons, is calcium dependent. Their data also suggest that calcium-dependent WD-associated proteases are present in Wld^s axons, but that these may require higher levels of calcium for their activation than in normal axons. However, Tsao et al. (1994) found that the levels of calcium-activated proteases in Wld^s axons were normal and Glass et al. (1998) found no evidence for a defect in the m-calpain 80 kDa subunit. The calpain system of neurofilament degradation therefore appears to be functioning normally, but the possibility that calpain activation by calcium is impaired in the Wld^s mouse has not been ruled out.

Recent evidence in support of the hypothesis that calcium ions may play a pivotal role in the regulation of the Wld^s phenotype is provided by experiments examining the role of NAD in intracellular Ca²⁺ regulation. Both cyclic ADP-ribose (cADPR) and nicotinic acid adenine dinucleotide phosphate (NAADP), derivatives of NAD and NADP, respectively, initiate release of Ca²⁺ from intracellular stores (Lee, 1999; Ziegler, 2000; Podesta et al. 2000). This has led to the suggestion that NAD derivatives are endogenous modulators of intracellular Ca²⁺ (White et al. 2000) and may also have a role in the modulation of neurotransmitter release from pre-synaptic nerve terminals (Mothet et al. 1998). This hypothesis is noteworthy given the identification of Nmnat in the Wld^s genotype (Conforti et al. 2000).

Intriguingly, a recent study by Benavides et al. (2000) suggests that the elevation of intracellular Ca²⁺ in Wld^s hippocampal neurones following depolarisation is significantly less than in control preparations. They suggest that this is evidence for abnormal calcium ion entry into Wld^s neurones. However, if the mutation produces abnormal calcium buffering, perhaps via its effects on NAD levels, this could perhaps explain their findings.

Finally, proteins targeted by ubiquitin (Laney & Hochstrasser, 1999) may play a critical role in the Wld^s phenotype. It is possible that the fragment of Ufd2 expressed in the chimeric Wld^s protein may act to competitively inhibit the action of native Ufd2 (a ‘dominant-negative’ effect).

Effect of axotomy on Wld^s NMJs

Axotomised Wld^s motor nerve terminals, as well as axons, undergo a process remarkably distinct from classical WD. Ribchester et al. (1995) showed that Wld^s NMJs are also preserved and retain their ability to release neurotransmitter and recycle synaptic vesicle membrane for at least 3 days, and in a few instances as long as 2 weeks, following nerve lesion.

Evidence for piecemeal withdrawal of axotomised Wld^s nerve terminals and retraction of axons has since been presented, based on studies using vital dye labelling, immunocytochemistry, electrophysiology and nerve/ muscle cultures (Fig. 3; Mattison et al. 1996; Parson et al. 1998a,b; Ribchester et al. 1999). These studies suggest that nerve terminals remove themselves from the endplate, bouton by bouton, until they form a characteristic bulbous swelling at the distal end of the axon, detached from the endplate (Mattison, 1999; D. Thomson, T. H. Gillingwater & R. R. Ribchester, manuscript in preparation). Physiological changes at these synapses include a reduced quantal content and the occasional appearance of ‘giant’ miniature endplate potentials (MEPPs). The progressive nature of this synapse withdrawal is quite distinct from the synchronous degeneration of terminals observed at wild-type denervated NMJs (Miledi & Slater, 1968, 1970; Winlow & Usherwood, 1975).

The ultrastructure of withdrawing terminals in axotomised Wld^s neurones is also distinctive (Fig. 3; Ribchester et al. 1995; Gillingwater et al. 2000). They appear morphologically intact and show none of the classical signs of degeneration such as mitochondrial disruption and swelling. Morphological alterations include dense packing of synaptic vesicles towards the pre-synaptic membrane, with some vesicles as large as 130 nm in diameter. The appearance of large synaptic vesicles suggests that the synaptic vesicle retrieval mechanism may be impaired during withdrawal of Wld^s nerve terminals. Neurofilaments also accumulate in the centre of withdrawing nerve terminal boutons.

Like axon degeneration, the degeneration of denervated NMJs in the Wld^s mouse is also age dependent. In mice older than 3-4 months, both the time course and morphology of degeneration revert to that seen in wild-type mice (Gillingwater et al. 2000, 2001).

Secondary characteristics associated with the Wld^s phenotype

Although the Wld^s mouse appears indistinguishable in appearance and behaviour from the C57Bl/6J mouse, both operated and unoperated Wld^s mice show some subtle, as yet unexplained, phenotypic features, suggesting that slow nerve degeneration is not the only effect of the Wld^s mutation.

Brown et al. (1991a) found that normal, unoperated Wld^s soleus muscles express greater intrinsic tension, fewer macrophages per muscle fibre and lower basal levels of acetylcholine sensitivity than their wild-type counterparts. Using high resolution ¹H magnetic resonance spectroscopy, Tsao et al. (1999) detected altered cerebral metabolism (indicated by decreased levels of glutamate and phosphocholine relative to total N-acetyl aspartate content) in Wld^s mice compared to wild-type controls.

Other studies describe specific axotomy-related secondary characteristics associated with the Wld^s phenotype. For example, axotomy-induced motoneurone death in neonates (Lapper et al. 1994) and retrograde degeneration of cell bodies in axotomised adult retinal ganglion cells (Perry et al. 1990a) are also significantly delayed. In denervated Wld^s muscles, the initiation of muscle atrophy and development of acetylcholine sensitivity have a slower onset, and the rise in serum creatine kinase levels is also delayed (Brown et al. 1991a).

Axon-glia signalling plays an important role in the maintenance and control of both cell types in vivo. Sprouting responses of non-myelinating terminal Schwann cells are delayed by several days (both in vivo and in vitro) following denervation in the Wld^s mouse (Barry et al. 1997; Parson et al. 1998b). The resident Schwann cells in peripheral nerve produce potential maintenance factors that are taken up by the axon: for example, ciliary neurotrophic factor (CNTF). Following axotomy in the Wld^s mouse, both mRNA and protein levels of CNTF remain normal for up to 4 days, whereas in wild-type animals they both decline rapidly and synchronously (Subang et al. 1997). However, by 10 days after axotomy, CNTF mRNA levels in the Wld^s mice have decreased to wild-type levels, but levels of CNTF protein remain unchanged. This suggests that the CNTF protein is relatively stable in axotomised Wld^s axons. Its retention could contribute some protection against degeneration.

The failure of macrophage recruitment after axotomy (see above) may be due to a failure of the nerve to produce a chemotactic signal, or there may be some form of blockade that prevents the myelomonocytic cells from entering the nerve (Perry et al. 1990b). Some candidates for this chemotactic signal are the monocyte chemoattractant protein-1 (JE), which fails to be induced in Wld^s mice following axotomy (Carroll & Frohnert, 1998), and granulocyte macrophage colony-stimulating factor (GM-CSF), whose levels are also deficient following axotomy in Wld^s mice. It is also possible that persisting distal axons produce a factor that inhibits macrophage recruitment, whereas production of this normally ceases in axotomised wild-type axons (Ludwin & Bisby, 1992). Whether any or all of these features are related to Ufd2-Nmnat overexpression in Wld^s mice remains to be tested.

Nerve regeneration in the Wld^s mouse

Following axotomy in the PNS, immediate-early gene expression is initiated within a few hours (see above), and after an initial ‘dying back’, the proximal axon stump is primed for regeneration, which begins 5-48 h after the lesion (Ramon y Cajal, 1928; Brecknell & Fawcett, 1996). In mice, this process gains momentum and proceeds over a number of days before achieving a reinnervation of skeletal muscle after about 2-3 weeks.

Previously WD was thought to play a functional role in generating an environment conducive to nerve regeneration (Ramon y Cajal, 1928). It was therefore hypothesised that in the Wld^s mouse, intact distal nerve stumps would be equivalent in their obstructive effect to intact nerves. Brown et al. (1991b) showed that severed axons would not grow into a completely undegenerated portion of nerve. Remarkably, however, regeneration of motor axons after a nerve crush injury in the Wld^s mouse is not prevented by the presence of axons in the remaining distal nerve stump (Lunn et al. 1989). This suggests that even though Wld^s axons persist after axotomy and remain capable of functioning normally, they may undergo some conformational changes that allow regenerating nerves to progress along the distal stump.

Other studies suggest that whilst motor axons appear able to regenerate almost as well as in wild-type mice, sensory nerve regeneration in Wld^s mice is significantly impaired (Bisby & Chen, 1990; Brown et al. 1992). Sommer & Schafers (1998) showed that the delay in sensory nerve regeneration leads to prolonged mechanical allodynia. One possible explanation for the disparity between the regeneration of motor and sensory nerves is that axotomised sensory nerve axons degenerate and regenerate more slowly because they are responsive to NGF levels (Bisby & Chen, 1990) and NGF levels do not increase as rapidly after nerve section in Wld^s mice compared to wild-type mice (Brown et al. 1991b). Motor axons do not respond to NGF. They also appear to be more opportunistic, using myelinating and non-myelinating Schwann cells to guide regeneration (Brown et al. 1992, 1993). Tenascin-C is upregulated selectively in Schwann cells of axotomised muscle nerves of Wld^s mice, and axons regenerate more rapidly in these nerves than in cutaneous nerves, where there is delayed expression of tenascin-C (Fruttiger et al. 1995). Conversely, myelin-associated glycoprotein (MAG) normally inhibits axon regeneration, but the rate of axon regeneration in Wld^s mice was partially restored when these mice were cross-bred with MAG knockout mice (Schafer et al. 1996).

Shi & Stanfield (1996) also reported differences in sprouting and regeneration patterns in the CNS of the Wld^s mouse. In wild-type mice, sprouting responses are detected in the septohippocampal and hippocampal commissural projections within 3 days of a lesion to the perforant path to the dentate gyrus. However, in Wld^s mice septohippocampal axonal sprouting was only observed 5-7 days after a perforant pathway lesion and changes in the distribution of commissural axons in the dentate gyrus were not observed before 12 days. It would clearly be of interest to establish the fate of axotomised synapses, and their associations with astrocytic neuroglia following axonal lesions in the brains of Wld^s mice.

Alterations in the rate and extent of both motor and sensory reinnervation occurring as a result of the Wld^s mutation are not due to problems with the cell body reaction to axotomy. The time course of expression of transcription factors such as c-Jun is very similar in Wld^s mice to that in BALB/c controls (Brown et al. 1993).

Neuronal apoptosis

The term ‘apoptosis’ was coined by Kerr et al. (1972), to describe a process involved in the normal turnover of hepatocytes: intrinsic cell suicide. There are three main regulators of apoptosis in neurones: the Bcl-2 family of proteins, an adaptor protein known as apoptotic protease-activating factor 1 (Apaf-1) and the cysteine protease caspase family (Yuan & Yanker, 2000). Neuronal apoptosis is inhibited by growth factors and by over-expression of genes such as Bcl-2. By contrast, WD appears to be independent of the same growth factors from the cell bodies, and is unaffected by the overexpression of Bcl-2 (Dubois-Dauphin et al. 1994; Sagot et al. 1995; Burne et al. 1996).

‘Cytoplastic apoptosis’ of axons

Whilst some of the cellular processes that occur during WD (for example ellipsoid body formation) are remarkably similar to those seen during apoptosis, WD obviously cannot involve all apoptotic cellular processes (eg. DNA degradation) because axotomised distal stumps do not include motoneurone cell bodies. However, it has previously been shown that apoptosis can occur in the absence of RNA or protein synthesis (Martin, 1993) and in cells lacking nuclei (Jacobson et al. 1994; Ellerby et al. 1997). Thus, a number of authors have suggested that axonal degeneration might be viewed as a form of ‘cytoplastic apoptosis’ (Ribchester et al. 1995; Buckmaster et al. 1995; Alvarez et al. 2000). To test this hypothesis, however, it will be necessary to discover the enzymes and signalling molecules that are involved.

Studies in which the role of Bcl-2 in axonal degeneration was examined also concluded that the molecular mechanisms involved are distinct from those activated during apoptosis. Burne et al. (1996) showed that overexpression of the human Bcl-2 protein in retinal ganglion cells protects the cell body as expected. However, the axons were not protected from WD. Similar findings were reported in a study by Sagot et al. (1995) who examined the fate of cell bodies and axons in a mouse model of motoneurone disease with Bcl-2 overexpression. The increased level of Bcl-2 rescued facial motoneurones and restored their soma size and choline acetyltransferase expression. However, there was no effect on the rate of axonal degeneration in facial and phrenic motoneurones. Thus it seems unlikely that Bcl-2 itself plays a significant role in axon degeneration.

Finn et al. (2000) recently examined whether the molecular machinery of WD depends upon the caspase family of cysteine proteases. They found that caspase-3 (which is thought to be important for apoptosis in neurones) was not activated in the axon during WD. Thus, they argued that WD is molecularly distinct from the classical caspase-dependent apoptotic process implicated in axonal degeneration.

‘Synaptosis’

The slow withdrawal of axotomised nerve terminals in Wld^s mice suggests a novel form of synapse-specific neurodegeneration that is unmasked when axonal degeneration (WD) is delayed. We tentatively designate this process ‘synaptosis’. Interestingly, nerve terminals withdraw in a similar fashion during synapse elimination, an essential step in the formation or reformation of normal neuromuscular innervation patterns, which takes place in postnatal development or following reinnervation (Sanes & Lichtman, 1999; Lichtman & Colman, 2000; Ribchester, 2001). Thus, nerve terminals undergoing withdrawal during synapse elimination are removed from the endplate bouton by bouton, ending with the formation of a characteristic ‘retraction bulb’ (Brown et al. 1976; Riley, 1977, 1981; Balice-Gordon et al. 1993; Gan & Lichtman, 1998). These events resemble those at axotomised Wld^s NMJs in both morphology and time course (Ribchester et al. 1995, 1999; Mattison et al. 1996; Parson et al. 1998a; Costanzo et al. 2000; Gillingwater et al. 2000).

Developmental synapse elimination takes place at a normal rate in Wld^s mice, although axotomy in the neonate still delays degeneration (Parson et al. 1997). Thus we are not suggesting that the mechanisms of synapse withdrawal are in any way abnormal in Wld^s mice. However, it is noteworthy that the role of ‘degeneration’ in synapse elimination was a matter of debate during the 1970s. For instance, ultrastructural analysis of neonatal synapse elimination by Rosenthal & Taraskevich (1977) led to their suggestion that the retraction of supernumerary inputs from developing motor endplates occurs by a mechanism similar to WD. Subsequent electron microscopic studies repudiated these findings, however, and provided evidence that the removal of excess inputs occurs by a distinctive process of retraction (Korneliussen & Jansen, 1976; Riley, 1977, 1981; Bixby, 1981). Since the withdrawal of nerve terminals in axotomised Wld^s and normal development appears to be morphologically similar, and yet distinct from WD, this raises the possibility of a common underlying mechanism of nerve terminal retraction. One way of rationalising these findings is to suggest that synaptic maintenance depends upon the supply of essential and specific maintenance factors or molecules. If the supply of these factors is compromised, then synapses withdraw at a rate that varies inversely with the concentration of such factors. During normal development, an adequate supply may depend on their trafficking into motor nerve terminals. In Wld^s mice disruption of this trafficking is synchronised by axotomy. Synapse withdrawal (rather than degeneration) is observed because WD is absent (or profoundly delayed). In wild-type animals axotomy induces additional, rapid degenerative mechanisms in axons, which therefore mask the slower process of synapse withdrawal. In normal development there is no axotomy as such, but disruptive trafficking of the same molecules could lead to selective loss of synaptic inputs to muscle fibres. Such a mechanism is consistent with earlier proposals of ‘sibling neurite bias’ or ‘intrinsic withdrawal’ as important processes in establishing the normal motor innervation pattern of skeletal muscle fibres (Brown et al. 1976; Thompson & Jansen, 1977; Smalheiser & Crain, 1984; Fladby & Jansen, 1987).

Disruption of axonal transport under experimental conditions also triggers synapse withdrawal. For instance, blocking fast axonal transport with batrachotoxin causes nerve terminals to withdraw from endplates within 18 h; but they then grow back (Hudson et al. 1984). Nerve terminals also retract in a ‘non-Wallerian’ fashion following a single subcutaneous injection of the organophosphate sarin (Kawabuchi et al. 1991). Rich & Lichtman (1989) observed comparable, reversible nerve terminal withdrawal from degenerating muscle fibres. Nerve terminals also retract in piecemeal fashion in a mouse model of myasthenia gravis (Rich et al. 1994).

The molecular mechanisms of synapse withdrawal, either during synapse elimination in development or following axotomy in Wld^s mice, remain unknown. However, further evidence in support of the hypothesis that the ‘degeneration’ of synapses is regulated independently from that of cell bodies or axons has been obtained in recent studies of CNS neurons. For example, Ivins et al. (1998) have shown that activation of caspases plays a crucial role in neurite degeneration in cultured hippocampal neurones exposed to an apoptotic stimulus (amyloid β-peptide). Using the same cell-death stimulus, Mattson et al. (1998b) found no caspase activation in axons, but instead reported that caspases were activated in cortical synaptosomes. Their data also provide evidence that apoptotic biochemical cascades (such as caspase activation) are selectively triggered at synaptic sites following exposure to staurosporine and Fe²⁺. The term ‘synaptic apoptosis’ was proposed (Mattson et al. 1998a,b; Mattson, 2000; Guo & Mattson, 2000).

In summary, analysis of the reactions of cell bodies, axons and synaptic terminals to axotomy in Wld^s mice suggests that mechanisms of degeneration are highly compartmentalised in neurones (Fig. 4). Neuronal cell bodies react to certain pathophysiological stimuli, or deprivation of growth factors by changing their pattern of gene expression, leading - at some stages in development, or after prolonged periods of deprivation in adults - to programmed cell death by apoptosis. Under normal circumstances, the death of the cell body is followed by rapid degeneration of the other neuronal compartments: dendrites, axons and synaptic terminals. Nerve injury disconnects axons and synapses from their cell bodies, and this normally triggers an independent mechanism - WD - in the isolated distal axon. The dissociation between neuronal cell body apoptosis and WD is revealed when WD is absent or delayed, as in the Wld^s mutant mouse. The slow pace of axon degeneration in this mutant (and in transgenic mice expressing the Wld^s chimeric protein) also reveals that disconnection of synapses from cell bodies can induce at least one additional mechanism of degeneration, namely withdrawal of synaptic terminals.