OUP user menu

The slow Wallerian degeneration gene, WldS, inhibits axonal spheroid pathology in gracile axonal dystrophy mice

Weiqian Mi, Bogdan Beirowski, Thomas H. Gillingwater, Robert Adalbert, Diana Wagner, Daniela Grumme, Hitoshi Osaka, Laura Conforti, Stefan Arnhold, Klaus Addicks, Keiji Wada, Richard R. Ribchester, Michael P. Coleman
DOI: http://dx.doi.org/10.1093/brain/awh368 405-416 First published online: 11 January 2005


Axonal dystrophy is the hallmark of axon pathology in many neurodegenerative disorders of the CNS, including Alzheimer's disease, Parkinson's disease and stroke. Axons can also form larger swellings, or spheroids, as in multiple sclerosis and traumatic brain injury. Some spheroids are terminal endbulbs of axon stumps, but swellings may also occur on unbroken axons and their role in axon loss remains uncertain. Similarly, it is not known whether spheroids and axonal dystrophy in so many different CNS disorders arise by a common mechanism. These surprising gaps in current knowledge result largely from the lack of experimental methods to manipulate axon pathology. The slow Wallerian degeneration gene, WldS, delays Wallerian degeneration after injury, and also delays ‘dying-back’ in peripheral nervous system disorders, revealing a mechanistic link between two forms of axon degeneration traditionally considered distinct. We now report that WldS also inhibits axonal spheroid pathology in gracile axonal dystrophy (gad) mice. Both gracile nucleus (P < 0.001) and cervical gracile fascicle (P = 0.001) contained significantly fewer spheroids in gad/WldS mice, and secondary signs of axon pathology such as myelin loss were also reduced. Motor nerve terminals at neuromuscular junctions continued to degenerate in gad/WldS mice, consistent with previous observations that WldS has a weaker effect on synapses than on axons, and probably contributing to the fact that WldS did not alleviate gad symptoms. WldS acts downstream of the initial pathogenic events to block gad pathology, suggesting that its effect on axonal swelling need not be specific to this disease. We conclude that axon degeneration mechanisms are more closely related than previously thought and that a link exists in gad between spheroid pathology and Wallerian degeneration that could hold for other disorders.

  • axon
  • axonal spheroid
  • gracile axonal dystrophy
  • ubiquitin
  • Wallerian degeneration
  • APP = amyloid precursor protein
  • gad = gracile axonal dystrophy
  • GFAP = glial fibrillary acidic protein
  • H & E = haematoxylin and eosin
  • NMJ = neuromuscular junction
  • PFA = paraformaldehyde
  • PNS = peripheral nervous system
  • WldS = slow Wallerian degeneration gene, mutation or mice
  • WldS = slow Wallerian degeneration protein
  • YFP = yellow fluorescent protein


Axonal dystrophy and spheroids are hallmarks of CNS axon pathology. Axonal spheroids are focal 10–50 µm diameter swellings, which are sometimes, but not always, terminal endbulbs, and are filled with disorganized neurofilaments, tubules, organelles or multi-lamellar inclusions. Dystrophic axons are usually smaller swellings often associated with continuity of the axon. One or both of these aberrant axon morphologies is found in a wide range of CNS neurodegenerative disorders, including stroke (Dewar et al., 1999), myelin disorders (Griffiths et al., 1998), tauopathies (Lewis et al., 2000; Probst et al., 2000), amyotrophic lateral sclerosis (Tu et al., 1996; Oosthuyse et al., 2001; Howland et al., 2002), traumatic brain injury (Cheng and Povlishock, 1988), Alzheimer's disease (Brendza et al., 2003), Parkinson's disease (Galvin et al., 1999), Creutzfeldt–Jakob disease (Liberski and Budka, 1999), HIV dementia (Raja et al., 1997; Adle-Biassette et al., 1999), hereditary spastic paraplegia (Ferreirinha et al., 2004) and Niemann–Pick disease (Bu et al., 2002). They also occur during normal ageing and secondarily in some serious illnesses (Sung et al., 1981). In contrast, peripheral nervous system (PNS) axons undergo ‘Wallerian-like’ or ‘dying-back’ degeneration, even in diseases where CNS axons form swellings (Miura et al., 1993; Lewis et al., 2000; Oosthuyse et al., 2001), although swellings do also occur in some rare PNS disorders (Miike et al., 1986; Bomont et al., 2000).

The roles of axonal swellings in disease are poorly understood, as illustrated by the following examples. First, in multiple sclerosis, many large spheroids are terminal endbulbs of transected axons but there are also a few ‘en passant’ swellings of similar shape and dimension (Trapp et al., 1998) and many small dystrophic swellings (Ferguson et al., 1997; Kornek et al., 2000, 2001). It remains unclear whether these different types of swelling have common or different origins. Secondly, it is not clear whether disease-specific mechanisms lead to a common final pathway of axonal dystrophy, as in Alzheimer's disease, stroke and multiple sclerosis, and if so how they do this. Thirdly, it is not known why swellings predominate in distal axons in some diseases, such as gracile axonal dystrophy (gad) (Yamazaki et al., 1988; Mukoyama et al., 1989), caused by loss of ubiquitin C-terminal hydrolase l1 (Uch-l1) (Saigoh et al., 1999), while in other diseases they occur in proximal axons, as in amyotrophic lateral sclerosis (Tu et al., 1996) and tauopathy (Probst et al., 2000). Finally, a better understanding is needed of the relationship between axon swelling and impaired axonal transport. Amyloid precursor protein (APP) accumulates in axonal swellings and spheroids in stroke (Dewar et al., 1999), traumatic brain injury (Gentleman et al., 1993), multiple sclerosis (Ferguson et al., 1997), Creutzfeld–Jakob disease (Liberski and Budka, 1999), HIV dementia (Raja et al., 1997; Adle-Biassette et al., 1999) and gad (Ichihara et al., 1995), indicating that axonal transport is impaired. However, it is not known whether axon swelling in these disorders is simply a consequence of impaired axonal transport, or whether it causes the transport defect, or both. These and other important questions remain unanswered largely because experimental methods to manipulate axonal swelling have not been available.

A mutant mouse gene, WldS, blocks a rate-limiting step common to Wallerian degeneration and diverse PNS axon disorders, including dysmyelination (Samsam et al., 2003), motor neuronopathy (Ferri et al., 2003) and Taxol toxicity (Wang et al., 2002). Recently, WldS was reported to be effective in acute CNS lesions modelling stroke (Gillingwater et al., 2004) and Parkinson's disease (Sajadi et al., 2004) but its effect in a chronic CNS disease has not been reported. WldS is a chimeric gene (Conforti et al., 2000) formed by a stable triplication (Coleman et al., 1998; Mi et al., 2003) encoding the N-terminus of multiubiquitylation factor Ube4b fused in-frame to nicotinamide mononucleotide adenylyltransferase (Nmnat1) plus a short novel sequence (Mack et al., 2001). Nmnat1 appears to be sufficient to confer the phenotype in vitro, but it is not yet clear whether this holds in vivo (Coleman and Perry, 2002; Araki et al., 2004). WldS protein appears to be restricted to the nucleus, so its effect on axons is mediated by other factors (Mack et al., 2001), which may include the NAD-dependent deacetylase SIRT-1 (Araki et al., 2004).

To study the relationship between axonal swelling and Wallerian degeneration, we crossed WldS and gad mice. WldS significantly reduced spheroid numbers without altering the first stages of gad pathogenesis, revealing a link between Wallerian degeneration and axonal spheroids in this disease that could extend to other disorders.


Origin, breeding and genotyping of mice

Homozygous C57BL/WldS spontaneous mutants were obtained from Harlan UK (Bicester, UK) and mated with heterozygous gad mice, kindly provided by Professor Keiji Wada and Dr Hitoshi Osaka (National Institute of Neuroscience, Tokyo, Japan), following a cross to C57BL/6 to ensure a more homogeneous genetic background. Thus, the genetic background of the experimental mice was 75% C57BL/6, 12.5% CBA/Nga, 12.5% RFM/Nga. Double heterozygotes were identified in the F1 generation by genotyping for gad (below) and intercrossed. gad homozygotes were identified by genotyping and selected for further study. WldS genotype was determined post mortem by pulsed-field gel electrophoresis of spleen DNA (Mi et al., 2002). Hemizygous yellow fluorescent protein (YFP) mice of line YFP-H were obtained from Jackson Laboratories (Bar Harbor, MN, USA) and mated with gad/WldS double heterozygotes. Triple heterozygotes were then mated to gad/WldS double heterozygotes to produce gad homozygotes that were heterozygous for both WldS and YFP-H. For gad genotyping, tail genomic DNA was extracted at 3 weeks using the Nucleon II kit (Amersham Pharmacia), digested with PvuII, and Southern blotted. It was then hybridized with a 32P-labelled 764-bp probe generated by PCR from gad homozygous genomic DNA using primers 5′-ATCCAGGCGGCCCATGACTC-3′ and 5′-AGCTGCTTTGCAGAGAGCCA-3′. Positively hybridizing fragments indicative of the gad (0.75 kb) and wild-type (1.6 kb) alleles were then identified by autoradiography. To genotype for inheritance of the YFP-H transgene, the skin of a 1–2 mm ear punch at 21 days was pulled apart and fluorescent axons identified using a Zeiss Axiovert S100 inverted fluorescent microscope through the FITC filter.

Assessment of Wallerian degeneration

gad homozygotes that were heterozygous for WldS and hemizygous for the YFP-H transgene were anaesthetized prior to the onset of hindlimb weakness using intraperitoneal Ketanest (100 mg/kg; Parke Davis/Pfizer, Karlsruhe, Germany) and Rompun (5 mg/kg; Bayer, Leverkusen, Germany). The right sciatic nerve (upper thigh) was transected and the wound closed with a single suture. Five days later the mice were killed by cervical dislocation, the swollen first 2 mm of distal sciatic nerve was discarded and the next 2 mm was used for western blotting for heavy neurofilament protein as previously described (Mack et al., 2001). The tibial nerve of the operated leg with a minimum of attached non-nervous tissue was processed for YFP fluorescence as follows. The nerve was stretched by ∼10% by pinning onto a Sylgard (Du Pont) dish and fixed with 4% paraformaldehyde (PFA) (BDH Laboratory, UK) in 0.1 M phosphate-buffered saline (PBS) in the dark for 1 h. It was then incubated in 1% Triton X-100 (Sigma, Germany) in 0.1 M PBS for 10 min and washed three times with PBS before mounting in Vectashield (Vector Laboratories, USA). The degree of fragmentation of the representative subset of motor and sensory axons that are YFP-labelled was determined. For more detail, see Beirowski et al. (2004).

Preparation of gracile tract sections

Mice aged 126–130 days were anaesthetized using Ketanest and Rompun (100 mg/kg and 5 mg/kg intraperitoneally, respectively) or a higher dose as required for deep terminal anaesthesia. After sternotomy mice were killed by cardiac puncture and instantly intracardially perfused first with a solution containing 10 000 IE/l heparin (Liquemin N 25000; Hoffmann-La Roche) and 1% procainhydrochloride in 0.1 M PBS for 30 s and then with fixative (4% paraformaldehyde in 0.1 M PBS) for 10 min. Brain and spinal cord were carefully removed, further fixed in 4% PFA/0.1 M PBS overnight and extensively washed in 0.1 M PBS. Fixed tissues were extensively rinsed in fresh 0.1 M PBS, dehydrated in an ascending ethanol series and subsequently embedded in paraffin (Paraplast; Sherwood Medical Co., St Louis, MO, USA) applying standard histology techniques. Coronal serial sections (6 µm) were made using a Type HM355 microtome (Microm GmbH) from the entire gracile nucleus in medulla oblongata and cervical gracile fascicle starting at level 535 (Sidman et al., 1971). Serial paraffin sections were mounted on conventional glass slides for use in haematoxylin and eosin (H & E) staining or on poly-l-lysine-coated slides for use in Luxol Fast Blue staining and immunocytochemistry, alternating normally every 2–3 sections. Distinction between gracile nucleus and cervical gracile fascicle was made by applying histomorphological criteria for the typical shapes of coronal sections.

H & E staining and spheroid quantification

Six-micrometre sections were deparaffinized in xylol (Carl-Roth, Germany) for 10 min, rehydrated in a descending ethanol series and rinsed in deionized H2O for 1 min. Sections were placed in haematoxylin for 5 min, rinsed in tap water for 1 min to allow stain to develop and then placed in eosin for 2 min, dehydrated and mounted in Entellan resin (Merck, Germany). The occurrence of clearly detectable eosinophilic spheroids, indicative of dystrophic axons (Yamazaki et al., 1988; Mukoyama et al., 1989; Kikuchi et al., 1990) was quantified in ∼90 sections uniformly dispersed throughout the gracile nucleus of each individual and ∼30 sections uniformly dispersed throughout the cervical gracile fascicle. Analysis of lateral columns was performed on these same 30 sections, counting the sum of spheroid numbers on both sides of the spinal cord. In this way, irregular results due to local deviations in spheroid numbers could be ruled out. H & E stained axonal spheroids were generally eosinophilic and appeared glassy or hyaline with a round or oval shape. They varied in diameter (5–50 µm) and sometimes reached a size larger than the nerve cells in gracile nucleus. All specimens were scored blind and agreed by two independent investigators.

Luxol Fast Blue staining and densitometric quantification

Six-micrometre sections from equivalent points in gad and gad/ WldS cervical spinal cord and medulla oblongata were processed simultaneously as follows. Sections were deparaffinized in xylol (Carl-Roth, Germany) for 15 min, and processed twice through 100% ethanol for 2 min and 96% ethanol for a few seconds. Slides were transferred to Luxol Fast Blue solution [0.1% Luxol Fast Blue MBS chroma (Merck), 10% acetic acid all made up in 96% ethanol] and incubated at 60°C for 5 h. Sections were then rinsed in 95% ethanol and distilled water for 1 min each, dipped in 0.05% lithium carbonate (Merck) for 1 min, and differentiated in 70% ethanol for a further 1 min. After rinsing in distilled water, sections were examined under light microscope for suitable differentiation between white and grey matter. Nuclear Fast Red staining was carried out for 10 min in 5% aluminium sulphate, 0.1% Nuclear Fast Red followed by rinsing in distilled H2O, 90% ethanol and 100% ethanol for 1 min each. Slides were incubated in xylol for 5 min and mounted in Entellan resin (Merck). Slides were examined under light microscopy (Nikon Eclipse E200) and evaluated using Bioscan OPTIMAS 6.0 software (Optimas Corp., WA, USA) according to the manufacturer's instructions. For densitometric quantitation, mean grey values were obtained for circumscribed areas of interest using a three-chip monochrome CCD camera, and the background grey value (tissue-free area) was subtracted. Since demyelination occurs selectively in the gracile tract and not in the cuneate tract of gad mice by 126–130 days (Mukoyama et al., 1989; our observations), we used cuneate fascicle as a reference area and expressed Luxol Fast Blue staining in gracile tract as a percentage of that in cuneate tract. We applied this procedure to representative Luxol Fast Blue-stained sections of cranial gracile tract: two sections from level C2/C3 representing the cervical gracile fascicle and two sections from level 535 representing the gracile nucleus (Sidman et al., 1971).

Immunocytochemistry of gracile tract

Six-micrometre paraffin sections from equivalent points in gad and gad/WldS cervical spinal cord and medulla oblongata were processed simultaneously as follows. Sections were deparaffinized, rehydrated in a descending ethanol series, washed several times in 0.05 M Tris-buffered saline (TBS), and treated with a solution of 6% H2O2 in methanol for 20 min to block endogenous peroxidase activity. They were then permeabilized with 0.1% Triton X-100 (Sigma) in 0.05 M TBS additionally containing 0.05 M NH4Cl, rinsed in fresh TBS three times and subsequently immuno-blocked with 5% bovine serum albumin (Sigma) in 0.05 M TBS for 1 h. First antibody was polyclonal guinea pig anti-glial fibrillary acidic protein (GFAP) (1 : 400 dilution) (Progen, Germany) at 4°C overnight, while negative control sections were incubated without primary antibody. Secondary antibody was goat anti-guinea pig biotin conjugate (1 : 400 dilution) (Sigma) for 1 h at room temperature, and was followed by streptavidin-coupled horseradish peroxidase complex (Vector Laboratories; 1 : 200 dilution) for 1 h. After extensive washing, sections were developed under identical conditions for all specimens with 3,3-diaminobenzidine tetrahydrochloride (Sigma–Aldrich) in 0.1 M phosphate buffer until a clear dark-brown labelling of astrocytes in the gracile tract could be detected. In all cases the control sections without primary antibody incubation showed no labelling of astrocytes. For microscopic examination and TV densitometry, sections were dehydrated and mounted in Entellan resin (Merck). Quantitation was similar to that described for Luxol Fast Blue densitometry. GFAP immunostaining intensities in cranial gracile tract sections were expressed as percentage of GFAP staining intensity in wild-type sections at the same coronal level. We applied GFAP densitometry on representative cranial gracile tract sections from each examined mouse: two sections from level C2/C3 representing the cervical gracile fascicle and two sections from level 535 representing the gracile nucleus (Sidman et al., 1971).

Immunocytochemistry of sciatic nerves

Sciatic nerves from 15-week-old gad, gad/WldS, or control mice were immersion fixed in 4% PFA/0.1 M PBS for 1 h and washed extensively in 0.1 M PBS before paraffin embedding. Twenty-micrometre sections were immunostained using rabbit polyclonal antibody to ubiquitin (Sigma–Aldrich U5379) and Cy3-conjugated secondary antibody. Confocal images were obtained using a PerkinElmer UltraView LCI confocal microscope coupled to a Nikon Eclipse TE200 microscope, and processed using UltraView software (Perkin-Elmer Life Sciences Ltd, Cambridge, UK).

Statistical analysis of histopathology results

All data (axonal spheroid numbers, TV densitometry intensities) are presented as mean ± SD for the examined genotype groups. Data analysis was performed using PRISM for Macintosh or SPSS for Windows, including Student's t-test calculations for paired and unpaired data where appropriate. Significance was considered at P < 0.05 and high significance at a P < 0.01.

Analysis of neuromuscular pathology

Mice were killed by cervical dislocation and lumbrical muscles immediately dissected under oxygenated Ringer solution. Fixation, immunocytochemistry and signal imaging were then carried out as described previously (Gillingwater et al., 2002). The denervation rate was determined by counting 100–200 endplates in each of two to three lumbrical muscles and the mean value taken for each mouse.

Behavioural tests

The foot splay test (Norreel et al., 2001) was used to estimate the reflex reaction speed of the hind limbs. Mice were gently taken by the neck and tail, the plantar surface of their hind feet painted using a non-toxic children's painting set, and the mouse released from a height of 15 cm to land on white paper. Wild-type mice bring their legs together during descent to land in a controlled manner like a gymnast, whereas gad mice fail to do this and land with their feet far apart. The distance between the two hind heels was averaged from 10 successive trials on each testing date (9 and 13 weeks).

In the clasping test, the mouse was suspended by the tail >50 cm from any surface. Clasping time within a 1 min test was scored as flexing or folding of the hind limbs tightly towards the trunk plus any spasmodic stretching. Mice were examined once per week through the period from 6 to 16 weeks. No wild-type mice clasped, regardless of the presence of the WldS mutation.


gad does not weaken the WldS phenotype

Before assessing the effect of WldS on gad pathology we first showed that WldS can protect axons, even in the presence of the gad mutation, by inducing Wallerian degeneration in gad/WldS mice. Before the lesion, there was no sign of axon degeneration in these nerves, confirming previous reports (Mukoyama et al., 1989). We bred gad mice that were heterozygous for WldS and hemizygous for a YFP-H transgene (Feng et al., 2000) to allow a rapid and quantitative assessment of Wallerian degeneration (Beirowski et al., 2004) and transected sciatic nerves before the onset of hindlimb weakness. Wallerian degeneration was assessed after 5 days both by western blotting to see degraded heavy neurofilament protein (Fig. 1A) and by fluorescence microscopy to see fragmented YFP-containing axons (Fig. 1B). Nerves unprotected by WldS degenerated as expected (Fig. 1A, middle lane, and Fig. 1B, lower panel) but a single allele of WldS was sufficient to prevent axon degeneration in both readout methods. Thus, gad does not significantly weaken the WldS phenotype and it is feasible to test the effect of WldS on gad pathology.

Fig. 1

A single allele of WldS is sufficient to delay Wallerian degeneration even in gad mice. (A) Western blot showing complete preservation of intact heavy neurofilament protein (NF200) in the distal stump of axotomized gad sciatic nerve by heterozygous WldS 5 days after lesion (lane 1). In contrast, no intact NF200 remains after 5 days in axotomized wild-type sciatic nerve (lane 2). Lane 3 is an uncut nerve showing the expected appearance of intact NF200 (gel loading differences probably account for the difference in intensity with lane 1). (B) Complete preservation of distal gad tibial nerve by heterozygous WldS 5 days after nerve lesion (upper panel), visualized using the YFP-H transgene. In contrast, no unfragmented axons remained in a tibial nerve lacking WldS 3 days after a lesion (lower panel). Unlesioned nerves appear exactly as in the upper image (Beirowski et al., 2004). Scale bar = 100 µm.

Axonal spheroid pathology is reduced by WldS

In order to determine the effectiveness of WldS on gad axonal spheroid pathology, we counted axonal spheroids in ∼90 H & E stained 6-µm paraffin sections from throughout the gracile nucleus and 30 sections from throughout the cervical spinal cord of each 18-week-old gad mouse and gad/WldS double homozygote. Fifty per cent fewer spheroids were found in gracile nuclei of gad/WldS mice than in gad mice (P = 0.0004) and 63% fewer in cervical gracile fascicle (P = 0.0011) (Fig. 2). Intermediate values were observed in WldS heterozygotes, further supporting the result and no spheroids were observed in control animals of this age (data not shown). Spheroids have also been reported in the cervical lateral columns of gad mice (Kikuchi et al., 1990). We found far fewer spheroids here than in cervical gracile tract and gracile nucleus, but the number was also significantly reduced by homozygous WldS (P = 0.046; n = 3) (Fig. 2). We also observed a reduction in axonal spheroids in lumbar spinal cord, from 42 to six in the ventral column and from 13 to four in the dorsal horn grey matter. Although lumbar regions of only a single gad and two gad/WldS mice were studied, these mice were independent of those used for the gracile tract analysis and 3 weeks younger, so these data independently support our conclusion that WldS reduces axonal spheroid pathology in several different regions of gad CNS well into late-stage disease.

Fig. 2

WldS reduces spheroid body numbers in the gracile tract and lateral columns of gad mice. (A and C) Representative sections from gracile nucleus of (A) gad and (C) gad/WldS mice stained with H & E, showing a large reduction in the number of axonal spheroids (large pink swellings, indicated by arrows) when WldS is present. (B and D) Representative sections from cervical gracile fascicle of (B) gad and (D) gad/WldS mice. Scale bar (AD) = 25 µm. (EG) Quantitation (mean ± SD) of spheroid counting data in (E) gracile nucleus (n = 6), (F) cervical gracile fascicle (n = 6) and (G) cervical lateral columns (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001.

A reduction in the number of axonal spheroids could result theoretically from either reduced axon pathology or pathology so extensive that the axons are completely destroyed. Kurihara et al. (2001) reported that when gad pathology was made worse by crossing with Uch-l3 null mice, extensive axon pathology became detectable at more caudal locations in cervical and thoracic gracile fascicle. We did not observe this in the WldS cross, and WldS homozygotes maintain a rostral–caudal gradient of axonal spheroid pathology (Fig. 2E and F; and thoracic data not shown), indicating that gad remains a ‘dying-back’ pathology in WldS mice but that its progress is delayed.

Secondary measures of axon pathology are also reduced by WldS

Further evidence of a reduced loss of axon-myelin units in gad/WldS mice came from a significant reduction (P = 0.018) in secondary myelin loss in cervical gracile fascicle in the same animals (Fig. 3A–C). A similar protective trend in the medulla oblongata did not reach statistical significance (P = 0.059), probably due to the naturally weaker myelination in this region, but WldS clearly did not cause any deterioration, so the reduction in axonal spheroid numbers (Fig. 2) must reflect reduced pathology and not wholesale axon loss. Furthermore, as the rescued axons remain myelinated, they potentially retain normal conductance properties, at least in these locations. It is unlikely that WldS has any direct effect on myelin, because expression of WldS in glia does not alter Wallerian degeneration (Glass et al., 1993). Thus reduced myelin loss in gad/WldS mice is likely to reflect the maintenance of functional axon-myelin units. WldS also decreased GFAP signal in immunocytochemistry in gad, indicating a lower level of astrocyte activation in response to axon damage (Yamazaki et al., 1988) (data not shown). Thus, both direct and indirect measures of spheroidal axon pathology in the gracile tract are reduced by the WldS gene.

Fig. 3

WldS reduces also secondary demyelination in the gracile tract. (A and B) Representative cervical gracile fascicles of (A) gad and (B) gad/WldS mice stained with Luxol Fast Blue and Nuclear Fast Red, showing the reduction in myelin loss when WldS is present. Scale bar (A and B) = 25 µm. (C) Densitometric quantification (mean ± SD) of Luxol Fast Blue staining (n = 5). *P < 0.05.

WldS operates downstream of axonal ubiquitin depletion in gad

gad causes axon degeneration through defective ubiquitin metabolism (Osaka et al., 2003), and WldS also interferes with ubiquitin metabolism (Mack et al., 2001; Coleman and Perry, 2002; Zhai et al., 2003). It was important to establish whether WldS blocks the ubiquitin defect in gad, an action that would suggest a protective effect restricted to gad and other ubiquitin defects, or whether it acts on a downstream step, raising the possibility of delaying axonal spheroid pathology in a wide range of CNS disorders (see above). Interpretation of any change in ubiquitin level in gracile tract would be complicated by the degeneration of those axon branches, so instead we carried out immunocytochemistry for ubiquitin epitopes in the peripheral branch of the same axons in sciatic nerve (Fig. 4). First, we confirmed that axonal ubiquitin was severely depleted in gad mice compared with wild-type controls (P = 0.014) (Osaka et al., 2003). We then found that a similar defect was present in gad/WldS mice compared with WldS controls (P = 0.0004) and that WldS did not significantly increase the ubiquitin signal either in the presence (P = 0.902) or absence (P = 0.807) of gad. Thus, WldS does not correct the depletion of axonal ubiquitin in gad and instead operates at a downstream point in spheroid pathology that could be common to other CNS disorders.

Fig. 4

WldS does not correct the severe depletion of axonal ubiquitin in gad. Ubiquitin immunostaining in both (A) wild-type and (B) WldS mice is greatly reduced in C and D, respectively, where gad is also present. Comparison of A with B and C with D also shows that WldS does not alter ubiquitin signal either in the presence of absence of gad. (E) Quantitation (mean ± SD) of ubiquitin signal. *P < 0.05; ***P < 0.001. 6 J, n = 4; WldS, n = 4; gad, n = 4; and gad/WldS, n = 10. Scale bar = 50 µm.

Motor pathology

Despite the reduction in axonal spheroids in the gracile tract, there was no apparent reduction in the severity of gad symptoms when WldS was present, with no significant difference in hindlimb clasping, (P = 0.82; n = 9) or splay test (P = 0.33; n = 7). Thus, either prevention of swelling in the gracile tract does not preserve the function of those axons, or pathology elsewhere limits any improvement in phenotype of gad/WldS mice. In the absence of any tests to specifically target the function of gracile tract axons, we investigated neuromuscular junction (NMJ) pathology, where dying-back of motor nerve terminals has previously been reported (Miura et al., 1993). At 15 weeks, the degree of denervation was similar between the two strains, with 56.0 ± 6.0% of lumbrical NMJ fully or partially denervated in gad mice and 53.5 ± 11.8% in gad/WldS (Fig. 5C and D). This may be because protection of motor nerve terminals at the NMJ by WldS after axotomy is weaker than that of the axon trunk, especially in older mice (Gillingwater et al., 2002). However, at 9 weeks, an age where WldS does protect axotomized motor nerve terminals, neither strain showed any denervation of NMJ in lumbrical muscles (Fig. 5A and B), so there was no time window when both WldS and gad exert their opposing effects at the NMJ. Thus, the fact that WldS does not alleviate NMJ pathology in the older mice could explain why gad symptoms are not reduced.

Fig. 5

Denervation at the NMJ. Presynaptic structures labelled with SV2 and neurofilament antibody are shown in green, and postsynaptic structures labelled with TRITC-α-bungarotoxin are in red. At 9 weeks, denervation has hardly begun in (A) gad or (B) gad/WldS. At 15 weeks, both strains show extensive denervation (C and D, respectively), with partial occupancy of endplates by motor nerve terminals occurring frequently. Scale bar = 25 µm.


We report that WldS reduces the occurrence of axonal spheroids in gad. This is the first indication that WldS can alleviate axon pathology in chronic CNS disease, thus extending observations made in the PNS that WldS protects axons not only after injury (Lunn et al., 1989) but also in disorders where no physical injury takes place (Wang et al., 2002; Ferri et al., 2003; Samsam et al., 2003). We conclude that axonal spheroid pathology in gad and Wallerian degeneration are not independent events and axon degeneration mechanisms are more uniform than morphology would suggest. It follows that Wallerian degeneration, or processes related to it, could contribute to many other CNS disorders where its involvement has not previously been suspected.

The mechanism by which WldS protects axons is still under investigation (Mack et al., 2001; Coleman and Perry, 2002; Zhai et al., 2003; Araki et al., 2004), but appears to involve nuclear WldS protein and a factor(s) that communicates its effect to the axon. What is already becoming clear, however, is that WldS directly or indirectly blocks a central step of axon pathology onto which various pathological mechanisms converge (Fig. 6). This is indicated both by the wide range of disorders in which WldS protects axons, as it is inconceivable that WldS blocks different initial events in each case, and by our direct evidence, that early steps of gad pathogenesis are unaltered (Fig. 4). Intriguingly, it now seems that a number of different pathological manifestations result from the step delayed by WldS. These are axonal spheroids in gad, dying-back axon loss without swelling in peripheral neuropathy and motor neuronopathy, and Wallerian degeneration in CNS and PNS injury. The divergent morphology and topology in these disorders previously suggested independent mechanisms, but the results of directly probing the mechanism using WldS challenge this interpretation.

Fig. 6

WldS delays a central step of axonal pathology that lies after the convergence point of multiple degenerative stimuli but upstream of the divergence of several pathological manifestations.

Many CNS disorders in which there is axonal swelling show accumulation of amyloid precursor protein in the swellings, indicating impairment of axonal transport in each case and suggesting that their axon degeneration mechanisms are to some extent related. gad is one of these disorders, and the others include brain trauma (Gentleman et al., 1993), stroke (Dewar et al., 1999) and other forms of ischaemia (Hughes et al., 2003), multiple sclerosis (Ferguson et al., 1997), and HIV dementia (Medana and Esiri, 2003). This similarity with gad suggests that axon degeneration in other disorders may also be related to Wallerian degeneration, a possibility that should now be tested using WldS mice or, where appropriate, the newly generated WldS rat model (Adalbert et al., in press). However, it is unlikely that WldS will stop all forms of axonal swelling, as it appears unable to do so in Plp null mice (Edgar et al., 2004). Thus, it should be possible to categorize CNS axonal swelling disorders into those that are altered by WldS and those that are not. This will then enable disorders to be grouped together for mechanistic studies rather than focusing on each disorder in isolation.

It is important to consider the spatial and temporal relationship between axonal swelling and axonal breakdown in the light of our data. The lack of good methods for longitudinal imaging of CNS axons has made it difficult to determine whether spheroids first occur as terminal endbulbs of axons whose distal ends have degenerated, or as localized swellings on otherwise morphologically normal axons. Preliminary data from our laboratory using axons of gad/YFP-H mice (Adalbert and Coleman., unpublished) suggest that many spheroids in gad are not terminal endbulbs, at least in the early stages of the disease. Thus, one model to account for the effect of WldS in gad is that an ‘en passant’ spheroid is the first step in pathology, leading to degeneration of the distal axon due to the blockage of axonal transport, a process that fixes the spheroid as a terminal endbulb. In this model, WldS might block the Wallerian-like degeneration of the distal end for long enough to allow the spheroid to resolve and the axon to recover. Thus, our data suggest that WldS could be used to address the question of whether swollen axons can recover or whether they are destined, inevitably, to degenerate. In a wider context, this is an important issue in several CNS disorders where axonal spheroids occur, including brain trauma and multiple sclerosis (Cheng and Povlishock, 1988; Ferguson et al., 1997).

The above model assumes that Wallerian-like degeneration and axonal swelling in gad are separated in space and time, with one causing the other. Alternatively, the mechanism of the axonal swelling itself in gad may be related to that of Wallerian degeneration. In support of this model, there are a number of disorders in which CNS axons swell and PNS axons of the same animal degenerate by Wallerian-like degeneration without extensive swelling. In gad mice, this occurs even within the same cell, as gracile tract central projections of lumbar primary sensory neurons have spheroids, while peripheral muscle spindles degenerate without swelling (Oda et al., 1992). Similarly, amyotrophic lateral sclerosis (ALS) in humans (Tu et al., 1996; Takahashi et al., 1997), mice (Tu et al., 1996; Oosthuyse et al., 2001) and rats (Howland et al., 2002), together with tauopathy in mice (Lewis et al., 2000; Probst et al., 2000), all show axonal swelling in spinal cord and other CNS areas, but extensive ‘Wallerian-like’ degeneration without swelling in ventral roots and peripheral nerves. Even injury-induced Wallerian degeneration shows different morphology depending on experimental circumstances. For example, when injured gracile tract axons undergo Wallerian degeneration they swell to up to 10 times their normal diameter, quite unlike Wallerian degeneration in the PNS (George and Griffin, 1994). Thus, a number of observations support a direct mechanistic link between axonal swelling and Wallerian degeneration.

It is not yet clear how related mechanisms might cause swelling in spheroids but axon fragmentation in Wallerian degeneration. Cytoskeletal changes are common to both, so a loosening of cytoskeletal structure could cause disorganized cytoskeleton to accumulate in spheroids but to undergo rapid granular disintegration in Wallerian degeneration. Wallerian degeneration of injured gracile tract axons displays elements of both processes, possibly having an intermediate mechanism: like spheroids, these axons dilate considerably but, typical of Wallerian degeneration, they also rapidly lose their cytoskeletal proteins (George and Griffin, 1994). In traumatic brain injury, observation of Wallerian degeneration and spheroids in the same transverse thin section has been interpreted as degenerating axons having a more proximal spheroid that blocks axonal transport (Cheng and Povlishock, 1988). In view of our findings, an additional explanation needs to be considered, that spheroids and Wallerian degeneration are alternative responses of different axons to the same lesion. Methods for real-time or long-range longitudinal analysis of individual spheroid-containing axons are required to resolve this, similar to new methods already applicable in PNS axons (Pan et al., 2003; Beirowski et al., 2004). What determines whether an axon develops a spheroid or undergoes Wallerian degeneration? Possible explanations include the different glial and haematopoietic cell content of the CNS and the lower rate of axonal transport there (Wujek and Lasek, 1983), but injury type may also be important. Finally, since the discovery of the WldS mouse, Wallerian degeneration is no longer considered a passive wasting of distal axons but a regulated self-destruction programme (Buckmaster et al., 1995; Raff et al., 2002). The reduction of axonal spheroid pathology in gad by the same gene raises similar questions: rather than being a passive consequence of blocked axonal transport axonal swelling could be, like Wallerian degeneration, a programmed response to axon damage.

Altered ubiquitin metabolism plays important roles in neurodegenerative diseases of the CNS. Genetic mutations in Parkinson's disease include an E3 ligase (Kitada et al., 1998) and possibly UCH-L1, the human homologue of the gene mutated in gad (Leroy et al., 1998). Ubiquitin-positive inclusions and other evidence indicate abnormal ubiquitylation in Alzheimer's disease (Mori et al., 1987; van Leeuwen et al., 1998), polyglutamine disorders (DiFiglia et al., 1997; Cummings et al., 1999; Bence et al., 2001) and ALS (Tu et al., 1996; Bruijn et al., 1997). Axons and synapses are particularly vulnerable, as proteasome inhibitors cause specific degeneration of distal neurites (Laser et al., 2003) and ubiquitin-related mutations alter synapse growth (DiAntonio et al., 2001) and stability (Wilson et al., 2002). As WldS can counter a downstream effect of defective ubiquitin metabolism, it now becomes important to study its effects on the above disorders.

WldS did not alleviate the symptoms of gad mice. Unfortunately, methods do not currently exist to assess the function of gracile tract axons, so we cannot rule out the possibility that blocking spheroid formation did not preserve axon function. However, it is likely that continued neuromuscular pathology in gad/WldS mice also contributes to the symptoms. These mice suffered extensive synapse loss by 15 weeks (Fig. 5), whereas axon pathology was still strongly reduced 3 weeks later (Fig. 2). This supports the hypothesis that different mechanisms underlie synaptic and axonal degeneration, with WldS affording only limited protection to synapses, particularly in older mice (Gillingwater and Ribchester, 2001; Gillingwater et al., 2002). Similarly, the synapses of gracile tract axons may have been lost even when those axons are preserved. Our data suggest that synapse pathology is a limiting factor when axons are protected by WldS, a finding likely to be important in other models (Ferri et al., 2003; Samsam et al., 2003).

In summary, we conclude that WldS alleviates chronic CNS axon pathology in gad mice and that formation of distal axonal spheroids in this disease shares features with Wallerian degeneration and ‘dying-back’ axon loss without spheroids. The effect of WldS on other CNS disorders with ubiquitylation deficits and CNS axonal swelling disorders should now be studied. Finally, our data emphasize the importance of finding a way to protect synapses as strongly as WldS protects axons.


We thank Professor Tateki Kikuchi for advice on gad pathology, Professor Rudolf Martini (University of Würzburg), Dr Mohtashem Samsam (University of Würzburg and Saba University School of Medicine), Dr Till G. A. Mack (Key Neurotek, Magdeburg, Germany), Dr Martin Bootman (The Babraham Institute, Cambridge) and Ms Jolanta Kozlowski (University of Cologne) for helpful discussion and technical advice. This work was supported by the Federal Ministry of Education and Research (FKZ: 01 KS 9502) and Center for Molecular Medicine, University of Cologne (CMMC) (to W.M., B.B., R.A., D.W., D.G. and M.P.C.), the Wellcome Trust (to T.H.G., plus Biomedical Collaboration Grant to R.R.R. and M.P.C), the Biotechnology and Biological Sciences Research Council (M.P.C., R.A., L.C.), ALSA (R.A.), the Koeln Fortune Programme (B.B.), the Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan (K.W.) and from the Ministry of Health, Labour and Welfare of Japan (K.W.).


  • * W. Mi and B. Beirowski contributed equally to this work


View Abstract