OUP user menu

Abnormally phosphorylated tau is associated with neuronal and axonal loss in experimental autoimmune encephalomyelitis and multiple sclerosis

J. M. Anderson, D. W. Hampton, R. Patani, G. Pryce, R. A. Crowther, R. Reynolds, R. J. M. Franklin, G. Giovannoni, D. A. S. Compston, D. Baker, M. G. Spillantini, S. Chandran
DOI: http://dx.doi.org/10.1093/brain/awn119 1736-1748 First published online: 21 June 2008


The pathological correlate of clinical disability and progression in multiple sclerosis is neuronal and axonal loss; however, the underlying mechanisms are unknown. Abnormal phosphorylation of tau is a common feature of some neurodegenerative disorders, such as Alzheimer's disease. We investigated the presence of tau hyperphosphorylation and its relationship with neuronal and axonal loss in chronic experimental autoimmune encephalomyelitis (CEAE) and in brain samples from patients with secondary progressive multiple sclerosis. We report the novel finding of abnormal tau phosphorylation in CEAE. We further show that accumulation of insoluble tau is associated with both neuronal and axonal loss that correlates with progression from relapsing–remitting to chronic stages of EAE. Significantly, analysis of secondary progressive multiple sclerosis brain tissue also revealed abnormally phosphorylated tau and the formation of insoluble tau. Together, these observations provide the first evidence implicating abnormal tau in the neurodegenerative phase of tissue injury in experimental and human demyelinating disease.

  • tau
  • secondary progressive multiple sclerosis
  • experimental autoimmune encephalomyelitis
  • axonopathy
  • neuronal loss


Multiple sclerosis is the commonest cause of acquired neurological disability in young adults. Although the cause is unknown, it is well-established that an interplay of genetic and environmental factors results in a multi-focal and multi-phasic disease defined histologically by inflammatory demyelination, axonal injury, astrocytosis and varying degrees of remyelination (Compston et al., 2006). The most common form of the disease has two distinct clinical phases, reflecting inter-related pathological processes: inflammation drives activity during the relapse–remitting stage, whereas neuronal and axonal degeneration represents the principal substrate for progressive disability seen in the secondary progressive stage (Davie et al., 1995; De Stefano et al., 1998; Wegner et al., 2006). The cause of progressive axonal loss is unknown, although accumulating disability is known to occur in the absence of significant inflammatory activity (Coles et al., 1999). Axonal swellings and accumulation of amyloid precursor protein have been observed in MS lesions suggesting that disruption of axonal transport may contribute to axonal pathology (Ferguson et al., 1997; Trapp et al., 1998; Shriver and Dittel, 2006). Altered axonal transport is known to play a role in a number of neurodegenerative conditions, such as Alzheimer's disease and hereditary fronto-temporal dementias, characterized by hyperphosphorylation and aggregation of tau (Mandelkow et al., 2003; Stokin et al., 2005; Goedert and Spillantini, 2006). Tau is a microtubule-associated protein involved in the assembly and stabilization of microtubules that are necessary for axonal transport. Abnormal phosphorylation of tau leads to decreased microtubule stability and can result in the formation of potentially neurotoxic aggregates (Allen et al., 2002; Johnson and Stoothoff, 2004; Bandyopadhyay et al., 2007).

Recent studies highlight the potential value of chronic-relapsing experimental autoimmune encephalomyelitis (CREAE) as a quantifiable histological and behavioural model that mirrors some aspects of progressive multiple sclerosis (Kornek et al., 2000; Pryce et al., 2005; Papadopoulos et al., 2006). In Biozzi ABH mice, repeated neurological insults occur in CREAE with a relatively predictable time course, starting with a relapsing–remitting phase (RL/RM) and later progressing to a chronic, typically non-relapsing, stage (CEAE) that exhibits progressive disability (Baker et al., 1990; Petzold et al., 2003; Pryce et al., 2005). In addition to paralysis, CEAE animals accumulate further clinical signs, including tremor and spasticity, which are comparable with patients with secondary progressive multiple sclerosis (Pryce et al., 2005).

In this study, we show that tau is hyperphosphorylated in progressive CEAE and multiple sclerosis. Furthermore, neuronal and axonal loss is associated with abnormally phosphorylated and insoluble tau in post-inflammatory CEAE mice that exhibit progressive disability, Importantly, we also demonstrate abnormal tau phosphorylation and insoluble tau formation in cerebral tissue from post-mortem cases of secondary progressive multiple sclerosis.

Material and Methods

Animals and surgery

All procedures were performed in compliance with national and institutional guidelines [UK Animals (Scientific Procedures) Act 1986 and the University of Cambridge and London Animal Care Committees].

Induction of chronic relapsing EAE (CREAE)

Biozzi ABH mice were inoculated as previously described (Baker et al., 1990). Briefly, spinal cord homogenate, emulsified in complete Freund's adjuvant, was injected subcutaneously in both hind-flanks and repeated 7 days later. Mice were monitored daily and scored according to motor disability and weight loss (typical at the onset of a relapse), as previously documented (Baker et al., 1990; Pryce et al., 2005). Mice were sacrificed during the relapsing–remitting phase at day 40–50 post-inoculation whilst in remission (typically after two attacks, referred to as RM2EAE) or during the chronic progressive phase at least 65 days post-inoculation (typically after four attacks, referred to as CEAE).

Immunohistochemistry and histology

Animals for immunohistochemistry were perfused with 4% paraformaldehyde/PB at the stated time-points, with age-matched control ABH mice (at least four animals per group). The brain and individual cervical spinal cord segments were dissected, post-fixed for 4 h and cryoprotected. Sagitttal cerebral and parasagittal spinal cord sections (16 μm) were processed by immunofluoresence as previously described (Hampton et al., 2007). Briefly, slides were washed in PBS and blocked for 1 h using 3% normal goat serum (NGS) in 0.2% Triton-X100/0.1M phosphate buffered saline (TX-PBS). Primary antibodies applied overnight in TX-PBS containing 1% NGS, were: mouse monoclonal antibody (mAb) anti-glial fibrillary acid protein (GFAP)-Cy3 conjugate (1 : 500; Sigma, Poole, UK), biotinylated mouse mAb specific for Neuronal Nuclei (NeuN) antigen (1 : 100; Chemicon, Chandlers Ford, UK), rat mAb specific for the macrophage/microglial F4/80 antigen (1 : 200; AbD Serotec, Kidlington, UK), rat polyclonal antibody specific for myelin basic protein (MBP) (1 : 100; AbD Serotec, UK), mouse mAb anti-tau phospho-dependent AT8 (1 : 1000; Ser202/Thr205) and rabbit polyclonal antibody specific for neurofilament 200 (NF200) antigen (1: 200; Sigma). Primary species antibody was detected with appropriate secondary antibodies (conjugated Alexa Ab: Molecular Probes/Invitrogen, Paisley, UK. 1 : 500) diluted in TX-PBS containing 1% NGS and Bis-benzamide (Sigma, 1 : 4000).

Sections for Luxol fast blue (LFB) staining were dehydrated before incubation in 50 : 50 chloroform–ethanol for 6 h. Sections were washed in ethanol and incubated in LFB solution [LFB-MBS 10% w/v, 95% ethanol (v/v), 87 mM glacial acetic acid] overnight at 40–55°C. Following water washes sections were differentiated in 6.8 mM lithium carbonate (Sigma) solution. The sections were counterstained with cresyl violet, dehydrated and placed under coverslips using DPX (RA Lamb, Eastbourne, UK).

Sections for modified Bielschowsky staining, which labels neuronal components, were washed in water and then incubated in pre-warmed 10% silver nitrate for 25 min at 40°C. Sections were then removed to permit addition of ammonium hydroxide to just clear the silver precipitate, after which they were returned to the silver ammoniacal solution and incubated for a further 45 min at 40°C. Sections were then placed in developing solution for 40–50 s until the stain was visualized. Staining was subsequently stopped by incubation in 1% ammonium hydroxide solution for 1 min. After washing, the sections were incubated in 5% sodium thiosulphate solution for 5 min, washed, dehydrated and placed under coverslips with DPX.

Semithin section processing

Animals for resin embedding (five per group) were perfused using 4% glutaraldehyde. The spinal cord was dissected and segments were post-fixed in 4% glutaraldehyde, before fixation with 2% osmium tetraoxide (Oxkem Limited, Reading, UK) overnight at 4°C. Tissue was dehydrated and embedded in resin (TAAB Labs, Aldermaston, UK). One micro metre semithin sections were cut on a Leica RM2065 ultra microtome and dried onto Polysine slides (VWR international, Lutterworth, UK). Sections were then stained with toluidine blue (5% in a Borax solution) and placed under coverslips with DPX.

Western blots

Animals were sacrificed and the spinal cord of control, RM2EAE and CEAE ABH mice were dissected. Tissue was immediately snap frozen. The cervical cord was used for biochemical analysis unless otherwise specified. Tissue was homogenized in a Tris-based saline extraction buffer containing: Mini Complete Protease Inhibitor (Roche, Burgess Hill, Sussex, UK) Phosphatase Inhibitor Cocktail I & II (Sigma), 100 mM sodium fluoride, 10 mM β-glycerophosphate and 2.5 mM sodium pyrophosphate. After centrifugation, the supernatant was harvested and the remaining pellet underwent further extraction using the above extraction buffer plus detergents (0.1% SDS, 1%NP40, 0.25% sodium deoxycholate). Protein concentration of the lysates was estimated using the Bicinchoninic acid (BCA) assay.

The presence of insoluble tau can be biochemically demonstrated by immunoblotting of sarcosyl insoluble protein extracts. Sarcosyl-insoluble tau preparation was performed as previously described (Goedert et al., 1992). Due to the volume of tissue required for an efficient extract, cervical cords of two animals were pooled for a single preparation (n = 3, i.e. six animals). Anterior forebrain was processed as individual animal extracts (n = 4, i.e. four CEAE animals). Tissue was homogenized in a 10× volume of high salt buffer of 10 mM Tris pH 7.4, 0.8 M sodium chloride, 1 mM EGTA, 10% sucrose (w/v). The extract was centrifuged at 45 000g for 30 min at 4°C, after which the supernatant was collected (S1), and the pellet exposed to a further extraction step with a 5× volume of the same buffer. The subsequent extract was centrifuged and the two supernatants pooled (S1 + S2). The supernatant fraction was incubated with 1% sarcosyl for 1 h and then centrifuged at 260 000g for 1 h at 4°C. The subsequent pellet represented the sarcosyl-insoluble tau preparation, potentially enriched for tau filaments. This was re-suspended in 50 mM Tris–HCl (pH 7.4), 0.8 μl/mg of starting tissue and stored at 4°C.

Samples were resolved by 10% sodium dodecyl sulphate (SDS) polyacrylamide gel electrophoresis (PAGE) and transferred to polyvinylidene difluoride (PVDF) membrane (Millipore, Watford, UK). Membranes were blocked in 5% bovine serum albumin (BSA), fraction IV (Sigma) and probed with either the phospho-independent mouse mAb Tau5 (1 : 1500, Calbiochem, Nottingham, UK;) or the following anti-tau phospho-dependent mouse mAbs: AT8 (1 : 1000; Ser202/Thr205), AT180 (1 : 1000; Thr 231/Ser 235), AT270 (1 : 1500; Thr 181), AT100 (dilution 1 : 1000; Ser 212/Thr 214)—all from Autogen Bioclear, Wiltshire, UK—and PHF-1 (1 : 2000; Ser396/Ser404; a kind gift from Dr P Davies, Albert Einstein College of Medicine, Bronx, NY). Membrane-bound primary antibody was probed with horse radish peroxidase (HRP)-conjugated, rabbit polyclonal anti-mouse immunoglobulin antibody (Dako, Ely, UK) and visualized with Amersham's basic or enhanced chemi-luminescence kits (Amersham, Buckinghamshire, UK).

Controls with secondary antibody alone or no antibody were performed for all blots and non-specific binding was negligible. In addition to the BCA, membranes were probed with mouse monoclonal β-actin antibody (1 : 10 000, Sigma), to control for the amount of protein loaded. All western blots were replicated three times.

Electron microscopy

Immunogold labelling of sarcosyl-insoluble material with AT8, MT1 (antiserum MT1 recognizes the three adult mouse tau isoform but fails to recognize human tau isoforms; a kind gift from Dr M. Goedert), BR134 (anti-serum BR134 recognizes the C terminus of both human and mouse tau), was performed as described previously (Goedert et al., 1992). Micrographs were recorded on a Philips EM208S Microscope at 40 000× magnification.

Human brain tissue

Samples of fronto-parietal cerebral tissue (4% paraformaldehyde fixed and snap frozen) were supplied by the UK Multiple Sclerosis Tissue Bank. This study was approved by the Multi Centre Research Ethics Committee (MREC). Human post-mortem material from seven pathologically confirmed cases of multiple sclerosis (four females and three males, aged 34–57 years, mean 43 years: disease duration 8–17 years, mean 12.9 years), known to have entered the secondary progressive phase in life, were studied. The cause of death in all cases was non-CNS sepsis. We also studied a rare acute relapsing–remitting female case, aged 32 years (2 years total disease duration). Fronto-parietal cerebral tissue from three female controls (aged 60–69 years, mean 64 years), with no known neuro-pathological disease, were also used. In addition, occipital cortex tissue from a patient with Alzheimer's disease was used as a positive control for abnormally phosphorylated insoluble tau. In the individuals with multiple sclerosis, the average time to post-mortem from death was 15.6 h, but 25.5 h in the controls.

Tissue used in this study was either 4% paraformaldehyde fixed or snap frozen, unfixed blocks. Snap-frozen sections (12 μm) were fixed in 90% ethanol or 4% paraformaldehyde for 10 min before use. Serial sections were immunohistochemically labelled with LN3 mAb [reactive with non-polymorphic human leucocyte antigen (HLA)-DR antigens on B cells, monocytes, macrophages and some activated T cells] and counterstained with LFB or modified Bielschowsky silver impregnation, in order to assess inflammation in association with demyelination and axonal pathology, respectively.

Sections requiring antigen retrieval (not used for LN3) were microwaved in 10 mM citrate buffer (pH6) (2 × 5 min at 600 W) after rehydration. Endogenous peroxidase activity was blocked using 3% hydrogen peroxide in 10% methanol and non-specific antibody binding blocked with 5% NGS. Primary antibodies used were: mouse mAb anti-human B lymphocyte (Clone LN3 : 1: 50; MP Biomedicals, Cambridge, UK), mouse mAb anti-human GFAP (1: 500; Covance, Harrogate, UK), mouse mAb anti-amyloid precursor protein (APP)-A4, (Clone 22C11; 1 : 100; Chemicon), mouse mAb anti-non-phopsphorylated neurofilament (Clone SMI32; 1 : 1000; Covance), aforementioned anti-tau phospho-dependent antibodies (AT8 and AT180 1 : 1000, PHF-1 1 : 1500) and NF200. Primary antibody binding was revealed using a biotinylated donkey anti-mouse IgG (1 : 200; Jackson Immuno Research Europe Ltd, Newmarket, UK) and a Vector Laboratories ABC Elite kit (Peterborough, UK). Labelling was visualized using a 3′3′-diaminobenzidine fast kit (Sigma). To confirm tau immunoreactivity specificity a further control was performed: the primary antibody mix was adsorbed against a small aliquot of insoluble tau extract prior to application (data not shown).

Snap-frozen tissue from the cases and controls was used for immunoblotting. Several white matter lesions were identified by systematic review of stained sections. This allowed a ‘lesion map’ to be created on the relevant tissue block enabling discrete lesions to be sampled.


Images were captured using Lucia or Leica software (Nikon, Kingston upon Thames, UK) via a digital camera. SigmaScan Pro (SPSS, Chicago, IL, USA) was used for GFAP and brain NeuN quantitative measurements. Four animals were used for each group. Immunopositive cell density measurements were made on overlays of transverse spinal cord segment images. This analysis used an automated thresholding procedure (Hampton et al., 2007) that measures the density of immunopositive objects independent of their individual intensities. For GFAP densitometric measurements, representative areas through the dorsal spinal tracts were sampled using a mask overlay of six 25 × 50 μm boxes. For NeuN analysis, irrespective of the density reading, it was possible to define how much of the area measured contained a positive pixel. In the brain, motor cortex regions M1 and M2 were analysed with a parallel 25 μm strip mask overlay, running through the lamina layers, 1, 1.5 and 2 mm lateral from the midline. Results were normalized by applying thresholds to the raw images prior to analysis.

Dorsal horn NeuN quantification was performed by formal cell counts on NeuN immunohistochemistry of C5 dorsal horns. Multiple counts (2–4 according to size of lamina) were performed at 60× magnification under a defined grid (160 μm × 16 μm) in each of the lamina, with the observer blinded to the animal group. Both dorsal horns were analysed using a minimum of three sections per animal. Four animals were analysed in the control and RM2EAE group and five in the CEAE group. Data was presented as number of NeuN-positive cells per 1 mm2 of tissue. Axonal counts on toluidine blue semithin sections were performed blinded. All healthy myelinated axons that dissected a grid line across dorsal spinal tracts (from surface to a depth of 480 μm) were counted.

Immunoblots were scanned and NIH image J software used to measure pixel intensity. The same CEAE cervical spinal cord lysate was used as a loading reference between independent immunoblots.

Statistical analysis was performed using SigmaStat 3.0 software. Unpaired two tailed t-tests were used to assess significance (P < 0.01) and linear regression analysis performed to measure association. Values expressed are the mean ± SEM, from a minimum of four animals, unless otherwise stated.


Demyelination, gliosis, neuronal and axonal loss in CEAE Biozzi-ABH mice

All studies were undertaken in post relapse-remission, chronic stage ABH mice (CEAE) that displayed progressive disability (Fig. 1A; day 70–90 post inoculation; Baker et al., 1990) unless specified otherwise. Analysis was restricted to the spinal cord, since earlier sampling showed that pathology was largely confined to this region of the nervous system (Baker et al., 1990). LFB staining of myelin was used to determine demyelination in RM2EAE and CEAE (Fig. 1D and E). Loss of myelin was most evident in CEAE throughout the white matter tracts of the spinal cord, including dorsal, lateral and anterior tracts. However, these multifocal lesions were variable except in the dorsal funiculus of the cervical spinal cord and so all further analysis was restricted to the dorsal funiculus. Reactive astrocytes were identified by positive GFAP staining (Fig. 1F–H). A significant increase in glial reactivity was present in CEAE compared to control sections and this was quantified by measuring the proportional density of GFAP immunoreactivity throughout the dorsal spinal tracts (98.67 ± 3.41% versus 52.56 ± 15.54%; P < 0.001; Fig. 1F and H). Microglia were immunolabelled with F4/80 antibody; an increased number were observed in the dorsal spinal tracts, particularly in RM2EAE compared with control (Fig. 1I–K).

Fig. 1

Biozzi ABH mouse CREAE is clinically characterized by an initial relapsing–remitting phase with subsequent progression to irreversible progressive disability with histopathological evidence of demyelination and gliosis. (A) Schema showing disease progression in CREAE with motor disability score of animals post induction of EAE (see ‘Material and Methods’ section; 0 = no deficit; 1 = tail paresis; 2 = impaired righting reflex; 3 = hind-limb paresis; 4 = hind-limb paralysis; 5 = moribund). Following the acute phase (AP) and relapses (RL), where disease episodes were associated with worsening in clinical score and associated weight loss, animals went into remission (RM) and were sacrificed during the post-relapse remission phase (RM2, day 40–50 post-inoculation) or the late chronic progressive disease stage (CEAE, after 65 days post-inoculation). Results represent the mean ± SEM for 11 animals. Histology and immunohistochemistry of spinal cord of CEAE and controls. (B and C) Myelin staining was identified using LFB in normal biozzi ABH mice. The boxed area in B highlights the dorsal funiculus and this is the region that the subsequent images focus on. (D and E) This white matter tract shows a lack of myelin staining (LFB), which is modest in RM2EAE (D) but marked in CEAE (E) C5 spinal cord. (F–H) GFAP immunohistochemistry of control, RM2EAE and CEAE C5 spinal cord shows increased immunoreactivity throughout the dorsal funiculus being most marked in CEAE (H) compared to control (F). (IK) F4/80 immunohistochemistry of C5 spinal cord, showing marked specific immunoreactivity in RM2EAE (J) compared to both CEAE (K) and control (I). Scale bars: (B) = 500 μm; (C–K) = 25 μm.

In order to evaluate neuronal and axonal loss, we quantified neurones (by NeuN immunohistochemistry) and axons in toluidine blue-stained semithin sections (Fig. 2). Neuronal quantification was performed throughout the C5 dorsal horn of CEAE spinal cord. Comparison with age-matched controls revealed a significant reduction in NeuN-labelled cells in CEAE spinal cord (31.04 ± 10.89% of that of controls; P < 0.001; Fig. 2A, C and D). Quantification of surviving axons in the C5 dorsal spinal tracts revealed significant loss in CEAE spinal cord (44.06 ± 6.59% of that in controls; P < 0.001; Fig. 2E, G and H).

Fig. 2

Neuronal and axonal loss increases with disease progression in CREAE. (A–C) NeuN immunohistochemistry of C5 spinal cord shows loss of NeuN staining within the dorsal horn in RM2EAE (B) and CEAE (C) compared to control (A) biozzi ABH mice. (D) Graph showing formal quantification by counts with significant loss of NeuN positive cells throughout Lamina I–IV of the dorsal horns in both the RM2EAE and CEAE phases, compared to control (mean NeuN cell count per 1 mm2 ± SEM shown; ***P < 0.001). See Supplementary Fig. 1 for NeuN quantification according to lamina. (E–G) Toluidine blue semithin stained section showing demyelination, gliosis and axonal loss in C5 spinal cord in RM2EAE (F) and CEAE (G) compared to control (E). (H) Graph showing significant decrease in the percentage of axonal survival in CEAE compared to both control and RM2EAE (mean percentage of axons ± SEM shown; ***P < 0.001). Scale bars: AC = 200 μm; EG = 50 μm.

Characterization of the temporal evolution of neuronal loss was next undertaken by analysing the cervical spinal cord during the relapse–remitting (RM2EAE) phase in ABH mice (day 40 post-inoculation). NeuN immunohistochemical analysis of the C5 dorsal horn at this stage showed significant neuronal loss in the RM2EAE group compared with age-matched controls (54.15 ± 11.94% of that of controls; P < 0.001; Fig. 2B and D). No significant axonal loss was present in RM2EAE cervical spinal cord (Fig. 2F and H).

Interestingly, when the NeuN counts were separated into their relevant spinal cord lamina (LI–IV were analysed) the neuronal loss based on NeuN counts was only significant throughout the first two lamina in RM2EAE phase mice (LI 40.21 ± 8.57% and LII 48.07 ± 9.96% of that of controls; P < 0.001) whereas in CEAE the NeuN loss was present throughout all four lamina (LI 23.81 ± 9.51%, LII 24.65 ± 10.50%, LIII 37.5 ± 11.03%, LIV 39.34 ± 12.26% of that of controls; P < 0.001; Supplementary Fig. 1).

Tau hyperphosporylation is associated with neuronal and axonal loss in CEAE

To investigate tau phosphorylation status in CEAE spinal cord, we performed immunohistochemistry using the phospho-dependent anti-tau antibody AT8. Confocal analysis showed that hyperphosphorylated tau colocalized with neurofilament detected in AT8-positive axons, and also in spheroids (Fig. 3A–F). Additional immunostaining for axonal amyloid precursor protein (APP)—which accumulates during axonal transport disruption and is a marker of axonal dysfunction—showed modest immunoreactivity within the spinal tracts (Supplementary Fig. 2), Immunoblot studies using the pan-tau antibody, Tau-5, indicated increased levels of tau in CEAE spinal cord compared with control tissue (Fig. 3G). To examine the characteristics of tau phosphorylation, replica blots were probed with a panel of antibodies against key phospho-dependent epitopes routinely used to assess tau hyperphosphorylation (AT8, AT180, AT270, AT100 and PHF-1) (Goedert et al., 1994). A substantial increase in immunoreactivity to all phospho-epitopes, especially AT8 and AT100, was demonstrated in lysates derived from CEAE compared with control animals (Fig. 3G).

Fig. 3

Hyperphosphorylated tau is present in axons and in a soluble fraction in CEAE. (A and B) AT8 (red) immunohistochemistry in longitudinal sections of CEAE thoracic spinal cord demonstrating focal areas of strong immunoreactivity (boxed area) with spheroid formation (arrow) in the dorsal white matter tract. Spheroid is distinct from cell bodies as revealed by DAPI (blue) counterstain. (C–F) Importantly, AT8 immunoreactivity colocalizes with NF200 immunopositive axonal compartment (CE; single channel images; F, merge with accompanying z plane images). Scale bars = 200 μm (A and B), 20 μm (CF). (G) Biochemical analysis of control and CEAE samples for soluble tau by western blot using phospho-dependent anti-tau antibodies AT8, AT180, AT270, AT100, PHF-1 and the phospho-independent pan anti-tau antibody Tau-5. β-actin confirms equal protein loading. Marked hyperphosphorylation of soluble tau in CEAE is observed at all phospho-epitopes.

Tau phosphorylation can result in disruption of microtubule stability and formation of insoluble tau aggregates. Analysis of RM2EAE and CEAE cervical spinal cord insoluble sarcosyl extract with AT8 antibody, revealed a strongly reactive band of 60 kDa and a weaker band of 64 kDa (Fig. 4A). A significant increase in insoluble hyperphosphorylated tau in CEAE was evident compared to RM2EAE (Fig. 4A: P < 0.005). This pattern of reactivity was uniform throughout the CEAE spinal cord (Supplementary Fig. 3). To confirm that insoluble tau is not a feature of all CNS regions in EAE, we next examined forebrain tissue from CEAE, since this is known to be comparatively spared in CEAE, consistent with our own finding of negligible neuronal loss (Supplementary Fig. 4) (Baker et al., 1990). Immunoblot analysis of insoluble sarcosyl extracts from CEAE forebrain revealed significantly less tau phosphorylation (Fig. 4B: 11.82-fold reduction compared to CEAE—cervical cord; P < 0.0001).

Fig. 4

Insoluble tau is regionally and temporally graded in CREAE. (A) Western blot analysis of sarcosyl-insoluble tau preparations from cervical spinal cord of control, RM2EAE and CEAE mice. AT8 immunoreactive insoluble tau is detected in CEAE and RM2EAE, but not in control samples. Positive control material derived from an Alzheimer's disease case corroborates strong immunoreactivity of the 60 kDa band and weaker reactivity of the 64 kDa band. Densitometric analysis revealed a significant increase of insoluble tau in CEAE samples compared to RM2EAE (P < 0.005). (B) Immunoblot analysis of sarcosyl-insoluble tau preparations from anterior forebrain (FB) shows a significant decrease in insoluble tau compared to cervical spinal cord (CSp) in CEAE (P < 0.0001). Densitometric analysis (A and B) represents mean ± SEM from three independent samples (six animals per group). (C) Graph showing a positive association between insoluble tau deposition and neuronal loss in cervical cord of RM2EAE and CEAE, with a linear regression analysis value of r2 = 0.8875.

To determine whether there was a relationship between the presence of insoluble tau and neuronal or axonal loss, we next compared levels of insoluble tau (semi-quantitative western blot analysis) with both neuronal and axonal quantification data from control, RM2EAE and CEAE cervical spinal cord. Samples were selected at specific time points. Linear regression analysis showed a clear association between accumulation of insoluble tau and the neuronal loss that accompanies progression of EAE (Fig 4C: r2 = 0.8875). In addition, there was an increase in insoluble tau between RM2EAE and CEAE which mirrored the axonal loss.

Tau phosphorylation in secondary progressive multiple sclerosis

We investigated whether chronic white matter lesions from cases with secondary progressive multiple sclerosis also display abnormal tau phosphorylation and formation of insoluble tau. Initial studies using post-mortem samples of fronto-parietal cerebral tissue confirmed the key pathological features of progressive multiple sclerosis. Active, chronic active or inactive lesions, plus shadow plaques were identified based on inflammatory cell infiltration (HLA-DR immunohistochemistry; LN3) and the extent of demyelination (LFB) (Fig. 5A–H). The majority of the lesions identified within the cohort and subsequently biochemically studied were chronically active. Gliosis, as revealed by GFAP-immunoreactive astrocytes, was prominent (Fig. 5P–R). Widespread axonal injury was confirmed by dystrophic axons and loss of neuronal elements on Bielschowsky staining (Fig. 5I and J). Numerous axons labelled with dephosphorylated neurofilament (SMI-32) showed axonal end-bulb accumulations or ‘axonal ovoids’ representing the terminal ends of transected axons (Fig. 5K–M; Trapp et al., 1998). Some partially demyelinated axons were noted to end in ovoids immunopositive for dephosphorylated neurofilament in their demyelinated sections (Fig. 5M). In addition, swellings and spheroids immunopositive for axonal amyloid precursor protein (APP) were abundant in active lesions and at the border of chronic active lesions (Fig. 5I–M).

Fig. 5

Immunohistochemical and histological characterization of multiple sclerosis tissue. (A–H) Characterization of secondary progressive multiple sclerosis lesions using LFB/LN3 immunohistochemistry and a modified Bielschowsky axonal stain. The four lesion types are demonstrated: (A and B) active—characterized by increased infiltration of inflammatory cells (HLA-DR/LN3) and a loss of myelin staining (LFB), associated with a decrease in axonal staining; (C and D) chronic-active—showed an expanding border of HLA-DR+ve inflammatory cells and a demyelinated core with significant reduction in axon staining; (E and F) chronic inactive—demyelinated throughout with fewer HLA-DR +ve inflammatory cells and marked loss of axons; (G and H) shadow plaques were also observed showing myelin pallor along with evidence of axonal disruption. Further characterization of chronic active lesions shows widespread axonal injury. (I and J) Axonal pathology occurs in the border, with evidence of significant loss at the centre of the plaque. Modified Bielschowsky staining (I and J) combined with SMI-32 (dephosphorylated neurofilament) (K–M) and APP (N and O) immunohistochemistry reveals disruption of axons and the presence of axonal swellings and spheroids. White arrows indicate axonal swellings (I, M and O); pink arrows indicate spheroids (I, K, L and N). SMI-32 labelled axons (green) also show regions denuded of myelin (MBP: red) that terminate in end-bulbs (K). (P–R) Gliosis is also an established feature within chronic lesions and GFAP immunohistochemistry (counterstained with LFB) reveals reactive enlarged multi-polar astrocytes (Q) at the borders of the lesion with dense fibrillary immunoreactivity within the lesion itself (R). Enlarged images (LM and QR) are high-power images of boxed regions shown in K and P, respectively. Scale bars = 500 μm (AH) and 25 μm (IL).

We next examined for the presence of hyperphosphorylated tau using phosphorylation-dependent anti-tau antibodies combined with immunohistochemical and western blot analysis. Immunohistochemistry of fronto-parietal cerebral tissue revealed labelling with PHF-1, AT180 and AT8 antibodies predominantly in diseased white matter with a predilection for the border zone of chronic active lesions (Fig. 6A–C, Supplementary Fig. 5). No positive labelling with these antibodies was present in control tissue (e.g. AT180, Fig. 6D; other data not shown). Furthermore, using confocal analysis, co-localization of phosphorylated tau (AT180 or PHF-1) with neurofilament (NF200)-positive spheroids confirmed that phosphorylated tau is present in neurons (Fig. 6F–H).

Fig. 6

Abnormal phosphorylated tau is present in secondary progressive multiple sclerosis. (A–D) Immunohistochemistry for phosphorylated tau using antibodies AT8/AT180/PHF-1 in fronto-parietal tissue from cases with multiple sclerosis reveals spheroid-like intensities not seen in age-matched controls [(D) control with AT180]. (F–H) AT180 (red) is also shown to colocalize with NF200 (green) in spheroids which compartmentalized to axons using confocal microscopy: [(H) merged images with accompanying z plane images]. Scale bars = 100 μm (AD); 20 μm (FH). (E) Western blot analysis was performed on white matter tissue lysates, from controls (Ctrl) and from several fronto-parietal white matter lesions (lanes A–G) from three representative cases with secondary progressive multiple sclerosis. Several phospho-dependent epitope antibodies, including AT8, AT100, AT180, AT270 and PHF-1, were used to show site-specific phosphorylation of tau in the lesions of multiple sclerosis. β-Actin confirmed equal protein loading. Significant phosphorylation is evident in two lesions (lanes C and F), with AT180- and PHF-1-directed epitopes being the most immunoreactive. Tau5 demonstrated the variability of total tau concentration within multiple sclerosis lesions. (I) Western blot of sarcosyl-insoluble extracts showing immunoreactivity against PHF-1, AT100 and Tau5. All six cases of secondary progressive multiple sclerosis studied were immunoblot positive. Representative samples from three separate cases are shown.

Tissue protein extracts from 19 of 24 lesions in fronto-parietal white matter, obtained from six cases, showed strong immunoreactive bands at 50–60 kDa of soluble hyperphosphorylated tau using the following anti-tau antibodies; AT8, AT180, AT270, AT100 and PHF-1 (representative immunoblots from three cases studied: Fig. 6E). Phosphorylation at these different phospho-epitopes was variable between lesions. Negligible phosphorylation was seen in control samples. Immunoblot using the pan anti-tau antibody Tau 5 demonstrated that, despite equal actin levels, there was either a marked reduction or an apparent increment in tau levels in lesion lysates compared with controls (Fig. 6E). Immunoblotting of insoluble sarcosyl protein extracts from cerebral white matter lesions was next performed to determine whether phosphorylation is associated with the formation of insoluble tau. Strong immunoreactive bands were present in insoluble extracts derived from six separate cases, but not from control samples (representative immunoblots from three cases studied: Fig. 6I). Whereas soluble extracts were immunoreactive to a broad spectrum of phospho-dependent antibodies (Fig. 6E), insoluble extracts had more limited immunoreactivity, preferentially to the PHF-1 and AT100 antibodies.

Sarcosyl insoluble preparations both from CEAE and multiple sclerosis tissue samples containing insoluble tau were next studied by electron microscopy. In neither instance were tau filaments observed although occasional accumulations of material with unclear specificity and structure did show immunogold labelling.


Tau is a microtubule stabilizing protein that is essential for efficient axonal transport (Johnson and Stoothoff, 2004). Abnormal tau phosphorylation and the formation of insoluble tau are defining features of common neurodegenerative diseases such as Alzheimer's and frontotemporal dementia. The pathological correlate of clinical disability and progression in multiple sclerosis is neuronal and axonal loss. Therefore, we hypothesized that tau pathology may also be important in the neurodegenerative phase of multiple sclerosis. We report the novel finding of soluble and specifically insoluble hyperphosphorylated tau both in chronic EAE and secondary progressive multiple sclerosis. We also show a clear temporal association between the accumulation of insoluble tau and the evolution of both neuronal and axonal loss in the progression of EAE.

Although reports of neuronal and axonal injury in multiple sclerosis date from the earliest pathological descriptions of the disease, it is only with the comparatively recent emergence of improved imaging techniques and more detailed histological analyses that axonal loss has been identified as the main pathological substrate for progressive disability (Davie et al., 1995; De Stefano et al., 1998; Kornek and Lassmann, 1999). Despite early documentation of cortical lesions (Brownell and Hughes, 1962), neuronal pathology has been comparatively understudied. An emerging body of evidence suggests that grey matter lesions are frequent and contribute to brain atrophy (Kidd et al., 1999; Peterson et al., 2001; Vercellino et al., 2005; Wegner et al., 2006). The relationship between neuronal and axonal degeneration is not clearly defined. However, indirect evidence supporting an association is suggested in a recent study of the anterior optic pathway in multiple sclerosis cases (Evangelou et al., 2001).

Improved understanding of the mechanisms underlying chronic neuronal and axonal loss in multiple sclerosis has been hampered by the relative absence of a suitable experimental model of secondary progression. Most experimental models of multiple sclerosis tend to mirror acute inflammation or primary progression (Gold et al., 2006). In contrast, ABH mouse CEAE models key aspects of secondary progression. The natural history is consistent with relapse–remission followed by a secondary phase associated with progressive clinical disability (Baker et al., 1990; Petzold et al., 2003; Pryce et al., 2005). Our findings of demyelination, gliosis, axonal and neuronal loss in ABH CEAE mice extend earlier studies and indicate that this model reproduces many of the pathological features of secondary progressive multiple sclerosis (Petzold et al., 2003). In addition, we show significant axonal loss in the dorsal spinal tracts including the ascending sensory fibres running in the dorsal funiculus in the chronic progressive phase of EAE. Recent reports of axonal loss in multiple sclerosis also identify significant axonal loss in the equivalent tracts of the cervical spinal cord (DeLuca et al., 2004). Neuronal loss is also evident in the cervical spinal cord and increases with disease progression from the relapse–remitting phase to chronic secondary progression (post relapse–remission phase). Whilst these observations document significant neuronal loss occurring earlier in the disease course than the axonal loss, this is likely to be a sampling effect rather than a reflection on the natural history. The axonal density analysis was restricted to the dorsal spinal columns, and therefore did not take into account any pathology in other axonal tracts which may have been differentially affected at RM2EAE. Further sub-analysis of neuronal loss according to spinal cord lamina demonstrated that NeuN cell loss in RM2EAE occurred predominantly in Lamina I and II of the dorsal horn whilst NeuN cell loss in CEAE was evident throughout all four lamina (Supplementary Fig. 1). Neurones in Lamina I and II are known to cross the mid-line and project to the upper cervical cord in the spinothalamic, spinoreticular and spinomesencephalic tracts (Willis and Coggeshall, 2004). These tracts run through the lateral spinal cord, not through the dorsal column–medial lemniscus pathway (DC–ML) and so the relevant axonal projections of these Lamina I–II neurones would not be identified in our analysis. However, the post-synaptic dorsal column pathway (which is comprised of tract cells resident in the dorsal horns of the spinal cord through lamina III–VI) does project to the dorsal column nuclei, through the DC–ML tracts. Therefore, at the later time-point of the CEAE mice, as the neuronal loss progresses so does the axonal loss that is now represented as loss observed through the dorsal funiculus. In contrast, negligible neuronal loss is observed in the anterior forebrain in CEAE.

Axonal injury in the context of acute inflammation is present early in the course of multiple sclerosis (Ferguson et al., 1997; Trapp et al., 1998; Bitsch et al., 2000). However, significant axonal loss in normal appearing white matter and ongoing loss in inflammatory quiescent chronic lesions suggest that additional mechanisms may contribute to chronic axonal pathology in multiple sclerosis (Evangelou et al., 2000; Kornek et al., 2000; Lovas et al., 2000). This is reinforced by the clinical observation of progressive disability in the absence of ongoing inflammation in humans (Coles et al., 1999) and animal models (Pryce et al., 2005) following immunomodulation strategies. The presence of axonal injury, including disruption of axonal transport, is well established from studies of active multiple sclerosis lesions and models of acute inflammatory demyelination (Ferguson et al., 1997; O’Neill et al., 1998; Trapp et al., 1998; Aboul-Enein et al., 2006; Shriver and Dittel, 2006). In addition, one study of hyper-acute EAE modelling acute inflammation, showed abnormal phosphorylation of soluble tau (Schneider et al., 2004). We focused on chronic lesions that do not display active inflammation and are dominated by gliosis, demyelination and neuronal injury. The direct contribution of demyelination to neuronal and axonal loss is uncertain, although increasing evidence suggests a role for oligodendrocyte-derived signals in maintaining neuronal and axonal health including modulation of axonal transport (Kornek et al., 2000; Lappe-Siefke et al., 2003; Wilkins et al., 2003; Yin et al., 2006).

Against this background, we first examined whether progressive neuronal degeneration in CEAE is associated with accumulation of soluble and insoluble hyperphosphorylated tau. It is well established that abnormal phosphorylation of tau can lead to neuronal injury through reduced microtubule stability and/or formation of neurotoxic protein aggregates (Gustke et al., 1992; Alonso et al., 1997). We demonstrated hyperphosphorylated tau localized within axons in CEAE. Soluble tau was found to be heavily phosphorylated at multiple epitopes using a panel of antibodies that detect proline-directed kinase phosphorylation sites (Goedert et al., 1994; Gustke et al., 1992). These same epitopes are known to be phosphorylated in degenerative tauopathies, characterized by a toxic gain of function and the formation of insoluble filamentous aggregates (Goedert et al., 1992; Forman et al., 2002; Zhukareva et al., 2002a, b; Goedert, 2004; Mott et al., 2005). Hyperphosphorylation of the soluble form causes tau to dissociate from the microtubule and increases the propensity for the formation of insoluble tau (Gustke et al., 1992; Alonso et al., 1997; Johnson and Stoothoff, 2004). In support of this, we report the presence of insoluble tau in the spinal cord of the ABH CEAE mouse. The graded accumulation of insoluble hyperphosphorylated tau in CEAE correlated positively with significant neurodegeneration. Importantly, this temporal trend of accumulation of insoluble tau parallels the development of progressive disability in CREAE (Baker et al., 1990; Pryce et al., 2005). Electron microscopy of preparations containing insoluble tau did not show tau filaments. Interestingly, occasional small amorphous aggregates of uncertain significance were observed. Insoluble, non-filamentous tau has been reported to cause axonopathy and muscle weakness in the single isoform human 4R tau transgenic mouse (Probst et al., 2000). Furthermore, mouse tau protein is not consistently present in filaments (Allen et al., 2002).

Previous studies investigating tau in multiple sclerosis have been restricted to cerebrospinal fluid analysis of total tau as a potential biomarker of neuronal injury (Kapaki et al., 2000; Teunissen et al., 2005). Our study is the first to implicate abnormal phosphorylation of tau in the progression of multiple sclerosis; fronto-parietal white matter plaques from six cases with established secondary progression all showed abnormal soluble tau phosphorylation. All cases were aged <57 years (mean age 51 years) making the presence of changes due to Alzheimer's disease highly improbable. The complex pattern of the axonal tau ‘phosphoprint’ in lesions, with variable phosphorylation at different epitopes, is perhaps not surprising given the heterogeneity of pathology observed in MS (Lassmann et al., 2001). Importantly, phosphorylated tau colocalized with neurofilament immunopositive spheroids in axons of chronic multiple sclerosis lesions.

Significantly, all six cases of secondary progressive multiple sclerosis displayed evidence of an insoluble fraction of tau within chronic lesions. One postulated mechanism of tau-mediated neurodegeneration involves formation of filamentous aggregates. Ultrastructural analysis revealed occasional accumulations of immunogold-labelled material but no filaments. It is unclear why the insoluble aggregates observed were not filamentous. However, our findings show that in the multiple sclerosis cases not all epitopes were phosphorylated—as they are in normal tauopathies—and this could contribute to different conformations. Emerging evidence from Drosophila melanogaster tauopathy models additionally suggests that abnormally phosphorylated tau may be neurotoxic in the absence of filaments (Wittmann et al., 2001; Jackson et al., 2002). Studies in the P301S tau transgenic mouse have also shown evidence for neuronal dysfunction, with synaptic loss and behavioural deficits that precede the emergence of insoluble filamentous tau pathology (Yoshiyama et al., 2007). Furthermore, a recent in vitro study has shown that aggregation of hyperphosphorylated tau reduces its ability to inhibit microtubule assembly, promoting sequestration of normal tau, and proposes that soluble hyperphosphorylated tau or tau oligomers are the more toxic species (Alonso et al., 2006).

Our finding of insoluble tau is unlikely to represent an invariant feature of neurodegeneration, given the absence of abnormal tau phosphorylation in other neurodegenerative conditions such as idiopathic Parkinson's disease (Trojanowski et al., 1998). Furthermore, the absence of insoluble tau in a rare acute active case of relapsing–remitting multiple sclerosis (data not shown) provides some support for the involvement of insoluble tau in the aetiopathogenesis of progression in multiple sclerosis. An interpretation of our observation is that there is an evolution of tau pathology from a soluble hyperphosphorylated state to an insoluble phosphorylated fraction that occurs with disease progression.

This study demonstrates a positive association between insoluble tau and both neuronal and axonal loss in the ABH mouse CREAE model of multiple sclerosis, and reports accumulation of insoluble tau in secondary progressive multiple sclerosis. Together, we report the first evidence implicating tau dysfunction in both experimental and human demyelinating disease. Our findings provide a platform for further evaluating the contribution of tau pathology to progression in multiple sclerosis.

Supplementary material

Supplementary material is available at Brain online.


These studies were supported by the Webb Trust Fund, Sir David Walker Trust Fund, Husky Foundation, Medical Research Council and MS Society of Great Britain and Northern Ireland and the National Multiple Sclerosis Society, USA. Tissue samples were supplied by the UK MS Tissue Bank, funded by the Multiple Sclerosis Society of Great Britain and Northern Ireland, registered charity 207495.


  • Abbreviations:
    chronic experimental autoimmune encephalomyelitis
    chronic relapsing experimental autoimmune encephalomyelitis
    Luxol fast blue
    monoclonal antibody
    normal goat serum
    phosphate buffer
    phosphate buffered saline
    second remission experimental autoimmune encephalomyelitis
    Triton-phosphate buffered saline


View Abstract