|Home | About | Journals | Submit | Contact Us | Français|
Tumour necrosis factor is linked to the pathophysiology of various neurodegenerative disorders including multiple sclerosis. Tumour necrosis factor exists in two biologically active forms, soluble and transmembrane. Here we show that selective inhibition of soluble tumour necrosis factor is therapeutic in experimental autoimmune encephalomyelitis. Treatment with XPro1595, a selective soluble tumour necrosis factor blocker, improves the clinical outcome, whereas non-selective inhibition of both forms of tumour necrosis factor with etanercept does not result in protection. The therapeutic effect of XPro1595 is associated with axon preservation and improved myelin compaction, paralleled by increased expression of axon-specific molecules (e.g. neurofilament-H) and reduced expression of non-phosphorylated neurofilament-H which is associated with axon damage. XPro1595-treated mice show significant remyelination accompanied by elevated expression of myelin-specific genes and increased numbers of oligodendrocyte precursors. Immunohistochemical characterization of tumour necrosis factor receptors in the spinal cord following experimental autoimmune encephalomyelitis shows tumour necrosis factor receptor 1 expression in neurons, oligodendrocytes and astrocytes, while tumour necrosis factor receptor 2 is localized in oligodendrocytes, oligodendrocyte precursors, astrocytes and macrophages/microglia. Importantly, a similar pattern of expression is found in post-mortem spinal cord of patients affected by progressive multiple sclerosis, suggesting that pharmacological modulation of tumour necrosis factor receptor signalling may represent an important target in affecting not only the course of mouse experimental autoimmune encephalomyelitis but human multiple sclerosis as well. Collectively, our data demonstrate that selective inhibition of soluble tumour necrosis factor improves recovery following experimental autoimmune encephalomyelitis, and that signalling mediated by transmembrane tumour necrosis factor is essential for axon and myelin preservation as well as remyelination, opening the possibility of a new avenue of treatment for multiple sclerosis.
Tumour necrosis factor (TNF) is a pleiotropic cytokine involved in the regulation of numerous physiological and pathological processes, such as inflammation, cancer, autoimmunity and infection. TNF exists in two biologically active forms, transmembrane TNF and soluble TNF. The cellular functions of TNF are mediated by TNF receptor 1 (TNFR1) and TNF receptor 2 (TNFR2), which differ in expression profiles, ligand affinity, cytoplasmic tail structure and downstream signalling pathway activation. TNFR1, unlike TNFR2, contains a death domain (DD) localized within the cytoplasmic segment. Soluble TNF and transmembrane TNF are capable of binding to both receptors, although with significantly different binding affinities. As a consequence, soluble TNF, which has higher affinity for TNFR1, primarily signals through TNFR1, mediating apoptosis and chronic inflammation (Holtmann and Neurath, 2004). Conversely transmembrane TNF, with higher affinity for TNFR2, preferentially signals through this receptor (Grell et al., 1995; Wajant et al., 2003), regulating gene programs important for cell survival, resolution of inflammation, maintenance of immunity to pathogens and myelination (Pasparakis et al., 1996; Arnett et al., 2001; Canault et al., 2004; Olleros et al., 2005, 2009; Alexopoulou et al., 2006; Ierna et al., 2009). It is therefore apparent that the two forms of TNF have specific and often opposing biological effects, which may explain the contrasting outcomes of TNF inhibition both in animal models of demyelinating diseases and human clinical trials. Indeed, non-selective anti-TNF biologics (e.g. infliximab, adalimumab, etanercept, certolizumab pegol and golimumab) are successfully used in treating TNF-mediated pathologies such as rheumatoid arthritis, psoriatic arthritis and Crohn's disease (Tracey et al., 2008); however, they are associated with serious adverse effects such as congestive heart failure, increased risk of infection and demyelination (Ramos-Casals et al., 2008; Fromont et al., 2009; Wallis, 2009; Sfikakis, 2010).
TNF is also linked to the development of various neurological/neurodegenerative disorders, including multiple sclerosis (McCoy and Tansey, 2008). Patients with multiple sclerosis manifest elevated concentrations of TNF in serum, CSF (Maimone et al., 1991; Sharief and Hentges, 1991; Spuler et al., 1996) and within active lesions (Hofman et al., 1989; Selmaj et al., 1991). Peaks of serum and CSF TNF are correlated with the presence of active MRI-detectable multiple sclerosis lesions (Spuler et al., 1996). Furthermore, overexpression of TNF in mice leads to demyelinating disease (Probert et al., 1995; Akassoglou et al., 1998; Dal Canto et al., 1999) and in animal models of multiple sclerosis, e.g. experimental autoimmune encephalomyelitis (EAE), TNF blockade prevents or ameliorates the pathology (Ruddle et al., 1990; Baker et al., 1994).
Such compelling evidence in human disease and animal models led, in 1997, to the first multiple sclerosis clinical trial with a non-selective TNF inhibitor, lenercept. Against all expectations, the trial had to be abruptly terminated due to dose-dependent increases in frequency, duration and severity of multiple sclerosis attacks (Lenercept-Group, 1999). The failure of the trial underscored the necessity for more in-depth studies to better understand the complex roles of TNF in multiple sclerosis pathology. Studies in animal models of multiple sclerosis showed that the absence of TNF or TNFR1/TNFR2 combined does not protect from the pathology, only delays it and induces exacerbation of chronic disease (Korner et al., 1997; Liu et al., 1998; Eugster et al., 1999; Kassiotis et al., 1999; Suvannavejh et al., 2000). TNFR1−/− mice, which lack the receptor preferentially activated by soluble TNF, are protected from EAE, show faster regression of the pathology and remyelinate normally (Eugster et al., 1999; Suvannavejh et al., 2000). Conversely, TNF−/− and TNFR2−/− mice (the latter lacking the receptor preferentially activated by transmembrane TNF) develop worse EAE and show significantly impaired remyelination (Liu et al., 1998; Eugster et al., 1999; Suvannavejh et al., 2000; Arnett et al., 2001). In an elegant study with knock-in mice expressing transmembrane TNF only, Alexopoulou et al. (2006) demonstrated that transmembrane TNF is sufficient to suppress both induction and chronic phases of EAE as well as anti-myelin autoimmune reactivity, while maintaining antimicrobial host defences (Alexopoulou et al., 2006). Collectively, these studies illustrate how the general understanding of TNF involvement in multiple sclerosis has profoundly evolved over the past decade: from the dogma that TNF, as a whole, is a key pathological mediator in multiple sclerosis, to recognizing that two distinct forms of TNF exist, soluble TNF and transmembrane TNF, and that multiple sclerosis pathology is associated with a detrimental effect of soluble TNF (Kassiotis and Kollias, 2001).
Given the contrasting and often confusing data in the literature, we sought to investigate the distinct roles of soluble TNF and transmembrane TNF in the development of EAE using a pharmacological approach, hypothesizing that if selective pharmacological inhibition of soluble TNF without interfering with transmembrane TNF proved efficacious in a model of multiple sclerosis, we could have a candidate for immediate translation into multiple sclerosis therapy. We employed a new biologic, XPro1595, which selectively inhibits soluble TNF without affecting transmembrane TNF. XPro1595 is the leading compound of a novel class of dominant-negative TNF biologics that, by rapidly exchanging subunits with native soluble TNF but not membrane-anchored transmembrane TNF, form inactive mixed heterotrimers eliminating soluble TNF activity without inhibiting transmembrane TNF (Steed et al., 2003; Zalevsky et al., 2007; Olleros et al., 2009). We compared the effects of XPro1595 to those of the non-selective TNF inhibitor etanercept [decoy TNFR2 which blocks soluble TNF, transmembrane TNF and lymphotoxin (Goffe and Cather, 2003)] and found that inhibition of soluble TNF by XPro1595 is therapeutic in myelin oligodendrocyte glycoprotein (MOG)-induced EAE, whereas non-selective inhibition of both soluble TNF and transmembrane TNF by etanercept is not. Mice treated with XPro1595 recover from paralysis and display significant axonal preservation and remyelination. Our studies underscore the opposing roles of soluble TNF and transmembrane TNF in the pathophysiology of EAE, and suggest that the protective effects of TNF in chronic disease are mediated by transmembrane TNF. We show that by inhibiting soluble TNF while preserving transmembrane TNF signalling, we can positively modulate two essential processes for maintaining and improving function in multiple sclerosis, neuroprotection and remyelination, not only opening a new avenue of therapy for the treatment of multiple sclerosis, but also helping to reconcile the apparently conflicting findings emerged from animal models and clinical trials (Caminero et al., 2011).
Experiments were performed according to the protocols approved by the Institutional Animal Care and Use Committee of the University of Miami. Female C57BL/6 mice 2- to 4-months-old were obtained from Jackson Laboratories. Animals were housed in a virus/antigen-free facility with a 12 h light/dark cycle, controlled temperature and humidity, and provided with water and food ad libitum.
Active EAE was induced with MOG(35–55) peptide purchased from BioSynthesis Inc., as previously described (Szalai et al., 2002; Brambilla et al., 2009). Clinical signs of EAE were assessed daily using a standard scale of 0–6 as follows: 0, no clinical signs; 1, loss of tail tone; 2, fully flaccid tail; 3, complete hind limb paralysis; 4, complete forelimb paralysis; 5, moribund; 6, dead. XPro1595 [formerly XENP1595 (Zalevsky et al., 2007), Xencor Inc.] and etanercept (Enbrel®, Amgen-Wyeth) were administered subcutaneously at a dose of 10 mg/kg every 3 days beginning at Day 16, when at least half of the mice had reached a clinical score ≥2.
Cells were isolated as previously described (Brambilla et al., 2009). Briefly, spinal cords and spleens were homogenized into single cell suspensions through a 70-µm mesh cell strainer. For isolation of infiltrating leucocytes into the spinal cord, spinal cord suspensions were spun at 400g for 5 min. After removal of supernatants, cells were collected, resuspended in 40% Percoll, layered on 70% Percoll and centrifuged at 400g for 25 min at room temperature. Cells at the gradient interface were removed, washed in fluorescence-activated cell sorting buffer (eBioscience) and stained as described below. For leucocyte isolation from the spleen, suspensions were spun at 400g for 5 min. After removal of supernatants, red blood cells were lysed in 1 ml lysis buffer (eBioscience) according to the manufacturer's instructions. Cells were then resuspended in fluorescence-activated cell sorting buffer and stained as described below. Spinal cords from seven to eight animals were pooled to obtain a sufficient number of cells for flow cytometry analysis. Spleens from individual animals were evaluated separately.
Cells were incubated on ice for 10 min with anti-CD16/32 (FcR block, eBioscience) to prevent non-specific staining and subsequently stained for 30 min at 4°C with: PE-Cy7-anti-CD45 (1:10 000); AlexaFluor488-anti-CD3, Pacific Blue- or PE-anti-CD4 (1:200); APC-anti-CD8 (1:200); APC-AlexaFluor750-anti-B220 (1:200); PE-anti-NK1.1 (1:200); and eFluor450-CD11b (1:200), from eBioscience. Cell suspensions were fixed overnight in 1% paraformaldehyde in fluorescence-activated cell sorting buffer and analysed with an LSR II flow cytometer equipped with FACSDiva 6.0 software (BD Biosciences).
Animals were perfused with 4% paraformaldehyde in 0.1 M phosphate-buffered saline. Tissues were cryoprotected in 0.1 M phosphate-buffered saline + 25% sucrose and cryostat-cut into 15 or 30 µm thick sections. After blocking for 1 h with 5% normal goat serum in 0.1 M phosphate-buffered saline + 0.4% Triton-X, sections were incubated overnight at 4°C with primary antibodies against GFAP (rabbit, 1:1000, Dako; mouse, 1:1000, BD Biosciences), CD45 (rat, 1:200, eBioscience), NG2 (rabbit, 1:500, Millipore), TNFR1 (rabbit, 1:50, or mouse, 1:50, Santa Cruz), TNFR2 (rabbit, 1:50, or mouse, 1:50, Santa Cruz), MAP2 (mouse, 1:1000, Sigma), CD11b (rat, 1:100, Serotec) and CC1 (mouse, 1:500, Calbiochem). Immunoreactivity was visualized either with secondary species-specific fluorescent antibodies (AlexaFluor-594 and -488, Invitrogen) or with 3-3′diaminobenzidine staining using the Vector Elite ABC kit (Vector Labs) according to the manufacturer's instructions. Images were obtained with an Olympus FluoView 1000 confocal microscope or with a Zeiss Axiovert 200 M fluorescence microscope.
Following immunolabelling with anti-NG2 antibody as described above, quantification of the total numbers of NG2+ cells was determined applying principles of stereology. Briefly, 10 serial sections taken at 150 µm intervals and with a thickness of 30 µm were analysed, and the cells of interest manually counted on each section at ×63 magnification. The total number of positive cells per cubic millimetre of tissue was estimated with Stereoinvestigator software (MicroBrightField, Inc.). For these assessments, six animals per group were used.
After perfusion, 1 mm segments of the thoracic spinal cord were fixed in 2% glutaraldehyde/100 mM sucrose and rinsed in 0.15 M phosphate buffer before post-fixing with 2% OsO4 for 1 h. Following dehydration in graded ethanol solutions, tissues were embedded in epoxy resin (Embed, Electron Microscopy Sciences). Semi-thin sections (1 μm thick) were obtained with a Leica Ultracut E microtome and stained with 1% toluidine blue solution. Samples were then examined by light microscopy or using a Philips CM-10 transmission electron microscope. The number of toluidine blue-stained myelinated axons was estimated using the software Stereoinvestigator. For this assessment, six animals per group were used.
For assessment of the g-ratio (axon diameter/fibre diameter) of axons in the thoracic spinal cord, electron microscopy micrographs were taken at a magnification of ×52 000 with a Philips CM-10 transmission electron microscope. A grid was placed over the section and pictures taken randomly from each quadrant corresponding to the lateral columns of the cord. One picture per quadrant was evaluated for a minimum of 30 images per mouse. On each micrograph, fibre diameter and axon diameter of each single axon were manually measured. Due to the irregular shape of the elements, the approximate diameter was calculated as the average of three diameters measured for each element. For this assessment, six animals per group were used.
Total RNA was extracted with TRIZOL® according to the manufacturer's instructions (Invitrogen). To ensure complete elimination of genomic DNA, RNA was further purified with RNeasy® MinElute Cleanup Kit (Qiagen) in combination with DNA digestion using RNase-free DNase (Qiagen). Reverse transcription was performed with Superscript® II (Invitrogen), according to the manufacturer's protocols. Complementary DNA equal to 10–50 ng of initial total RNA was used as a template in each polymerase chain reaction. Real-time polymerase chain reaction was performed in the Rotor-Gene 3000 Real Time Cycler (Corbett Life Science) with QuantiTect® SYBR® Green PCR MasterMix (Qiagen). Relative expression was calculated by comparison with a standard curve, after normalization to glyceraldehyde-3-phosphate dehydrogenase and β-actin gene expression. Primers for gene amplification are listed in Supplementary Table 3.
Proteins were resolved by sodium dodecyl sulphate–polyacrylamide gel electrophoresis on 6–15% gels, transferred to nitrocellulose and blocked in 5% non-fat milk. Membranes were probed with antibodies against: CD45 (rat, 1:1000, eBioscience); neurofilament-H (rabbit, 1:1000, Serotec); neurofilament-H non-phosphorylated (SMI32; mouse, 1:1000, Covance); GFAP (mouse, 1:1000, BD Pharmingen); glyceraldehyde-3-phosphate dehydrogenase (mouse, 1:2000, Imgenex), followed by horseradish peroxidase-conjugated secondary antibodies. Proteins were visualized with ECL (GE Healthcare/Amersham) and bands quantified with Quantity One® software (Biorad). Data were normalized against glyceraldehyde-3-phosphate dehydrogenase and expressed as per cent of vehicle.
The study was performed on post-mortem spinal cord tissue from two cases with secondary progressive multiple sclerosis and one case with primary progressive multiple sclerosis obtained from the UK Multiple Sclerosis Tissue Bank at Imperial College, London (see Supplementary Table 2 for demographic, clinical and neuropathological data). All multiple sclerosis tissues were obtained via a UK prospective donor scheme with full ethical approval (08/MRE09/31). One tissue block (4 cm2 surface area) was randomly selected for each case based on images taken at the time of dissection, without prior knowledge of the presence of white matter or grey matter lesions. Serial sections were then processed by immunohistochemistry in order to determine cell-specific expression of TNFR1 and TNFR2.
Air dried, acetone fixed, 10 µm-thick cryosections (cut from the 4% paraformaldehyde fixed tissue blocks) were rehydrated with phosphate-buffered saline and subjected to antigen retrieval procedure (microwave treatment in 10 mM citrate buffer, pH 6.0). Sections were then incubated for 30 min with 0.1% H2O2 in phosphate-buffered saline to eliminate endogenous peroxidase activity, for 1 h with 10% normal serum, and overnight at 4°C with primary antibodies against TNFR1 (mouse, 1:50; Serotec) or TNFR2 (rabbit, 1:50; Sigma) diluted in phosphate-buffered saline containing 0.2% Triton 100-X and 1% bovine serum albumin or 1% normal serum. Binding of biotinylated secondary antibodies (Jackson Immunoresearch Laboratories) was visualized with the avidin–biotin horseradish peroxidase complex (ABC Vectastain Elite kit, Vector Laboratories) followed by 3,3′-diaminobenzidine (Sigma) as substrate. All sections were counterstained with haematoxylin, sealed with DePeX polystyrene and viewed under a Zeiss Axiophot microscope equipped with a digital camera (Axiocam MRC). Images were acquired using Axiovision 4 AC software. Negative controls included IgG isotype antibodies or preimmune serum or omission of the primary antibody.
After post-fixation in cold acetone, sections were blocked with 5–10% normal serum in phosphate-buffered saline, and incubated overnight at 4°C with primary antibodies diluted in phosphate-buffered saline containing 0.2% Triton X-100 and 1% normal serum. Double-immunofluorescent labelling was conducted with anti-TNFR1 or anti-TNFR2 antibodies in combination with antibodies against: MAP2 (1:300, Dako), CNPase (1:100, Chemicon), GFAP (1:300, Dako), MHC class II (1:50, Dako) or IBA1 (1:300, Abcam). Sections were then incubated with Cy3-, rhodamine- or fluorescein-conjugated species-specific secondary antibodies in phosphate-buffered saline containing 1% normal serum for 1 h. Some sections were counterstained with DAPI (Sigma) for localization of cell nuclei. Finally, sections were coverslipped with Vectashield® mounting medium (Vector Laboratories). For negative controls, the primary antibodies were replaced with preimmune serum or IgG isotype controls. Slides were viewed under epifluorescence microscope (Leica Microsystem).
Statistical analysis of the clinical course of EAE was carried out with the Mann–Whitney test. Real-time polymerase chain reaction and western blot data were analysed with one-way ANOVA followed by Tukey's test for multiple comparisons. For single comparisons, Student's t-test was applied. P ≤ 0.05 were considered statistically significant.
To dissect the roles of soluble TNF and transmembrane TNF in EAE, we analysed in parallel the effect of XPro1595, a selective soluble TNF inhibitor (Zalevsky et al., 2007; Olleros et al., 2009), and etanercept, a non-selective TNF inhibitor (Goffe and Cather, 2003; Tracey et al., 2008), on MOG(35–55)-induced EAE in C57BL/6 female mice. Both biologics have comparable efficacy and potency (Zalevsky et al., 2007) and contrary to other anti-TNF options, which are structurally anti-human antibodies (e.g. infliximab, adalimumab, certolizumab pegol), they are effective in murine models. We chose a clinically relevant treatment protocol starting administration at 16 days post-induction of disease, when at least half the animals displayed acute symptoms of EAE (score ≥2). Treatment at a dose of 10 mg/kg was repeated every 3 days until sacrifice (55 days post-induction). All groups reached peak disease in a similar fashion, but animals treated with XPro1595 exhibited markedly improved functional recovery compared with vehicle- and etanercept-treated mice (Fig. 1A), with maximum improvement achieved around Day 25 and maintained throughout the remainder of the experiment. From Day 25 onwards, XPro1595-treated mice scored consistently <1.5, which signifies only mild flaccidity of the tail without any sign of hindlimb or forelimb paralysis. The overall attenuation of the pathology is also reflected in the significant reduction in cumulative disease index (Table 1). On the contrary, etanercept-treated mice did not show amelioration but rather chronic exacerbation of disease similarly to control (vehicle) mice (Fig. 1A). Chronically, both groups scored well >2.0, showing hindlimb and forelimb paralysis at various degrees of severity. It is noteworthy that treatment with either compound did not alter TNFR1 or TNFR2 protein expression in the CNS (Supplementary Fig. 1), indicating that no compensatory expression of these receptors was induced as a consequence of the pharmacological inhibition.
In order to evaluate whether the therapeutic effect of XPro1595 could be associated with changes in the immune response to MOG immunization in the periphery or in the spinal cord, we analysed the distribution of leucocyte populations by flow cytometry in spleen and spinal cord preparations at acute (28 days post-induction) and chronic (55 days post-induction) disease (Table 2). In the spleen at both time points, XPro1595 and etanercept had no effect on the relative percentages of B cells (B220+), T cells (CD3+) and macrophages (CD11b+) compared with the vehicle group. The only significant differences, although minimal, were a small reduction in CD4+ and increase in CD8+ T cells. In the spinal cord, at acute disease, we detected a marked attenuation in CD45highCD11b+ infiltrating macrophages, as well as in CD45lowCD11b+ microglia both in XPro1595- and etanercept-treated groups compared with vehicle (Fig. 1B, Table 2). This effect appeared comparable between XPro1595- and etanercept-treated groups, suggesting that the long-term functional recovery observed with XPro1595 cannot be solely dependent on inhibition of acute macrophage infiltration in the CNS or of microglia activation. At chronic disease, leucocyte infiltrates into the spinal cord were almost undetectable in all groups, rendering flow cytometry analysis experimentally unfeasible. Immunohistochemically, we detected the presence of sporadic CD45+ cells in subpial regions of the white matter in all conditions, without appreciable differences between treatments (Fig. 1C). This was confirmed by western blot analysis of CD45 protein expression (Fig. 1D), showing no significant differences among groups. Interestingly, in XPro1595-treated mice we measured a significantly reduced expression of GFAP by immunohistochemistry and western blot (Fig. 1C and D), indicating attenuated astroglial reactivity compared with vehicle- and etanercept-treated mice. It has been shown with genetic models that inhibition of astrogliosis and astrocyte-dependent inflammation is protective against EAE (van Loo et al., 2006; Brambilla et al., 2009), hence it is plausible that by blocking soluble TNF signalling with XPro1595 we can elicit a dampening effect, direct or indirect, on astrocyte activation, obtaining with our pharmacological approach a similar effect to that observed in transgenic mice.
To further investigate whether the therapeutic effect of XPro1595 was dependent on a reduction of the inflammatory response, we evaluated the expression of proinflammatory cytokines and chemokines at chronic disease (55 days post-induction). We found that both XPro1595 and etanercept significantly reduced expression of IFNγ, TNF, IL1β, IL6 and CCL2 compared with the vehicle-treated mice (Fig. 2), providing further demonstration that the therapeutic outcome of soluble TNF inhibition is not exclusively dependent on attenuation of the inflammatory response. Indeed, a reduction of the inflammatory response occurs similarly in etanercept-treated mice, but is not sufficient to produce functional improvement. XPro1595 treatment also reduced expression of CCL5, CXCL10 and CCR2 compared with vehicle, while etanercept did not, suggesting that transmembrane TNF signalling could be associated with a specific anti-inflammatory and repair signature.
Histological analysis of the thoracic spinal cord of naïve and EAE-induced mice (55 days post-induction) showed a more prevalent axonal damage in the lateral and posterior funiculi. Clear axon preservation was observed in the XPro1595 group compared with vehicle and etanercept (Fig. 3A), which both displayed a high frequency of enlarged, swollen and hypomyelinated axons, as well as fully collapsed (black arrows) axons. When counting the number of myelinated axons in the white matter, we found a significantly higher number in the XPro1595 group (24800 ± 2477 axons/mm2) compared with vehicle (16681 ± 1528 axons/mm2) and etanercept (17556 ± 1295 axons/mm2) (Fig. 3B). Collapsed axons (obliterated, hence non-functional) were also reduced in XPro1595-treated mice (992 ± 314 collapsed axons/mm2) compared with vehicle- (1958 ± 399 collapsed axons/mm2) and etanercept-treated (2842 ± 585 collapsed axons/mm2) mice (Fig. 3C). Interestingly, no differences were found among all groups in the number of infiltrating leucocytes (Fig. 3D), in agreement with previously shown CD45 immunohistochemistry and protein expression data (Fig. 1C and D).
To further substantiate the observation that XPro1595 treatment promotes axonal preservation, we evaluated expression of the axon-specific molecule neurofilament-H. Neurofilament-H gene and protein expression were increased in the XPro1595-treated group compared with vehicle (Fig. 3E), and vehicle and etanercept (Fig. 3F), respectively. On the other hand, the non-phosphorylated form of neurofilament-H, commonly associated with axonal damage (Trapp et al., 1998), was significantly reduced in XPro1595-treated mice compared with etanercept (Fig. 3G). Together these data point to a neuroprotective effect of XPro1595 following EAE.
To address whether XPro1595-mediated neuroprotection was associated with a direct effect on myelin preservation, we analysed myelin compaction by electron microscopy. As expected, all EAE-induced groups (Fig. 4B–D and F–H) showed signs of ‘loose’ myelin compared with naïve mice (Fig. 4A and E). Significant damage was especially evident in vehicle- and etanercept-treated mice, where blebbing, loss of adaxonal contact and unravelling of the myelin sheaths were prevalent. Lipid inclusions in the axoplasm, a sign of axonal damage, were also found. In the XPro1595-treated group, however, myelin compaction appeared rather preserved, as indicated by more tightly associated myelin layers (Fig. 4H) compared with vehicle- and etanercept-treated groups (Fig. 4F and G). These observations were confirmed by quantification of the g-ratio (axon diameter/fibre diameter), which was significantly higher for all diameter sizes in XPro1595-treated mice compared with vehicle and etanercept groups, and similar to naïve mice (Fig. 4I).
We used electron microscopy to investigate whether inhibition of soluble TNF could have an effect on remyelination, increasing the potential for properly functioning axons. We counted remyelinating axons based on morphological characteristics (Fig. 5A) and found a significantly higher number in the XPro1595-treated group compared with vehicle and etanercept (Fig. 5B). As expected, all groups had more remyelinating axons than uninjured naïve mice. Interestingly, not only did XPro1595 promote remyelination, but it supported remyelination of larger diameter axons. Indeed, the percentage of remyelinating axons with diameter >1 µm was significantly higher following treatment with XPro1595 (Fig. 5C). Because these larger axons are primarily motor fibres, rather than sensory, their remyelination is highly relevant for motor function recovery. Increased remyelination was also accompanied by higher expression of myelin-specific proteins, particularly proteolipid protein, the predominant myelin protein in the CNS, which was upregulated in XPro1595-treated animals compared with both vehicle- and etanercept-treated mice (Fig. 5D). Myelin basic protein and myelin-associated glycoprotein levels were significantly higher in XPro1595-treated mice compared with etanercept, and vehicle and etanercept, respectively (Fig. 5D).
Finally, to determine if the pro-myelinating effect of XPro1595 could be associated with a modulation of the number of oligodendrocyte precursor cells, we counted NG2+ oligodendrocyte precursor cells in the spinal cord white matter of EAE-induced mice and found a significant increase in the XPro1595-treated group compared with the others. This opens the possibility that transmembrane TNF activates mechanisms leading to proliferation and/or survival of precursors, increasing the pool of cells ready to mature into new myelinating oligodendrocytes.
To begin evaluating the possible mechanisms associated with the beneficial effects of transmembrane TNF in EAE, we performed an in-depth expression study analysing by confocal microscopy, TNFR1 and TNFR2 cell-specific distribution in the mouse spinal cord at the acute phase of EAE. TNFR1 was found to be highly expressed in neurons, particularly in cluster-like structures along MAP2+ dendrites in the grey matter (Fig. 6). Expression was also detected on GFAP+ astrocytes and CC1+ oligodendrocyte cell bodies, mainly in the white matter. Conversely, NG2+ oligodendrocyte precursor cells and CD11b+ microglia/macrophages appeared devoid of TNFR1 (Fig. 6). As for TNFR2 (Fig. 7), no specific immunoreactivity was found on MAP2+ neurons, contrary to TNFR1. TNFR2 expression was instead detected on GFAP+ astrocytes (with high levels in the white matter), CC1+ oligodendrocytes and NG2+ oligodendrocyte precursor cellss. Finally, strong TNFR2 immunoreactivity was found on CD11b+ macrophages and microglial cells, particularly in and around subpial infiltrates (Fig. 7). This study represents the first exhaustive analysis of TNF receptor expression in the mouse spinal cord, and demonstrates the presence of TNFR1 and TNFR2 on most CNS cell populations. Interestingly, both receptors are found on oligodendrocytes (although only TNFR2 appears to be in oligodendrocyte precursor cells), underscoring the critical role played by TNF signalling in oligodendrocyte function. The cell-specific expression of TNF receptors is summarized in Supplementary Table 1.
In parallel, we analysed TNFR1 and TNFR2 expression in the post-mortem spinal cord of patients with progressive multiple sclerosis (three cases; see Supplementary Table 2). Diaminobenzidene staining for TNFR1 in the white matter, in particular at the edge of the examined chronic inactive lesions characterized by low degree of perivascular inflammatory infiltrates, showed reactivity in the cell body of cells morphologically resembling neurons (Fig. 8A). Neuronal-specific expression was confirmed with double-immunofluorescent labelling of TNFR1 with MAP2 (Fig. 8B), in agreement with our data in the mouse. Interestingly, in the same lesions, TNFR1 expression was detected surrounding CNPase+ myelin rings (Fig. 8C and D), within myelin lamellas (Fig. 8C and E), and on occasional oligodendrocyte processes (Fig. 8C and F), but not on oligodendrocytes cell bodies. This suggests that TNFR1 is present in myelin-containing membrane structures wrapped around axons, and possibly on the axons as well, at the axon–oligodendrocyte interface. We can exclude that TNFR1 immunoreactivity on CNPase+ myelin rings is derived from TNFR1-expressing microglia/macrophages or astrocytes in contact with the myelinated axons, since double-immunofluorescent labelling of TNFR1 with IBA1 or GFAP showed absence of TNFR1 expression on either microglia/macrophages or astrocytes (data not shown). The localization of a TNF-responsive receptor in myelin lamellas, specifically within lesions, provides strong evidence of the involvement of TNF signalling in the direct regulation of oligodendrocyte function and, perhaps, myelination. In contrast with TNFR1, TNFR2 expression was localized in activated microglia/macrophages within white matter lesions (Fig. 8G and H) and in reactive astrocytes (Fig. 8I and J). TNFR2 was not detected in neurons, and also no definitive evidence of TNFR2 expression was found in oligodendrocytes (data not shown). The cell-specific expression of TNF receptors is summarized in Supplementary Table 1.
Our study represents a comprehensive investigation of the dual functions of soluble TNF and transmembrane TNF in the pathophysiology of EAE carried out with pharmacological inhibitors. By comparing the effects of XPro1595, a selective soluble TNF inhibitor, and etanercept, a non-selective TNF inhibitor, we directly demonstrate that transmembrane TNF signalling is necessary for functional recovery, axon preservation and, most importantly, remyelination in a murine model of multiple sclerosis. Our findings build upon genetic evidence showing that soluble TNF is acutely proinflammatory, whereas transmembrane TNF is chronically involved in repair, and suggest that pharmacological inhibition of soluble TNF may be effectively translated into a therapy for multiple sclerosis. Furthermore, this work includes a complete investigation of TNF receptor cellular localization in the spinal cord of EAE mice and of patients affected by progressive multiple sclerosis, providing a basis for a better understanding of the role of TNF signalling in the human pathology, as well as its experimental model.
The evidence emerging from our study, in addition to extensive data in the literature, leads us to formulate the hypothesis that selective blockade of soluble TNF, which has the highest affinity for TNFR1, and preservation of signalling mediated by transmembrane TNF, which conversely has the highest affinity for TNFR2, causes a shift in the balance of TNF receptor activation towards TNFR2, and the protective functions associated with transmembrane TNF would likely be attributed to TNFR2-mediated processes (Caminero et al., 2011). However, mindful that transmembrane TNF is also capable of activating TNFR1, we cannot exclude the possibility that transmembrane TNF-mediated TNFR1 activation could also result, at least in part, in protection.
We show that in MOG-induced EAE, inhibition of soluble TNF significantly improves the clinical outcome when administered after disease onset, hence with a clinically relevant regimen. After overcoming the acute phase of disease, which develops similarly to etanercept-treated and control mice, XPro1595-treated mice recover from paralysis and maintain long-term function. Although it is well documented that activation of TNFR1 signalling is associated with a proinflammatory effect that sustains the initiation phase of EAE and causes demyelination (Korner et al., 1997; Akassoglou et al., 1998; Eugster et al., 1999; Kassiotis et al., 1999; Arnett et al., 2001; Kassiotis and Kollias, 2001; Gimenez et al., 2006), our data suggest that long-term improvement is not dependent, at least not exclusively, on the ability of XPro1595 to antagonize inflammation by inhibiting this pathway. Indeed, both XPro1595 and etanercept equally attenuate inflammation, as demonstrated by reduced macrophage infiltration and expression of proinflammatory cytokines and chemokines (Figs 1B and and2).2). Nevertheless, the anti-inflammatory effect of etanercept is not sufficient to stimulate recovery, indicating that it is the protective effect of transmembrane TNF signalling, which is unaffected by XPro1595, to ultimately drive the positive outcome in chronic disease. It is also noteworthy that, except for the reduction in macrophage infiltrates into the cord, treatment with either drug after disease onset did not alter the immune cell profile in spleen and spinal cord, particularly encephalitogenic CD4 and CD8 T cell subsets. It has been recently shown that prophylactic treatment with a TNFR1-selective antagonistic TNF mutant reduces the severity of EAE by suppressing Th1 and Th17 responses, as well as immune cell infiltration into the cord (Nomura et al., 2009). This is not the case in our experimental paradigm, as we begin administration after disease onset, when activation of the immune response is already underway. Based on our data, we can conclude that delayed therapeutic inhibition of soluble TNF in fully developed disease is independent of immune cell modulation, and is rather associated with a direct effect on the CNS compartment. Indeed, our findings show that XPro1595 treatment increases the number of myelinated axons and the expression of neuronal-specific molecules in EAE spinal cords, while drastically diminishing the number of degenerated axons, suggesting that transmembrane TNF may play a role in axonal sparing. Although we did not assess neuronal survival per se, the reduction in axonal pathology could be the consequence of the ability of transmembrane TNF to directly or indirectly activate neuroprotective cascades. Based on our hypothesis of an XPro1595-mediated balance shift towards TNFR2 activation, we can speculate that TNFR2-mediated neuroprotection is occurring, while at the same time TNFR1-mediated neurotoxic events are minimized. This concept is supported by published evidence showing that activation of TNFR2 promotes neuronal survival under a variety of pathological conditions, such as ischaemia reperfusion, glutamate- and β-amyloid-induced cytotoxicity (Shen et al., 1997; Fontaine et al., 2002; Marchetti et al., 2004), whereas TNFR1-mediated signalling is associated with neuronal cell death (McCoy et al., 2006; Wen et al., 2006; He et al., 2007; McCoy and Tansey, 2008; McAlpine et al., 2009). Our own immunohistochemical data show that TNFR2 is absent in neurons, both in mouse EAE and human multiple sclerosis (Figs 7 and and8),8), suggesting that transmembrane TNF/TNFR2-mediated neuroprotective effects occur via indirect mechanisms. On the contrary, TNFR1 is highly expressed in neurons (Figs 6 and and8);8); therefore, by reducing the tone of TNFR1 activation as a consequence of blocking soluble TNF, we may be directly dampening TNFR1-mediated neurotoxicity and pro-apoptotic function. This could be especially important in the human pathology, where administration of a selective soluble TNF inhibitor, by virtue of a limited neuronal TNFR1 activation, could play a role in delaying or reducing neurodegeneration, and hence prevent patient disability. On the other hand, it has also been described that NF-ΚB signalling engaged downstream of TNFR1 activation in neurons is responsible for counteracting TNFR1-mediated pro-apoptotic effect (Kaltschmidt et al., 1999; Taoufik et al., 2007, 2010), therefore we cannot exclude that transmembrane TNF-mediated activation of TNFR1 may be acting, at least partially, as anti-apoptotic signal preventing neuronal death. This possibility is further corroborated by the findings of Taoufik et al. (2011) which, in agreement with our study, demonstrate that selective inhibition of soluble TNF by XPro1595 is protective in EAE and such effect is dependent, at least in part, upon maintenance of neuroprotective neuronal NF-ΚB activity.
A crucial finding of our study is the demonstration that transmembrane TNF signalling is essential in preserving myelin integrity and compaction and, most importantly, in promoting remyelination. This, combined with the immunohistochemical evidence on EAE models that TNFR1 and TNFR2 are both expressed on oligodendrocytes and TNFR2 is expressed on oligodendrocyte precursor cells, suggests that TNF plays a direct role in modulating properties and functions of the oligodendrocyte compartment. In our EAE model, and in agreement with previous reports (Arnett et al., 2001, 2003), remyelination is accompanied by a significant increase in the number of NG2+ oligodendrocyte precursor cells (Fig. 5F); therefore, a plausible scenario is that signalling mediated by transmembrane TNF via TNFR2 induces the proliferation and/or survival of oligodendrocyte precursor cells, thereby increasing the pool of cells readily available to remyelinate spared axons. We also found that oligodendrocytes express TNFR1. Hovelmeyer et al. (2005) have documented that TNFR1-mediated oligodendrocyte apoptosis is a key event in the induction of EAE. More recently, Paintlia et al. (2011) demonstrated that synergistic action of TNF and IL17 can also cause oligodendrocyte apoptosis via a TNFR1-mediated mechanism. Based on this evidence we can speculate that, by shifting the balance towards a preferential activation of TNFR2, XPro1595 may also contribute to prevent oligodendrocyte apoptosis.
Another possibility in explaining the protective effect of Xpro1595 is that transmembrane TNF is required for establishing immune tolerance. The inhibition by etanercept of transmembrane TNF signalling on specific cell populations, e.g. dendritic cells, which have tolerogenic potential (Fu and Jiang, 2010), could disrupt cross-talk with effector T cells and prevent the formation of tolerizing T cells, allowing for continued immune attack of oligodendrocyte precursor cells and oligodendrocytes resulting in ongoing demyelination.
Another important finding of our study is that TNFR2 is highly expressed in microglia following EAE (Fig. 7). It has been recently shown by Veroni et al. (2010) that TNFR2 stimulation in microglia promotes the expression of anti-inflammatory and neuroprotective genes such as granulocyte colony-stimulating factor (GCSF), adrenomedullin and IL-10, suggesting that microglia may contribute to the counter-regulatory response activated in neuropathological conditions. On this basis, we can speculate that microglial TNFR2 activated by transmembrane TNF may be promoting the synthesis of protective molecules participating in the overall positive outcome of XPro1595 treatment. We also found expression of TNFR2 in microglia and macrophages localized within and around multiple sclerosis lesions in the spinal cord, suggesting that similar transmembrane TNF/TNFR2-mediated protective events could also be taking place in the human pathology. To our knowledge, this is the first report of TNFR2 expression in microglia in multiple sclerosis tissue, since thus far microglial TNFR2 was only identified in the brains of patients with AIDS (Sippy et al., 1995).
It should also be noted that an interesting pattern of TNFR1 expression was found in spinal cord multiple sclerosis lesions, specifically around and within CNPase+ myelin rings (Fig. 3C). The presence of TNFR1 in oligodendrocyte myelin lamellas wrapped around axons, and possibly at the axon–oligodendrocyte interface, provides strong anatomical evidence of the involvement of TNF signalling in the direct regulation of oligodendrocyte function. Our finding complements previous works showing TNFR1 expression in oligodendrocytes in the brain of patients with multiple sclerosis (Raine et al., 1998). Similarly to the mouse, this may suggest that TNFR1 activation could lead to oligodendrocyte damage and apoptosis. Therefore, in view of a possible therapeutic application in multiple sclerosis, we can speculate that by shifting the balance towards a preferential TNFR2 activation, XPro1595 administration could result in reduction and/or inhibition of oligodendrocyte apoptosis, hence protection from demyelination.
Lastly, it is worth mentioning that, unlike etanercept, XPro1595 does not directly block lymphotoxin (Zalevsky et al., 2007), a cytokine belonging to the tumour necrosis factor superfamily and capable of binding to both TNFR1 and TNFR2. Lymphotoxin is secreted by activated T cells (Zipp et al., 1995) and is expressed in multiple sclerosis plaques (Selmaj et al., 1991). Although the role of lymphotoxin in EAE is controversial, most studies seem to indicate that inhibition of the lymphotoxin pathway may be protective against EAE (Aktas et al., 2006). Our work with XPro1595 and etanercept, although it was not specifically directed at investigating the role of lymphotoxin, seems to indicate that lymphotoxin plays a marginal role in EAE, at least in sustaining the chronic phase of the disease. Indeed XPro1595, which does not block lymphotoxin, has a chronic protective effect, while etanercept, which blocks lymphotoxin, does not show protection. We cannot exclude, however, that XPro1595 may indirectly reduce the tone of lymphotoxin signalling by reducing lymphotoxin expression, which could contribute to the protective effect of the compound in EAE.
Collectively, our data suggest that, in the balance between demyelination and remyelination, it is the positive effect of transmembrane TNF in the remyelination process that ultimately accounts for the observed functional recovery and resolution of EAE. This carries important clinical implications for multiple sclerosis therapy. Since human multiple sclerosis is primarily a relapsing–remitting disease characterized by alternating damage and repair phases, the challenge in therapies targeting multiple sclerosis is to block the damage while still allowing repair to take place. Using a pharmacological approach, here we show that soluble TNF is responsible for the damage, whereas transmembrane TNF drives the repair process. XPro1595 maintains the protective properties of TNF and allows remyelination to occur. Therefore, unlike non-selective TNF inhibitors, which are associated with demyelination when administered in human therapy, XPro1595 represents a promising new candidate to be added to the limited repertoire of multiple sclerosis modulating drugs, finally opening the door to the introduction of a TNF inhibitor into multiple sclerosis therapy.
National Institutes of Health (NS051709, NS065479 to J.R.B.); The Miami Project To Cure Paralysis.
Supplementary material is available at Brain online.
We thank the UK MS Tissue Bank (www.ukmstissuebank.imperial.ac.uk) for providing the post-mortem multiple sclerosis samples analysed in the study. We are also grateful to Dr Mary Bunge (The Miami Project To Cure Paralysis, University of Miami Miller School of Medicine) for her invaluable advice on morphological evaluation and interpretation of EM micrographs. The authors do not have any conflicting financial interests. It should be noted, however, that D.E.S. is an employee of Xencor, which provided the compound XPro1595 used in the study.