Progressive multiple sclerosis is associated with metabolic failure of the axon and excitotoxicity that leads to chronic neurodegeneration. Global sodium-channel blockade causes side effects that can limit its use for neuroprotection in multiple sclerosis. Through selective targeting of drugs to lesions we aimed to improve the potential therapeutic window for treatment. This was assessed in the relapsing-progressive experimental autoimmune encephalomyelitis ABH mouse model of multiple sclerosis using conventional sodium channel blockers and a novel central nervous system-excluded sodium channel blocker (CFM6104) that was synthesized with properties that selectively target the inflammatory penumbra in experimental autoimmune encephalomyelitis lesions. Carbamazepine and oxcarbazepine were not immunosuppressive in lymphocyte-driven autoimmunity, but slowed the accumulation of disability in experimental autoimmune encephalomyelitis when administered during periods of the inflammatory penumbra after active lesion formation, and was shown to limit the development of neurodegeneration during optic neuritis in myelin-specific T cell receptor transgenic mice. CFM6104 was shown to be a state-selective, sodium channel blocker and a fluorescent p-glycoprotein substrate that was traceable. This compound was >90% excluded from the central nervous system in normal mice, but entered the central nervous system during the inflammatory phase in experimental autoimmune encephalomyelitis mice. This occurs after the focal and selective downregulation of endothelial p-glycoprotein at the blood–brain barrier that occurs in both experimental autoimmune encephalomyelitis and multiple sclerosis lesions. CFM6104 significantly slowed down the accumulation of disability and nerve loss in experimental autoimmune encephalomyelitis. Therapeutic-targeting of drugs to lesions may reduce the potential side effect profile of neuroprotective agents that can influence neurotransmission. This class of agents inhibit microglial activity and neural sodium loading, which are both thought to contribute to progressive neurodegeneration in multiple sclerosis and possibly other neurodegenerative diseases.
Multiple sclerosis is the major cause of non-traumatic disability in young adults (Compston and Coles, 2008). It is clear that relapsing multiple sclerosis can be controlled by the action of peripherally-active immunosuppressive agents that reduce CNS atrophy (Gunnarsson et al., 2011; Marta and Giovannoni, 2012). These serve to limit waves of inflammatory cells entering the CNS from the blood that are manifest in relapsing disease in humans and in experimental autoimmune encephalomyelitis (EAE) animal models of multiple sclerosis (Allen et al., 1993; Lassmann, 2013). These cause focal lesions that are characterized by demyelination and variable degrees of axonal loss, gliosis and microglial activation. The microenvironment in lesions is believed to progressively damage axons and make them susceptible to chronic degeneration (Pryce et al., 2005; Hampton et al., 2008), which underpins the accumulation of chronic disability (Lassmann, 2013). This secondary progression appears to be unresponsive to peripheral immunosuppression in both multiple sclerosis and EAE and is currently untreatable (Coles et al., 1999; Pryce et al., 2005; Dutta and Trapp, 2011; Al-Izki et al., 2012).
Although, the precise mechanisms of axonal pathology are unproven, axonal pathology in acute multiple sclerosis lesions is associated with macrophage and glial cell activation, excitotoxicity and loss of metabolic support (Dutta and Trapp, 2011; Lee et al., 2012; Parpura and Verkhratsky, 2012). Demyelination also leads to neuronal sodium and calcium channel upregulation and increased energy demands that leave demyelinated axons vulnerable to further damage (Craner et al., 2004). Voltage-gated sodium channels (Nav) are redistributed from the nodes of Ranvier along demyelinated axonal segments to maintain neural conduction (Craner et al., 2004). This facilitates sodium loading that is compounded by loss of activity of energy-dependent sodium pumps following mitochondrial dysfunction secondary to a complex inflammatory cascade (Dutta and Trapp, 2011). This can lead to the formation of toxic levels of axoplasmic calcium (Kapoor et al., 2003; Dutta and Trapp, 2011), which trigger further neuroaxonal damage. This, in conjunction with and altered astrocytic control of neurotransmitter balance, leads to progressive disability that no longer responds to systemic immunosuppression (Dutta and Trapp, 2011; Lee et al., 2012; Parpura and Verkhratsky, 2012). Although these mechanisms highlight potential therapeutic targets (Fern et al., 1993; Waxman, 2008), many of them are involved in normal nerve function; inhibiting their function may therefore be associated with side effects that can limit their clinical use (Kapoor et al., 2010; Gnanapavan et al., 2013). This is particularly a problem in progressive multiple sclerosis where accumulating nerve loss reduces compensatory or reserve capacity and lowers the threshold for side effects from drugs that alter neurotransmitter function. There is an urgent, unmet need to find new treatments for progressive multiple sclerosis.
Studies in myelin oligodendrocyte glycoprotein (MOG) peptide-induced EAE in C57BL/6 and Biozzi ABH mice have shown that Nav blockers can ameliorate the development of neurological disease (Lo et al., 2003; Black et al., 2006, 2007). In many instances the treatment benefit of Nav blockers could be attributed to immunosuppression that prevented lesion formation and the development of disease (Craner et al., 2005; Black et al., 2007). However, in Biozzi ABH mice, the myelin peptide used to induce disease to examine the influence of carbamazepine (Black et al., 2007) was subdominant even within the MOG protein, which in itself is not the major encephalitogenic auto-antigen (Amor et al., 1994; Pryce et al., 2005; Smith et al., 2005). After active sensitization with spinal cord homogenate in Freund’s adjuvant the initial paralytic attack is inflammatory, but is not associated with significant demyelination and neuroaxonal loss. The latter occurs following the development of spontaneous or induced-relapsing disease (Baker et al., 1990; Hampton et al., 2008; Jackson et al., 2009; Al-Izki et al., 2012). This stage of disease in the ABH model allows one to discriminate between the effects of therapies that target both immunosuppression and/or neuroprotective mechanisms. Neuroprotection is seen by the slower accumulation of residual neurological deficits and slower loss of motor control on RotaRod activity, which have a good correlation with spinal nerve content (Al-Izki et al., 2012).
As demyelinated lesions centre around the vasculature (Baker et al., 1990; Lassmann, 2013), we hypothesized that it would be possible to use blood–brain barrier activities, notably exploiting the actions of energy-dependent, adenosine triphosphatase binding cassette (ABC) transporters, such as p-glycoprotein (Miller et al., 2008), to selectively target potentially neuroprotective agents to multiple sclerosis lesions. We show this activity using a novel CNS-excluded Nav blocker, which is a p-glycoprotein substrate, and was targeted to lesions after the selective loss of p-glycoprotein that occurs in both multiple sclerosis and EAE lesions.
Materials and methods
Multiple sclerosis tissue
Brain tissue samples were obtained from the Netherlands Brain Bank. The Netherlands Brain Bank received permission to perform autopsies for the use of tissue and for access to medical records for research purposes from the Ethical Committee of the VU University Medical Centre, Amsterdam, The Netherlands. A total of 10 paraffin-embedded tissue blocks from patients with multiple sclerosis, containing active, chronic active and chronic inactive lesions, and one white matter block from six different donors without neurological disease were selected. Normal-appearing white matter and lesions in white matter were selected on the basis of post-mortem MRI. All patients with multiple sclerosis and control subjects, or their next of kin, had given informed consent for autopsy and use of their brain tissue for research purposes. Analysis of tissue was supported by ethical review panels within The Netherlands and UK (REC 08/H0722/26).
Adult female (6–8 weeks), pathogen-free, Biozzi ABH mice and C57BL/10.RIII mice were purchased from Harlan UK Ltd, or male and female ABH mice were bred at Queen Mary University of London. C57BL6.Cg-Tg (Thy1-CFP)23Jrs.Tg (Tcra2D2, Tcrb2D2)1Kuch/J transgenic mice, which express a neuronally-restricted, Thy1-promoter driven expression of cyan fluorescent protein (CFP) that is limited to retinal ganglion cells within the retina (Feng et al., 2000) and a H-2b-restricted MOG-specific T cell receptor (Bettelli et al., 2003; Guan et al., 2006), were from stock bred at Queen Mary University of London (Lidster et al., 2013). Methods of housing, allocation to groups and other reporting elements including blinding, relevant to the ARRIVE guidelines have been reported previously (Al-Izki et al., 2012). All procedures were approved by the local ethical review processes and were performed in accordance with the UK Animal (Scientific Procedures) Act 1986.
Carbamazepine, oxcarbazepine, memantine hydrochloride and minocycline hydrochloride were purchased from Sigma Aldrich. CFM6104 [N-((5-(2-(2-(1h-imidazol-1-yl)ethyl)-2h-indazol-3-yl)-1,2,4-oxadiazol-3-yl)methyl)-3-methylbenzamide] is a novel, non-selective, state-dependent sodium channel blocker. The synthesis will be described elsewhere (Browne et al., submitted for publication). This has no detectable activity on active receptor subtypes (IC50 > 30 µM), but had modest activity (IC50 = ∼20–25 µM) on inactive Nav1.1, Nav1.2, Nav1.5, Nav1.6, Nav1.7 receptors (Browne et al., submitted for publication). CFM6104 was dissolved in a 1:1 solution of dimethyl sulphoxide and PBS and allowed to cool before administration intraperitoneally, or dissolved in ethanol:cremophor:PBS 1:1:18. Minocycline hydrochloride was adjusted to pH 7, using 10 N sodium hydroxide before use. These were administered either daily just before or after the anticipated onset of relapsing disease.
In vivo T cell proliferation assays
The contact sensitizer 4-ethoxymethylene-2-phenyl-2-oxazolin-5-one (oxazolone, OX, Sigma) was dissolved (25 mg/ml) in 4:1 acetone:olive oil. Mice (n = 3 per group) received epicutaneous application of either 25 µl of 2.5% oxazolone or acetone:olive oil on the dorsum of the ear on Day 0 (O'Neill et al., 1992). The draining auricular lymph nodes were removed 3 days later and the induced proliferative response was assessed as previously described. Briefly, 5 × 105 cells per well were cultured in RPMI-1640 medium with glutamate (Gibco®, Invitrogen), supplemented with 0.5 mM sodium, in round-bottomed 96-well plates overnight at 37°C in a humidified atmosphere of 5% CO2. T cell proliferation was assessed using the CellTiter 96® AQueous Non-Radioactive Cell Proliferation Assay according to the manufacturer’s instructions (Promega) and the absorbance was recorded at 490 nm. Animals received daily intraperitoneal injections of either vehicle or CFM6104 from Day 0–3.
Induction of relapsing-progressive experimental autoimmune encephalomyelitis
Mice were injected subcutaneously with 1 mg freeze-dried mouse spinal cord homogenate in Freund’s adjuvant on Days 0 and 7 as described previously (Al-Izki et al., 2012). After the initial paralytic disease and subsequent remission, a relapse was induced by a further injection of spinal cord homogenate in Freund’s adjuvant on Day 28 to induce a relapse 7 days later (Al-Izki et al., 2012). Studies were randomized, blinded and powered as described previously (Al-Izki et al., 2012). Neurological scores were graded as: 0 = normal; 1 = limp tail; 2 = impaired righting reflex; 3 = hindlimb paresis; 4 = complete hindlimb paralysis; and 5 = moribund/death (Al-Izki et al., 2012). Motor control and co-ordination was assessed on an accelerating (4–40 rpm, accelerating at 6 rpm/25 s) RotaRod (ENV-575 M, Med Associates Inc.) as described previously (Al-Izki et al., 2012). This was performed 1 day before induction of relapse and 1 day before termination of the experiment on Day 49–50. RotaRod assessment was performed blinded to treatment. Animals were randomized to vehicle or treatment based on their RotaRod scores.
At the end of the experiment the spinal cord was removed and an ELISA for heavy chain neurofilament on spinal cord was performed and total nerve content of each spinal cord was estimated following calibration against neurofilament protein standards as described previously (Jackson et al., 2005).
Induction of experimental autoimmune uveitis
C57BL/10.RIII mice were injected subcutaneously with 25 µg interphotoreceptor binding peptide residues 161–182 in Freund’s adjuvant on Days 0 and 7 as described previously (Hankey et al., 2001). On Day 21 post-induction animals were killed and eyes were removed, fixed, wax embedded and 5 μm sections through the optic nerve head were stained with haematoxylin and eosin. Sections were scored blinded to treatment on a 0–6 scale to assess the degree of cellular infiltration and a 0–5 scale to assess retinal nerve damage as described previously (Hankey et al., 2001).
Induction of optic neuritis mouse model
Optic neuritis was induced in C57BL/6-.Cg-Tg (Thy1-CFP)23Jrs.Tg (Tcra2D2, Tcrb2D2)1Kuch/J transgenic mice after administration of 150 ng Bordetella pertussis toxin (Sigma Aldrich) in PBS on Days 0 and 2 (Bettelli et al., 2003; Lidster et al., 2013) Subclinical EAE and optic neuritis within the optic nerve was augmented following the injection of 250 µg Z12 rat IgG2b mouse myelin oligodendrocyte-specific demyelinating antibody (Piddlesden et al., 1993; Lidster et al., 2013) on Day 14. Animals were injected daily on Days 0–21 with the compounds. Animals were killed on Day 21 post-induction and eyes were enucleated and immersed in 4% paraformaldehyde (Sigma Aldrich) in PBS pH 7.4 overnight. The retinas were dissected and the cornea, sclera, lens, hyaloid vasculature and connective tissue were removed in 2× normal strength PBS. Four radial incisions were cut around the retinas and flat mounts were mounted onto slides and cover-slipped with anti-fade glycerol (CitiFluor Ltd). The retinal flat mounts were imaged by fluorescent microscopy and retinal ganglion cells density was calculated by counting CFP expressing retinal ganglion cells using stereology software with a fractionator probe and the number cells in the four quadrants around the optic nerve head counted (Lidster et al., 2013). The results are expressed as mean ± standard error of the mean (SEM).
Liquid chromatography–mass spectroscopy
Pharmacokinetic studies on plasma and brain samples using liquid chromatography-mass spectroscopy (LC-MS/MS) with a sensitivity of 3 ng/ml, was performed by Merck Serono. In addition plasma and spinal cords were harvested from ABH mice 30 min after the intraperitoneal injection of 10 mg/kg CFM6104 in ethanol:cremophor:PBS 1:1:18. Spinal cords were homogenized in 2 ml eppendorf tubes using a 5 mm steel bead and TissueLyzer-II (Qiagen) set at 25 Hz for 5 min. Twenty microlitres of the homogenate or plasma was extracted by 100 µl acetonitrile (with internal standard, coumaphos). Samples were then incubated on ice for 30 min, centrifuged and the supernatant was analysed using UPLC-MS/MS. Analytes were resolved using an Accela UPLC (Thermo Scientific) equipped with 1.7 µm PFP Kinetex 2.1 × 50 mm UPLC column (Phenomenex) and a mobile phase gradient of buffer A (water + 0.1% formic acid) and buffer B (acetonitrile + 0.1 formic acid) at a flow rate of 500 µl/min. Eluting compounds of interest were detected using a TSQ Vantage mass spectrometry system (Thermo Scientific). The optimum transitional daughter ions mass of each analyst were as follows: CFM6104 m/z 414.1 → 239.1 and coumaphos m/z 363.3 → 227. The sensitivity was 1 ng/ml.
Assessment of vascular permeability was performed using the Pgp-Glo™ assay (Promega) containing recombinant human ABCB1 in membranes according to the manufacturer’s instructions. Briefly, samples were preincubated with ATP before incubation with 100 μM compound or positive control. The residual ATP was assayed by luciferin bioluminescence. In addition, monolayer permeability and efflux was assessed in a bi-directional transport assay by Merck Serono using CacoReady™ (ReadyCell SL). Briefly, differentiated and polarized Caco-2 cell barriers were grown on polycarbone microporous filters in a 24-well format. One micrometre compounds were incubated for 120 min pH 7.4/7.4 apical/basolateral after apical or basolateral application and basolateral or apical sampling, respectively, in the presence or absence of 1 μM cyclosporin A, which is a p-glycoprotein inhibitor. Drug concentrations were assessed by LC-MS/MS using an API 5500-Trap.
Formalin-fixed paraffin-embedded mouse and human tissue was sectioned at 5 µm. Human tissue was stained for proteolipid protein (PLP, clone plpc1, Serotec Ltd), HLA-DR (clone LN3, Dako), p-glycoprotein/CD243 (clone C219, Abcam), breast cancer resistance protein (BCRP/CD338, clone 6D170, US Biological) and the endothelial marker glucose transporter-1 (Glut-1; a kind gift from Jack van Horssen, VUMC). Sections were deparaffinized in xylene and rehydrated through graded alcohol into distilled water. Endogenous peroxidase activity was quenched using 0.3% hydrogen peroxide in methanol. For HLA-DR, p-glycoprotein, BCRP (now known as ABCG2) and Glut-1 detection, slides were rinsed with distilled water and transferred to citric acid for heat-induced antigen retrieval. After microwave irradiation for 5 min on the high setting and for 10 min on the medium setting, slides were cooled to room temperature and rinsed in PBS. Serial sections were incubated overnight with either anti-HLA-DR (1:100), proteolipid protein (1:500), p-glycoprotein (clone C219, 1:50); BCRP (1:400); Glut-1 (1:1000). Slides were incubated with Envision Kit horseradish peroxidase-labelled anti-mouse/rabbit Ig (Dako) for 30 min at room temperature, and finally diaminobenzidine tetrachloride. Between incubation steps, sections were thoroughly washed with PBS. After a short rinse in tap water sections were incubated with haematoxylin for 1 min and extensively washed with tap water for 10 min. Finally, sections were dehydrated with ethanol followed by xylol and mounted with Entellan (Merck). All antibodies were diluted in PBS containing 0.1% bovine serum albumin (Boehringer-Mannheim), which also served as a negative control. Immunofluorescence was carried out on sections from three patients with multiple sclerosis to depict Glut-1, p-glycoprotein and cell nuclei with DAPI. After antigen retrieval and blocking for 1 h with 5% normal goat serum, the sections were incubated with p-glycoprotein (1:20) and Glut-1 (1:50) overnight at room temperature. The sections were then rinsed in PBS and the secondary antibodies (goat anti-rabbit Ig Alexa Fluor® 488 and goat anti-mouse Ig Alexa Fluor® 594) applied for 1 h. After washing in PBS, sections where incubated with 0.5% Sudan Black (Sigma) and nuclei visualized with DAPI, washed and mounted in glycerol/PBS. This protocol was also used on formalin fixed-snap-frozen mouse tissues from chronic secondary progressive EAE tissues with the exception of a mouse-on-mouse blocking step using CleanVision (Immunologic).
The neurological scores are presented as the mean daily neurological score ± SEM. Differences in scores in EAE and experimental autoimmune uveitis were assessed using Mann-Whitney U statistics. The parametric data, including RotaRod and retinal ganglion cells number, were assessed using a student’s t-test or one-way ANOVA incorporating tests for equality of variance using Sigmastat/Sigmaplot software (Systat Software, Inc.).
Nav blockade is neuroprotective in central nervous system autoimmunity
Spinal cord homogenate-induced relapsing-progressive disease in Biozzi ABH mice was used to examine the neuroprotective influence of Nav blockage. When drug administration was avoided during the sensitization periods, there was no evidence of any marked immunosuppressive effect of administration intraperitoneally of either 1 mg/kg or 10 mg/kg carbamazepine. All animals developed EAE with comparable high severity (Fig. 1A). This indicated marked inflammatory cell recruitment into the CNS (Allen et al., 1993; Jackson et al., 2009), in contrast to that occurring with T cell immunosuppressive agents (Watson et al., 1991; Baker et al., 1992; O'Neill et al., 1992, 1993). The maximum disease score of animals in relapse was comparable between vehicle and carbamazepine-treated animals (Fig. 1A and B); however, the residual deficit during remission from relapse was significantly reduced (P < 0.001). Vehicle-treated mice remitted to a score of 3.1 ± 0.2 compared with 0.8 ± 0.1 (1 mg/kg) or 0.8 ± 0.3 (10 mg/kg) for carbamazepine-treated animals. Although there was no statistically significant difference in the scores of groups before treatment onset it appeared that following randomization to the RotaRod score, there were more animals with residual limb paresis in the vehicle group after unblinding (Fig. 1A). As a consequence of this, for subsequent experiments we set an exclusion criterion such that animals with residual limb paresis after the initial attack were not entered into the randomization or treatment processes. Applying this criterion to remove animals from the experiment (Fig. 1A) such that the score before treatment onset was 0.5 ± 0.0, the data, however, still demonstrated that there was no difference in the severity of the relapse [3.6 ± 0.4 (vehicle) verses 3.6 ± 0.2 (10 mg/kg) and 3.7 ± 0.1 (1 mg/kg) carbamazepine] but carbamazepine-treated animals demonstrated a significantly better recovery [2.6 ± 0.7 (vehicle) verses 0.8 ± 0.3 (10 mg/kg) and 0.5 ± 0.0 (1 mg/kg) carbamazepine (P < 0.05)]. This recovery was mirrored in the motor functional analysis using the RotaRod whereby animals treated with carbamazepine were able to stay on the RotaRod significantly (P < 0.001) longer (1 mg/kg 191.6 ± 25.3 s, 10 mg/kg 131 ± 17.1 s) than those treated with vehicle (35.4 ± 11.6 s) after 14 days of treatment.
These data suggest that treatment does not stop the cause of disease or lesion formation indicating a lack of lymphoid cell-induced immunosuppression (Baker et al., 2011). In an additional model (Fig. 1C), experimental autoimmune uveitis in C57BL/10.RIII mice (Hankey et al., 2001), daily intraperitoneal injection from Day 8–22 of 10 mg/kg carbamazepine limited the accumulation of retinal nerve damage in the eye (P < 0.02), without significantly influencing the degree of immune infiltration with the retina (Fig. 1C). This was more neuroprotective than similar delivery of 10 mg/kg memantine (a weak N-methyl-D-aspartate glutamate receptor antagonist) or minocycline (microglial-inactivating antibiotic).
Time window for Nav blockers to influence the inflammatory penumbra
While treatment was therapeutic as animals were already sensitized and disease had manifested before therapy was initiated (Fig. 1A and B), the protective effect of administration after disease onset was investigated (Fig. 2). Carbamazepine and oxcarbazepine were administered at various times after the onset of signs and the influence on recovery assessed. In the first set of experiments, treatment with carbamazepine, oxcarbazepine or vehicle was initiated at the onset of signs (Day 0) or 2 days later and continued until the end of the experiment ∼Day 50 post-induction. Animals treated with carbamazepine and oxcarbazepine from Day 0 (day of relapse onset) had significantly less disability (P < 0.01 and P < 0.05, respectively) than those treated with vehicle control as assessed using the neurological scoring system. This improvement was also seen in the RotaRod, in which animals treated with both sodium channel blockers performed significantly (P < 0.001) better than those treated with vehicle control. When treatment was delayed a further 2 days after onset of relapse, the beneficial effect on the neurological score was not as pronounced, but there was some improvement in RotoRod scores (Fig. 2B and D). In an additional experiment, treatment with oxcarbazepine was delayed until Day 4 post disease-onset (Fig. 2E and F). At this time point, 4 days post disease-onset, the drug had no effect on clinical score or RotaRod performance (Fig. 2E and F). Animals to be treated were selected before randomization to treatment on the basis that their weight had stabilized or improved from the previous day. Paralysed animals that are gaining weight are beginning to remit and have a reduction in blood–brain barrier dysfunction manifest as cellular recruitment and plasma protein extravasation (Al-Izki et al., 2012). This suggests that the Nav blocking compounds were most active during lesion formation and periods of blood–brain barrier dysfunction. Once the natural regulatory mechanisms that limit relapse were initiated, Nav blockage failed to augment the recovery response.
Although RotoRod activity strongly correlates with spinal nerve loss (Al-Izki et al. 2012), unfortunately it was not possible to measure spinal cord nerve content for technical reasons. Therefore, to demonstrate the neuroprotective effects of these compounds histologically, additional experiments were performed using a transgenic mouse model of multiple sclerosis that develops optic neuritis, where EAE typically remains subclinical after stimulation of MOG-specific transgenic T cell receptors (Lidster et al. 2013). As a consequence of autoimmunity to MOG within the optic nerve, consequent retinal ganglion cells loss within the retina can subsequently be monitored using fluorescence microscopy to detect CFP expressing retinal ganglion cells (Fig. 3A–D). Sodium channel blockage with either 10 mg/kg carbamazepine or oxcarbazepine (Fig. 3A, B and D) both significantly (P < 0.05) limited the amount of nerve loss compared to the influence of vehicle (ECP) occurring as a consequence of optic neuritis (Fig. 3C and D).
Selective loss of energy-dependent efflux transporters during neuroinflammatory disease
The clinical use of Nav channel blockers in multiple sclerosis can be associated with the development of neurological side effects at doses below those used for anti-convulsive activity in epilepsy (Kapoor et al., 2010). Therefore, by selectively targeting potential neuroprotective drugs to active lesions, it may be possible to improve the therapeutic window for this class of drugs. Although the nature of blood–brain barrier dysfunction is varied, it is well-known that this is associated with leakage of cells and plasma proteins in EAE and multiple sclerosis (Butter et al., 1991; Al-Izki et al., 2012). One aspect that contributes to this—as shown here—is the selective loss of p-glycoprotein (ABCB1) within lesions. This occurs both in EAE (Fig. 4) and multiple sclerosis (Fig. 5) lesions. In comparison with vascular p-glycoprotein expression in normal mice (Fig. 4A), p-glycoprotein was selectively lost in active lesions during relapsing disease (Fig. 4B and C) in comparison with BCRP (now known as ABCG2, Fig. 4D) and compared with non-inflamed tissues (Fig. 4C and E–K). P-glycoprotein was lost from the vasculature (Fig. 4H–K) as shown using double-labelling of p-glycoprotein with Glut-1 on endothelia (Fig. 4L–N). This vascular loss of p-glycoprotein remained during post-relapsing secondary progressive EAE (Fig. 4L–N) and in comparison with normal brain tissue (Fig. 5A) and normal-appearing white matter in multiple sclerosis (Fig. 5B), p-glycoprotein was also absent from the vasculature in chronic inactive in multiple sclerosis lesions (Fig. 5C and D). Interestingly there was an upregulation of astrocytic p-glycoprotein in such chronic lesions (Fig. 5D) and non-vascular p-glycoprotein expression seemed to occur in EAE lesions (Fig. 4B and L–N). Again, in humans there was selective loss of p-glycoprotein in contrast to vascular expression of BCRP in multiple sclerosis lesions (Fig. 5E). Double-labelling of p-glycoprotein with endothelial Glut-1 (Fig. 5F–K) demonstrated vascular expression of p-glycoprotein in vessels in normal-appearing white matter (Fig. 5F–H), which was reduced or lost in lesions (Fig. 5I–K) although it was detected in non-Glut-1-positive cells even in active multiple sclerosis lesions (Fig. 5J and K).
Therefore local delivery to injured neurons of p-glycoprotein substrates that are normally excluded from the CNS may allow selective targeting of drugs to EAE and multiple sclerosis lesions.
Production of a lesion-targeting central nervous system-excluded, sodium-channel blocking agent
A series of sodium channel blockers were synthesized (Clutterbuck et al., 2009), and to test our hypothesis, we generated a number of molecules, of which, CFM6104 [N-((5-(2-(2-(1h-imidazol-1-yl)ethyl)-2h-indazol-3-yl)-1,2,4-oxadiazol-3-yl)methyl)-3-methyl-benzamid] exhibited moderately favourable pharmacokinetic properties (Browne et al., submitted for publication). This molecule was predicted to be traceable using fluorescence because of the presence of an indazole group in the structure and had a blood/brain partition coefficient (Rishton et al., 2006) of −0.72 that would predict ∼80% CNS exclusion. Using membranes transfected with human ABCB1 (Fig. 6A) and a Caco-2 bioassay (Table 1) the compound was found to be a p-glycoprotein substrate and, accordingly, its transport was facilitated by the p-glycoprotein inhibitor cyclosporin A. These findings indicated that the transporter would normally actively remove CFM6104 from the CNS. When administered intravenously, CFM6104, failed to exhibit marked brain penetration at Cmax indicative of CNS exclusion following intravenous administration (Table 2). The intraperitoneal route gave a better duration of action than the oral route and was used for in vivo studies. Following intraperitoneal injection of CFM6104 to ABH mice there was >90% exclusion in the spinal cord (Table 3). In EAE there was no apparent enhancement of CNS penetration, despite a significantly (P < 0.001) higher plasma level than in healthy animals (Table 3). This is probably related to a slowing of drug metabolism following the development of hypothermia that develops in the mice during periods of immobility during active EAE (Al-Izki et al., 2012). However, as CFM6104 was a naturally fluorescent compound, it could be traced. It was found to be excluded from the CNS in normal animals (Fig. 6B and C) but was able to access inflammatory lesions in the spinal cords of EAE animals (Fig. 6D and E). This indicated that the drug was indeed targeting and penetrating lesions, presumably by a mechanism involving downregulation of p-glycoprotein at the blood–brain barrier in the vicinity of lesions.
|Efflux ratio (external versus internal)|
|Drug||Drug alone||Drug + CacoReady™ assay|
|Efflux ratio (external versus internal)|
|Drug||Drug alone||Drug + CacoReady™ assay|
Caco2 cells were grown on a membrane; compounds were added to the apical or basolateral compartments and assayed in those compartments by LC-MS 2 h later. Drug concentrations were quantified by LC-MS/MS.
|Mean CFM6104 levels (pg/µl)|
|Time (min)||Plasma (intraperitoneal)||Plasma (intravenous)||Brain (intravenous)|
|5||445 ± 71||1107 ± 32||<3.0 ± 0.0|
|15||373 ± 38||212 ± 6||<3.0 ± 0.0|
|30||129 ± 12||63 ± 24||<3.0 ± 0.0|
|Mean CFM6104 levels (pg/µl)|
|Time (min)||Plasma (intraperitoneal)||Plasma (intravenous)||Brain (intravenous)|
|5||445 ± 71||1107 ± 32||<3.0 ± 0.0|
|15||373 ± 38||212 ± 6||<3.0 ± 0.0|
|30||129 ± 12||63 ± 24||<3.0 ± 0.0|
The results show the mean (n = 3) amount of compound detected.
|Mean CFM6104 levels (pg/µl)|
|Normal||Vehicle||4||<1.0 ± 0.0||<1.0 ± 0.0|
|Normal||CFM6104||4||124.0 ± 34.2||10.0 ± 6.4|
|EAE||CFM6104||3||985.0 ± 90.4*||18.9 ± 6.8|
|Mean CFM6104 levels (pg/µl)|
|Normal||Vehicle||4||<1.0 ± 0.0||<1.0 ± 0.0|
|Normal||CFM6104||4||124.0 ± 34.2||10.0 ± 6.4|
|EAE||CFM6104||3||985.0 ± 90.4*||18.9 ± 6.8|
The results represent mean ± SEM levels. * P < 0.05 compared to normal mice.
Neuroprotective effect of a central nervous system-excluded, lesion targeting sodium-channel blocker
The dominant antigen in spinal cord-induced EAE in ABH mice is within the proteolipid protein that gives a poor in vitro proliferative response (Pryce et al., 2005; Smith et al., 2005). Therefore, the immunosuppressive effect of CFM6104 on T cell function was investigated using a simple contact-hypersensitivity paradigm that has been used to identify in vivo immunosuppressive doses (Watson et al., 1991; Baker et al., 1992; O'Neill et al., 1992). Compared with vehicle treatment, CMF6104 failed to inhibit lymph node proliferative responses after daily intraperitoneal administration of 5 mg/kg, 25 mg/kg and 50 mg/kg from a day before sensitization to the time of collecting the lymph nodes (Browne et al., submitted for publication). Therefore, it was felt unnecessary to dose-titrate CFM6104 to avoid any immunosuppression-induced neuroprotection that could mask any real neuroprotective action of this drug. To limit animal usage required for a full-dose-response study, we selected 10 mg/kg CFM6104 administered intraperitoneally as a proof of principle dose. This dose was within the range used in humans for sodium channel-blocking agents (1–50 mg/kg per os), particularly when scaled to mouse and was well below the maximum tolerated dose (>50 mg/kg/daily). Intraperitoneal administration of 10 mg/kg CFM6104 was used to investigate its neuroprotective potential in an induced-relapse during EAE. The lack of immunosuppressive activity of CFM6104 was indicated further by the observation that CMF6104 did not significantly (P > 0.05) inhibit the incidence, onset, or maximum severity of relapsing EAE (Fig. 7A and B). Although CFM6104 did not inhibit the severity of paralysis during relapse (Fig. 7A), consistent with a lack of overt lymphoid cell immunosuppression, it did promote a significantly (P < 0.01) greater degree of recovery of motor-function during remission (Fig. 7B). This could be seen using the subjective clinical scoring system (Fig. 7B) and objective motor-function using the RotaRod (P < 0.001, Fig. 7C). These improved motor outcomes were paralleled by the ability of CMF6104 to reduce spinal cord nerve loss, a consequence of relapsing disease, as assessed by an ELISA specific for neurofilament, (Fig. 7D). Therefore, CMF6104 did exhibit a neuroprotective effect and slowed the loss of nerves due to inflammatory attack.
This study provides further evidence that Nav blockers have the potential to limit nerve loss in the progressive phase of EAE. Although this finding is not novel, this current study differs from many in that the clinical activity is not simply secondary to an immunosuppressive action exerted within the CNS (Lo et al., 2003; Black et al., 2007). In the present study the compounds were delivered to sensitized animals with disease at a time that does not to interfere with priming of T cells and at a dose with no overt T cell immunosuppression; based on the clinical profile and ex vivo T cell assays as shown with carbamazepine, oxcarbamazepine and CFM6104. Interestingly, inhibition of T cell responses was not seen either with boluses of 1, 10 and 30 mg/kg phenytoin (50 mg/kg was toxic) (unpublished observation) or in the rat with lamotrigine (Bechtold et al., 2006) and contrasts with that seen with peripherally-active immunosuppressive agents (Baker et al., 1992; O'Neill et al., 1993). Although we have not seen immunosuppressive activity in the Nav blockers we examined, such an activity could be of additional benefit in arresting autoimmune-dependent neurodegeneration. It has been shown that prophylactic peripheral immunosuppression can account for the protection of EAE (Black et al., 2007), however, therapeutic doses of agents used here were not overtly T cell immunosuppressive. The degree of inflammation was not directly measured in the EAE as clinical course was the primary outcome measure; however, the degree of inflammation directly correlates with the severity of clinical disease during disease development and is dynamic such that inflammatory infiltrates are lost as animals begin to remit (Baker et al., 1990, 2011; Allen et al., 1993; Wujek et al., 2002; Jackson et al., 2009). In this study there was no indication of immunosuppression of the generation of inflammation as assessed histologically in the uveitis model and by the fact that the onset and severity of clinical disease was unaffected. However, inhibition of peripheral immunity preventing lesion formation (Black et al., 2007) in addition to anti-inflammatory effects within the CNS once lesions form (Bechtold et al., 2006; Morsali et al., 2013) and direct neuroprotection by an action on nerves (Fern et al., 1993; Garthwaite et al., 1999; Kapoor et al., 2003) will contribute to the overall neuroprotective potential of this class of agents.
Clinical trials in multiple sclerosis of sodium channel blockers have been halted because of concern about a rebound effect following cessation of treatment that occurred in animals (Black et al., 2007; Waxman, 2008). This can occur when immunosuppressive doses of compounds that prevent disease development are removed, which allows CNS autoimmunity to return (Greenwood et al., 2003; Black et al., 2007). As the compounds reported here were found not to be immunosuppressive and did not inhibit the generation of relapse, we would not anticipate, and did not look for, a rebound effect following cessation of treatment. However, in the absence of peripheral immunosuppression, sodium channel blockade also induces a neuroprotective anti-inflammatory effect through inhibition of microglial function (Bechtold et al., 2006; Morsali et al., 2013). Lamotrigine treatment resulted in what was thought to be pseudoatrophy in brain volume because of some anti-inflammatory effect reducing tissue oedema (Kapoor et al., 2010). Following cessation of lamotrigine treatment brain volume increased suggestive of some return of disease activity, but this was not associated with a rebound exacerbation in secondary progressive multiple sclerosis (Kapoor et al., 2010).
We have previously demonstrated that the neurofilament ELISA provides a good correlate to histological assessment of axon numbers (Jackson et al., 2005, 2009; Al-Izki et al., 2012) and neurofilament release into biological fluids as a consequence of nerve damage is a reliable biomarker for neurodegeneration in people with multiple sclerosis and other disorders (Gunnarsson et al., 2011; Gnanapavan et al., 2013). Neurofilament measurement offers the advantage of rapid analysis of the whole spinal cord and provides an outcome measure that can be translated into human disease (Gunnarsson et al., 2011; Gnanapavan et al., 2012, 2013). This technique lacks the bias associated with the use of histological analysis, which is not only time-consuming but invariably focuses analysis on restricted-CNS tracts. Any marked neuroprotection by Nav blockers in sensory tracts may not be reflected in the motor outcomes measures (Lo et al., 2003; Bechtold et al., 2006). Clinical scoring systems commonly used in EAE have been developed to assess active paralysis similar to the Expanded Disability Status Scale (EDSS) that is used for multiple sclerosis. These are subjective, user-dependent and are subject to bias of analysis, as treatment effects are greater in studies not reporting blinding (Vesterinen et al., 2010). Furthermore, they are non-linear scales that are insensitive to change. This is especially the case in assessing functional deficits in animals that occur as a consequence of paralytic attacks (Al-Izki et al., 2012). However, residual neurological deficit after attack can be used to detect neuroprotective effects (Baker et al., 2011). More importantly the accelerating RotaRod appears to be a sensitive, objective and quantitative outcome measure that can detect motor deficits and functional improvements during remission from paralytic attacks (Al-Izki et al., 2011, 2012). This shows a strong correlation with spinal nerve content (Al-Izki et al., 2012). These two outcomes together showed significant treatment responsiveness to Nav blockade. In addition, the use of the novel optic neuritis model provides a further tool to confirm the neuroprotective effects of sodium channel blockers.
Although it has been reported that the Nav blocker lamotrigine is ineffective in progressive multiple sclerosis (Kapoor et al., 2010), it is premature to discard this class of agents on the basis of this result. Lamotrigine did not appear to influence the course of multiple sclerosis, except perhaps by slowing the decline in the 25 m walk (Kapoor et al., 2010). However, lamotrigine in comparison with other compounds is a weak neuroprotectant in vitro (Garthwaite et al., 1999). In preclinical EAE studies (group sizes of n = 70), lamotrigine demonstrated a marginal (P < 0.05) clinical improvement and histological analysis of selected sensory nerve tracts showed a 30% reduction in axonal loss in a subset (50% of animals) of the experimental animals (Bechtold et al., 2006). Moreover, in a trial in patients with multiple sclerosis, the primary MRI outcome measure of cerebral volume was complicated by accelerated atrophy in the first year as mentioned previously. Unfortunately, the trial was too short to accommodate pseudoatrophy in a positive fashion (Kapoor et al., 2010). Furthermore, the patient cohort may not have been optimal for this study. The disease in the placebo group was slowly progressing with few clinical changes from baseline. Importantly, the drug was also poorly tolerated (Kapoor et al., 2010). This was indicated by the high number of drop-outs during the study, the inability of many to reach the target anti-epileptic dose because of side effects and most importantly, the observation that ∼40–50% of patients in the active arm were not drug-compliant as defined by serum levels (Kapoor et al., 2010; Gnanapavan et al., 2013; Raftopoulos and Kapoor, 2013). Therefore, it is not surprising that the main study failed to show a drug effect. Nevertheless, there was a significant reduction in serum neurofilament levels between placebo and drug-compliant individuals, indicating that Nav blockage was neuroprotective (Gnanapavan et al., 2013), as found here in EAE. Therefore, study of people with multiple sclerosis at earlier stages of their disease, when they can better tolerate Nav blockage is warranted.
We demonstrate that Nav blockade is neuroprotective in a novel experimental optic neuritis model that exhibits subclinical EAE and offers a less severe, 3Rs-refined alternative to classical paralytic EAE. Both oxcarbazepine and carbamazepine inhibited the consequences of optic neuritis in the optic nerve that is reflected by loss of retinal ganglion cells in the retina (Bettelli et al., 2003; Guan et al., 2006). This is consistent with the neuroprotective effect of 5 mg/kg intraperitoneal of CFM6104 in the same model and treatment paradigm that produced significant (P < 0.05) neuroprotection of retinal ganglion cells (Browne et al., submitted for publication). This concept is currently being tested in an optic neuritis trial (Raftopoulos and Kapoor, 2013).
Although a desired mechanism for treatment of progressive multiple sclerosis is through prevention of sodium loading in nerves (Dutta and Trapp, 2011), the neuroprotective influence found here is probably anti-inflammatory within the CNS. While the data may suggest that such sodium channel blockade may have some benefit in the control of progression, the results of the current study are best translated to treatment during relapsing multiple sclerosis. Further studies to examine the influence in post-relapsing secondary progressive EAE (Al-Izki et al., 2011) are warranted to explore the neuroprotective potential of sodium channel blockade for progressive multiple sclerosis. That p-glycoprotein is lost in chronic inactive lesions, where blood–brain barrier dysfunction is reduced, may allow neuroprotection of demyelinated nerves to be targeted. These are particularly vulnerable to damage due to metabolic demands required to maintain conduction following rearrangement of Nav1.6 from the nodes of Ranvier during demyelination (Craner et al., 2004). Low levels of these receptors are expressed along the length of demyelinated axons and the novel expression of Nav1.2 occurs on these nerves to maintain conduction (Craner et al., 2004). Local targeting of axonal conduction in areas of myelin-denuded axon may, however, still cause some negative symptomatic deficits because of impedance mismatch and through influences of low density of sodium channels on demyelinated axons (Waxman and Brill, 1978). Although the risk of this is perhaps reduced compared to the globally active fully penetrant CNS compounds, side effects were not noticed during this study; however, the aim was not to induce toxicity and therefore low doses of compounds were used. Nevertheless, it was noted that although doses of 50 mg/kg intraperitoneal CFM6104 were well tolerated, phenytoin induced toxicity and caused fatality at the same dose, supporting the concept that CNS-excluded compounds may have a better tolerability profile.
It is known that Nav blockers inhibit macrophage and microglial activation (Black and Waxman, 2012; Morsali et al., 2013). Through this mechanism, the neurodegenerative effect of the inflammatory-penumbra that is formed around the perivascular lesions by T cell invasion correlates in time with the period of active blood–brain barrier dysfunction. In mice, this ‘inflammatory penumbra’ lasts for a few days and correlates with the period of time that the blood–brain barrier is disrupted. In humans, it is likely to be longer because acute inflammatory multiple sclerosis lesions are known to enhance with gadolinium on MRI, signifying blood–brain barrier dysfunction for a few weeks (Butter et al., 1991; Cotton et al., 2003; Al-Izki et al., 2012). For this reason we are recruiting and dosing volunteers for the acute optic neuritis study (NCT01451593; Raftopoulos and Kapoor, 2013) within 14 days of disease onset, when lesions would have focal disruption of the blood–brain barrier as detected with gadolinium-enhanced MRI. Furthermore, as there are many more MRI lesions than clinical attacks the approach of selectively targeting lesions is important even when the disease is in apparent clinical remission. Although our studies may indicate an anti-inflammatory effect, which may occur in multiple sclerosis as seen by a reduction in tissue oedema (Kapoor et al., 2010), this is unlikely to be because of a lymphoid cell immunosuppressive effect that is hypothesized to drive relapsing disease. Therefore, an optimal therapeutic strategy may require a combination of a systemic immunosuppressive agent combined with a neuroprotective agent. We initially examined carbamazepine, which has side effects such as drowsiness, dizziness, ataxia and nausea in humans. However, oxcarbazepine, which is an analogue of carbamazepine, was slightly more active as a neuroprotectant in our hands and has a lower incidence of side effects and is therefore better tolerated (Elger and Bauer, 1998). We therefore plan to use oxcarbazepine in another novel trial (Protective Role of OXcarbazepine In MUltiple Sclerosis or PROXIMUS study) that will use neurofilament levels in serial CSF analyses as the primary outcome measure. This provides another example of how animal studies can help inform the development of new treatments for people with multiple sclerosis.
Loss of p-glycoprotein expression in multiple sclerosis has been independently confirmed and it has been shown that p-glycoprotein is lost in active, chronic active and chronic inactive multiple sclerosis lesions (Kooij et al., 2010, 2011). Blockade of p-glycoprotein enhances brain penetration of p-glycoprotein substrates as shown using PET (Nojiri et al., 2008) and therefore loss of vascular p-glycoprotein may potentially allow targeting of acute multiple sclerosis lesions similarly to the way in which active EAE lesions were targeted here. Despite this potential benefit for pharmacotherapy, loss of p-glycoprotein expression may contribute to the neurotoxicity that drives progression of the disease. Downregulation of p-glycoprotein occurs after CD54 cross-linking associated with leucocyte diapedesis (Kooij et al., 2010). However, there are other factors that regulate p-glycoprotein expression (Miller et al., 2008) and loss of p-glycoprotein is observed in both chronic active and importantly chronic inactive lesions, which are not all associated with perivascular cuffing as shown here. Therefore, Nav blocking agents that are p-glycoprotein substrates would be able to inhibit both the progressive neurodegenerative processes in nerves and the glial cell activation that drive neurodegeneration in vivo. We have produced a Nav blocker that was selectively CNS-excluded and traceable. However, LC-MS studies failed to demonstrate CFM6104 penetration during EAE, which was surprising because active disease is associated with blood–brain barrier dysfunction in the spinal cord (Al-Izki et al., 2012). This suggests that either only small amounts were penetrating the lesions or that CNS CFM6104 was not being detected because of technical reasons such as binding of the hydrophobic compound to CNS structures, (e.g. myelin) or it was rapidly metabolized in the brain. However, the compound could be seen to penetrate the lesions and this had an action consistent with CNS penetration and may indicate that only small amounts of compound are required for a biological effect.
Current Nav blockers, given their general hydrophobicity, may be expected to show some affinity for ABC efflux transporters and thus may partially exploit this targeting approach (Crowe and Teoh, 2006; Nakanishi et al., 2013). CFM6104 is a CNS-excluded p-glycoprotein substrate and yet it clearly targeted lesions during EAE and induced neuroprotection. Although CFM6104 was sufficient to provide proof of principle, further development of this class of compound is warranted to improve potency at the target and predicted pharmacokinetic profile. This selective targeting strategy could be developed for other drugs, where global CNS penetration and receptor stimulation may be problematic, and provides new avenues for future treatment of multiple sclerosis.
Dr Edmund Hoppe Merck Serono, Grafing, Germany and Merck Serono and Adriano Soares de Souza, Merck Serono, Geneva, Switzerland are thanked for performing and supporting some of the pharmacokinetic studies. Dr Ann Wheeler is thanked for her help with microscopy.
The authors would like to thank the support of the National Multiple Sclerosis Society (USA) and the Multiple Sclerosis Society of Great Britain and Northern Ireland.
Conflict of interest
D.B., L.C., J.G., C.P. and D.L.S. have filed patents related to the compounds used in this study. G.G. has provided consultancies for many pharmaceutical companies in the multiple sclerosis field. This had no influence on the design and execution of this study.
experimental autoimmune encephalomyelitis
- sodium channel blockers
- immunosuppressive agents
- central nervous system
- blood-brain barrier
- experimental autoimmune encephalomyelitis
- mice, transgenic
- multiple sclerosis
- nerve degeneration
- neuroprotective agents
- optic neuritis
- sodium channel
- nerve injury