Complement-dependent pathogenicity of brain-specific antibodies in cerebrospinal fluid

Abstract

The specificity and potential pathogenicity of autoantibodies vary between neurological diseases. It is often unclear whether their detection in cerebrospinal fluid (CSF) is a consequence or a cause of pathology. The goal was to test whether administration of brain-specific antibodies into CSF would be sufficient for pathology. Purified immunoglobulin G from a neuromyelitis optica patient was injected intrathecally with complement to naïve mice. Histopathological analysis at 7 days revealed damage to the ependyma, disruption of the CSF parenchymal barrier and pathologic lesions, distant from the site of injection. In the absence of complement there was no pathology. Autoantibody and complement in CSF are thus sufficient to initiate a pathologic cascade.

Introduction

Inflammatory disorders of the central nervous system (CNS) are frequently accompanied by increased levels of immunoglobulin in cerebrospinal fluid (CSF). The inflammatory diseases may have infectious causes such as Lyme disease, or non-infectious causes, possibly including systemic inflammatory diseases e.g. systemic lupus erythematosus (SLE), or multiple sclerosis (MS). Elevated levels of CSF-IgG and the presence of oligoclonal IgG bands in CSF are characteristics of some inflammatory diseases, notably MS (Kabat et al., 1942), and a useful diagnostic tool. However, despite correlation to disease progression there is no consensus for MS-associated antibody specificities in CSF that would indicate pathogenicity of this oligoclonal IgG. Antibody specificities such as anti-glutamate receptor antibodies have been associated with a clinical encephalitis syndrome as a paraneoplastic phenomenon (Dalmau et al., 2008). Even where autoantibodies are useful biomarkers in such cases (Dalmau and Rosenfeld, 2008), their pathogenic importance has not been clarified (Bernal et al., 2002). By contrast, anti-aquaporin-4 (AQP4) antibody–positive IgG from neuromyelitis optica (NMO) patients is a distinct biomarker for NMO‐spectrum disease (Lennon et al., 2004, Weinshenker et al., 2006) and immunopathological evidence implicates anti-AQP4 antibodies/NMO-IgG in the pathogenesis of NMO (Lucchinetti et al., 2002, Roemer et al., 2007, Bennett et al., 2009, Bradl et al., 2009, Kinoshita et al., 2009, Saadoun et al., 2010). Although oligoclonal bands and increased titers of IgG in the CSF occur significantly less often in NMO than in MS (Wingerchuk et al., 2007), anti-AQP4 antibodies/NMO-IgG are detectable in the CSF of 68% of NMO patients who have acute disease relapse and high NMO-IgG serum titers (Takahashi et al., 2007, Jarius et al., 2010). Exacerbation of NMO is related to high AQP4 antibody levels in serum and presence of such antibodies in CSF (Takahashi et al., 2007, Jarius et al., 2010). As for CSF antibodies in MS and other neuroinflammatory diseases, the pathogenic potential of these CSF antibodies, although sometimes presumed, is unknown.

Transfer of disease with autoantibodies or induction of pathology in animals by antibody transfer would seem a useful criterion for pathogenicity. However, this approach is often compromised by lack of access, particularly due to the fact that the intact blood–brain barrier (BBB) prevents antibodies entering the CNS parenchyma. Thus, many studies have shown that anti-myelin antibodies cannot transfer MS-like disease to mice, whereas in conjunction with an autospecific T cell response, such antibodies promote demyelination (Schluesener et al., 1987, Iglesias et al., 2001). Similarly, co‐administration of anti-AQP4 or of NMO-IgG to animals in which myelin-specific T cell responses were ongoing showed pathogenicity of those antibodies (Bennett et al., 2009, Bradl et al., 2009). An alternative approach has been to bypass the BBB by directly injecting the antibody into parenchymal tissue (Saadoun et al., 2010). Antibodies with specificity for elements of the BBB might be expected to have greater potential for direct induction of CNS pathology but this has been difficult to test due to considerations of antibody titer in blood and the complexity of the cellular barriers and basement membranes that form the BBB.

To address these questions we utilized the intrathecal route (into the cisterna magna) for direct administration of antibody with complement into CSF. We used disease-derived purified antibody that included specificity for AQP4. AQP4 is densely localized in membranes of ependymal cells and astrocytes that form the glia limitans of the BBB and the CSF-parenchymal barrier (Nielsen et al., 2002, Amiry-Moghaddam and Ottersen, 2003, Asgari et al., 2011b) and is highly expressed in ependymal cells lining the ventricles and to a lesser degree in the central canal of the spinal cord (Oshio et al., 2004, Goren et al., 2006). The aim of our study was to investigate potential for direct pathogenicity of such antibody in CSF, and to determine the distribution of antibody-induced lesions in the CNS. The results show that antibody accessed the parenchyma via complement‐mediated disruption of the CSF-parenchymal barrier leading to loss of AQP4 and glial fibrillary acidic protein (GFAP, an astrocytic marker), and to deposition of C9neo (a marker of membrane attack complex, i.e. activated complement). Demyelination was observed at topographically restricted sites in the cerebellum, brainstem and in periventricular areas. Pathology was dependent on co-injection of human complement. Thus, under appropriate circumstances, autoantibodies within the CSF can induce pathogenesis and must be included in consideration of etiology of neuroinflammatory disease.