Paraneoplastic and non-paraneoplastic autoimmunity to neurons in the central nervous system

Autoimmune central nervous system (CNS) inflammation occurs both in a paraneoplastic and non-paraneoplastic context. In a widening spectrum of clinical disorders, the underlying adaptive (auto) immune response targets neurons with a divergent role for cellular and humoral disease mechanisms: (1) in encephalitis associated with antibodies to intracellular neuronal antigens, neuronal antigen-specific CD8+ T cells seemingly account for irreversible progressive neuronal cell death and neurological decline with poor response to immunotherapy. However, a pathogenic effect of humoral immune mechanisms is also debated. (2) In encephalitis associated with antibodies to synaptic and extrasynaptic neuronal cell surface antigens, potentially reversible antibody-mediated disturbance of synaptic transmission and neuronal excitability occurs in the absence of excessive neuronal damage and accounts for a good response to immunotherapy. However, a pathogenic effect of cellular immune mechanisms is also debated. We provide an overview of entities, clinical hallmarks, imaging features, characteristic laboratory, electrophysiological, cerebrospinal fluid and neuropathological findings, cellular and molecular disease mechanisms as well as therapeutic options in these two broad categories of inflammatory CNS disorders.


Introduction
The human central nervous system (CNS) can be targeted by aberrant cellular and humoral immune responses, which can either be triggered by systemic infections, and vaccinations (''postinfectious/-vaccinal autoimmune encephalitis'') and a variety of cancers (''paraneoplastic autoimmune encephalitis'') or occur without an (yet) identifiable cause (''non-paraneoplastic autoimmune encephalitis'') [25]. The scope of neurological disorders, in which such misguided adaptive immune responses are directed towards the oligodendrocyte and myelin-sheath is well described [24]. However, in a variety of immune-mediated CNS disorders, neurons seem to be targeted by adaptive cellular and humoral (auto) immune responses of both paraneoplastic and non-paraneoplastic origin [35]. Here, we summarize important clinical phenotypes together with their typical paraclinical measures, putative disease mechanisms, and therapeutic options in this emerging class of inflammatory CNS disorders.

Prerequisites for neuron-directed autoimmunity in the CNS
Antigen-specific cellular and humoral immune responses directed towards CNS neurons are believed to develop as a multi-step process [73]. Soluble or cell-bound neuronal or neuronal-like antigens are engulfed and presented in the context of MHC II and co-stimulatory molecules to CD4 ? T cells by professional antigen-presenting cell (APCs) within secondary lymphatic organs (e.g., cervical lymph nodes). This in turn permits CD4 ? T cells via cytokine secretion and ligation of CD40 to license APCs to crosspresent these antigens in the context of MHC I and costimulatory molecules to naive CD8 ? T cells, which then become activated and may acquire cytotoxic effector functions (cellular effectors). A lack of such CD4 ? T cell help usually results in anergy of CD8 ? T cells. Depending on the local cytokine milieu mainly provided by CD4 ? T cells, CD8 ? T cells with different functional polarization may develop [86]. Tc1 cells are differentiated in the presence of IL-2 and IL-12, which induce the transcription factor T-bet. They produce IFN-c and TNF-a and exert strong cytotoxicity. Tc2 cells are differentiated in the presence of IL-4, which induces the transcription factor GATA-3. They produce IL-4, IL-5, and IL-13 and exert a less robust cytotoxicity. Tc17 cells are differentiated in the presence of TGF-b and IL-6, which mainly induce the transcription factor RORct. They produce IL-17 and have been reported to exert only weak cytotoxicity.
Naive B cells produce both IgM (and IgD) that are anchored in their plasma membrane and function as BCRs [48]. Naive B cells that encounter, ingest, and present their cognate so-called ''thymus-dependent (TD) antigen'' in the context of MHC II and co-stimulatory molecules to CD4 ? T cells are in turn activated via cytokine secretion and ligation of CD40 and become antibody-secreting plasma cells (humoral effectors). Thereby, B cells can further diversify their Ig-genes by two DNA-modifying mechanisms [48]; somatic hypermutation and class switch recombination generate highly specific and adapted humoral responses. Somatic hypermutation introduces in a transcriptiondependent manner non-templated point mutations in the variable (V) region of Ig genes, thereby enabling the selection of antibodies with increased affinity for the antigen. In contrast, class switch recombination modulates antibody effector function by replacing one constant (C) region with another, while retaining the binding specificity of the BCR. Depending on the cytokine milieu mainly provided by the CD4 ? T cells, activated B cells undergo antibody class switching to produce IgG, IgA, or IgE antibodies [48]. Switch to IgG1 and IgG3 promoting complement activation and antibody-mediated cellular cytotoxicity by NK cells occurs in the presence of IFN-c. In contrast, switch to IgG2, IgG4, and IgA promoting antigen-neutralizing effects occurs in the presence of IL-4 and IL-5. Although some pathogen-derived ''thymus-independent (TI) antigens'' may induce somatic hypermutation and class switch recombination in B cells independent from CD4 ? T cell help, a lack of such help usually results in persistent secretion of complement-activating IgM (and IgD; [48]).
Following peripheral activation, both antibody-secreting plasma cells and cytotoxic CD8 ? T cells (together with CD4 ? T cells) may enter the CNS to attack neurons and cause functional and structural impairment [69,97]. Moreover, under such inflammatory conditions, even antibodies produced in the periphery may permeate the blood-brain barrier (BBB) by various paracellular and transcellular mechanisms and thus contribute to neuron-directed immunity, whereas under physiological conditions, the BBB is usually impermeable for antibodies [23] (see Fig. 1).
In general, both effector arms of the adaptive immune response may be activated irrespective of the cellular localization of the neuronal antigen or its antigenic epitope (plasma membrane vs. interior cellular compartments). In terms of relevant effector mechanisms, plasma cell-derived antibodies usually recognize discontinuous conformational epitopes composed of segments of the respective neuronal plasma membrane protein antigen that are brought together in its three-dimensional structure and exposed on the neuronal plasma membrane. Antibodies may thus specifically impact the function and expression of theses antigens. Whether antibodies may also bind to and impact the function or expression of intracellular neuronal antigens, either by passive uptake into the neuron or by active binding to intracellular antigens that are transiently exposed to the plasma membrane is currently a matter of debate [30,96]. Moreover, peptides derived from both intracellular and plasma membrane neuronal antigens might potentially be recognized by antibodies when exposed on the surface membrane in complex with MHC I molecules, although this is usually performed by CD8 ? T cells.
Cytotoxic CD8 ? T cells usually recognize continuous linear peptide epitopes consisting of 8-10 amino acids that are derived from intracellular neuronal proteins by extensive antigen processing and presented in the context of MHC I molecules on the cell surface membrane. Whether peptides derived from neuronal surface membrane antigens are also presented to cytotoxic CD8 ? T cells in the context of MHC I molecules is unclear at present. In both cases, CD8 ? T cells cannot directly impact the function or expression of their cognate antigens, but recognize their expression by the respective neuron. This enables them to contribute to neuronal dysfunction and cell death by the antigen-dependent release of effector molecules (perforin, granzymes) from cytotoxic granules. Indeed, we could show that two separate functional consequences result from a direct cell-to-cell contact between antigen-presenting neurons and antigen-specific CD8 ? T cells. (1) An immediate impairment of electrical signaling in single neurons and neuronal networks occurs as a result of massive shunting of the membrane capacitance after insertion of channel-forming perforin (and probably activation of other transmembrane conductances), which is paralleled by an increase of intracellular Ca 2? levels. (2) Antigen-dependent neuronal apoptosis may occur independently of perforin and members of the granzyme B cluster, suggesting that extracellular effects can substitute for intracellular delivery of granzymes by perforin. Thus, electrical silencing is an immediate consequence of MHC I-restricted interaction of CD8 ? T cells with neurons. Of course, these changes in neuronal excitability are not induced specifically in response to a certain antigen, but apply to all antigen-presenting neurons encountered by activated cytotoxic CD8 ? T cells [69,71].
Paraneoplastic autoimmune encephalitis is probably mediated by cytotoxic CD8 1 T cells specific for intracellular neuronal antigens An ever-growing number of paraneoplastic CNS disorders are defined by the presence of IgG antibodies in the serum and CSF directed against intracellular neuronal antigens aberrantly expressed also by tumor cells (''onco-neuronal antibodies'') [67]. These tumors often contain neuronally differentiated tissue (germ cell tumors), express certain neuroendokrine peptides (SCLC, neuroblastoma), or occur in organs with a role in immune regulation (thymoma). However, due to the intracellular localization of the antigens, the humoral immune response is considered a non-pathogenic ''epiphenomenon'' solely indicating neuron-directed immunity and defining its antigen. In contrast, a variety of findings suggest a pathogenic role of cytotoxic CD8 ? T cells for neuronal damage in these disorders: (1) neuronal damage often correlates with the number of CD8 ? T cells, (2) CD8 ? T cells are found in the CNS parenchyma in close spatial proximity to neuronal target cells, (3) CD8 ? T cells show an activated phenotype with substantial expression of the effector molecules (perforin and granzymes) in cytotoxic granules with a polar orientation towards the target cell membrane, (4) CD8 ? T cells stain positive for CD107 indicating recent exocytosis of cytotoxic granules (i.e., degranulation), (5) neuronal target cells exhibit substantial cell surface expression of MHC I molecules allowing for cognate antigen-recognition by CD8 ? Fig. 1 Putative immunopathogenesis of paraneoplastic and nonparaneoplastic auto-immunity to neurons in the CNS (modified and extended from Melzer et al. [70]). To set up an adaptive humoral and cellular immune response directed towards intracellular (left, yellow circles) or surface membrane (right, bright blue circles) neuronal antigen both professional antigen-presenting cells (APC, green) and naive antigen-specific B cells (BC, orange) need to encounter and engulf soluble or cell-bound neuronal or neuronal-like antigens within secondary lymphatic organs. Subsequently, APCs may process and present this antigen in the context of MHC II molecules to naive antigen-specific CD4 ? T cells (CD4 TC, blue). MHC II-dependent recognition by these CD4 ? T cells is then required for full activation of B cells to become antibody-secreting plasma cells (PC, red). In addition, CD4 ? T cells may license APCs to present neuronal antigen peptides in the context of MHC I molecules also to naive CD8 ? T cells (CD8 TC, blue) to acquire cytotoxic effector function. Antibodysecreting plasma cells and cytotoxic CD8 ? T cells may invade the CNS to exert humoral and cellular neuron-directed autoimmunity J Neurol (2013) 260:1215-1233 1217 T cells, (6) CD8 ? T cells exhibit a restricted T cell receptor (TCR) repertoire (i.e., oligoclonal expansions) suggesting that they have expanded from a few precursors locally responding to a distinct neuronal antigen [6,69]. These criteria, however, have not yet been demonstrated entirely for all entities.
In clinical terms, inflammatory CNS disorders associated with IgG antibodies against intracellular neuronal antigens are characterized by a multifocal presentation of CNS-related symptoms involving the neocortex, the limbic system, basal ganglia, brainstem, cerebellum, and spinal cord as well as PNS-related symptoms involving radices, plexus, and peripheral nerves in a variable extent ( Table 1). The clinical presentation partially reflects the pattern of expression of the respective neuronal antigen: ANNA-1 targets nuclear ELAVL (''Hu'') proteins expressed in central and peripheral neurons, and the corresponding clinical syndrome typically includes CNS and PNS manifestations [58]. In contrast, ANNA-2 targets nuclear NOVA-1 and -2 (''Ri'') proteins expressed in central, but not peripheral neurons and the clinical syndrome is usually restricted to the CNS [75] (Table 1). MRI findings include T2/FLAIR hyperintense, occasionally contrast-enhancing lesions in the cortex, medial temporal lobes, basal ganglia, brainstem, cerebellum, and spinal cord. Inflammatory changes are usually found in CSF studies including lymphocytic pleocytosis, mildly elevated protein together with intrathecal IgG synthesis and oligoclonal bands, but normal glucose and lactate levels [19]. The disease entities usually exhibit a chronic progressive clinical course and poor response to immunotherapy, especially to antibodydepleting therapies. Even successful removal of the tumor considered to drive the pathogenic immune response is usually not associated with disease amelioration [19].
However, there exists a group of CNS disorders with antibodies against intracellular neuronal antigens located mainly at presynaptic (GAD65) or postsynaptic (GABA-RAP, Gephyrin) sites of inhibitory GABAergic and glycinergic synapses, which less frequently associate with tumors. In these entities, there is often no evidence for cellular or humoral neuronal cytotoxicity, although some patients show neuroaxonal swelling, chromatolysis and vacuolization of neurons, microglial proliferation, as well as infiltration and apposition of cytotoxic CD8 ? T cells to neurons [6,40]. Further, there are reports of potentially pathogenic humoral mechanisms probably targeting inhibitory CNS neuronal networks in anti-GAD encephalitis [29,57,61], but until now the specificity of possible pathogenic antibodies has not been elucidated. These findings together with the wide spectrum of clinical presentations suggest that anti-GAD encephalitis comprises of a quite heterogenous group of CNS disorders with regard to their etiologies and disease mechanisms.
Paraneoplastic and non-paraneoplastic autoimmune encephalitis is probably mediated by antibodies to neuronal surface membrane antigens Autoimmune inflammatory CNS disorders associated with IgG antibodies in the serum and CSF directed against neuronal surface membrane antigens [54,97] occur both in a paraneoplastic and non-paraneoplastic context. Tumors assumed to drive the pathogenic immune response usually contain neuronally differentiated tissue expressing the respective neuronal antigen or occur in organs with a role in immune regulation, such as the thymus.
In principle, depending on the IgG subtype, antibodies may (1) specifically activate or block the function of their target molecules (GABA B -, Glycine-nAch-receptors, VGCC, VGKC), (2) crosslink and internalize the receptors (AMPA-and NMDA-receptors), (3) activate the complement cascade with subsequent formation of the terminal membrane attack complex and target cell lysis (probably mGluR1/5 receptors, VGCC, VGKC) and (4) activate Fc-receptors with subsequent antibody-dependent cellmediated cytotoxicity (ADCC) mainly by NK cells [23]. However, the effector mechanisms involved in the pathogenic effect of the autoantibodies within the CNS are not yet fully understood. In fact, autoantibodies in anti-NMDA-R and -AMPA-R, GABA B -R encephalitis are of the IgG1 and IgG3 type and are thus capable of activating complement in the presence of patient plasma (containing high concentrations of complement factors). However, in none of the reported autopsy or biopsy studies complement depositions could be detected on neurons suggesting that in the presence of patient CSF (containing low concentrations of complement factors) these IgG 1 and 3 autoantibodies do not lead to relevant complement activation [6,32,47,50,63,93]. In contrast, autoantibodies in VGKC-complex encephalitis are predominantly of the IgG4 (and IgG1) type and thus are unable to activate complement in the presence of patient plasma. However, in the only biopsy study reported thus far, complement depositions could be detected on neurons in VGKC-complex encephalitis suggesting that in the presence of patient CSF these IgG4 (or    [11,55] the IgG1) autoantibodies are capable of activating complement [6]. These conflicting results suggest that effector mechanisms of distinct autoantibody classes and subtypes may differ between the CNS and peripheral organs. Moreover, with the growing clinical awareness of the autoantibodies and improved diagnostic possibilities due to cell-based detection systems more patients are currently identified with persisting IgM antibodies without subsequent class switch to IgG and rarely with IgA antibodies against neuronal plasma membrane antigens [78]. This class-diversity of the neuron-directed adaptive humoral immune response in CNS may indeed reflect different milieus (e.g., absence or presence of different tumors), in which B cell activation takes place. Moreover, lack of class switch might indicate insufficient CD4 ? T cell-mediated help in B cell activation. However, these issues need further experimental investigation. Further, it remains to be determined, whether cytotoxicity by CD8 ? T cells contributes to functional and structural neuronal impairment.
Inflammatory CNS disorders associated with IgG antibodies against neuronal surface membrane antigens are characterized by a presentation of CNS-and sometimes PNS-related clinical symptoms representing the expression distribution and function of the respective target antigen [54,97]. On cerebral MRI cortical and subcortical gray matter regions may display mild and often transient T2/ FLAIR hyperintense signals. CSF findings may initially be normal, but often include inflammatory changes (lymphocytic pleocytosis, mildly elevated protein, intrathecal IgG synthesis with oligoclonal bands, but normal glucose and lactate levels). The clinical course is usually either monophasic or relapsing-remitting, but rarely progressive and clinical symptoms and paraclinical measures usually display a good response to immunotherapy, especially to antibody-depleting therapies. Upon detection of a tumor, sufficient tumor therapy is crucial to halt the pathogenic immune response [54,97].

Conclusions
A growing number of immune-mediated CNS disorders of paraneoplastic and non-paraneoplastic autoimmune origin have recently emerged, in which neurons are the target of AMPA both adaptive cellular and humoral immune responses. In autoimmune encephalitis associated with antibodies to neuronal surface membrane antigens, potentially reversible mechanisms of antibody-mediated impairment of synaptic transmission and neuronal excitability prevail. Hence, these disorders offer unique insight and provoke further investigation into the consequences of immune-mediated disruption of distinct neuronal signaling pathways within the living CNS. In contrast, paraneoplastic autoimmune encephalitis associated with antibodies to intracellular neuronal antigens seems to be mediated by cytotoxic CD8 ? T cells that cause functional and structural neuronal impairment in a way not specific for the respective antigen.