NeuroImmunology

Abstract Despite the well‐recognized importance of immunoglobulin therapy individualization during the treatment of chronic inflammatory demyelinating polyneuropathy (CIDP), the pathway to best achieve optimization is unknown. There are many pharmacokinetic and immunobiologic variables that can potentially influence the appropriateness of any individual therapy. Although identification of specific autoantibodies and their targets has only been accomplished in a minority of patients with CIDP, already the diagnostic and treatment implications of specific autoantibody detection are being realized. Individual variability in IgG pharmacokinetic properties including IgG catabolic rates and distribution, as well as the IgG level necessary for disease control also require consideration during the optimization process. For optimization to be successful there must be a measure of treatment response that has a clinically meaningful interpretation. There are currently available well‐defined and validated clinical assessment tools and outcome measures that are well suited for this purpose. While there remains much to learn on how best to manipulate immunopathology and immunoglobulin pharmacokinetics in the most favorable way, there currently exists an understanding of these principles to a degree sufficient to begin to develop rational and evidence‐based treatment optimization strategies. Introduction Intravenous immunoglobulin (IVIG), plasmapheresis (PE), and corticosteroids (CS) are first line CIDP treatment options (Joint Task Force of the EFNS and the PNS, 2010). Existing trials have focused on the efficacy of immunotherapy in groups of patients with CIDP (Hughes et al., 2008), with less emphasis placed on understanding the variability of treatment responses in individual patients (Pestronk et al., 1988; Donaghy et al., 1994; Attarian et al., 2011; Eftimov et al., 2014; Léger, 2014). What does the existing literature tell us about the immunobiologic and immunoglobulin pharmacokinetic determinants of treatment response, and how might we manipulate these in our favor? This review discusses how our increasing knowledge of specific autoantibodies and individual variations in IgG metabolism can be utilized with disease‐specific clinical outcome assessments to individually tailor therapy. The concept of optimizing immunotherapy for individual patients with inflammatory neuropathies is not new. For IVIG in CIDP, current guidelines recommend using the lowest effective maintenance dose and suggest that stable patients undergo periodic dose reduction or interval lengthening trials to establish the need for ongoing therapy (Joint Task Force of the EFNS and the PNS, 2010). The observation that in one clinical trial 44% of patients treated with placebo in addition to previous medications were able to reduce mean weekly doses of CS or IVIG by greater than 20% (RMC Trial Group, 2009) and in another trial 40% of placebo‐treated patients did not relapse even though an IVIG dependency test was required prior to randomization (van Schaik et al., 2018) indicates that overtreatment is common and not well addressed in clinical practice. The importance of treatment individualization as highlighted in current guidelines and supported by the findings in prior clinical trials is indisputable. What is less certain, however, is the best strategy by which to achieve that goal. There is much to be learned on how to escalate or de‐escalate IVIG therapy in such a way that maximizes clinical efficacy and minimizes over‐exposure and cost. One important determinant of IVIG optimization is how best to manage and measure wear‐off. Wear‐off is the cyclic or periodic occurrence of clinical deterioration at an interval following an IVIG infusion. It is unknown if reducing wear‐off prevents permanent damage or if some degree of end of cycle deterioration is tolerated without accumulating disability. From a pathobiologic perspective, wear‐off implies that disability, at least to some extent, cannot be solely attributed to inflammatory damage and destruction of myelin or axon itself. There is a growing appreciation that dysfunction at the nodes of Ranvier plays an early pathologic role in some patients (Pollard and Armati, 2011; Berger et al., 2013; Querol et al., 2017) and that rapid clinical improvement following IVIG or PE may be driven by reversible nodal or paranodal immune‐mediated dysfunction (Pollard and Armati, 2011; Berger et al., 2013). Electrophysiologically, reversible conduction block and decreased compound muscle action potential amplitudes have been demonstrated within the context of autoantibodies that bind to nodal and paranodal proteins or ganglioside complexes (Harschnitz et al., 2014; Notturno et al., 2014; Delmont et al., 2017). These observations imply that although IVIG may temporarily reduce the effects of autoantibodies, as the concentration of therapeutic IgG dwindles the pathogenic process re‐emerges which in turn leads to the clinical occurrence of wear‐off (Berger et al., 2013; Rojavin et al., 2016). In this review, we discuss how increasing knowledge of specific autoantibodies and individual variations in IgG metabolism, together with frequent quantitative strength and disability measurements, might be used to develop rational individualized treatment approaches. Although the term “immunoglobulin” when used herein refers to immunoglobulin administration intravenously (IVIG) or subcutaneously (SCIG), given the wealth of IVIG clinical data relative to SCIG, our focus is on optimization of IVIG during the treatment of CIDP. Immunology of CIDP: Will Knowledge of Antibodies Change Treatment? The diagnosis of CIDP relies on the combination of characteristic symptoms and signs, together with abnormalities on nerve conduction studies indicative of peripheral nerve demyelination or conduction block (Joint Task Force of the EFNS and the PNS, 2010). Albuminocytologic dissociation in cerebrospinal fluid, nerve enlargement or enhancement by magnetic resonance imaging (MRI), and characteristic histopathologic changes on nerve biopsy can add confidence in cases in which the electrophysiologic abnormalities are equivocal, but are not themselves diagnostic. The observation that almost half of the US patients that carry a diagnosis of CIDP are misdiagnosed (Allen and Lewis, 2015; Allen et al., 2018) underscores the challenges encountered in defining a disorder without a reliable biologic marker. In a tissue‐specific autoimmune disease like CIDP, an improved understanding of the characteristics of pathogenic antibodies and their target antigens has the capacity to not only improve CIDP diagnostic accuracy but may be critical when determining disease‐specific immunotherapy approaches. In order to understand how specific autoantibodies influence immunoglobulin treatment it is useful to understand what is known about IgG's mechanism of action. IgG therapy is thought to act by reducing the amounts and/or effector functions of pathologic autoantibodies (Berger et al., 2013). Mechanisms dependent on competition between therapeutic (normal) and pathogenic IgG include saturating neonatal Fc receptors (FcRn), anti‐idiotype neutralization, inhibition of complement deposition, and possibly feedback inhibition of B‐cell antibody production (Kuitwaard et al., 2009; Berger et al., 2013; Schwab and Nimmerjahn, 2013; Espéli et al., 2016). Each of these mechanisms has the potential for manipulation, which may be influenced by the specific autoantibody. For example, autoantibody Fc domain structure is defined by class (IgG, IgM, or IgA) and subclass (IgG1‐4). If this structure was known for any particular autoantibody, it might help predict the likelihood that FcRn blocking or saturating agents will be effective. Knowing the different Fc domain structures and the effector mechanisms they recruit (Table 1) may also help determine which immunotherapy is most likely to be effective in any given disease subset. There is much to learn on how specific autoantibodies influence immunoglobulin treatment in any specific patient. Given the observations that some patients with CIDP carry on for years with cyclic IVIG treatment‐related fluctuations, one underlying principal of IVIG therapy is that it competes with but does not fundamentally alter autoantibody production or function (Dacci et al., 2010; Pollard and Armati, 2011; Kokubun et al., 2013; Debs et al., 2017). Ig type % of serum IgG Half‐life (days) Complement classical activation pathway Binding to FcRn Binding to myeloid FcRs IgA (0.60–3.50)* 4–6 − − (FcαRs) IgM (0.40–2.30)* 5–6 ++++ − − IgG1 67–75% 23–28 +++ ++++ +++ IgG2 22–25% 23–28 + ++ ++ IgG3 7–10% 5–10 +++ ++ +++ IgG4 3–5% 23–28 − +++ + IgA, immunoglobulin A; IgG, immunoglobulin G; IgM, immunoglobulin M; FcαR, IgA Fc receptor; FcR, receptors specific for the Fc region of immunoglobulins; FcRn, neonatal Fc receptor. * For IgA and IgM, serum concentration in g/l is given. For IgG subclasses, % of total IgG is given. Normal total IgG concentrations are 7.25–12.50 g/l in most laboratories. There are scenarios in which autoantibody recognition may already directly influence individualization of immunotherapy. Anti‐neurofascin 155 (NF155) antibodies have been detected in 3.3%–7.1% of patients with chronic inflammatory polyneuropathies (Ng et al., 2012; Querol et al., 2014; Querol and Illa, 2015; Devaux et al., 2016) and between 2.4% and 7.5% of patients have been found to have anti‐contactin 1 (CNTN1) antibodies (Querol et al., 2013; Yan et al., 2014; Miura et al., 2015). NF155 and CNTN1 are localized to the paranodal region and both antibodies are of the IgG4 isotype. One emerging feature in patients that harbor these antibodies is a reportedly poor response to IVIG but more favorable responses to CS and B‐cell depletion therapy. One potential explanation for why IVIG may be less effective in these subsets of CIDP patients is that the IgG4 isotype does not efficiently activate complement and has a low affinity for Fc receptors on effector cells (Labasque et al., 2014; Vidarsson et al., 2014). Although these anecdotal treatment observations currently based on only small case series need confirmation in larger groups of patients, detection of NF155 and CNTN1 autoantibodies might shift treatment paradigms for autoantibody‐defined subsets of patients in consideration of their specific pathophysiology. Future treatment strategies may take earlier advantage of CS and anti‐B‐cell therapies such as rituximab, with a deemphasized role of IVIG (Huijbers et al., 2015; Halder et al., 2016). Yet‐to‐be identified autoantibodies and target antigens might provide similar insight into other groups of CIDP patients. Furthermore, knowledge of the specificity and affinity of autoantibodies and recognition of the target antigen will likely facilitate utilization of new and improved therapies. Agents in clinical development include FcRn blockers (Zuercher et al., 2016; Kiessling et al., 2017), complement inhibitors (Fitzpatrick et al., 2011), antigen‐specific adsorbents/binding site blockers (Herrendorff et al., 2017), and blockers of pro‐inflammatory Fc receptors (Schwab and Nimmerjahn, 2013). IgG Metabolism and its Impact on Individualized Therapy When considering strategies for optimization of immunoglobulin therapy, an understanding of the individual variations in the distribution and metabolism of therapeutic IgG may be equal in importance to the characteristics of pathological autoantibodies. The speed with which serum IgG levels rise depends on the method of IgG delivery. When immunoglobulin is administered intravascularly, serum IgG levels peak immediately (Fig. 1) and then drop by about 50% over 2–4 days as IgG is distributed into the total extracellular fluid volume (Bonilla, 2008). In contrast to IV administration, following subcutaneous immunoglobulin administration, serum IgG levels rise more slowly, peaking at 48–72 h (Fig. 2) (Berger, 2004; Bonilla, 2008). Regardless of the route of delivery, once IgG reaches the intravascular space, catabolism proceeds as a first‐order process with a half‐life of 21–30 days (Bonilla, 2008). The relatively slow catabolism of IgG, as compared to other plasma proteins, is due to a saturable endothelial cell receptor, FcRn, which protects endocytosed IgG from lysosomal degradation (Fig. 2). The principle of first‐order kinetics is one fundamental concept that requires consideration when attempting to optimize immunoglobulin therapy. First‐order kinetics means that the higher the serum IgG level, the faster it will be catabolized (Fig. 3) (Waldmann and Strober, 1969). As illustrated in the figure, if the IgG concentration is raised from 8 g/l to 30 g/l the half‐life drops from 28 to 11 days. High‐dose IVIG infusions can raise serum IgG concentrations to peaks of 30–50 g/l or higher (Reinhart and Berchtold, 1992; Dalakas, 1994). In contrast, the peak serum concentration achieved with SCIG is on average only 61% of the peak achieved with IV infusions of the same dose (Berger et al., 2011). Whereas IVIG bolus infusions may result in trough‐to‐peak serum IgG differences greater than 100% of the overall mean, frequent SCIG administration may result in peak‐trough differences of only about 5% of the overall mean (Berger, 2004; Bonilla, 2008; Berger et al., 2011). Regardless of whether the route of admistration is by IV or SC, with more frequent administration of fractional doses, the peaks and troughs will approach each other and ultimately near “steady‐state” concentrations will be reached (Berger, 2004; Buclin et al., 2009). The impact that these findings have on immunoglubulin optimization becomes apperent when considering that in Guillian–Barre Syndrome (Kuitwaard et al., 2009) and perhaps CIDP (Debs et al., 2017) there appears to be a serum IgG level which must be reached in order to achieve a desirable clinical outcome. These findings indicate that high and stable trough levels may have particular importance in immunoglobulin optimization such that a satisfactory minimum IgG level is achieved while avoiding the accelerated catabolism that comes with high immunoglobulin peak levels. A clinical study exploring IgG levels and clinical efficacy in patients subjected to frequent low dose vs. less frequent higher dose IVIG is currently underway (Kuitwaard, personal communication, 2017). Beyond the normal physiological characteristics of IgG catabolism that may be manipulated to achieve treatment optimization, there are also patient to patient differences that influence optimization. Different patients vary in the expression and function of FcRn, which can also be affected by cytokines such as tumor necrosis factor (Liu et al., 2007). The variability in FcRn is one explanation for why the half‐life of the catabolic phase has a fairly broad distribution of 21–30 days. The reasons for this variability are currently unknown. Genetic polymorphisms in the FcRn receptor structural gene have been studied in only relatively few patients, but have not yet been shown to correlate with serum IgG levels or the response to IVIG (Fokkink et al., 2016; Vlam et al., 2014). Wear‐Off with IVIG and What it Means in Clinical Practice As might be predicted based upon patient to patient immunoglobulin pharmacokinetic heterogeneity and immunopathogenic diversity, treatment responses to IVIG can be variable between patients and even in the same patient at different stages of disease. One challenge that arises during IVIG treatment optimization is how best to manage wear‐off and other treatment‐related fluctuations, a rather dramatic example of which is shown in Fig. 4. In clinical practice the extent of wear‐off can be highly variable. In order to continuously maintain strength and disability, one study found that 60% of subjects needed IVIG at intervals ≤14 days (Kuitwaard et al., 2013). Several other studies (Table 2) have shown that many patients do better when the IVIG dosing interval is less than the expected half‐life of IgG (Broyles et al., 2013; Rajabally et al., 2013). Clinical experience tells us that other patients do not have clinically meaningful treatment‐related fluctuations and may maintain stability with less frequent IVIG dosing at 4 to 6‐week intervals. If different IgG levels are required to counter different autoantibodies and IgG pharmacokinetic variability influences speed of IgG catabolism in different patients then it becomes clear that dose and interval of immunoglobulin therapy should be individualized and one important determinant of optimization is quantification of treatment‐related fluctuations. Reference Patients (n) Mean Ig dose (g/kg) Mean interval (weeks) Patients (%) receiving IgG at interval ≤ 14 days Lunn et al., 2016 CIDP 39 1.4 4.3 17 MMN 24 Kuitwaard et al., 2013 CIDP 25 0.43 2.4 60 Rajabally et al., 2006* CIDP 15 0.69* 6.9 13 Broyles et al., 2013 CIDP 31 0.87 4.0 25 CIDP, chronic inflammatory demyelinating polyneuropathy; IgG, immunoglobulin G; MMN, multifocal motor neuropathy; SEM, standard error of the mean. Note that common criteria for defining endpoints and/or selecting maintenance doses were not employed across the different studies. * Final or lowest dose per course. One of the major aims of the Peripheral Neuropathy Outcome Measures Standardization (PeriNomS) study was to better define the metrics used to assess and follow patients with inflammatory neuropathies (Draak et al., 2014). The authors emphasized the importance of measuring disability (i.e., activities and participation), strength and sensory impairment, and quality of life in patients with CIDP. The Inflammatory Rasch‐built Overall Disability Scale (I‐RODS), INCAT disability score, and grip strength measurements using the Martin Vigorimeter® (Martin GmbH & Co., Tutlingen, Germany) have emerged as strongly validated and valuable assessment tools. Grip strength in particular provides a quantitative, immediately available result that can be easily performed during physician office visits (Berger and Allen, 2015). In a randomized controlled trial, grip strength was shown to provide objective documentation of global neurologic status in patients with CIDP, not just limited to the upper limb or exclusively motor function (Vanhoutte et al., 2013). Grip strength and I‐RODS have also been shown to correlate with each other (Draak et al., 2016). The observation that daily grip strength measurements can vary dramatically during a single IVIG treatment cycle with some patients showing maximum strength more than 50% higher than minimal strength and other patients showing no significant changes (Hadden, personal communication, 2017) highlights the potential for grip strength monitoring to be a useful adjunct assessment when determining the IgG dose and interval most appropriate for any individual patient. Further research is needed to better understand what magnitude of intracycle grip strength and disability fluctuations are acceptable, and what magnitude should trigger therapy changes to maximize short‐term function and minimize long‐term disability. Such questions are currently being explored in a study which records daily grip strength and weekly disability outcomes over a 6 month period in a cohort of IVIG‐treated CIDP patients (Allen, personal communication, 2017). One practical consideration that commonly arises during IVIG optimization is whether it is preferable to adjust IVIG dose or interval. There is no best answer to this question, and ultimately the approach to IVIG optimization should blend individual patient treatment responses, tolerability, and convenience of administration. However, when optimizing IVIG, it is important to recall that (1) IVIG half‐life can vary between 21 and 30 days, (2) there is likely a minimum serum IgG threshold that must be reached in order to achieve efficacy, and (3) IgG catabolism accelerates with very high serum IgG levels. With these points in mind, consider three scenarios. In the first, clinically meaningful treatment‐related fluctuations are present, necessitating escalated immunoglobulin therapy. Increasing IVIG dose while keeping the interval constant may keep the serum IgG level above the necessary therapeutic minimum to achieve the desired clinical effect, but at the cost of accelerated IgG catabolism which in turn may result in an overall higher total amount of IgG needed (Gouilleux‐Gruart et al., 2013). Alternatively, fractionating the treatment into smaller doses at shorter intervals may minimize clinical fluctuations and immunoglobulin exposure as the serum IgG level reaches a stable steady state with higher trough level (Berger, 2004; Buclin et al., 2009). In the second scenario, treatment‐related fluctuations are unequivocally present but are minimal without an impact on disability. In this setting no treatment modification may be needed, since a minor degree of wear‐off may provide assurance that treatment is needed and not over‐utilized. In the third scenario, no treatment‐related fluctuations are present. In that situation, immunoglobulin de‐escalation is encouraged. The optimal strategy to wean IVIG is unknown. In cases in which one important determinant of optimization is maintaining a stable IgG steady‐state trough‐level, then periodic IVIG dose reduction with stable infusion frequency may be preferable to interval lengthening with stable dosing. Regardless of whether treatment is escalated or de‐escalated, or which optimization approach is taken, collection of validated outcomes during optimizing is of critical importance when interpreting the response to treatment changes. As opposed to the general optimization comments above that might be applied to patients already on maintenance therapy, a useful treatment optimization protocol designed for IVIG‐naïve patients was developed by Lunn et al. (2016) (Fig. 5). In this approach the patient with CIDP is given a loading dose of 2 g/kg IVIG, which is repeated 6 weeks later if the patient is not fully normalized. No further IVIG is given unless deterioration occurs, in which case regularly scheduled IVIG is started. The time interval from IVIG treatment to deterioration is used as the individualized interval for all subsequent IVIG. After two further 2 g/kg loading doses, the dose is reduced by 20% per course. When relapse or wear‐off emerges, that dose is set as the lowest dose without wear‐off. In a cohort of 71 patients (including 39 with CIDP and 24 with MMN) utilization of this protocol resulted in a mean (± standard error) dose of 1.4 ± 0.6 g/kg at a mean interval of 4.3 weeks (range 0.5 to 10 weeks) (Lunn et al., 2016). A similar, but not identical, approach has been proposed by the Department of Neurology Erasmus MC (Kuitwaard et al., 2017). Neither the Lunn nor the Erasmus MC approaches were designed to be dogmatic, but both highlight important principles that should be considered in any optimization approach. These principles include structured tapering protocols to determine IgG dependency, close clinical monitoring for wear‐off, and an appreciation for treatment response variability. The presence or absence of any specific autoantibody will also provide rational for further refinement of the optimization approach or immunotherapy of choice. Conclusion Presently there is no “best” approach to treatment optimization, and there is much to be considered when attempting to manipulate immunotherapy and immunoglobulin pharmacokinetics in the most favorable way (Table 3). Differences in IgG catabolic rates and IgG level necessary for disease control are major determinants that require consideration in every individual patient. An evolving understanding of the heterogeneity of FcRn expression and function may further facilitate personalized treatment approaches. Ultimately an improved understanding of the specificity and affinity of pathogenic autoantibodies in what is now broadly considered to be idiopathic CIDP, combined with what is known about immunoglobulin pharmacokinetics, may provide clinicians the most precise pathway to optimal therapy selection and the roadmap to maximize the effects of that therapy. Utilization of validated outcome measures during routine clinical care is of great importance in order to objectify the response to treatment, quantify treatment‐related fluctuations, and detect relapse in patients undergoing therapy weaning or suspension. As our ability to rationally design and assess personalized therapeutic regimens improves so too does our capacity to explore other yet‐to‐be answered questions, including understanding the long‐term impact of short‐term optimization and understanding which optimized interventions are not only the most efficacious but also the most cost‐effective. Frequent application of outcome measurements, such as strength impairment with grip strength dynamometry and disability with I‐RODS, can be useful to document treatment response, quantify treatment‐related fluctuations, and guide maintenance therapy. Some degree of “wear‐off” may be acceptable and provide reassurance that continued immunotherapy is necessary. However, if clinically meaningful and unacceptable IVIG treatment‐related fluctuations are present, shortening treatment intervals (more so than increasing IVIG dose) is likely to be beneficial. All patients should undergo periodic structured immunotherapy taper trials to assess disease activity. CS should be used only with caution in the motor‐predominant and perhaps Lewis Sumner Syndrome atypical CIDP variants. Worsening following CS exposure has been described in these subsets of patients. Patients with predominantly IgG4 autoantibodies (against nodal and/or paranodal antigens) are unlikely to benefit from IVIG. Treatment with CS or PE should be considered first line in these individuals, with B‐cell depletion therapy reserved for patients refractory or intolerant to CS. CIDP, chronic inflammatory demyelinating polyneuropathy; CS, corticosteroid; I‐RODS, Inflammatory Rasch‐built Overall Disability Scale; IgG4, immunoglobulin G4; IVIG, intravenous immunoglobulin; PE, plasmapheresis. Acknowledgements The authors wish to thank Jason Heckler of Meridian HealthComms, Cheshire, UK for editorial assistance, formatting, and processing the manuscript. References

ABSTRACT An increasing number of movement disorders are associated with autoantibodies. Many of these autoantibodies target the extracellular domain of neuronal surface proteins and associate with highly specific phenotypes, suggesting they have pathogenic potential. Below, we describe the phenotypes associated with some of these commoner autoantibody‐mediated movement disorders, and outline increasingly well‐established mechanisms of autoantibody pathogenicity which include antigen downregulation and complement fixation. Despite these advances, and the increasingly robust evidence for improved clinical outcomes with early escalation of immunotherapies, the underlying cellular immunology of these conditions has received little attention. Therefore, here, we outline the likely roles of T cells and B cells in the generation of autoantibodies, and reflect on how these may guide both current immunotherapy regimes and our future understanding of precision medicine in the field. In addition, we summarise potential mechanisms by which these peripherally‐driven immune responses may reach the central nervous system. We integrate this with the immunologically‐relevant clinical observations of preceding infections, tumours and human leucocyte antigen‐associations to provide an overview of the therapeutically‐relevant underlying adaptive immunology in the autoantibody‐mediated movement disorders. © 2018 The Authors. Movement Disorders published by Wiley Periodicals, Inc. on behalf of International Parkinson and Movement Disorder Society. The spectrum of autoantibody‐mediated movement disorders includes a broad and clinically heterogeneous group of conditions. The movement disorders occur either in isolation or, more commonly, as prominent and often distinctive manifestations of autoimmune encephalitides. Patients typically present with a subacute onset and multifocal neurological features involving the cortex, basal ganglia, brain stem, and/or spinal cord (Table 1). Although formal epidemiological data are still emerging, it is clear that both sexes and patients of all ages can be affected by this spectrum of disorders. The detection of neuronal autoantibodies in serum and the CSF can help to guide the diagnostic process, prognosis, and the treatment of these disorders. In addition, the autoantibody specificity may predict an underlying tumour association (Table 1). Perhaps most important, many of these conditions respond to immunotherapies, making them one of the earliest therapeutic considerations in the correct clinical context.1-5 Antigen Movement disorders Additional features Tumour association Extracellular antigens NMDA receptor Orobuccolingual dyskinesia, catatonia, limb dystonia, stereotypies, chorea Amnesia, psychiatric features, seizures, dysautonomia, coma Ovarian teratoma (especially if > 18 years old) LGI1 Faciobrachial dystonic seizures, myoclonus, chorea, parkinsonism LE, hyponatremia Thymoma, SCLC CASPR2 Chorea, ataxia LE, Morvan's syndrome, neuromyotonia, neuropathic pain Thymoma GABAB receptor Ataxia, OMS, chorea LE SCLC GABAA receptor OMS, SPS, chorea Status epilepticus, LE Thymoma, SCLC mGluR1 Ataxia Seizures, cognitive impairment Hodgkin Lymphoma, renal cancer VGCC Ataxia Lambert‐Eaton syndrome SCLC DPPX PERM, OMS, tremor, ataxia Behaviour changes, cognitive decline, seizures, dysautonomia, diarrhoea, weight loss B cell neoplasms IgLON5 Chorea, parkinsonism, ataxia, limb stiffness, dystonia Non‐REM and REM‐sleep disorder, stridor, bulbar symptoms, cognitive impairment, eye movement abnormalities Not reported Glycine receptor SPSD Seizures, encephalopathy Thymoma, lymphoma, SCLC, breast cancer Dopamine 2 receptor Chorea, dystonia, parkinsonism, tics Psychiatric disturbances Not reported Neurexin‐3α Orofacial dyskinesias Confusion, seizures, decrease level of consciousness Not reported Intracellular antigens Amphiphysin SPSD SCLC, breast cancer GAD65 SPSD, ataxia LE, epilepsy Rare: thymoma, lymphoma, breast cancer, other CRMP5 Chorea, ataxia, OMS LE, encephalomyelitis, neuropathies SCLC, thymoma Ma2 OMS, parkinsonism LE, brain stem encephalopathy Testicular cancer Ri Jaw dystonia, ataxia, OMS, parkinsonism Brain stem encephalopathy SCLC, breast cancer Yo Ataxia Ovarian cancer, breast cancer Hu Ataxia LE, polyneuropathy, brainstem encephalopathy, pseudoathetosis SCLC Tr/DNER Ataxia Hodgkin Lymphoma GFAP Tremor, ataxia Encephalopathy, meningitis, myelopathy, seizures, dysautonomia, psychiatric Ovarian teratoma, prostate adenocarcinoma NMDA, N‐methyl‐d‐aspartate; LGI1, leucine‐rich glioma‐inactivated 1; CASPR2, contactin‐associated protein‐like 2; GABAA/B, gamma‐aminobutyric acid A/B; mGluR1, metabotropic glutamate receptor type 1; VGCC, voltage gated calcium channel; DPPX, dipeptidyl‐peptidase‐like protein‐6; GAD, glutamic acid decarboxylase; CRMP5, collapsin‐response mediated protein 5; GFAP, glial fibrillary acidic protein; SCLC, small cell lung cancer; LE, limbic encephalitis; SPS, stiff person syndrome; SPSD, stiff‐person syndrome spectrum disorder; PERM, progressive encephalomyelitis with rigidity and myoclonus; OMS, opsoclonus myoclonus syndrome; REM, rapid eye movement. The practical importance of both the antigenic specificity and the effects of immunotherapies make our understanding of the underlying immunopathology critical to managing patients with these conditions.4, 6-10 Therefore, in this review, we focus on the immunological mechanisms that are likely to initiate and propagate the diseases and outline roles for the autoantibody‐producing plasma cells, their precursor B cells, and Tcells. We also discuss the relevance of antigen‐drainage, tumors, the blood‐brain barrier and theprinciple pathogenic mechanisms by which the autoantibodies may induce disease at a molecular level(Figs. 1 and 2).1, 4, 6 This permits us to consider methods to tailor immunotherapies toward the underlying immunology. However, to ensure the accurate administration of immunotherapies, we describe these alongside the key clinical features that permit early recognition of immunotherapy‐responsive autoantibody‐mediated movement disorders.3, 5, 7 Throughout, we focus on the most common likely pathogenic autoantibodies and review those autoantibodies which potentially challenge the paradigm that targetting of surface epitopes implies causation. Inside or Outside the Plasma Membrane? Location Matters The major factor that governs the likely pathogenicity of an autoantibody response is whether it targets the intracellular versus extracellular domain of the autoantigen (Table 1 and Fig. 1). Autoantibodies directed against the extracellular domains of surface proteins (NSAbs; often termed neuroglial surface autoantibodies) are able to exert an effect on their target antigen in vivo and are therefore considered to have pathogenic potential. In contrast, autoantibodies that target intracellular antigens may never have the opportunity to bind their target.8-10 Many such autoantibodies (including Hu, Yo, Ri, CRMP5, and Ma2) are considered bystanders of an immunological process that is often mediated by pathogenic CD8 (cluster of differentiation) T cells.11, 12 These autoantibodies are frequently associated with tumors and are summarized in Table 1 and previous reviews.3, 13 Indeed, unlike NSAb‐related disorders,14 the passive transfer of intracellular‐directed autoantibodies to experimental animals has failed to reproduce features of the disease.15 Furthermore, it may be predicted that if driven by a dominant CD8 T cell response, the human disease should benefit from drugs directed to inhibit T cell function and, indeed, sirolimus has been shown to somewhat improve functional outcomes in these disorders that typically have a very poor prognosis.16 In addition, a hinterland category exists of antigens that may transiently reach the cell surface, such as those directed at the synaptic vescicle protein glutamic acid decarboxylase (GAD, Fig. 3). Here, exposure to the extracellular compartment may occur during vescicle fusion and potentially account for an effect on the intracellular antigenic target.8 Alternatively, they may represent an immune epiphenomenon that is sometimes detectable alongside, currently largely unidentified, coexistent NSAbs.17 In addition, antibodies to the voltage‐gated potassium channel complex that do not target leucine‐rich glioma‐inactivated‐1 (LGI1) or contactin‐associated protein‐like 2 (CASPR2) are directed against intracelluar epitopes and, associate with a broad, diverse, and seemingly unrelated set of neurological conditions.18 Therefore, these clinically‐irrelevant ‘double‐negative’ voltage gated potassium channel complex antibodies will not be discussed further herein. Treatment Principles Unlike most of those with solely intracellular‐directed autoantibodies, patients with NSAbs often show a good response to first‐line therapies such as corticosteroids, intravenous immunoglobulins, and plasma exchange.19-21 Patients with GAD antibodies can also show a response to many of these treatments, but overall this cohort is more refractory to available agents. However, in all patients, the upscaling of immunotherapies to second‐line treatments—such as cyclophosphamide and rituximab—should be instigated in those refractory to the first‐line therapies. This up‐titration should be quicker in patients whose disease is more severe, often within 2 weeks in N‐methyl‐d‐aspartate receptor (NMDAR)‐antibody encephalitis. Also, there are increasingly strong data to support the generic notion that early treatments improve clinical outcomes19, 21-25; hence, therapies should be ideally instigated on the basis of a clinical diagnosis while awaiting confirmatory serology. Therefore, in this review, we emphasize highly distinctive clinical features within the autoantibody‐associated movement disorders and use the aforementioned framework to reflect on the underlying mechanistic neuroimmunology as a basis to guide current and future immunotherapy options. The NSAb‐Mediated Syndromes and Their Related Immunology NMDAR Antibodies Clinical Features The discovery of autoantibodies against the GluN1 (NR1) subunit of the NMDAR identified a diffuse encephalitis with early psychiatric and cognitive features.19, 24-26 The associated characteristic hyperkinetic movement disorder is typically recognized after about 1 to 2 weeks, commonly involves the face, limbs, and trunk and has been variably described as dyskinetic orchoreoathetoid.24, 25, 27, 28 However, a recent study has suggested a more complex, combinatorial nomenclature may be most appropriate with expert raters noting the highly‐distinctive combination of dystonia, chorea and stereotypies with a paucity of tremor or myoclonus in many patients (Fig. 3C).29 Accompanying agitation may alternate with periods of catalepsy and catatonia and sometimes a hypokinetic phenotype can predominate, resembling endophenotypes of encephalitis lethargica.30 Treatment In NMDAR‐antibody encephalitis, around 50% of patients respond to first‐line medications, usually with a good recovery over several weeks to months.19, 24 However, in this condition, their use may be limited by agitation and behavioral difficulties. This is especially true of plasma exchange, which requires significant patient compliance. Hence, it is sometimes necessary to sedate patients to permit administration of these first‐line therapies. For patients who are refractory to these drugs, rituximab and/or cyclophosphamide are recommended second‐line options, and there are data to suggest that their administration is associated with improved outcomes.19, 22, 24, 25 Throughout, early removal of the ovarian teratoma should be a therapeutic goal. As recent laboratory observations may explain some of these clinical findings,31-33 we next discuss the immunology in the context of the therapeutic data. Immunology NMDAR‐antibody encephalitis is associated with 2 known immunological triggers: an ovarian teratoma and preceding herpes simplex virus encephalitis (HSVE). Although in patients with intracellular directed autoantibodies tumors are often malignant, the ovarian teratoma in patients with NMDAR antibodies is typically benign. The teratoma, seen in about 20% of adults and few children,24, 25, 30 is likely to be a site of immunization as it contains dense infiltrations of T cells and B cells31 and its removal can hasten recovery.19, 24 Indeed, a recent paper showed that lymphocytes—both B cells and plasma cells—within the teratoma have the capacity to produce NR1‐directed autoantibodies, and the cystic teratoma fluid contains higher levels of NMDAR‐antibodies than serum.31 By contrast, the mechanism of NMDAR‐antibody encephalitis post‐HSVE is less clear.34-36 Typically, this disorder begins around 4 to 8 weeks after onset of HSVE at a time where patients, mainly children, are improving from the HSVE. Children often present with prominent choreoathetosis, abnormal behavior, and cognitive impairment: this syndrome appears identical to a primary NMDAR‐antibody encephalitis and distinctive from a relapse of HSVE.37 In adults, a similar pathophysiological phenomenon is observed but is not associated with a clear clinical relapse, rather a prolonged cognitive syndrome associated with a lower rate of abnormal movements.38, 39 Mechanistically, the necrotic disease process of HSVE may release a variety of neuronal antigens, including neuronal surface proteins. This may be more prominent after HSVE and other viruses, by comparison to traumatic brain injury, stroke, or neurodegeneration40 due to the more inflammatory environment or direct effects of viruses on lymphocytes.41 Subsequently, released antigen may be soluble or taken up by antigen‐presenting cells that migrate to cervical lymph nodes, the secondary lymphoid organs known to drain CNS lymphatics.42 Presentation of this antigen to T cells can lead to consequent B cell activation and antibody production in lymph node germinal centres (Figure 2). Consistent with this interaction, interruption of germinal center reactions with ongoing T cell and B cell interactions may explain the benefits of early and rapidly escalated immunotherapies including corticosteroids, cyclophosphamide, and rituximab.19, 24, 25, 43 Indeed, circulating B cells from patients with NMDAR antibody encephalitis can produce NMDAR antibodies in vitro, especially under conditions that mimic T cell help.31 Methods to determine the degree and nature of T and B cell involvement may in future help predict the value of cell specific therapies: for example, the autoantibodies can be transiently removed with plasma exchange, B cells deleted with rituximab, and the T cells inhibited with drugs such as cyclophosphamide. In terms of autoantibody generation mechanisms, molecular mimicry between HSV‐associated antigens and the NMDAR seems unlikely as other CNS viruses, such as varicella zoster, have been shown to trigger NMDAR‐antibody encephalitis.44 By analogy, we have also observed NMDAR‐antibody encephalitis after other viral and idiopathic neurological inflammatory illnesses (Irani and Leite, 92). In addition, there is often the concomitant presence of other antigen‐specific NSAbs, such as those against the dopamine 2 receptor, gamma‐aminobutyric acid type A receptor (GABAAR) and other unknown targets, after HSVE.35, 45 Finally, no viral epitope has been reported with sequence homology to the NMDAR. Rather, this array of autoantibody specificities post‐HSVE is likely to reflect the concept of epitope spread, where, in an inflammatory milleu, there is a polyclonal immune response against a range of antigens exposed after a single inciting event. However, given that most patients have neither preceding HSVE nor a teratoma and this idiopathic group have the highest relapse rate,19, 24 the most common immunological triggers of this condition have clinical importance and await discovery. One such trigger may be genetic, perhaps a HLA predisposition, particularly given the known nonwhite racial bias of this condition.24, 46 By contrast to emerging data about the cellular immunology, the autoantibodies themselves have been relatively well characterized. Their principle mechanism of action appears to be the downregulation of surface NMDARs. This leads to a direct reduction infunctional NMDARs and, in addition, may have consequences for the stability and function of other neighboring synaptic and extrasynaptic proteins.25, 47 Furthermore, although the NMDAR antibodies are of the complement‐fixing IgG1 subclass, the available brain pathology does not show complement deposition.24, 48 Complement induction often causes tissue necrosis. Hence the absence of complement deposition, alongside established functional effects of the autoantibodies, may explain the substantial reversibility and limited atrophy observed in this condition after immunotherapy.19, 24, 25 LGI1 Antibodies Clinical Features Faciobrachial dystonic seizures (FBDS) are stereotyped, frequent, and brief dystonic movements consistently associated with LGI1 antbodies (see Supporting Information Video). They predominantly involve the arm and the ipsilateral face, and less commonly the leg or the trunk.20, 21, 49 As the attacks are rarely associated with disturbance of consciousness or ictal EEG changes, they may be considered to lie in a borderland between movement disorders and seizures.50 However, and consistent with seizures, FBDS can be preceded by sensory auras and automatisms, and both agitation and speech arrest are described during or after the episodes.20, 49 Nevertheless, the semiology of FBDS is very different to that of more typical frontal and temporal lobe epilepsies. The attacks show a limited response to antiepileptic drugs, and multimodal radiological involvement of the basal ganglia is frequently observed in patients with FBDS (Fig. 3A).20, 49, 51, 52 Therefore, the origin of FBDS, and their preferred classification as a movement disorder or an epilepsy, is still debated, but pragmatically FBDS certainly continue to present to movement disorder neurologists, amongst others. Importantly, although FBDS were originally observed in the context of marked cognitive impairment (as part of an “encephalitis”), several patients present with FBDS alone. Furthermore, the relatively dramatic response of FBDS to immunotherapies, and their onset preceding the development of cognitive impairment, led to the hypothesis that their effective cessation may prevent the occurrence of cognitive impairment associated with limbic encephalitis.20, 49 Indeed, in a recent cohort of 103 patients with FBDS, this appeared to be the case, with cognitive impairment appearing frequently, and almost exclusively, in patients with ongoing FBDS.21 This may also be true of other seizure semiologies in patients with LGI1‐antibodies, which are well‐recognised and also very frequent.53-55 Moreover, as generalized chorea can sometimes precede the onset of LGI1‐antibody encephalitis, perhaps a similar paradigm also operates in this clinical scenario.56, 57 Treatment Therefore, after a clinical diagnosis is made, and by analogy to NMDAR antibodies, early immunotherapy appears to be key to outcome optimization.21 Timing is especially critical in this condition as there has been a demonstrable reduction in the probability of seizure cessation with each day of delay to immunotherapy, and because the effective treatment of FBDS may prevent cognitive impairment. In this condition, there are surprisingly limited data to suggest a benefit of rituximab.58 However, to date, early treatment with rituximab has not been systematically reported, nor have studies using medications including cyclophosphamide or bortezomib. Immunology The highly consistent association between a distinctive clinical phenotype and the presence of LGI1 antibodies strongly suggests that they have a pathogenic role. This has been strengthened by in vitro data that implicate alpha‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazolepropionic acid (AMPA)‐receptors and potassium channels in the downstream mechanisms of LGI1‐modulation induced neuronal dysfunction.10, 59, 60 Other functional effects of the LGI1 autoantibodies include the downregulation of the LGI1 complex, which includes LGI1's natural binding partners a disintegrin and metalloproteinase domain‐containing protein 22 and 23 (ADAM22/23).21 In addition, and by contrast to the NMDAR‐directed antibodies, LGI1 antibodies are mainly of the IgG4 subclass.21, 61 However, the LGI1‐IgG1 antibodies appear to correlate with disease severity, perhaps as they have the potential to deposit complement in the brain, as observed from some postmortem tissues.21, 62 Although tumors and prodromal infections have not been consistently observed in patients with LGI1 antibodies, the recent description of an almost universal HLA‐DRB1*07:01 allele strongly implicates a role for T cells in disease pathogenesis.63, 64 These LGI1‐specific T cells are likely to interact with B cells in peripheral germinal centers. CASPR2 Antibodies Clinical Features Antibodies against the juxtaparanodal protein CASPR2 are associated with a variety of movement disorders including neuromyotonia, chorea, ataxia, and a syndrome of orthostatic myoclonus.3, 65 Many of these typically occur in the context of an encephalopathy that is similar to that associated with LGI1 antibodies. Indeed, although neuromyotonia is often thought to occur as an isolated phenomenon, it is intriguing that many patients have additional autonomic and CNS features, suggesting a frequently more diffuse neuronal disease process.66 Treatment The management of CASPR2‐antibody conditions has received little attention to date.67-69 Our experience suggests that the ataxia and encephalopathy are usually responsive to similar therapeutic approaches as for patients with LGI1‐antibody encephalitis (Irani, Jacob, & Leite, 2017). Immunology As with LGI1 antibodies, one plausible mechanism of CASPR2‐antibody‐induced hyperexcitability is interference with the tightly CASPR2‐complexed juxtaparanodal potassium channels.10 This mechanism has been suggested by human models of CASPR2 mutations and was recently confirmed in animals receiving CASPR2‐IgG.70 In terms of autoantibody generation, CASPR2 antibodies are often associated with a thymoma, particularly in patients with neuromyotonia and Morvan's syndrome,61 and this frequent clinical observation may implicate defective central tolerance checkpoints that permit autoreactive T cells to escape into the periphery and facilitate antigen‐specific autoimmunity. In addition, 50% of patients with CASPR2‐antibodies have a recently described HLA (human leucocyte antigen)‐DRB1*11:01 association, which contrasts to the HLA‐DRB1*07:01 allele observed in patients with LGI1‐antibodies.64 Aquaporin‐4 (AQP4) and Myelin Oligodendrocyte Glycoprotein (MOG) Antibodies Clinical Features Another distinctive paroxysmal phenomenon is the tonic spasms observed in neuromyelitis optica spectrum disorders (NMOSD). Tonic spasms occur more frequently in NMOSD than multiple sclerosis71 and consist of recurrent, painful, asymmetrical dystonic posturing, typically in one or more limbs, that usually last a few seconds to minutes and occur at high frequency.72 Occasionally, they can be preceded by a sensory aura and frequently they are triggered by hyperventilation, tactile stimuli, or voluntary movements. Typically, patients have a favorable course with a rapid response to anticonvulsant drugs. Patients with the more recently described MOG antibodies typically associate with phenotypes of optic neuritis, longitudinally extensive myelitis, and acute disseminated encephalomyelitis.73 Although the latter often shows basal ganglia and thalamic imaging changes, there are only rare descriptions of movement disorders in patients with MOG antibodies in addition to the few with ataxia.74 Treatment AQP4‐antibody‐mediated NMOSD is a chronic, naturally relapsing condition. Although immunotherapy efficacy has not been explored alongside a placebo arm, to date it appears that rituximab, azathioprine, and mycophenolate mofetil all reduce relapse rates by around 60% to 70%.75, 76 Furthermore, the avoidance of several agents with proven efficacy in multiple sclerosis is important in NMOSD as they can promote NMOSD relapses.77 The longer term treatment of MOG‐antibody‐mediated diseases has only recently been investigated, and it was revealed that a corticosteroid duration of less than 6 months is associated with a higher rate of relapses.73 Immunology The astrocytopathy associated with NMOSD is characterized by autoantibodies directed against AQP4, a water channel expressed on the astrocyte end‐foot processes. These autoantibodies can induce complement deposition, AQP4 internalization, and cointernalization of the glutamate transporter excitatory amino acid transporter‐2 (EAAT2).78 The deposition of complement is marked in tissue from patients with NMOSD and is a likely explanation as to why the relapses can produce severe disability. Indeed, a recent study using the monoclonal antibody eculizumab, which neutralizes the C5 complement component, has shown striking efficacy.79 By contrast to AQP4 antibodies, the initial pathology induced by MOG antibodies occurs on oligodendrocytes, and the downstream mechanisms are currently under active investigation. IgLON5‐Antibody Associated Neurodegeneration Clinical Features Patients with Iglon5‐autoantibodies mostly present with a chronic history of a rapid eye movement sleep behavior disorder with a distinctive non–rapid eye movement parasomnia plus bulbar involvement, dysautonomia, stridor, and hypoventilation.4 The main movement disorder described is chorea, but a few cases with postural instability and a supranuclear vertical gaze palsy had a phenotype resembling progressive supranucelar palsy (PSP), and the clinical spectrum continues to expand with the recent inclusion of myoclonus, myorhythmia, and dystonia.80, 81 Immunology IgLON family member 5 (IgLON5) antibodies are found in the serum and CSF of patients with postmortem evidence of a tauopathy. This finding highlights a novel relationship between autoimmune and degenerative disorders.81 IgLON5 antibodies bind the extracellular domain of their target neuronal protein and avidly label live neurons in culture. Furthermore, these patients have a consistent HLA‐DQB1*0501 and HLA‐DRB1*1001 genotype association. However, the phenotype and histology are highly suggestive of a neurodegenerative process. Brain pathology shows neuronal loss and extensive deposits of hyperphosphorylated tau protein predominantly in the hypothalamus and the brain stem tegmentum, with a different distribution from other tauopathies. Treatment Although initial reports described a universally poor outcome after immunotherapy, with death a common outcome,81, 82 more recent work reports a relatively frequent response to immunotherapies.80 It may be that earlier recognition of this condition portends a promising outcome with immunotherapies. In summary, this distinctive tauopathy is associated with NSAbs and a clear HLA association and may be a paradigm for the future study of the immune system leading to neurodegeneration. If so, it may yet be that all known NSAbs are pathogenic. Indeed, in the other NSAb‐associated diseases discussed previously, the autoantibodies are also very likely to be causative. Hence, knowledge of the location and subsets of B cells that produce these autoantibodies has biological and therapeutic relevance.31, 83 Therefore, next we use conventional immunological paradigms to help describe and model mechanisms of antigen‐specific autoantibody generation in these conditions and link these to current and future immunotherapies. The Therapeutically Relevant Immunology Identifying the autoantibody‐producing cells has important potential implications for considering future individualized therapies because, as shown in Figure 2, through the B cell lineage there are different sets of expressed surface markers. These markers alter as originally naïve B cells encounter T cell help and antigen, and differentiate into class‐switched IgG‐positive memory B cells.84, 85 Spatial requirements for these interactions are met in germinal centers, and this interaction involves a number of T and B cell molecules—both inhibitory and stimulatory—that regulate the intensity of this reaction.86 In humans, this balance is well exemplified by the appearance of autoantibody‐mediated neurology after administration of T cell–directed checkpoint inhibitors.87, 88 Therefore, even in the most well‐established NSAb‐mediated diseases, an isolated contribution of B cells is unlikely, and there should be more consideration given to therapies targeting both T cells and B cells. Successfully activated B cells will typically undergo successive rounds of interactions with antigen and T cells until they acquire high‐affinity antibodies. Subsequently, these activated B cells may differentiate into antibody secreting cells in circulation (plasmablasts), where importantly they downregulate CD20, the target of rituximab.89 The plasmablasts which reach bone marrow niches have also often downregulated CD19 and expressed CD138.90 One important question is the degree to which these now long‐lived bone marrow resident plasma cells produce the autoantibodies.85 If they are major producers of autoantibodies, patients should be sensitive to drugs such as bortezomib, which target the proteasome—an organelle that is highly active in plasma cells. Indeed, early observational studies suggest a possible, albeit limited, response to bortezomib.91 Conversely, rituximab should have little impact as these cells are CD20 negative. Yet there is evidence that several such syndromes can respond to CD20‐targeted medications.19, 76, 58 Therefore, further careful clinic‐immunological studies are required to highlight the relative roles of these therapies and their effect on B cell subsets and autoantibody levels: these will better inform our understanding of disease immunobiology.31-33, 83 It may be that combinations of plasma cell depleting agents plus removal of precursor B cells are required to effectively remove autoreactive capacities from the B cell lineage. However, it is not solely the lymphocyte surface markers that provide therapeutically tractable drug targets. There are increasing examples where pharmacological manipulation of the cytokine and chemokine signalling pathways produce therapeutic efficacy. For example, B cells and antibody‐secreting cells are known to show dependence on Interleukin 6 (IL‐6) for survival. Indeed, inhibition of the IL‐6 receptor with tociluzimab has some proven benefit in both NMOSD and NMDAR‐antibody encephalitis.93, 94 Perhaps this line of inquiry will generate further targets to which there are already available modulators. As causative autoantibodies are consistently, albeit not universally, found in the CSF, one possibility is disease initiation and propagation by a primary intrathecal response: a hypothesis that does not involve peripheral lymph nodes. This is strongly mitigated by the consistent observation of serum autoantibodies in all of these conditions, including NMDAR‐antibody encephalitis post‐HSVE, typically at concentrations far in excess of the CSF autoantibodies.1, 10, 24, 34, 95, 96 In addition, the universal observation of peripheral, likely immunizing, systemic tumors also mitigates this possibility. More plausible is the notion of an initiating peripheral immunization. Perhaps in a manner akin to HSVE, many CNS‐restricted antigens reach the periphery via CNS lymphatics and draining cervical lymph nodes (Fig. 2).42 The constitutive versus active nature of this drainage requires further investigation as it may determine the probability of a CNS lesion initiating a peripheral immunization. Subsequent to the immunization, at least some of the peripheral response must transfer to the CNS to mediate a brain disease. One outstanding question is whether this is principally mediated by migration of the B cells or the soluble antibodies across the blood‐brain or blood‐CSF barrier. This distinction may alter clinical management strategies; for example, natalizumab will block lymphocyte trafficking across the blood‐brain barrier. Several clues exist to help us understand the nature of the brain‐based immune response. Animal studies lend support to the intuitive notion that the brain‐expressed antigens can act as a “sink” for CNS‐transferred serum autoantibodies resulting in undetectable CSF autoantibody levels.97 Indeed, although CSF LGI1 antibodies are present in most patients, they are undetectable in some cases,96, 98 and this balance may represent a saturation point of the “sink.” By contrast, in patients with NMDAR‐antibody encephalitis, the presence of autoantibodies in CSF is a requirement for definitive diagnosis.99 However, serial measurements of CSF or serum autoantibody levels in all of these conditions only broadly correlate with clinical outcomes,19, 20, 100 and perhaps other, many as yet undetermined, factors including cytokines, complement and even other autoantibody reactivities,33 together contribute to the overall clinical status. In addition to diffusion alone, there is often marked intrathecal synthesis of the NMDAR antibodies and, indeed, NMDAR‐autoantibody secreting cells within the CNS have recently been isolated by single‐cell cloning techniques.33 Therefore, the intrathecal retention of antigen‐specific cells appears necessary for generation of NMDAR‐antibody encephalitis and may be mediated by the high CSF levels of the B cell and plasma cell chemokine (C‐X‐C motif) ligand 13 (CXCL13).32 These observations, coupled with the highly heterogeneous phenotypes of patients with NMDAR antibodies in serum but not CSF,40, 101, 102 may suggest that the predicted natural diffusion of NMDAR‐IgG into the CNS is insufficient to generate this encephalitis phenotype. Yet, maybe the soluble autoantibodies do play a role in ongoing disease as plasma immunoabsorption of IgG leads to a fall in CSF autoantibodies and correlates with improvements in clinical status.103 The relative contributions of serum autoantibody transfer into the CNS and the degree of immune cell infiltration may vary across diseases but also within diseases, and could determine the likelihood of amelioration with plasma exchange or predict the need for future intrathecal‐directed therapies. Autoantibodies Against Glutamic Acid Decarboxylase 65 (GAD65), an Intracellular Synaptic Protein: Syndromes, Immunology, and Treatments In contrast to the NSAb‐mediated disorders, those associated with autoantibodies directed against intracellular targets are generally considered nonpathogenic. Indeed, as discussed previously, passive transfer and active immunization experiments have proven that some of these autoantibodies do not cause neurological diseases.104 However, the spectrum of disorders associated with autoantibodies against GAD65 and amphiphysin may challenge this notion from clinical and laboratory perspectives.105, 106 Here, we discuss the GAD‐antibody syndromes, with prominent movement disorders, in greater detail. Clinical Features Antibodies against the intracellular enzyme GAD65 are very frequently detected in stiff person syndrome and related disorders (stiff person spectrum disorders [SPSD]). This group of disorders share the hallmark features of fluctuating muscle stiffness with superimposed spasms and hyperekplexia (an excessive startle response to acoustic or tactile stimuli).107 Classic stiff person syndrome involves stiffness of the lower back and proximal leg muscles with characteristic hyperlordotic posturing. In focal forms of SPSD, stiffness may be restricted to 1 limb (stiff limb syndrome).108 Other variants of SPSD are defined by the presence of additional neurological signs such as cerebellar ataxia. Progressive encephalomyelitis with rigidity and myoclonus typically designates the severe end of the spectrum, characterized by prominent hyperekplexia and myoclonus, generalized stiffness, brain stem signs, and dysautonomia.109 Immunology Including Coexistent NSAbs High concentrations of GAD65 antibodies associate with a limited set of clinically distinctive phenotypes, namely: SPSD, cerebellar ataxia, epilepsy and limbic encephalitis (Fig. 3B), suggesting some syndrome specificity.107, 110, 111 In addition, several patients with GAD65 antibodies do respond to immunotherapy. Also, by comparison to patients with Hu‐ or Ma2‐antibody‐associated encephalitis, patients with GAD65‐antibody encephalitis showed lower CD8/CD3 ratios, indicating an appropriate designation of GAD65 antibodies between other intracellular autoantibodies and NSAbs.62 These collective observations lead to the intriguing notion that GAD65 antibodies may have some causative potential. Indeed, the closely related amphiphysin antibodies, typically associated with paraneoplastic stiff person syndrome, have been shown to both reproduce disease upon transfer to experimental animals, and they may gain access to their intracellular antigenic target.8, 106, 112 Alternatively, maybe the GAD65 antibodies coexist with NSAbs that target the extracellular domains of antigens at GABAergic and glycinergic inhibitory synapses, such as the alpha 1 subunit of the glycine receptor (GlyRα1), and GABAAR.3, 105-107 Furthermore, another NSAb found in some SPSD patients is directedagainst dipeptidyl‐peptidase‐like protein‐6 (DPPX),113, 114 a regulatory subunit of Kv4.2 potassium channels. Perhaps these coexistent autoantibodies, often with antigens expressed in the same neurons, implicate epitope spread as a mechanism to diversify the polyclonal immune responses after a triggering event that exposes several antigens to the immune system. Indeed, it appears that these coexistent NSAbs confer even more disease specificity: for example, Glycine Receptor (GlyR) antibodies frequently associate with progressive encephalomyelitis with rigidity and myoclonus and DPPX antibodies with a distinct phenotype of truncal stiffness, prominent hyperekplexia, and cerebellar ataxia.113 Both GlyR and DPPX are surface expressed, and the respective NSAbs are likely to be pathogenic, with in vitro evidence they induce antigen internalization.109, 114 Treatment Consistent with the aforementioned paradigm of intracellular antibodies versus NSAbs, from patients with SPSD the treatment responses appear better in patients with GlyR antibodies than GAD65 antibodies alone.115 The treatment with the only proven randomized clinical data within the conditions discussed in this review is intravenous immunoglobulins in SPSD.116 Patients often show a moderate benefit from this drug, but longer term alternatives are yet to be satisfactorily explored. Future Directions Given that our understanding of the immunology underlying NSAb‐mediated diseases remains in its infancy, this appears to be an important avenue for future study. Available data have led researchers to consider drugs that appear biologically intuitive, but it is yet to be seen if we can achieve disease specificity. There are some potential reasons to maintain optimism in the potential for precision medicine. First, the generation of patient‐derived monoclonal antibodies in some of these conditions offers a method to directly out‐compete the endogenous patient antibodies.117 Of course, this comes with a series of potential immunological hazards, but it would form an elegant method to test the hypothesis of whether the antibodies are the major disease perpetrators. Other options in the pipeline include the use of selective cytokine and chemokine blockade. For example, in NMDAR‐antibody encephalitis, raised CSF levels of CXCL13 have been proposed as an intrathecal lymphocyte chemoattractant. Their neutralization may inhibit lymphocyte crossing. However, if this were effective, it may yet require a parallel peripheral depletion of B cells to adequately extinguish the disease. Indeed, this combinatorial approach may be a future theme in these increasingly complex diseases that require collaborations of at least T cells and B cells. One such vision may by combinatorial assessment of the autoantibody production from patient lymphocytes under a large number of cytokine conditions31, 83 and then block the dominant culprit pathways with targeted monoclonal therapies. Such an approach would be patient specific, especially if complemented by evaluation of endogenous cytokine levels, but if generic stimuli expanded these B cells, this approach may be intrinsically limited. Given recent advances in the clinical immunology,31, 83, 93 future studies should be able to answer these possibilities rapidly. Conclusions The increasing numbers of identified neuronal autoantibodies are associated with a broadening clinical spectrum of autoantibody‐mediated movement disorders. This contemporary expansion makes the field increasingly important for the movement disorder specialist and for the general neurologist. This is particularly the case given the recognition that early immunotherapy is likely to improve prognosis and prevent the ongoing pathogenic effects of the autoantibodies. However, to understand the root causes of these illnesses, the field requires an improved future understanding of the varied roles of immune components—including T cells, B cells, plasma cells—and their associated surface markers. Indeed, many of the conventional immunological paradigms as outlined require confirmation with the direct study of these diseases. In addition to biological insights, this may offer a method to specifically target causative cell types in these diseases. The relative roles of these different immune components may vary in conditions with intracellular versus surface autoantibodies and depend on the inciting factor in the autoantibody‐mediated conditions. Furthermore, the degree to which reduction in T cell function or autoantibody level is required to achieve clinical improvement should be considered as many of these autoantibodies can persist for years despite good clinical remission.20, 24, 118 Therefore, to move toward precision medicine in these conditions, there is an urgency to better appreciate the immunological mechanisms that underlie the generation and perpetuation of the autoantibodies, and this may lead to novel therapeutic strategies that could be addressed in clinical trials. Author Roles 1) Research project: A. Conception, B. Organization, C. Execution; 2) Statistical Analysis: A. Design, B. Execution, C. Review and Critique; 3) Manuscript: A. Writing of the first draft, B. Review and Critique. V.D.: 1A, 1B, 1C, 3A, 3B B.B.: 1A, 1B, 1C, 3A, 3B A.K.K.: 1B, 1C, 3B S.R.I.: 1A, 1B, 1C, 3A, 3B Full financial disclosure for the previous 12 months SRI is a coapplicant and receives royalties on patent application WO/2010/046716 (U.K. patent no., PCT/GB2009/051441) entitled ‘Neurological Autoimmune Disorders’. The patent has been licensed to Euroimmun AG for the development of assays for LGI1 and other VGKC‐complex antibodies. Supporting Information References

Research ArticleAutoimmunityInflammation Free access | 10.1172/JCI121901 TLR-stimulated IRAKM activates caspase-8 inflammasome in microglia and promotes neuroinflammation Cun-Jin Zhang,1,2,3 Meiling Jiang,1,3,4 Hao Zhou,3 Weiwei Liu,3 Chenhui Wang,3,5,6 Zizhen Kang,3 Bing Han,3 Quanri Zhang,3 Xing Chen,3 Jianxin Xiao,3 Amanda Fisher,7 William J. Kaiser,7 Masanori A. Murayama,8,9 Yoichiro Iwakura,9 Ji Gao,10 Julie Carman,10 Ashok Dongre,10 George Dubyak,11 Derek W. Abbott,12 Fu-Dong Shi,2,13 Richard M. Ransohoff,14 and Xiaoxia Li3 First published October 29, 2018 - More info Abstract NLRP3 inflammasome plays a critical spatiotemporal role in the pathogenesis of experimental autoimmune encephalomyelitis (EAE). This study reports a mechanistic insight into noncanonical NLRP3 inflammasome activation in microglia for the effector stage of EAE. Microglia-specific deficiency of ASC (apoptosis-associated speck-like protein containing a C-terminal caspase-activation and recruitment [CARD] domain) attenuated T cell expansion and neutrophil recruitment during EAE pathogenesis. Mechanistically, TLR stimulation led to IRAKM–caspase-8–ASC complex formation, resulting in the activation of caspase-8 and IL-1β release in microglia. Noncanonical inflammasome-derived IL-1β produced by microglia in the CNS helped to expand the microglia population in an autocrine manner and amplified the production of inflammatory cytokines/chemokines. Furthermore, active caspase-8 was markedly increased in the microglia in the brain tissue from patients with multiple sclerosis. Taken together, our study suggests that microglia-derived IL-1β via noncanonical caspase-8–dependent inflammasome is necessary for microglia to exert their pathogenic role during CNS inflammation. Introduction Numerous studies indicate that the inflammatory process in multiple sclerosis (MS) and experimental autoimmune encephalomyelitis (EAE) is initiated by T cells that are reactive against myelin (1–6). Importantly, a growing body of literature has suggested that IL-1β produced by the multimolecular complex known as the inflammasome is critical for the pathogenesis of a variety of CNS inflammatory processes, including MS and EAE (7–10). The NLRP3 inflammasome consists of NLRP3 linked via a homotypic pyrin domain interaction to the inflammasome adaptor molecule apoptosis-associated speck-like protein containing a C-terminal caspase-activation and recruitment (CARD) domain (ASC). The canonical NLRP3 inflammasome is activated in macrophages and dendritic cells via procaspase-1 self-cleavage to generate active caspase-1, which processes pro–IL-1β, resulting in release of mature IL-1β (11–15). NLRP3 inflammasome activation in macrophages and dendritic cells is required for activation and expansion of Th17 cells during priming in the periphery (16–18). Interestingly, we found that caspase-8 instead of caspase-1 is critical for ASC-dependent IL-1β production in Th17 cells (3). Notably, we recently reported that T cell–intrinsic ASC is essential for the effector stage of EAE (3). Microglia in the CNS continuously scrutinize their environment for damage or infection. Microglia are equipped with TLRs and P2X7R to sense TLR ligands and ATP released by dying CNS resident cells during CNS inflammatory tissue injury (19–23). We now report that microglia-intrinsic ASC is required for the effector stage of EAE, acting in the context of a noncanonical inflammasome complex. In particular, IL-1β is processed and produced in an ASC-NLRP3–caspase-8–dependent manner in primary microglia. Mechanistically, TLR stimulation induces IRAKM–caspase-8–ASC complex formation, resulting in the activation of caspase-8 in the presence of ATP and IL-1β production in microglia. We also found that microglia-intrinsic IRAKM–caspase-8 inflammasome activation is required for microglia survival and proliferation in the CNS of EAE mice. Consistently, active caspase-8 was markedly increased in the microglia in brain tissue from patients with MS. Notably, while IL-1R was highly induced on microglia during EAE, transfer of WT but not IL-1β–deficient microglia was able to enhance the EAE phenotype in microglia-diminished recipient mice (Cx3cr1Cre-Csf1rfl/fl mice). Furthermore, Ki67+ microglia were much reduced in Cx3cr1Cre-Csf1rfl/fl mice receiving IL-1β–deficient microglia compared with mice transferred with WT microglia. Taken together, our results suggest that noncanonical inflammasome-derived IL-1β produced by microglia in the CNS helps expand the microglia population via an autocrine manner and amplifies the production of inflammatory cytokines/chemokines. As a consequence, the activated microglia are permitted to exert their pathogenic role during EAE, providing an unparalleled opportunity to develop therapeutic strategies for the treatment of MS. Results Microglia-intrinsic ASC is required for the effector stage of Th1- and Th17-induced EAE in the CNS. In the brain, antigen-specific T cells are restimulated by antigen presenting cells (APCs), leading to disease induction and progression (24–26). While dendritic cells (DCs) are professional APCs, microglia in the CNS are also potential APCs (24–26). In addition to Ascfl/flCd11cCre mice, we also bred Ascfl/fl mice to Cx3cr1Cre-ER. To delete ASC specifically in microglia, we transferred WT bone marrow to lethally irradiated Ascfl/flCx3cr1Cre-ER-EYFP and control Ascfl/+Cx3cr1Cre-ER-EYFP mice, which are referred as AscΔmicroglia and littermate control mice. Using Cx3cr1Cre-ER-EYFP bone marrow chimeric mice (transferred with WT bone marrow cells), we found that over 97% of the CD45lo microglia from the brain tissue gained eYFP expression, implying the high efficiency of Asc deletion in microglia of the AscΔmicroglia mice (Figure 1A and Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI121901DS1). To further confirm these results, we also crossed Cx3cr1Cre-ER-EYFP mice with Rosa26-stop-DsRed to trace the efficiency of Cre recombinase in microglia. Consistently, over 95% of CD45lo CD11b+ microglia gained DsRed expression (Supplemental Figure 1B). We tested the effect of microglia-specific Asc deletion on neuroinflammation and demyelination by immunizing AscΔmicroglia and littermate control mice with the neuroantigen myelin oligodendrocyte glycoprotein (MOG35–55) peptide. AscΔmicroglia mice had attenuated disease severity compared with controls (Figure 1B). Inflammatory mononuclear cell infiltration in the brain, including by CD4+ T cells, B cells, neutrophils, and macrophages, was similarly decreased in mice with microglia-specific Asc ablation compared with controls (Figure 1, C and D; and Supplemental Figure 1C), and the expression of inflammatory cytokines and chemokines in the spinal cord was also significantly decreased (Figure 1E). Histopathological analysis showed decreased infiltrating immune cell accumulation and resultant demyelination in spinal cords of AscΔmicroglia mice compared with controls (Figure 1F). Together, these data indicate that deletion of ASC from microglia protects mice from the pathogenesis of EAE, with marked attenuation of disease severity. We found that Asc deletion in microglia (AscΔmicroglia) substantially reduced Th1- and Th17-induced EAE pathogenesis (Figure 1, G and H), with markedly reduced inflammation in the brain and spinal cord (Supplemental Figure 1, D–G). Figure 1 Microglia-specific ASC deficiency attenuated EAE disease. Analysis of results for WT→ASCfl/+Cx3cr1Cre-ER (ASCΔWT) and WT→ASCfl/flCx3cr1Cre-ER (ASCΔmicroglia) bone marrow chimera mice in EAE disease. (A) FACS analysis of CreER-EYFP expression in microglia of ASCΔWT mice with or without tamoxifen administration on day 16 of EAE induced by active immunization with MOG35–55. (B) Mean clinical score for EAE in ASCΔWT (n = 6) and ASCΔmicroglia (n = 5) bone marrow chimera mice induced by active immunization with MOG35–55. Absolute numbers (C) and gating strategy (D) of immune cell infiltration determined at the peak of disease in brains of EAE mice by flow cytometry (n = 3/group). (E) Inflammatory gene expression in the lumbar spinal cords as assessed at the peak of disease (n = 4). (F) Luxol Fast Blue and H&E staining of lumbar spinal cords harvested at the peak of disease. Scale bars: 200 μm. (G and H) Mean clinical score for EAE in ASCΔWT and ASCΔmicroglia bone marrow chimera mice induced by adoptive Th17 (G) (n = 7 and n = 5, respectively) or (H) Th1 (n = 5/group) transfer. Data are representative of 2 independent experiments; mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 (unpaired 2-tailed Student’s t test). EAE clinical score by 2-way ANOVA. Microglia processes IL-1β in an NLRP3-ASC–caspase-8–dependent manner. NLRP3 inflammasome was activated by LPS+ATP stimulation in primary microglia, as evident by pro–IL-1β cleavage and IL-1β secretion in an NLRP3/ASC-dependent manner (Figure 2, A, B, and E). Interestingly, we found that caspase-8 instead of caspase-1 was activated in an NLRP3/ASC-dependent manner in primary microglia (Figure 2, A and B). LPS+ATP stimulation triggered the recruitment of caspase-8 to NLRP3-ASC in microglia (Figure 2C). Consistently, caspase-8 was recently identified as a noncanonical inflammatory caspase in myeloid-lineage cells and T cells (3), and it has been shown to be capable of processing pro–IL-1β (21, 27–29). We thus speculated that in microglia, caspase-8 might participate in an ASC-dependent noncanonical inflammasome capable of processing pro–IL-1β. In response to ATP stimulation, IL-1β production was significantly reduced in microglia from Casp8−/−Ripk3−/− mice compared with those from Casp8+/−Ripk3−/− mice, implying a role for caspase-8 in microglia for the ASC-dependent processing of pro–IL-1β (Figure 2, D and E). In support of this, peptide inhibitor of caspase-8 (IETD-fmk) but not of caspase-1 (YVAD-fmk) specifically blocked IL-1β production in microglia (Figure 2F). Taken together, our results suggest that caspase-8 is critical for ASC-dependent IL-1β production in primary microglia, which is analogue to the NLRP3-ASC–caspase-8 cascade in Th17 cells (3). This NLRP3-ASC–caspase-8 axis was operative in microglia in vivo, since FACS analysis showed caspase-8 (instead of caspase-1) was indeed activated in microglia from the brain tissue of EAE mice in an NLRP3/ASC-dependent manner (Figure 2, G and H). Importantly, endogenous TLR ligands (derived from tissue damage) could also activate this caspase-8–dependent noncanonical NLRP3 inflammasome in both human and mouse microglia (Figure 2, I and J). Additionally, genetic ablation of caspase-8 specifically in microglia attenuated EAE, indicating the importance of this pathway for the pathogenesis of EAE (Figure 2K). Figure 2 Microglia processes IL-1β in an ASC-NLRP3–caspase-8–dependent manner. (A and B) Primary microglia from 6-week-old mice were primed with LPS for 4 hours prior to stimulation with ATP for the indicated times. Cell lysate and supernatants were collected together and immunoblotted with the indicated antibodies. (C) Results of immunoblotting with antibodies to the indicated proteins for microglia stimulated with LPS (4 hours) plus ATP (30 minutes) and immunoprecipitated with anti-ASC. (D) Primary microglia from adult mice were primed with LPS for 4 hours prior to stimulation with ATP. Cell lysate and supernatants were collected together and immunoblotted with the indicated antibodies. (E) IL-1β ELISA of cell-free supernatants from adult mice–derived primary microglia treated with LPS for 4 hours and ATP for 15 or 30 minutes (n = 3/group). (F) IL-1β ELISA of primary microglia from adult mice treated with LPS and indicated inhibitors for 4 hours prior to stimulation with 0.2 mM ATP for the indicated times. UN, untreated; YVAD-fmk, caspase-1 inhibitor; IETD-fmk, caspase-8 inhibitor (n = 4/group). (G and H) FACS analysis of caspase-8–FLICA (fluorescent labeled inhibitors of caspases) in microglia of EAE mice at peak disease in vivo (n = 6/group). (I) Primary microglia from 6-week-old mice were stimulated with TLR agonists HMGB1 and CpGb for 8 hours prior to stimulation with 0.2 mM ATP. Cell lysate and supernatants were collected together and immunoblotted with the indicated antibodies. (J) Human microglia cell line SV40 was stimulated with LPS (100 pg/ml) for 4 hours and ATP (0.2 mM) for 30 or 60 minutes. Cell lysate and supernatants were collected together and immunoblotted with the indicated antibodies. (K) Clinical score for WT→Casp8fl/+Cx3cr1Cre-ER Ripk3–/– (Caspase8ΔWT) and WT Casp8fl/flCx3cr1Cre-ER Ripk3–/– (Caspase8Δmicroglia) bone marrow chimera mice (n = 6/group). Data are representative of 2 independent experiments; mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (unpaired 2-tailed Student’s t test). EAE clinical score by 2-way ANOVA. IRAKM controls the activity of caspase-8 and IL-1β production in microglia. We wondered how TLR signaling leads to caspase-8 activation. Upon TLR stimulation, IRAK proteins (IRAK1–4) are recruited to the receptor complex to mediate downstream signaling pathways. By mass spectometry analysis, we found that IRAKM, but not IRAK1 or IRAK2, pulled down caspase-8 (Supplemental Figure 2). Consistently, we found that in response to LPS or LPS+ATP stimulation, caspase-8 indeed specifically interacted with IRAKM in primary microglia (Figure 3A). IRAKM (but not IRAK1/2) formed a complex with ASC/NLRP3 in primary microglia (Figure 3, B and C). Based on these biochemical findings, we hypothesized that IRAKM may play a critical role in activating the NLRP3-ASC–caspase-8 cascade in microglia. We indeed found that IRAKM deficiency abolished LPS+ATP-induced caspase-8 and IL-1β cleavage and IL-1β production in microglia (Figure 3, D–F). Nevertheless, LPS-induced NF-κB activation was not affected in IRAKM-deficient microglia (Figure 3G). These results suggest that IRAKM is specifically required for the assembly of caspase-8–dependent NLRP3 inflammasome. In support of this hypothesis, we found that ASC–caspase-8 interaction was ablated in IRAKM-deficient microglia (Figure 3H), whereas IRAKM was still able to form a complex with caspase-8 in the absence of ASC (Figure 3I). These results indicate that IRAKM first engages caspase-8, followed by their interaction with ASC for activation and cleavage of caspase-8. Interestingly, IRAKM-ASC interaction was still intact in the absence of caspase-8 (Figure 3J), indicating that IRAKM serves as a bridging molecule for ASC and caspase-8. Figure 3 IRAKM controls the activity of caspase-8 and IL-1β production in microglia. (A–C) Results of immunoblotting with antibodies to the indicated proteins for primary microglia stimulated with LPS (4 hours) plus ATP (30 minutes) and immunoprecipitated with anti–caspase-8 (A), anti-IRAKM (B), or anti-ASC (C). (D) Primary microglia from 6-week-old mice with indicated genotyping were primed with LPS (100 pg/ml) for 4 hours prior to stimulation with 0.5 mM ATP for the indicated times. Cell lysate and supernatants were collected together and immunoblotted with the indicated antibodies. (E) Primary microglia from 6-week-old mice with indicated genotypes was stimulated with LPS (0.1 μg/ml, 4 hours) plus ATP (30 minutes) and stained with caspase-8–FLICA in the last hour, followed by flow analysis of caspase-8 activation (n = 4/group). (F) IL-1β ELISA of cell-free supernatants from adult mice–derived primary microglia treated with LPS (100 pg/ml) for 4 hours and ATP (0.2 mM) for 15 or 30 minutes (n = 4/group). (G) Primary microglia from mice with indicated genotypes was primed with LPS (100 pg/ml) for 4 hours prior to stimulation with 0.2 mM ATP for the indicated times. Cell lysate and supernatants were collected together and immunoblotted with the indicated antibodies. (H–J) Results of immunoblotting with antibodies to the indicated proteins for primary microglia stimulated with LPS (100 pg/ml, 4 hours) plus ATP (30 minutes) and immunoprecipitated with anti-ASC (H) or anti-IRAKM (I and J). Data are representative of 2 independent experiments; mean ± SEM. *P < 0.05 (unpaired 2-tailed Student’s t test). IRAKM is required for the activation of caspase-8 and IL-1β production in microglia during EAE. We then examined the importance of IRAKM in caspase-8–dependent noncanonical inflammasome activation in microglia in vivo. To achieve this goal, we generated a mouse strain in which exon 3 of the gene Irak3 (which encodes IRAKM) is flanked by loxP sites (Figure 4, A and B). To delete IRAKM in microglia, we transferred WT bone marrow to lethally irradiated Irakmfl/flCx3cr1Cre-ER-EYFP and control Irakmfl/+Cx3cr1Cre-ER-EYFP mice, which are referred to as IrakmΔmicroglia and littermate control mice. We examined the effect of microglia-specific IRAKM deletion on neuroinflammation and demyelination by immunizing IrakmΔmicroglia and littermate control mice with the neuroantigen MOG35–55 peptide. IrakmΔmicroglia mice had attenuated disease severity compared with controls (Figure 4C). Inflammatory mononuclear cell infiltration in the brain, including by CD4+ T cells, B cells, neutrophils, and macrophages, was similarly decreased in mice with microglia-specific IRAKM ablation compared with controls (Figure 4D). Histopathological analysis showed decreased infiltrating immune cell accumulation and resultant demyelination in spinal cords of IrakmΔmicroglia mice compared with controls (Figure 4E). Together, these data indicate that deletion of IRAKM from microglia protects mice from the pathogenesis of EAE, with marked attenuation of disease severity. Consistent with the ex vivo data, caspase-8 activation and IL-1β production in primary microglia from the EAE mice were substantially reduced by IRAKM deficiency (Figure 4, F and G). Figure 4 IRAKM controls the activity of caspase-8 and IL-1β production in microglia in vivo. (A and B) Targeting vector design for generation of a mouse strain with Irak3 exon 3 flanked by loxp sites (A), and Western blot analysis of IRAKM expression in FACS-sorted microglia from indicated mice (B). (C) Mean clinical score for EAE in WT→Cx3cr1creIrakmfl/+ (IRAKMWT, n = 5) and WT→Cx3cr1creIrakmfl/fl (IRAKMΔmicroglia, n = 7) bone marrow chimera mice induced by active immunization with MOG35–55. (D) Absolute numbers of immune cell infiltration as well as resident microglia determined at the peak of disease in brains of EAE mice by flow cytometry (n = 3/group). (E) H&E and Luxol Fast Blue staining of lumbar spinal cords harvested at the peak of disease. Scale bars: 200 μm. (F) Flow cytometry analysis of caspase-8 activation in microglia of EAE mice at peak disease (n = 4). (G) Primary microglia of EAE brains were isolated and cultured for 24 hours ex vivo. Cell-free supernatant was collected for IL-1β ELISA (n = 4). Data are representative of 2 independent experiments; mean ± SEM. *P < 0.05, **P < 0.01 (unpaired 2-tailed Student’s t test). EAE clinical score by 2-way ANOVA. Microglia-intrinsic caspase-8 inflammasome activation is required for microglia survival and proliferation in the CNS. We next investigated how caspase-8–dependent noncanonical inflammasome activation in microglia contributes to EAE pathogenesis. Caspase-8 was indeed activated in the expanded microglia in the CNS during the course of EAE (Figure 5, A and B). Importantly, active caspase-8 was increased in the microglia in brain tissue from patients with MS (Figure 5, C and D, and Supplemental Table 1). Furthermore, we detected IL-1β production in sorted microglia from brain tissue of EAE mice (Figure 5E). Interestingly, we found IL-1R was highly induced on microglia from brain tissue of EAE mice (Figure 5F). In the brain tissue of EAE mice, we noted that 46% of the IL-1R+caspase-8+ microglia were Ki67+, whereas only 18% of IL-1R–caspase-8– microglia were Ki67+, suggesting that caspase-8–dependent IL-1β production by microglia in the CNS might directly act on microglia in an autocrine manner (Figure 6A). In support of an autocrine action of IL-1β in this ex vivo system, we found that IL-1RA treatment also reduced cell survival of LPS+ATP-treated microglia (Figure 6B). The removal of the IRAKM–caspase-8–ASC-NLRP3 axis increased cell death in LPS+ATP-treated microglia (Figure 6B). While caspase-8+ microglia as well as Ki67+ and Brdu+ microglia were substantially reduced in the spinal cord tissues of EAE mice with microglia-specific deletion of ASC, IRAKM, or caspase-8 (Figure 6, C and D), establishing the IRAKM–caspase-8–ASC axis in promoting microglia survival and proliferation contributed to EAE pathogenesis. Flow cytometry analysis of the infiltrated cells in brain tissues indicated that the microglia have enhanced cell proliferation (Ki67+ cells) in the WT mice as compared with those in microglia-specific ASC-, IRAKM-, or caspase-8–deficient mice (Figure 6E). It was reported that specific deletion of CSF1R in microglia (Csf1rfl/flCx3cr1Cre) diminishes the microglia population in the CNS (Supplemental Figure 3A). We transferred WT and IL-1β–deficient microglia into the brains of Csf1rfl/flCx3cr1Cre mice after tamoxifen-induced deletion of CSF1R in microglia (Supplemental Figure 3, A and B). We found that WT but not IL-1β–deficient microglia were able to enhance the EAE phenotype, indicating the pathogenic role of microglia-derived IL-1β in EAE (Figure 6F). Consistently, we showed that Csf1rfl/flCx3cr1Cre mice (treated with tamoxifen) without microglia transfer displayed reduced EAE disease compared with mice transferred with WT microglia (Supplemental Figure 3C), further supporting the pathogenic role of microglia-derived IL-1β in EAE. Furthermore, the microglia numbers were reduced in the brains of mice transferred with IL-1β–deficient microglia compared with mice that received WT microglia (Figure 6G), supporting the impact of microglia-derived IL-1β on microglia survival during EAE. Moreover, Ki67+ and Brdu+ microglia were much reduced in Cx3cr1Cre-Csf1rfl/fl mice receiving IL-1β–deficient microglia compared with mice that were transferred with WT microglia (Figure 6, H–J). Compared with WT microglia, Il1b–/– microglia from EAE brains produce less inflammatory cytokines and chemokines, indicating IL-1β signaling is critical for promoting microglia activation (Supplemental Figure 3D). Figure 5 Dynamics of caspase-8 activation, IL-1β production, and IL-1R expression in microglia. (A and B) Analysis of results for WT→ASCfl/+Cx3cr1Cre-ER (ASCΔWT) bone marrow chimeric mice in EAE disease. Flow cytometry analysis of activated caspase-8 in primary EYFP+ microglia isolated from brains of EAE mice at the indicated time points (n = 4). The quantification of activated caspase8+ microglia is shown in B. (C and D) Brain specimens of 7 controls and 11 MS cases were provided by the Multiple Sclerosis Tissue Bank at Imperial College London. The edges of cortical active lesions in brain slices were stained with the indicated antibodies against microglia markers TMEM119 and activated caspase-8–FLICA. Scale bars: 50 μm (C). The activated caspase-8+ cells were quantified (D). (E) EFYP+ primary microglia sorted from EAE mice (ASCΔWT) at the indicated times was cultured overnight and supernatant was subjected to ELISA analysis of IL-1β production (n = 4). (F) FACS analysis of IL-1R expression in EFYP+ primary microglia isolated from brains of WT naive and EAE mice (ASCΔWT) (n = 4). Data are representative of 2 independent experiments; mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (unpaired 2-tailed Student’s t test). Figure 6 Microglia-intrinsic caspase-8 inflammasome activation is required for microglia survival and proliferation in the CNS. (A) The proliferation of microglia from WT EAE mouse brain at peak disease was analyzed by flow cytometry after staining with antibodies against CD45, CD11b, IL-1R, Ki67, and caspase-8–FLICA. (B) Primary microglia isolated from mouse brain with indicated genotypes were stimulated with LPS+ATP in the presence of IL-1Rα followed by PI staining and flow cytometry analysis (n = 3). (C and D) Spinal cord from EAE mice with indicated genotypes at peak disease were stained with antibodies against caspase-8–FLICA, Ki67, and Brdu. EYFP indicates the microglia (n = 6/group). WT→Cx3cr1Cre-ER (Cx3Cr1ΔWT). Scale bars, 50 μm. (E) Flow cytometry analysis of Ki67+ microglia in brains of EAE mice with indicated genotypes (n = 6). (F–J) WT and Il1b–/– microglia isolated from adult mouse brain were transferred to Csf1rfl/flCx3cr1cre mice after the fourth tamoxifen injection (5 mg/mouse/week, i.p.) and the mice were subjected to EAE induction 2 weeks after the last tamoxifen injection (n = 6). Clinical score was presented as mean ± SEM (F). Total microglia (G) and Ki67+ microglia (H) in brains at EAE peak were analyzed by flow cytometry. Brdu+ microglia were analyzed in brain slices (I and J). Scale bar, 50 μm. Data are representative of 2 independent experiments; mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 (unpaired 2-tailed Student’s t test). EAE clinical score by 2-way ANOVA. Microglia-intrinsic IRAKM–caspase-8 inflammasome activation is required for neutrophil recruitment in the CNS during EAE. We next investigated the consequence of IRAKM–caspase-8–dependent microglia survival/proliferation on EAE pathogenesis. In addition to IL-1β production, we also detected release of other cytokines and chemokines from sorted microglia from brain tissue of EAE mice (Figure 7A), including IL-23, IL-6, CXCL-1, CXCL2, and CXCL5. Notably, microglia expansion was readily datable at day 10, peaking at day 16 (peak of disease) upon immunization (Figure 7B). Interestingly, both IL-17– and IFN-γ–producing cells were reduced at day 10 and day 16 (the peak of the disease) in the brain tissue of microglia-specific ASC- or IRAKM-deficient mice compared with that of WT mice (Figure 7C). Likewise, microglia-specific deficiency in ASC or IRAKM also had a substantial impact on the neutrophils in the brain tissue as early as day 10 upon immunization (Figure 7D), which coincided with the microglia activation and expansion in the WT cells (Figure 7A). Consistently, IL-17/IFN-γ–producing cells and neutrophils were reduced in the brain tissue of mice transferred with IL-1β–deficient microglia compared with mice that received WT microglia (Figure 7E). Figure 7 Microglia-intrinsic IRAKM–caspase-8 inflammasome activation is required for neutrophil recruitment and T cell expansion in the CNS. (A) Microglia were isolated from brains of EAE mice with indicated genotypes at peak disease and cultured for 24 hours. The supernatant was analyzed by ELISA for cytokine and chemokine production. (B–D) Flow cytometry analysis of microglia (B), IL-17+CD4+, IFN-γ+CD4+ cell (C) and neutrophil (D) in brains of EAE mice with indicated genotypes at different time points after active immunization with MOG35–55. (E) IL-17+CD4+, IFN-γ+CD4+ cells, and neutrophil in brains of mice indicated in Figure 6F were analyzed by flow cytometry. Data are representative of 2 independent experiments (n = 4/group). Mean ± SEM. *P < 0.05, **P < 0.01 (unpaired 2-tailed Student’s t test). Discussion Microglia have been shown to sense TLR ligands and ATP released by dying CNS resident cells during neuroinflammation (19–23). This study reports a mechanism for noncanonical NLRP3 inflammasome activation in adult microglia for the effector stage of EAE. Our data demonstrated that TLR stimulation leads to IRAKM–caspase-8–ASC complex formation, resulting in the activation of caspase-8 and IL-1β release in microglia. Notably, active caspase-8 was also markedly elevated in microglia in the brain tissue from patients with MS. We found that IL-1β produced by microglia in the CNS facilitates the expansion of the microglia population in an autocrine manner, resulting in the production of inflammatory cytokines/chemokines. Consequently, microglia-specific ASC deficiency attenuated T cell expansion and neutrophil recruitment during EAE pathogenesis. This study helps to define a molecular mechanism responsible for the noncanonical inflammasome activation in a critical cell type of EAE pathogenesis, providing an unparalleled opportunity to develop therapeutic strategies for the treatment of MS. While the microglial compartment is quite stable during homeostasis, microglial proliferation and expansion are detected during CNS pathology (30, 31). Ajami et al. reported microglia proliferation in the spinal cords of EAE mice (30). However, due to the design of the study, they were unable to completely exclude the possible contribution of infiltrated monocytes to the BrdU+Iba-1+ cells in the spinal cords of EAE mice. More recently, Bruttger et al. showed microglia repopulated in response to depletion by diphtheria toxin was exclusively derived from CNS-resident cells, without contribution from bone marrow–derived cells (32). Furthermore, inhibition of IL-1R signaling abolished microglia repopulation whereas high IL-1R expression was detected on proliferating microglia in the microglia depletion model (32), suggesting the critical role of IL-1 signaling in driving microglia survival/proliferation. However, the mechanism by which microglial proliferation is regulated in pathophysiological context has not been established. Our current work extends the previous studies and represents the first demonstration that IL-1β affects microglial proliferation during the pathogenesis of EAE. The study by Ofengaim et al. reported the activation of necroptosis in multiple sclerosis (33). It is well-documented in the literature that caspase-8 has to be inhibited for necroptosis to take place. Compared with those of age-matched controls, tissue samples from cortical lesions of the patients with MS showed a reduction in the levels of active caspase-8. After induction of EAE, the spinal cord level of the active caspase-8 subunit was decreased. While total tissue lysates were used in assessing the levels of active caspase-8 by Western blot analysis, Ofengaim et al. showed that activation of necroptosis was mostly detected in oligodendrocytes in MS (CC1+MLKL+ cells) and EAE lesions (MBP+RIPK1+ cells) (33). Interestingly, we detected more active caspase-8 in microglia of MS lesions compared with those of normal tissue. Future studies are required to investigate the mechanism and consequence of differential caspase-8 activation in various CNS resident cells during the pathogenesis of EAE and MS. Notably, we previously reported that caspase-8 instead of caspase-1 is also critical for ASC-dependent IL-1β production in Th17 cells (3). T cell antigen receptor (TCR) activation induced pro–IL-1β expression and nuclear-to-cytosolic translocation of ASC; polarized Th17 cells expressed IL-1R (3), and they produced mature IL-1β in response to ATP via ASC- and NLRP3-dependent caspase-8 activation. While IRAKM is a signaling molecule under the TLR-mediated pathway, IRAKM expression was not detected in T cells (our unpublished data). Consistently, IRAKM deficiency did not have any impact on ASC-dependent IL-1β production in Th17 cells (our unpublished data). Thus, IRAKM-mediated ASC–caspase-8 activation and IL-1β production represents a signaling pathway unique to microglia (Figure 8). In support of our study, it was previously reported that caspase-8 activation in microglia leads to IL-1β production without causing cell death, and depletion or inhibition of caspase-8 attenuated microglia activation accompanied by reduced neurotoxicity (21, 34). Figure 8 Model of microglia-intrinsic IRAKM–caspase-8 noncanonical inflammasome pathway in EAE/MS pathogenesis. Upon MOG immunization, IL-1R expression was detected in IL-17+ CD4+ T cells, but not in IFN-γ+ CD4+ T cells (3). Th17 cells were converted to IL-17–IFN-γ+ cells, so-called ex-Th17 cells, which actually became the dominant IFN-γ–producing cell population in the spinal cord (1). The fact that both IL-17– and IFN-γ–producing cell populations were reduced in the brain tissue of microglia-specific ASC- and IRAKM-deficient mice implicated a possible role of microglia for the recruitment and activation of Th17 cells and/or ex-Th17 cells in the CNS. Additionally, the neutrophils were also substantially reduced at day 10 in brain tissue of microglia-specific ASC- and IRAKM-deficient mice. These results suggest the critical impact of the IRAKM-ASC axis in microglia for the recruitment/activation of T cells and neutrophils during EAE pathogenesis (Figure 8). Notably, in addition to NLRP3, a recent study also reported the involvement of NLRC4 inflammasome activation in microglia and astrocytes during neuroinflammation induced by lysophosphatidyl choline (35). Future studies are required to investigate the possible involvement of this IRAKM–caspase-8 axis in other inflammasomes such as NLRC4 contributing to neuroinflammation. Methods Mice. WT, CX3Cr1-CreER-EYFP, CX3Cr1GFP, NLRP3–/–, and Casp8fl/fl on C57BL/6 mice were purchased from The Jackson Laboratory. ASCfl/fl mice were generated in our lab as previously reported (3). IRAKMfl/fl mice were generated in our lab (Figure 4, A and B). Casp8–/– and Rip3–/– mice were a gift from William J. Kaiser, Department of Microbiology and Immunology, Emory Vaccine Center, Atlanta, Georgia, USA. Il1β–/– mice were a gift from Yoichiro Iwakura, Institute for Biomedical Sciences, Tokyo University of Science, Yamazaki, Noda, Japan. All the mice were on C57BL/6 background and maintained at Cleveland Clinic. Age- and sex-matched littermate control mice were used for each experiment. Human brain and spinal cord specimens. The human brain and spinal cord specimens from patients with primary progressive multiple sclerosis were provided by Richard Reynolds, Imperial College London, London, United Kingdom. Cell line. The human microglia cell line SV40 was provided by Theoharis C. Theoharides, Department of Integrative Physiology and Pathobiology, Tufts University School of Medicine, Boston, MA, USA. Plasmocin was used to rule out contamination during the cell culture. Reagents. Anti-CD19 (1:400, 1D3) was from eBioscience. Anti-CD45 (1:500, 30-F11), anti-F4/80 (1:200, BM8), anti-CD4 (1:200, L3T4), anti-Ly6C (1:300, HK1.4), anti–IFN-γ (1:200, XMG1.2), anti–IL-1R/CD121a (1:500, JAMA-147, 113505), lgG isotype control of IL-1R (1:500, HTK888), and ly6G (1:300, 1A8) were from BioLegend. Anti-ASC (22514-R, 04-147 Millipore), anti-NLRP3 (1:500, H-66), and actin (β-actin antibody [C4], sc-47778) were purchased from Santa Cruz Biotechnology and Millipore. Anti–IL-1β (1:1,000, AF-401-NA), ELISA kit (IL-1b, IL-23, IL-6, CCL20, CXCL1, CXCL5), anti–IL-17A (1:300, 559502), anti-CD8 (1:300, 553041), Brdu (550891), anti-Ly6G (1:300, 1A8), and CD8 (1:300, 553041) were purchased from BD Biosciences. Cleaved caspase-8 (9496 and 9429 from CST; ALX-804-448 from Enzo), P-IKBa (1:1,000, 2859), IKBa (1:1,000, 4812), cleaved caspase-1 (4199), IRAK1 (1:1,000, 4504), IRAK2 (1:1,000, 4367), and anti–human IL-1b (1:1,000, 2021) were from Cell Signaling Technology (CST). IRAKM (1:1,000, PAB7483) was from Abnova. Anti–caspase-1 (p20) mouse monoclonal antibody (AG-20B-0042) was from Adipogen. Anti-Ki67 (1:1,000, ab15580), anti-Brdu (ab1893, 1:200), and HMGB1 (ab81876 and ab181949) were from Abcam. Caspase-1/8 FLICA (97 and 99) were purchased from ImmunoChemistry. YVAD-fmk (ALX-260-154-R100) and IETD-fmk (550380) were purchased from Enzo and BD Biosciences. Luxol Fast Blue MBS Solution (26681) was purchased from Electron Microscopy Sciences. Protein A and G sepharose (17-046901, 17-061801) were from GE Healthcare. ATP (A1852) and TERM119 (1:200, HPA051870) were from Sigma-Aldrich. IBA1 (1:200, 019 19741) was from Wako. CPGB (ODN 2006 and ODN 1826) was from Invivogen. Escherichia coli LPS serotype O1101:B4 (201) was from List Biological Laboratories. IL-1Ra (Anakinra) was purchased from AMGEN. Quantitative real-time PCR. EAE mice were euthanized and perfused with 1× PBS to remove the blood. The spinal cords were harvested and total RNA was extracted using TRIzol reagent according to the manufacturer’s instructions (Invitrogen). The cDNA was synthesized with Oligo (dT) primers and superscript II reverse transcriptase (Invitrogen) by using 1 μg total RNA. The cDNA was resuspended in 100 μl H2O, and 2 μl cDNA samples were used for real-time PCR in a total volume of 20 μl SYBR Green reagent (Invitrogen) and specific primers. All gene expression results are expressed as arbitrary units relative to the expression of β-actin. The following real-time PCR (RT-PCR) primers were used for mouse genes: IL-17A, 5′-gtccagggagagcttcatctg-3′ and 5′-cttggcctcagtgtttggac-3′; IFN-γ, 5′-ctcatggctgtttctggctg-3′ and 5′-ccttttgccagttcctccag-3′; TNF-α, 5′-ccaccacgctcttctgtcta-3′ and 5′-gatctgagtgtgagggtctgg-3′; Ccl20, 5′-cacaagacagatggccgatg-3′ and 5′-cccttttcacccagttctgc-3′; Cxcl1, 5′-ccagagcttgaaggtgttgc-3′ and 5′-tgaaccaagggagcttcagg-3′; IL-1b, 5′-ccatcctctgtgactcatggg-3′ and 5′-tcagctcatatgggtccgac-3′; IL-6, 5′-GACAAAGCCAGAGTCCTTCAGAGAG-3′ and 5′-CTAGGTTTGCCGAGTAGATCTC-3′; β-actin, 5′-AGATGTGGATCAGCAAGCAG-3′ and 5′-GCGCAAGTTAGGTTTTGTCA-3′; IL-18, 5′-gccgacttcactgtacaacc-3′ and 5′-gtctggtctggggttcactg-3′; IL-23/p19, 5′-TCCCTACTAGGACTCAGCCAACTC-3′ and 5′-ACTCAGGCTGGGCACTG-3′; STEAP4, 5′-gaacactagatgcaagccaa-3′ and 5′-gagagatccttggtccagtgg-3′. Cell surface and intracellular staining. For cell surface staining, single-cell suspensions were incubated with Fc blocking buffer (Fc block diluted in FACS buffer at 1:50 ratio) on ice for 30 minutes at 4°C. After washing, the cells were stained with cell surface antibody for 30 minutes at 4°C. For intracellular staining, single-cell suspensions were cultured for 5 hours with phorbol 12-myristate 13-acetate (PMA, 20 ng/ml, MilliporeSigma) plus ionomycin (500 ng/ml, MilliporeSigma). GolgiStop (1:500, BD Biosciences) was added during the final 2 hours of incubation. After washing, cells were fixed by fixation/permeabilization solution (BD Cytofix/Cytoperm Kit, BD Biosciences) according to the manufacturer’s instructions. After incubation with Perm/Wash buffer (BD Biosciences) for 1 hour, antibodies for intracellular staining were added for 1 hour in 4°C and cells were analyzed on a FACSCalibur (BD Biosciences). Primary microglia culture. Adult 6-week-old mice were euthanized and perfused with PBS to remove the blood, followed by brain collection and digestion with the Adult Brain Dissociation Kit (130-107-677, MACS) following the manufacturer’s instructions. The digested brain was homogenized and primary microglia were isolated by a discontinuous Percoll gradient. Isolated microglia were seeded on the plate and cultured with DMEM/F12 supplemented with 100 U/ml penicillin and streptomycin, 10% FBS, and 5 ng/ml mouse GM-CSF (415-ML/CF, R&D Systems) After 14 days, microglia were harvested by collecting the culture media and centrifuging for 5 minutes at 412 g. The purity of the microglia were confirmed by flow cytometry analysis after being stained with antibodies against CD45 and CD11b (>99% purity). Generation of chimeric mice. The chimera mice were generated as previously reported (36). Briefly, 6-week-old recipient mice were subjected to 9 Gy of total body lethal irradiation and allowed for 4 hours of recovery before reconstitution with bone marrow cells. Six-week-old donor mice were sacrificed by CO2 asphyxiation and bone marrow cells were flushed from femurs and tibias. The recipient mice were transferred with 1.5 × 107 WT bone marrow cells in 150 μl 1× PBS intravenously through retro-orbital venous sinus. Mice were fed with gentamicin-containing (10 μg/ml) drinking water for an additional 4 weeks before subjected to MOG immunization. Tamoxifen injection. Tamoxifen-induced Cre expression was performed as previously reported (2) (Supplemental Figure 1A). One week before bone marrow transplantation, tamoxifen was administered i.p. at 5 mg/mouse for 4 consecutive weeks (once per week). EAE was induced after tamoxifen treatment (4 weeks after bone marrow transplantation). Induction and assessment of EAE. Active EAE and adoptive transfer (passive) EAE were induced and assessed as previously described by our group (2, 3, 37). Briefly, for the adoptive transfer, donor mice were immunized with MOG35–55 subcutaneously and spleen/draining lymph nodes were harvested 10 days after immunization. The cells were cultured for 5 days with MOG35–55 at a concentration of 25 μg/ml under either Th1 cell–polarizing conditions (20 ng/ml IL-12, R&D Systems; 2 μg/ml anti–IL-23p19, eBiosciences) or Th17 cell–polarizing conditions (20 ng/ml IL-23, R&D Systems). We injected recipient mice with 3.0 × 107 polarized MOG35–55-specific Th1 or Th17 cells 4 hours after exposing them to 5 Gy sublethal irradiation. Clinical scores were assessed in a double-blinded manner every other day. BrdU incorporation assay. The animals were administrated BrdU (50 mg/kg, once per day) for 3 consecutive days by i.p. injection. Brains and spinal cords were fixed with 4% PFA overnight followed by dehydration with 15% and 30% sucrose for an additional 2 days. Tissues were quickly frozen in OCT and cut as 20 μm/slice. Tissue sections were treated with 2 M hydrochloric acid for 30 minutes at room temperature followed by treatment with 0.01 M borate buffer (pH 8.2) for 30 minutes at room temperature. Slices were then subjected to immunohistochemistry analysis. Intracerebroventricular injection of microglia. Microglia injection was performed as previously reported (22). Briefly, microglia isolated from WT or IL-1b–/– adult naive mice were injected (5.5 × 105 in 4 μl PBS per mouse) into the cerebrospinal fluid (mediolateral, −0.8 mm; anteroposterior, −0.4 mm; dorsoventral, −2.5 mm) of Csf1rfl/flCx3cr1cre mice after Cre induction by tamoxifen injection followed by immunization with MOG35–55 peptide. Histological analysis. EAE mice were perfused with PBS and PFA before harvesting spinal cords. The spinal cord was quick-frozen in OCT and cut to approximately 10- to 20-μm thickness. Sections were stained with H&E and with Luxol Fast Blue (Electron Microscope Sciences) for evaluation of inflammation and demyelination, respectively. Enzyme-linked immunosorbent assay (ELISA). The media from cultured microglia was harvested and centrifuged. The supernatant was subjected to ELISA (all kits from R&D Systems) according to the manufacturer’s instructions. The concentration of cytokines in the media was adjusted based on the standard curve. Immunofluorescence staining. Immunofluorescence staining was performed as previously reported (2). Briefly, sections were washed and blocked, followed by incubating with first antibodies overnight at 4°C. After washing and incubating with a second antibody for 1 hour at room temperature, the slices were either stained with DAPI or mounted for the microscope. All images were captured with a confocal microscope (Artemis or Keyence BZ-X710). Immunoblotting and immunoprecipitation. Cells were lysed in lysis buffer (0.5% Triton X-100, 20 mM HEPES, pH 7.4, 150 mM NaCl, 12.5 mM β-glycerophosphate, 1.5 mM MgCl2, 10 mM NaF, 2 mM dithiothreitol, 1 mM sodium orthovanadate, 2 mM EGTA, 20 mM aprotinin, 1 mM phenylmethylsulfonyl fluoride). Protein lysate (20 μg) was run per lane on an approximately 8%–12% SDS-PAGE gel and subjected to immunoblotting with different antibodies. Coimmunoprecipitation experiments were performed as previously described (2, 38). To avoid the heavy chain band, we used a different source (such as immunoprecipitation with rabbit-derived antibody and probe with mouse-derived antibody) to target the same protein. Statistics. The P values of clinical scores were determined by 2-way multiple-range ANOVA for multiple comparisons. Other P values were determined by unpaired 2-tailed Student’s t test. All results are mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. All data were from at least 2 independent experiments. Study approval. All animal experiments were conducted and approved by the Cleveland Clinic Institutional Animal Care and Use Committee. The human tissue experiments were approved by the ethics committees at Imperial College London and the Cleveland Clinic. Written informed consent was received from participants prior to inclusion in the study. Author contributions XL and CJZ conceived and coordinated the study, and wrote the manuscript. CJZ conducted the majority of experiments and analyzed the data. MJ, WL, QZ, and CW helped with experiments. XC, JX, BH, and AF provided technical support. HZ, JG, JC, and AD performed the mass spectometry and analyzed the data. ZK, WJK, MAM, YI, GD, DWA, FDS, and RMR provided critical discussion. Supplemental material View Supplemental data Acknowledgments This investigation was supported by grants from the National Multiple Sclerosis Society (RG-1707-28180 and RG5130A2/1 to XL) and National Institutes of Health (5R01NS071996-05, 1RO1AA023722, and MSTP-T32GM007250 to XL). The study was partially supported by a grant from the National Science Foundation of China (81701235 to CZ). Footnotes Conflict of interest: The authors have declared that no conflict of interest exists. License: Copyright 2018, American Society for Clinical Investigation. Reference information: J Clin Invest. https://doi.org/10.1172/JCI121901. References Hirota K, et al. Fate mapping of IL-17-producing T cells in inflammatory responses. Nat Immunol. 2011;12(3):255–263. View this article via: PubMed CrossRef Google Scholar Wang C, et al. IL-17 induced NOTCH1 activation in oligodendrocyte progenitor cells enhances proliferation and inflammatory gene expression. Nat Commun. 2017;8:15508. View this article via: PubMed Google Scholar Martin BN, et al. T cell-intrinsic ASC critically promotes T(H)17-mediated experimental autoimmune encephalomyelitis. Nat Immunol. 2016;17(5):583–592. View this article via: PubMed CrossRef Google Scholar Fletcher JM, Lalor SJ, Sweeney CM, Tubridy N, Mills KH. T cells in multiple sclerosis and experimental autoimmune encephalomyelitis. Clin Exp Immunol. 2010;162(1):1–11. View this article via: PubMed CrossRef Google Scholar Rothhammer V, et al. Th17 lymphocytes traffic to the central nervous system independently of α4 integrin expression during EAE. J Exp Med. 2011;208(12):2465–2476. View this article via: PubMed CrossRef Google Scholar Iwakura Y, Ishigame H. The IL-23/IL-17 axis in inflammation. J Clin Invest. 2006;116(5):1218–1222. View this article via: JCI PubMed CrossRef Google Scholar Lévesque SA, et al. Myeloid cell transmigration across the CNS vasculature triggers IL-1β-driven neuroinflammation during autoimmune encephalomyelitis in mice. J Exp Med. 2016;213(6):929–949. View this article via: PubMed CrossRef Google Scholar Lin CC, Edelson BT. New insights into the role of IL-1β in experimental autoimmune encephalomyelitis and multiple sclerosis. J Immunol. 2017;198(12):4553–4560. View this article via: PubMed CrossRef Google Scholar Ronchi F, et al. Experimental priming of encephalitogenic Th1/Th17 cells requires pertussis toxin-driven IL-1β production by myeloid cells. Nat Commun. 2016;7:11541. View this article via: PubMed Google Scholar Burm SM, et al. Expression of IL-1β in rhesus EAE and MS lesions is mainly induced in the CNS itself. J Neuroinflammation. 2016;13(1):138. View this article via: PubMed CrossRef Google Scholar Aglietti RA, et al. GsdmD p30 elicited by caspase-11 during pyroptosis forms pores in membranes. Proc Natl Acad Sci USA. 2016;113(28):7858–7863. View this article via: PubMed CrossRef Google Scholar He WT, et al. Gasdermin D is an executor of pyroptosis and required for interleukin-1β secretion. Cell Res. 2015;25(12):1285–1298. View this article via: PubMed CrossRef Google Scholar Shi J, et al. Cleavage of GSDMD by inflammatory caspases determines pyroptotic cell death. Nature. 2015;526(7575):660–665. View this article via: PubMed CrossRef Google Scholar Ding J, et al. Pore-forming activity and structural autoinhibition of the gasdermin family. Nature. 2016;535(7610):111–116. View this article via: PubMed CrossRef Google Scholar Boucher D, et al. Caspase-1 self-cleavage is an intrinsic mechanism to terminate inflammasome activity. J Exp Med. 2018;215(3):827–840. View this article via: PubMed CrossRef Google Scholar Inoue M, Shinohara ML. NLRP3 inflammasome and MS/EAE. Autoimmune Dis. 2013;2013:859145. View this article via: PubMed Google Scholar Inoue M, Williams KL, Gunn MD, Shinohara ML. NLRP3 inflammasome induces chemotactic immune cell migration to the CNS in experimental autoimmune encephalomyelitis. Proc Natl Acad Sci U S A. 2012;109(26):10480–10485. View this article via: PubMed CrossRef Google Scholar Shaw PJ, Lukens JR, Burns S, Chi H, McGargill MA, Kanneganti TD. Cutting edge: critical role for PYCARD/ASC in the development of experimental autoimmune encephalomyelitis. J Immunol. 2010;184(9):4610–4614. View this article via: PubMed CrossRef Google Scholar Chang M, Jin W, Sun SC. Peli1 facilitates TRIF-dependent Toll-like receptor signaling and proinflammatory cytokine production. Nat Immunol. 2009;10(10):1089–1095. View this article via: PubMed CrossRef Google Scholar Biber K, Möller T, Boddeke E, Prinz M. Central nervous system myeloid cells as drug targets: current status and translational challenges. Nat Rev Drug Discov. 2016;15(2):110–124. View this article via: PubMed CrossRef Google Scholar Burguillos MA, et al. Caspase signalling controls microglia activation and neurotoxicity. Nature. 2011;472(7343):319–324. View this article via: PubMed CrossRef Google Scholar Xiao Y, et al. Peli1 promotes microglia-mediated CNS inflammation by regulating Traf3 degradation. Nat Med. 2013;19(5):595–602. View this article via: PubMed CrossRef Google Scholar Monif M, Reid CA, Powell KL, Smart ML, Williams DA. The P2X7 receptor drives microglial activation and proliferation: a trophic role for P2X7R pore. J Neurosci. 2009;29(12):3781–3791. View this article via: PubMed CrossRef Google Scholar Chastain EM, Duncan DS, Rodgers JM, Miller SD. The role of antigen presenting cells in multiple sclerosis. Biochim Biophys Acta. 2011;1812(2):265–274. View this article via: PubMed Google Scholar Kivisäkk P, et al. Localizing central nervous system immune surveillance: meningeal antigen-presenting cells activate T cells during experimental autoimmune encephalomyelitis. Ann Neurol. 2009;65(4):457–469. View this article via: PubMed CrossRef Google Scholar Goverman J. Autoimmune T cell responses in the central nervous system. Nat Rev Immunol. 2009;9(6):393–407. View this article via: PubMed CrossRef Google Scholar Antonopoulos C, El Sanadi C, Kaiser WJ, Mocarski ES, Dubyak GR. Proapoptotic chemotherapeutic drugs induce noncanonical processing and release of IL-1β via caspase-8 in dendritic cells. J Immunol. 2013;191(9):4789–4803. View this article via: PubMed CrossRef Google Scholar Gringhuis SI, et al. Dectin-1 is an extracellular pathogen sensor for the induction and processing of IL-1β via a noncanonical caspase-8 inflammasome. Nat Immunol. 2012;13(3):246–254. View this article via: PubMed CrossRef Google Scholar Bossaller L, et al. Cutting edge: FAS (CD95) mediates noncanonical IL-1β and IL-18 maturation via caspase-8 in an RIP3-independent manner. J Immunol. 2012;189(12):5508–5512. View this article via: PubMed CrossRef Google Scholar Ajami B, Bennett JL, Krieger C, McNagny KM, Rossi FM. Infiltrating monocytes trigger EAE progression, but do not contribute to the resident microglia pool. Nat Neurosci. 2011;14(9):1142–1149. View this article via: PubMed CrossRef Google Scholar Ding Z, et al. Antiviral drug ganciclovir is a potent inhibitor of microglial proliferation and neuroinflammation. J Exp Med. 2014;211(2):189–198. View this article via: PubMed CrossRef Google Scholar Bruttger J, et al. Genetic cell ablation reveals clusters of local self-renewing microglia in the mammalian central nervous system. Immunity. 2015;43(1):92–106. View this article via: PubMed CrossRef Google Scholar Ofengeim D, et al. Activation of necroptosis in multiple sclerosis. Cell Reports. 2015;10(11):1836–1849. View this article via: PubMed CrossRef Google Scholar Viceconte N, Burguillos MA, Herrera AJ, De Pablos RM, Joseph B, Venero JL. Neuromelanin activates proinflammatory microglia through a caspase-8-dependent mechanism. J Neuroinflammation. 2015;12:5. View this article via: PubMed Google Scholar Freeman L, Guo H, David CN, Brickey WJ, Jha S, Ting JP. NLR members NLRC4 and NLRP3 mediate sterile inflammasome activation in microglia and astrocytes. J Exp Med. 2017;214(5):1351–1370. View this article via: PubMed CrossRef Google Scholar Cardona AE, et al. Scavenging roles of chemokine receptors: chemokine receptor deficiency is associated with increased levels of ligand in circulation and tissues. Blood. 2008;112(2):256–263. View this article via: PubMed CrossRef Google Scholar Qian Y, et al. The adaptor Act1 is required for interleukin 17-dependent signaling associated with autoimmune and inflammatory disease. Nat Immunol. 2007;8(3):247–256. View this article via: PubMed CrossRef Google Scholar Wang C, et al. The psoriasis-associated D10N variant of the adaptor Act1 with impaired regulation by the molecular chaperone hsp90. Nat Immunol. 2013;14(1):72–81. View this article via: PubMed CrossRef Google Scholar Version history Version 1 (October 29, 2018): Electronic publication

The glymphatic (glial-lymphatic) pathway is a fluid-clearance pathway identified in
the rodent brain in 2012. This pathway subserves the flow of CSF into the brain along
arterial perivascular spaces and subsequently into the brain interstitium, facilitated
by aquaporin 4 (AQP4) water channels.