Seizures and epilepsy are common in autoimmune encephalitis (1) and can even represent the main complaints, which led to the clinical entity of “immune epilepsy” that has been added to the International League Against Epilepsy (ILAE) classification of the epilepsies in 2017 (2) as an etiologic subgroup categorized as “immune disorder in which seizures are a core symptom.” In 2020, the ILAE Autoimmunity and Inflammation Taskforce proposed conceptual definitions for two main diagnostic entities in this field: (a) acute symptomatic seizures secondary to autoimmune encephalitis, and (b) autoimmune-associated epilepsy, the latter of which suggests an enduring predisposition to seizures (3) and corresponds to “immune epilepsy” in the 2017 ILAE classification.

Research on the clinical phenotypes and the underlying disease mechanisms has become a major focus over the past decade (4–7). In many patients, well-established autoantibodies are detected in clinical routine, such as autoantibodies against NMDA receptors (NMDAR) or leucine-rich glioma inactivated protein 1 (LGl1) (8, 9). In antibody-positive patients, new immunotherapeutic approaches became available (5), thus going beyond symptomatic seizure control with anti-seizure medication.

However, the exact proportion of immune epilepsies and the spectrum of underlying autoantibodies are still changing with novel findings. A recent comprehensive meta-analysis (10) revealed estimations between 0 and 24% depending on design and cohort selection. Differences partially relate to availability of bio-samples (often restricted to serum), the limited number of examined autoantibodies, the type and epileptogenic potential of underlying autoantibodies (e.g., targeting neuronal surfaces), and lack of follow-up and inconsistencies when distinguishing seizures from epilepsies (10). Likewise, the role of CSF testing in etiologically unexplained seizures and epilepsies is not fully clear, even though some studies suggest routine CSF diagnostics to not overlook autoimmune etiologies (4). This also considers the presence of some autoantibodies only in CSF and the correlation of clinical courses with CSF titers, such as with anti-NMDAR autoantibodies. However, only few studies screened CSF specimens for autoantibodies in patients with new-onset seizures of unknown etiology (10).

The present study therefore aimed at retrospectively analyzing the presence of autoantibodies in CSF and serum of well-phenotyped patients with seizures or epilepsy of unknown etiology. A particular focus was the potential benefit of CSF analysis, the long follow-up of 6–8 (mean 7.8) years to increase diagnostic accuracy, and the frequency of less well-established autoantibodies using murine brain immunofluorescence imaging.

In this retrospective analysis, (i) we showed that more than one in four patients with new-onset seizures of unknown etiology had autoantibodies in CSF targeting established and novel brain antigens, which resulted in re-classification of the etiology in two patients [2 of 39 (5.1%)]. (ii) Moreover, autoantibodies were much more common in CSF compared to serum, indicating that CSF analysis may be considered more frequently in clinical routine diagnostics. (iii) Also, tumors were significantly more common in the autoantibody-positive cohort.

Of the five patients with established CSF autoantibodies against NMDAR (n = 2), GAD65 (n = 1), and GFAP (n = 2), only one (A23) received the diagnosis of autoimmune encephalitis (NMDAR encephalitis) and appropriate immunotherapy. One further patient (A24) with CSF anti-NMDAR autoantibodies was lost to follow-up, thus precluding reassessment and final diagnosis. Two patients (A36, A32) remained categorized as etiologically unknown epilepsy based on the clinical course (anti-GFAP autoantibodies, but no further seizures; A36) and low-level antibodies (anti-GAD65; A32; Table 2). However, based on the presence of anti-GFAP autoantibodies and the otherwise unexplained epilepsy cause in the fifth patient (B1), we argue that this patient should also be re-classified as autoimmune-associated epilepsy. While in patient A23 with NMDAR encephalitis, the underlying ovarian teratoma represents a known strong association (13), the tumors in patients A8 and A19 are not clearly linked to the neurological syndrome as the underlying antigen is still unclear, thus they were not classified as paraneoplastic (14).

In general, NMDAR encephalitis is diagnosed by the presence of well-established pathogenic anti-NMDAR autoantibodies in CSF together with the compatible clinical picture. Seizures occur in 75% of cases and can be the predominant symptom (9, 15, 16). Also, anti-NMDAR autoantibodies were frequent (4.8%) in some cohorts of epilepsy patients (10). We therefore assume that both cases (A23, A24) with anti-NMDAR autoantibodies in our cohort had autoimmune-associated epilepsy based on the underlying anti-NMDA receptor encephalitis (3).

The causal relationship of anti-GAD65 or GFAP autoantibodies with seizures is much less clear. Although high-level anti-GAD65 autoantibodies (in particular when found in CSF) are strongly associated with a form of autoimmune encephalitis (17) and were detected in 3.7% of epilepsy patients including chronic treatment-refractory cases (10, 18, 19), low-level anti-GAD65 antibodies are also present in the general population (20). Also, their direct pathogenicity has not been demonstrated, in contrast to anti-NMDAR autoantibodies, which may relate to the intracellular localization of GAD65. Therefore, patient A32 of this study was not categorized as autoimmune-associated epilepsy despite of low-titer GAD65 antibodies in CSF.

Likewise, autoantibodies against GFAP target an intracellular epitope, but are strongly linked to a relatively new form of autoimmune encephalitis. First established in 2016 (21), diagnosis of anti-GFAP astrocytopathy was not possible in 2012–2013, when both anti-GFAP antibody-positive patients of our cohort entered the study. It became clear with further studies that detection of anti-GFAP autoantibodies in CSF is required for diagnosis, that seizures occur in 11–37% of these patients and that immunotherapy often has beneficial effects (22–28). Therefore, we re-classified one anti-GFAP antibody-positive patient (from our cohort with seizures refractory to anti-seizure medication as having autoimmune-associated epilepsy) (Figure 1). Even though detailed experimental studies clarifying the direct pathogenicity of human anti-GFAP autoantibodies are still needed (29), the antibodies may already serve as valuable biomarkers for early detection of immunotreatment-responsive epilepsy patients.

Our findings pointed to the usefulness of CSF analysis for increased detection of potentially pathogenic autoantibodies in patients with unknown seizures and unknown epilepsy. A recent meta-analysis (10) revealed that most published studies screen for autoantibodies in serum only for epilepsy patients, with only 3.4% of all patients undergoing an additional CSF testing. Different to routine procedures in most hospitals, we here systematically performed CSF analysis in all patients and tested for autoantibodies in CSF and serum independent of the presence of inflammatory signs (i.e., also in the absence of pleocytosis, high CSF protein, or oligoclonal bands). Using indirect immunofluorescence on both, transfected cells and unfixed brain sections, may explain why our cohort showed a higher frequency of autoantibodies compared to previous epilepsy studies that also included CSF specimens (30–35), in which autoantibody frequency was up to 16.8% (33). For the yet uncharacterized autoantibodies, future studies will have to identify the underlying antigens and structural epitopes, the pathogenicity of the antibodies in functional assays and the immunological mechanisms leading to humoral autoimmunity (1, 29).

Our study has some limitations. First, our sample size is relatively small thus complicating solid conclusions about the frequency of autoimmune-associated epilepsy. Future prospective studies with higher statistical power are therefore needed, however, inclusion of CSF analysis is beyond clinical routine in most centers. Second, several autoantibodies bind to as yet undetermined autoantigens, identification of which will require further work including antigen identification by protein arrays or immunoprecipitation/mass spectrometry. Furthermore, incomplete clinical information in the follow-up of some cases hinders clear correlation with antibody test results.

Taken together, our data support the notion that autoimmune-associated epilepsy may be more common than previously assumed, in particular with screening for a broad panel of established and novel autoantibodies. Given the retrospective nature of the study, autoantibody detection did not result in immediate treatment decisions. Together with the number of subjects lost to follow-up, we cannot infer solid conclusions on the relationship between autoantibody levels, clinical courses, and treatment options. CSF analysis seems to be of relevant additional help in the assessment of patients with new-onset seizures or epilepsy of otherwise unknown etiology, in particular as early autoantibody detection is a prerequisite for early treatment in confirmed cases of autoimmune-associated epilepsy. Re-classification of one patient with anti-GFAP autoantibodies as having autoimmune-associated epilepsy is a clear example of the fascinating ongoing progress in the field, given that these autoantibodies were first described in the literature not until the patient had already entered our study. Although further studies are needed to understand the autoantibody repertoire in epilepsy patients and the proportion of patients that might benefit already from available immunotherapies, it is tempting to speculate that even low-level autoantibodies, which are currently considered unspecific, may become more relevant once antibody-selective immunotherapies reach the clinical level.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants and animal study were reviewed and approved by Charité University Hospital Institutional Review Board. The patients/participants provided their written informed consent to participate in this study.

PS, MH, and HP contributed to the conception and design of the study. PS, AL, and HP contributed to the acquisition, analysis of data, drafting the text, and preparing the figures. All authors contributed to the article and approved the submitted version.

This work was supported by grants from the German Research Foundation (DFG; grants FOR3004, PR1274/3-1, PR1274/5-1, and PR1274/9-1), by the Helmholtz Association (HIL-A03 BaoBab), and by the German Federal Ministry of Education and Research (Connect-Generate 01GM1908D) to HP. Dr. Lütt is participant in the BIH Charité Junior Digital Clinician Scientist Program funded by the Charité - Universitätsmedizin Berlin, and the Berlin Institute of Health at Charité (BIH).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.