The study comprised sera from 18 controls and 36 patients, along with cerebrospinal fluid (CSF) from five controls and 12 patients. The patients exhibited diverse neuropsychiatric disease phenotypes, including neurodegenerative disorders and autoimmune encephalitis. Two patients (ID 20 and 21), included in this work, have been previously published (20); their CSF samples were not available. Data of all patients were collected at the Charité, Berlin, and all participants gave an informed written consent. All patient sera underwent cell-based screening at Euroimmun (EUROIMMUN Medizinische Labordiagnostika AG). Immunoglobulin G (IgG) detection was conducted using a biochip array with acetone-fixed recombinant human embryonic kidney 293 (HEK293) cells separately expressing the following autoantigens: Aquaporin-4 (AQP4), Rho GTPase activating protein 26 (ARHGAP26), sodium-potassium ATPase catalytic subunit alpha-3 (ATP1A3), Contactin-1 (CNTN1), collapsin response mediator protein 5 (CRMP5 or CV2), Carbonic Anhydrase-Related Protein VIII (CARP VIII), contactin-associated protein-like 2 (CASPR2), Tr/Delta/Notch-like epidermal growth factor-related receptor (DNER/Tr), Dipeptidyl-Peptidase-like Protein-6 (DPPX), Flotilin 1/2, γ-aminobutyric acid type B (GABA_B) receptor, glutamic acid decarboxylase 65 (GAD65), γ-aminobutyric acid type A (GABA_A)receptor, glycine receptor, glutamate receptors [types α-Amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptor, Glutamate Receptor ionotropic delta-2 (GLURD2), Metabotropic glutamate receptor 1 (mGluR1), Metabotropic glutamate receptor 5 (MGluR5), N-methyl-D-aspartate (NMDA)], Homer 3, Hu, Inositol 1,4,5-Trisphosphate Receptor Type 1 (ITPR1), IgLON family member 5 (IgLON5), K_v1.2, leucine-rich glioma inactivated 1 (LGI1), myelin oligodendrocyte glycoprotein (MOG), Neurochondrin (NCDN), Neurofascin 155 (NF155), Neurofascin 186 (NF186), Paraneoplastic antigen Ma2 (PNMA2), recoverin, Ri paraneoplastic antigen, Seizure Related 6 Homolog Like 2 (Sez6L2), Yo paraneoplastic antigen, Zic Family Member 4 (ZIC4). The initial laboratory pre-screening encompassed a cohort of 96 healthy controls, from which four individuals exhibited positive K_v1.2 autoantibodies (4.2% positive healthy donors). Subsequently, these samples were subjected to detailed peptide microarray–based evaluation.

The complete K_v sequences (UniProtKB: Q09470, P16389, P22001, and P22459) were displayed in microarray format as 15-mer overlapping peptide libraries. Peptide arrays were synthesized using µSPOT (25), a SPOT-based (26) synthesis approach. In brief, custom-prepared discs containing 9-fluorenylmethyloxycarbonyl(Fmoc)-β-alanine linkers (average loading: 130 nmol/disc, 4 mm in diameter) were loaded in a MultiPep rSi robot (CEM GmbH, Kamp-Lintford, Germany) together natural amino acid (AA) building blocks and phospho-building blocks from IRIS (IRIS Biotech GmbH, Marktredwitz, Germany) and the freshly prepared reagents. Synthesis was carried out by deprotecting the Fmoc-group using 20% piperidine in dimethylformamide (DMF). Peptide chains were elongated using a coupling solution consisting of aAs (0.5 M) with Oxyma (1 M) and diisopropylmethanediimine (1 M) in DMF (1:1:1). Coupling steps were carried out three times (30 min), followed by capping (4% acetic anhydride in DMF). Discs were transferred into 96–deep-well plates for the workup. Side chains were deprotected using 90% trifluoracetic acid (TFA), 2% dichloromethane (DCM), 5% H₂O, and 3% triisopropylsilane (TIPS) (150 μL per well) for 1 h at room temperature (RT). Afterwards, the deprotection solution was removed, and the discs were solubilized overnight (ON) at RT while shaking, using a solvation mixture containing 88.5% TFA, 4% trifluoromethanesulfonic acid, 5% H₂O, and 2.5% TIPS (250 μL per well). The resulting peptide-cellulose conjugates were precipitated with ice-cold ether (700 μL per well) and spin down at 2,000×g for 10 min at 4°C, followed by two additional washes of the formed pellet with ice-cold ether. The resulting pellets were dissolved in DMSO (250 μL per well).

Liquid chromatography–mass spectrometry (LC-MS) (27) was carried out using peptide quality controls that were cleaved from the solid support. To ensure cleavage, a Rink amide linker (Iris) suitable for Solid Phase Peptide synthesis (SPPS) on cellulose support was introduced during the first coupling cycle. In an acidic environment, the quality controls were cleaved off the solid support. To isolate the quality controls, 150 µL of the supernatant was transferred to 1.5-mL reaction tubes, followed by the addition of 700 µL of diethyl ether. The samples were then vortexed, and the peptides were allowed to precipitate by incubation at −20°C overnight. After centrifugation at 13,300×g and 4°C for 10 min, the supernatant was discarded, and 500 µL of diethyl ether was added. The mixture was vortexed and centrifuged for 10 min, and the supernatant was decanted. This process was repeated twice, and the peptides were left to dry for 60 min. Finally, the Rink amides were dissolved in 50 µL of 50% acetonitrile and 0.1% formic acid (v/v) and vortexed briefly before centrifugation at 13,300×g and RT. For analysis, the quality controls were diluted 1:3 and analyzed via LC-MS (Agilent technologies).

Peptide-cellulose conjugate (PCC) solutions were mixed 2:1 with saline-sodium citrate buffer [150 mM NaCl and 15 mM trisodium citrate (pH 7.0)] and transferred to a 384-well plate. For transfer of the PCC solutions to white-coated CelluSpot blank slides (76 mm × 26 mm, Intavis AG Peptide Services GmbH and Co. KG), a SlideSpotter (CEM GmbH) was used. After completion of the printing procedure, slides were left to dry overnight.

The microarray slides were blocked for 60 min in 5% (w/v) skimmed milk powder (Carl Roth) 0.05% Tween 20 phosphate-buffered saline [PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO₄ (pH7.4)]. After blocking, the slides were incubated for 30 min with either positive and negative serum or cerebrospinal fluid (dilution ranging from 1:300 to 1:3,000) or K_v1.2 monoclonal (NeuroMab clone: K14/16 at 0.5 µg/mL) in the blocking buffer and then washed three times with PBS 0.05% and Tween 20 for 1 min. IgG antibodies were detected using goat anti-human IgG-Horseradish peroxidase (HRP) (Thermo Fisher, Cat. No. 31410; 1:2,500) or goat anti-mouse IgG-HRP (Thermo Fisher, Cat. No. 31430; 1:5000). The chemiluminescence readout was detected with an Azure imaging system c400 (lowest sensitivity, exposure time of 60 s for serum and of 30 s for monoclonal) using SuperSignal West Femto maximum sensitive substrate (Thermo Scientific, GmbH, Schwerte, Germany). Microarray binding intensities were quantified using the MicroArray Rastering Tool – MARTin (28). Neutralization assays were performed by pre-incubating the sera in neutralizing solution, containing cleavable peptides or buffer for 30 min. The binding assay followed in the same manner.

K_v1.1, K_v1.2, and K_v1.6 were expressed in HEK293 cells, following a previously described protocol (Miske et al., 2023). To summarize, genomic DNA was extracted from HEK293 cells and utilized as a template for K_v1.2 coding sequence amplification via polymerase chain reaction (PCR). Respective DNA oligonucleotides were employed to introduce the required enzyme restriction sites ( Table 1 ). Indicated enzymes were used to digest the resulting PCR fragments and subsequently ligated with NcoI/XhoI-linearized pTriEx-1 (Merck). Prior to transfection, HEK293 cells were plated on sterile poly-L-lysine–treated coverslips. Transient expression of K_v1-encoded proteins was accomplished through Polyethylenimine (PEI)-mediated transfection (PEI 25K™) following the manufacturer’s instructions (Polysciences, Europe). After 48 h of transfection, cells were fixed, permeabilized with acetone, and rinsed with PBS before conducting the immunofluorescence described below.

All animal procedures were approved by the Landesamt für Gesundheit und Soziales (LaGeSo) Berlin, Germany (approval numbers T-CH 0009/22), and conducted in compliance with the German and international guidelines for care and humane use of animals. Unfixed brain processing and sciatic nerve teased fiber preparations were performed as previously described (29). In brief, male C57BL/6 mice were used at an age of 10–12 weeks. Unfixed brains were dissected and frozen in 2-methylbutan. Cryostat-cut 20-µm sections were mounted on glass slides and used for tissue-based immunofluorescence. Murine sciatic nerves were dissected from hind limbs, fixed in 4% paraformaldehyde, and washed in PBS. The epineurium was removed prior to teasing. Teased fibers were air-dried and stored at −20°C until further usage.

Slides and cells were post-fixed with acetone and washed with PBS. Serum, CSF, and/or commercial K_v1.1 (Abcam ab65790), K_v1.2 (K14/16 Neuromab, supplier Antibodies Incorporated 75-008), and K_v1.6 antibody (Antibodies Incorporated 75-012) incubation was done overnight at +4°C in a dilution of 1:300 (serum), 1:2 (CSF), and 1:200 (commercial) in PBS-T (containing 0.2% Tween 20 and 0.1% Triton X). Secondary antibodies against human IgG (Alexa488-labeled Dianova, 109-545-003; and Cy3-labeled Dianova, 109-165-003), mouse IgG (Alexa549-labeled goat anti-mouse, Jackson Research, 115-585-03; and Alexa-488-labeled, Dianova, 115-546-003) and rabbit IgG (Alexa488-labeled, Dianova, 111-546-003) were added in a dilution of 1:1,000 in PBS-T for 1 h at RT. Cell nuclei were stained using 4',6-diamidino-2-phenylindole (DAPI). Coverslips and glass slides were mounted in Fluoroshield and visualized using a widefield fluorescence microscope (LeicaSPE).

Neutralization assays were adapted from established protocols (Miske et al., 2023). In brief, peptides were solubilized in PBS at a final concentration of 1 mg/mL. Neutralization assays were performed on transfected HEK293 cells and on mouse tissues. Sera were diluted in PBS-T to the abovementioned concentrations. The peptide antigen was added in a dilution of 1:10. Sera and peptides were thereby pre-incubated for 1 h at RT and subsequently added to slides and cells incubating for 1 h at RT. After washing with PBS-T, cells and slides were stained with secondary Alexa488-labeled antibodies for 30 min at RT. After washing with PBS-T, cell nuclei were stained with DAPI. Co-staining with the commercial K_v1.2 antibody (K14/16 NeuroMab) and secondary Alexa594-labeled antibody was accomplished through sequential staining rounds, thereby avoiding potential cross-reactivity of secondary antibodies to primary mouse or human antibodies.

Data are available upon reasonable request to qualified investigators for the purposes of replicating procedures and results.

Screening, mapping, and validation of K_v1.2 epitopes was based on 36 samples from patients (74.3% men, median disease onset age of 65, median high sera titer of 1:1,000, with four out of the 96 positive healthy controls positive in this readout) positive for K_v1.2 but negative for 33 common neurological autoantigens in cell-based assays. Patients showed heterogeneous neuropsychiatric phenotypes ranging from dementia, to epilepsy, autoimmune encephalitis, ischemic strokes, and peripheral neuropathies. Patient ID 19 had anti-AP3B2 IgG (titer 1:32) and patient ID 30 anti-NMDAR IgM (1:1,000) autoantibodies in serum; in the rest of the cohort, no other co-existing antibodies were detected.

To identify anti-K_v1.2 epitopes, the qualified patient samples were screened in peptide microarray format. Here, the entire primary sequence of K_v1.2 was displayed in the form of 20-mer peptides with 17-residue overlap ( Figure 1A ). First, the array approach was validated using the commercial anti-mouse K_v1.2 (NeuroMab clone K14/16) antibody. The microarray defined the sequence NEDFRE as the core motif ( Figure 1B ), thereby recapitulating the expected (⁴⁶³ EGVNNSNEDFREENLKTA ⁴⁸⁰) epitope. Next, 22 seropositive ( Figure 1C ) and 13 seronegative (control) ( Figure 1D ) sera were probed on the K_v1.2 array. Here, autoantibody binding was detected using anti-human IgG coupled to HRP for chemiluminescence detection. The screening identified two prominent intracellular ( Figure 1E ) epitopes: Epitope 1 (E1) (⁴⁶⁹ NEDFREENLKTANCTLA⁴⁸⁵ ) and Epitope 2 (E2) (⁴⁸¹ NCTLANTNYVNITK⁴⁹⁵ ). Notably, none of the tested sera shared both epitopes. An additional array library with a single-AA shift of 15-mer peptides ( Figure 1F ) defined ⁴⁷⁸ KTANCTLA ⁴⁸⁵ as the E1 core motif, which was shared among 30 patients (Serum IDs 1–6, 8–10, 14–21, 23–34, and 36) and ⁴⁸⁵ ANTNYVNITK ⁴⁹⁵ as the minimal required E2 core motif, which was shared by two patients (Serum IDs 11 and 24) ( Table 2 ).

Remarkably, the identified epitopes both overlap and include a validated phosphorylation site (³⁸⁹Tyr; PhosphoSitePlus: P16389). In order to delineate the differences in the mode of binding between these two antibodies and specifically their dependency on phosphorylation, we next probed the corresponding phospho-tyrosine K_v1.2 library. Whereas autoantibody binding of E1 was not affected, E2 binding was completely abolished upon phosphorylation ( Figure 1G ).

To further resolve the binding requirements for autoantibodies, we conducted a deep mutational scan of the newly defined minimal binding epitopes. These libraries comprised all possible single-point variants of 18-mer peptides that harbored the minimal core motifs. The subsequent fingerprint analysis of the resulting 2 × 342 peptide variants ( Figure 2A ) confirmed both the previously defined minimal core motifs ( Figure 1C ) and the impact of the Tyr⁴⁸⁹ phosphorylation ( Figure 1G ).

Notably, for E1, all patient samples displayed a seemingly identical binding requirement for their K_v1.2 autoantibodies ( Figures 2B, C ). More precisely, within the mapped core motif ⁷⁸ KTANCTLA ⁴⁸⁵ ( Figure 1C ) residues, ⁴⁷⁸Lys, ⁴⁸¹Asn, and ⁴⁸³Thr were characterized by strict conservation with no tolerance toward any AA exchange ( Figure 2B ) for all tested patients ( Figure 2C ). For E2, both patients were fingerprinted ( Figure 2D ). Here, the fine-mapped core motif ⁴⁸⁵ ANTNYVNITK ⁴⁹⁵ ( Figure 1C ) was also recapitulated by both patients in the same way. In addition to the mapping, the fingerprint further highlights a strong conservation of the C-terminal part of the core motif ⁴⁸⁹Tyr to ⁴⁹⁵Lys and strict conservation of ⁴⁹⁰Val, ⁴⁹¹Asn, and ⁴⁹⁴Lys. In line with the analysis of the phosphorylated peptides ( Figure 1G ), the phophomimetic exchange of Tyr⁴⁸⁹ with Glu resulted in complete loss of autoantibody recognition, thereby confirming our previous finding of orthogonal recognition of this Tyrosine depending on its phosphorylation status. In summary, the profiling highlights and substantiates a shared autoimmune response toward two distinct K_v1.2 epitopes. The unexpectedly high similarities in the relative binding responses within the patient cohort and the strict conservation of identical residues suggest a homogenous autoantibody repertoire within the tested patient group and further hint toward a shared molecular origin of the observed epitopes.

The resolved binding profiles prompted us to next explore K_v-isoform specificity of the immunodominant antibody. Comparing the alignment of K_v subfamily members 1.1 to 1.4 with the previous binding requirements highlights the conservation of critical residues within these four subfamily members ( Figure 2E ), specifically between K_v1.1, K_v1.2, and K_v1.3. We therefore displayed and probed overlapping C-terminal peptides of all four members of this subfamily in microarray format and found that autoantibody binding to E1 is maintained within K_v1.2 and K_v1.3 but not K_v1.1 and K_v1.4 ( Figure 2F ). The additional binding capacity for K_v1.3 is in line with the previously resolved binding requirements of E1 ( Figure 2C ). Vice versa, the lack of binding of K_v1.1 that shares high homology in this region highlights the need for experimental validation of putative binders.

To resolve the predictive value of the specific epitope signals, we correlated seropositivity with the observed disease phenotype. To this end, the Euroimmun pre-screening determined 4% of the samples from healthy individuals as anti–K_v1.2-positive (four out of the 96 samples, data not shown). Our microarray readout recapitulated the reactivity and deep mutational scan analysis and attributed it toward E1, thus suggesting a similar broad occurrence (4%) of these mono-reactive autoantibodies within healthy individuals ( Supplementary Figure 1 ). In this light, we focused on the CSF samples obtained from 12 patients exploring both the diagnostic value and the sensitivity of the array screening in comparison to cell-based screening and its dependence of key parameters of the peptide display ( Figure 3 ). The CSF of first five patients was analyzed with longer peptides and larger offset (20-mers, offset of 3) ( Figure 3A ), and CSF from seven patients was screened using shorter peptides with shorter offset (15-mers, offset of 1) ( Figure 3B ) ( Table 2 ). Complementation of the array-based screening with cell-based screens showed equal in several cases even superior sensitivity for both display variants (7/12 CBA and 10/12 array detected patients) ( Figure 3C ). Specifically, CSFs from patients 2, 25, and 36 were tested positive in microarray and negative in CBA under the tested conditions. In addition, patient 12, negative from the arrays, resulted positive for K_v1.2 and K_v1.1 in cell-based assay (data not shown), potentially bearing a conformational epitope. Importantly, the CBA-determined positivity showed an association with cognitive impairment (30). Here, the array showed improved sensitivity over the CBA-based assay. We therefore conclude that, for K_v autoimmunity patients, clinicians should consider combining array- and CBA-based testing for autoantibody confirmation, when an autoimmune pathogenesis is suspected.

A commonly assumed limitation of array-based antibody screenings is the risk of missing autoantigen contributions from certain conformational and discontinuous epitopes not represented by the linear peptide display. In addition, anti-K_v1.2 autoimmune response was previously observed to co-occur with unrelated neurological diseases (30, 31) as well as other K_v1 complex–directed (7, 8) autoantibodies. To clarify the isoform specificity, K_v1.1, K_v1.2, and K_v1.6 were expressed in HEK293 cells and probed against the serum IDs 1, 3, and 27 and then compared with the respective control antibodies ( Supplementary Figure 2 ). In line with the deep mutational scans ( Figure 2 ), K_v1.1 and K_v1.6 showed no binding, thus confirming K_v1.2 specificity of the serum antibodies. To resolve contributions from possible additional antibodies binding via discontinuous epitopes that could not be resolved using peptide microarrays, we next conducted neutralization experiments of the identified epitope/autoantigen on chips and on cells. Soluble peptides were synthesized and purified on a preparative scale. On-chip neutralization resulted in a strongly reduced signal, thereby supporting a specific and exclusive binding through the identified epitopes toward K_v1.2 in this format ( Figures 4A, B ). To search for additional, e.g., conformational, epitopes, we expressed K_v1.2 in HEK293 cells ( Figures 4C, D ) and tested the neutralization of the observed immune response for the two identified epitopes and the commercial control antibody ( Figure 4C ). Transfected HEK293 cells were stained with serum pre-treated with neutralizing peptides, non-neutralizing peptides, and buffer only ( Figure 4C, D ). Patient’s sera were selected on the basis of their different epitopes for neutralization. Consequently, no residual binding was detected for either neutralized sample bearing E1 ( Figure 4C ) or E2 ( Figure 4D ). The non-neutralizing peptide had no impact on autoantibody binding. Notably, because no residual binding was detected, we conclude that K_v1.2 autoimmunity is primarily mediated by autoantibodies that recognize the previously highlighted linear motif, without any contributions from additional linear or conformational epitopes.

Prompted by the confirmation of the identified epitopes as sole driver of the observed immune response in transfected recombinant cells, we next explored autoantibody binding toward the native autoantigens in their cellular context. Here, we applied the neutralized sera on teased fibers ( Figure 5A ). Commercial antibody (K14/16) signals recapitulate the expected juxtaparanodal binding on teased fibers ( Figure 5B ). Sera applied together without peptide ( Figure 5C ) and with non-neutralizing peptide ( Figure 5D ) recapitulated the same labeling. In stark contrast, sera pre-treated with neutralizing peptide display a complete loss of binding signal ( Figure 5E ). Thus, complete anti-K_v1.2 epitope-specific neutralization was achieved using only the minimal K_v1.2 epitope. Remarkably, no residual autoantibody binding was detected in the nerve tissues, and the tested sera did not cross-react with additional autoantigens co-expressed peripherally. This corroborates the hypothesis of selective K_v1.2 autoimmunity without co-existing peripheral autoantibodies, thus contrasting previous observations (7, 8, 30, 31).

The neutralization in peripheral nerves was complemented by neutralization in tissue slices from the central nervous system, specifically mouse cerebellum, where K_v1.2 expression is high at the axon initial segment (AIS) of Purkinje cells. MAb K14/16 served as a K_v1.2 antibody control for specific binding of autoantibodies. Here, the typical axonal initial segment stainings were observed on the Purkinje cell layer. E1 ( Figure 5F ) and E2 ( Figure 5G ) positive serum was applied in presence of several neutralizing and non-neutralizing peptide variants.

Taken together, the neutralization data in cell and tissues confirmed the high selectivity for the identified peptide epitopes, showing no detectable residual binding. Thus, leading to the exciting conclusion that a single, broadly shared epitope may contribute significantly to the often-reported K_v1 autoimmunity (7, 8, 30, 31), in some cases, even without coexisting autoantigens or other conformational epitopes. In addition, the disease association of the detection of the here defined immunodominant epitope in CSF suggests implications for diagnosis, possibly even the pathology of a subgroup of autoimmune neuropsychiatric phenotypes.