HLA-DQB1*05 subtypes and not DRB1*10:01 mediates risk in anti-IgLON5 disease

Abstract

Introduction

Anti-IgLON5 disease is a recently described autoimmune syndrome, with distinctive and polymorphic clinical features.¹ It affects both sexes equally and has a late age of onset, with a mean age at diagnosis of >60 years.² Symptoms may include a complex form of sleep disorder, with REM/non-REM parasomnia, sleep apnoea, and stridor, gait and other movement abnormalities, as well as bulbar and cognitive impairment. All patients are positive for serum, and in most cases, CSF autoantibodies targeting the extracellular domain of IgLON5.² Early on in the disease, CSF pleocytosis is observed, consistent with CNS inflammation.² Also, a neurodegenerative biology is supported by neuropathological^3,4 and PET⁵ studies, showing the accumulation of hyperphosphorylated tau in the brains of patients. However, neuropathological studies indicate rare cases may present without a tauopathy.^6,7

A clinical diagnosis can be difficult, as late onset and symptomatology overlap with that of neurodegenerative disorders and the only very recent description of this condition suggests it remains likely under-recognized. Importantly, early and aggressive immunotherapy can stabilize or even ameliorate symptoms, although treatment remains poorly effective in >50% of patients.⁸ However, a better understanding of the underlying pathophysiology of this devastating disorder may pave the way for alternative and improved therapeutics.

IgLON5 belongs to a family of five highly homologous cell surface immunoglobulin-like cell adhesion molecules that are widely expressed in the CNS^9,10 and myoblasts,¹¹ whereby murine IgLON5 was found to be expressed at higher levels in the brainstem, thalamus, ventral striatum and olfactory bulb,¹² thus mirroring tauopathy-affected distribution.³ Studies of cultured rat hippocampal neurons¹³ and human neural stem cells¹⁴ have shown that IgGs from patient sera can lead to disrupted cytoskeletal organization and neurodegenerative changes, while in vivo evidence from mouse models further supports disruptions to synaptic homeostasis and irreversible neuronal damage.¹⁵

Altogether, these studies suggest that autoimmunity operates upstream of neurodegeneration, triggering neuronal structural damage and tau accumulation later in the course of the disease. Identification of key, early autoimmune pathways involved in triggering the generation of cytotoxic tau and neurodegeneration, therefore, may provide both early therapeutic targets for the successful treatment of this disorder and, beyond that, provide insights into crucial autoimmune-neurodegenerative mechanisms that are relevant beyond this disease too.

An autoimmune origin for the disease is further supported by a unique human leukocyte antigen (HLA) class II association,^1,16 similar to those found in other autoimmune diseases such as coeliac disease (DQ2),¹⁷ rheumatoid arthritis (DR4)¹⁸ and, in CNS conditions, narcolepsy (DQ0602)¹⁹ and other autoimmune encephalitis syndromes associated with LGI1 and CASPR2 antibodies.²⁰ HLA class II molecules play a key role in initiating and driving autoimmunity by presenting (auto)antigens to CD4⁺ T cells, leading to subsequent T- and B-cell activation.^21,22 However, T-cell contributions, and the precise HLA molecules involved in T-cell activation, have received limited attention. Although the mechanisms underlying autoreactivity are poorly understood, molecular mimicry with foreign antigens is strongly suggested in some diseases, such as Epstein-Barr virus in multiple sclerosis^23,24 or influenza in narcolepsy.^25,26 Interestingly, autoantigens are often post-translationally modified (PTM); for example, they are citrullinated in rheumatoid arthritis^18,27 or cryptically fused in type 1 diabetes,^28,29 in line with reduced negative selection for PTMs in the thymus.³⁰

The objective of this study was to conduct an HLA-association analysis in a large sample of anti-IgLON5 disease patients to examine candidate autoimmune HLA-peptide binding and probe CD4⁺ T-cell reactivity to candidate binders, thereby ultimately allowing us to disentangle triggers and key events in the pathophysiology of this disease.

Materials and methods

Participants

A total of 87 patients with anti-IgLON5 disease enrolled across Spain (51), Italy (2), France (18; French samples were banked in NeuroBioTec Hospices Civils de Lyon BRC, France, AC-2013-1867, NFS96-900), Germany (4), Switzerland (5), Brazil (1), Japan (1), USA (4) and the UK (1) were used (Table 1 and Supplementary Fig. 1). All patients were positive for IgLON5 autoantibodies in serum. Data from 44 of these patients were published previously^1,16,31-37 (Supplementary Table 1). A subset of 75 patients had DNA available for a genome-wide association study (GWAS); these were used for HLA analysis versus controls matched by ethnicity using principal component analysis (PCA; case control analysis). An additional 12 subjects had HLA typing and clinical data available; these were used in addition to the 75 to maximize power for analyses comparing symptoms across HLA type categories within anti-IgLON5 disease patients (phenotypic analysis) (Supplementary Fig. 1). Clinical data were collected from clinical records or assessed as previously described.^1,31

HLA and genotyping

DNA was available from 75 patients (Supplementary Fig. 1) and genotyped using the Affymetrix Precision Medicine Research Array (AKESOgen). All quality control and genotype single nucleotide polymorphism (SNP) operations such as creating Manhattan plots, zoom plots and using FINEMAP (Supplementary Fig. 2) were conducted using PLINK.³⁸ Genotypes with an imputation probability <0.3 were removed. Haplotypes were phased and merged using QCTOOL and imputed to 1000 Genome Phase III. Patients were matched to controls by PCA (Eucleadian distance-based measure³⁹) at a 1:8 ratio from a set of 2503 controls, giving rise to 232 matched controls (1:3.1; Fig. 1A). HLA Genotype Imputation with Attribute Bagging (HIBAG)⁴⁰ was used to perform HLA imputation. Calls with a probability <0.75 (n = 4) were additionally HLA-typed using next-generation sequencing (NGS)^41-43 (Supplementary Figs 1 and 3). Prior 4-digit resolution HLA typing data were available from 50 patients. To further validate imputation, an additional 14 randomly selected participants were typed using 8-digit resolution NGS HLA genotyping (Supplementary Fig. 1 and Supplementary Table 1).

Figure 1

HLA-association with three conserved HLA-DQ5 haplotypes. (A) The first two principal components (PC1, PC2) are shown to highlight the ethnic diversity of subjects included in the study. Cases were matched at a 1:8 ratio to controls from a set of 2503 controls by principal component analysis. [B(i)] HLA association of anti-IgLON5 disease across HLA class I and II is shown unconditioned, (ii) under conditioned exclusion of HLA-DRB1*10:01, (iii) HLA-DRB1*10:01, HLA-DRB1*01:01 and HLA-DRB1*01:02 and (iv) HLA-DRB1*10:01, HLA-DRB1*01:01, HLA-DRB1*14:01, HLA-DRB1*14:54 and HLA-DRB1*14:04. (C) Sankey plot shows conservation of risk-associated HLA-DRB1-DQA1-DQB1 haplotypes among cases.

Open in new tabDownload slide

Statistical analyses

For genetic analyses, categorical variables are presented as percentages (%) and total number of subjects (n) for any group reported in brackets. For allele carrier frequencies, a generalized logistic model (GLM) was fit to each allele, controlled by the first three principal components. The analysis was repeated after conditioning for specific haplotypes, as previously described.⁴⁴ For haplotype analyses, a count of heterozygous and homozygous cases and controls was conducted, whereby controls were matched to cases by PCA (see earlier). χ² statistics and odds ratios (ORs) are reported. Age at disease onset between different HLA-carrier groups (Fig. 2) was compared using Mann–Whitney U-test, with data-points shown for all groups together with a box plot showing interquartile range (IQR) mean and whiskers extending to the smallest and largest data-points within 1.5 times the IQR and violin plots showing data distribution. Total number n is reported for all clinical and demographic parameters. For peptide binding experiments, the mean of data and standard error of the mean (SEM) is shown for all data-points. All analyses were performed using R and Python. Significance levels were set at *P < 0.05 and **P < 0.01. All P-values are false discovery rate (FDR) corrected.

Figure 2

Delayed age of onset correlates with ranked HLA-risk and HLA-DQ5 dosage. (A) Age at disease onset, (B) sex demographics and (C) major clinical features reported are shown for cases carrying (i) HLA-DQA1*01:05-DQB1*05:01, (ii) HLA-DQA1*01:01-DQB1*05:01, (iii) HLA-DQA1*01:04-DQB1*05:03 or (iv) none of the aforementioned haplotypes. In A, coloured, filled circles indicate individual cases, whereas black filled markers denote homozygous carriers of the given haplotypes (Supplementary Table 2). In C, the numbers in the centres of the pie charts indicate the total numbers of cases reporting the given clinical feature. (D) Age at disease onset is shown for carriers (i) homozygous for HLA-DQ5, (ii) heterozygous for HLA-DQ5 without HLA-DQ6, (iii) heterozygous for HLA-DQ5:∼ with HLA-DQ6:∼ and (iv) not carrying HLA-DQ5:∼. For D(i), age range = 49–80 years (x͂ = 60.5 years); D(ii), age range = 34–91 years (x͂ = 62.0 years); D(iii), age range = 54–70 years (x͂ = 65.0 years); D(iv), age range = 59–81 years (x͂ = 71.5 years). n = total number of cases, μ_age = mean age, x͂ = median age, *P < 0.05, **P < 0.01.

Open in new tabDownload slide

Transfection and purification of HLA class II monomers

Plasmids encoding HLA-DQA1*01:01, HLA-DQA1*01:04, HLA-DQA1*01:05, HLA-DQB1*05:01, HLA-DQB1*05:03, HLA-DRA1*01:01 and HLA-DRB1*10:01 were purified according to manufacturer’s instructions (Qiagen, Cat. No. 12662) and transfected into Sf9 cells using BestBac2.0 (Cat. No. 91-002, Expression Systems), with P0 stock virus harvested after 5 days. Successful transfection was confirmed using a Baculovirus Titering Kit (Cat. No. 97-101, Expression Systems) and flow cytometric analysis. Following this, 100 ml of Sf9 cells were seeded at a concentration of ∼1 × 10⁶ cells/ml together with 100 μl of P0 in a 27°C shaking incubator. On Day 3, cells were resuspended in fresh medium (Sf-900™ III SFM; Thermo Scientific, Cat. No. 12658019), supplemented with 1% GlutaMAX™ (Thermo Fisher Scientific, Cat. No. 35050061) and 10% fetal bovine serum (FBS; Gibco, Cat. No. 26140079) and subsequently set down for 7 days more. P1 was harvested and tested using a Baculovirus Titering Kit (Cat. No. 97-101, Expression Systems), flow cytometric analysis and a Lentivirus Titer Kit (Cell Biolabs, Cat. No. VPK-107).

Amplification of virus titres was repeated for four to six generations and tested using a minimum protein expression protocol, whereby 300 ml of Hi5 cells at a density of 2 × 10⁶ cells/ml were incubated with 1.5 ml of amplified P virus at 27°C in a shaking incubator for 3 days. Following the 3-day incubation, the supernatant was harvested and proteins purified using a Protino Ni-NTA Agarose for His-tag protein purification kit (Machery-Nagel, Cat. No. 745400.100). The eluted protein was mixed with 4× Laemmli buffer (Bio-Rad, Cat. No. 1610747) and boiled for 10 min at 95°C before loading onto a mini-PROTEAN^® TGX^® precast gel for SDS-PAGE (Bio-Rad, Cat. No. 4561093) and stained with Coomassie Blue (Bio-Rad, Cat. No. 1610786). Presence of protein was confirmed using 10–180 kDa Prestained Protein Ladder (Fisher Scientific, Cat. No. 26616).

Purification of HLA-DQA1*01:01/DQB1*05:01 (DQ0101), HLA-DQA1*01:04/DQB1*05:03 (DQ0104), HLA-DQA1*01:05/DQB1*05:01 (DQ0105) and HLA-DRA1*01:01/DRB1*10:01 (DR10) was repeated in 5 l, as previously described.⁴⁵ Protein concentration was measured using a NanoDrop A280 and proteins were purified further by injection into a Superdex^® 200 Increase 10/300 GL (GE Healthcare, Cat. No. GE17-5174-01) and Akta Pure fast protein liquid chromatography.

Identification of strong binder ligands specific to each HLA molecule

To identify possible reference ligands for HLA binding competition assays, we computationally predicted binding of different viral peptides to DQ0101, DQ0104, DQ0105 and DR10⁴⁶: HIV Tat-specific factor 1, Plexin-C1, Epstein-Barr nuclear antigen 1 and 2, varicella-zoster virus and herpes simplex virus type 1. Seventeen 15-mer peptides were short-listed from this analysis as possible strong binders and synthesized biotin-conjugated at GenScript (NJ, USA) with >90% purity. Binding of these peptides to each HLA molecule was tested as described elsewhere.⁴⁵ In brief, CLIP peptides of purified HLA-monomers were removed using thrombin cleavage (Millipore, Cat. No. 69671). In each well of a 96-well plate (Bio-Rad, Cat. No. MLP9601), 25 nM HLA-monomer was incubated with 10 μM biotin-conjugated ligand and 1× protease inhibitor (Sigma Aldrich, Cat. No. P8849), and the final reaction volume was increased to 30 μl using reaction buffer [100 mM acetate, pH 4.6, 150 mM NaCl, 1% bovine serum albumin (BSA) and 0.5% Nonidet P-40]. All reactions were tested both with and without the addition of 25 nM HLA-DM, and all reactions were tested at least in duplicate. Controls were run for each plate by substituting biotin-conjugated ligands for dimethyl sulphoxide (DMSO). Plates were sealed and incubated for 3 days at 37°C. The reaction was neutralized with 60 μl neutralization buffer (100 mM Tris-HCl, pH 8.6, 150 mM NaCl, 1% BSA, 0.5% Nonidet P-40 and 0.1% NaN₃) and transferred to a plate coated with monoclonal anti-DQ (Biotium, Cat. No. BNUM0200-50) or anti-DR antibody (BioLegend, Cat. No. 327002). Plates were incubated at room temperature for 1 h, washed five times with 300 µl 1× PBS supplemented with 0.05% Tween-20 (pH 7.4) and subsequently incubated for 1 h with 100 µl europium-labelled streptavidin (Cat. No. 1244-360, PerkinElmer) at a 1:1000 dilution in PBS with 1% BSA (pH 7.4). Plates were washed as described above, 100 µl enhancement solution (Cat. No. 1244-105, PerkinElmer) were added to each well and plates were placed on a shaker for 5 min. DELFIA^® time-resolved fluorescence (TRF) intensity was detected using a Tecan Infinite^® M1000 system. Following repetition of candidate binders using the above protocol in titrations of 100, 10 and 1 μM, KNIYIYLTAGKEVRR (bio-PLXC1, derived from Plexin C1) was confirmed as a strong-binding ligand to DQ0101, DQ0104 and DQ0105, and EDEIRGYKLHVEVAK (bio-HTSF1, derived from HIV Tat-specific factor 1) as a strong-binding ligand to DR10.

Competition binding assay

Fifteen-mer peptides with an 11-amino acid overlap covering the full length of IgLON5 and selected PTMs predicted using MusiteDeep,⁴⁷ NetPhos3.1,⁴⁸ NetNGlyc1.0⁴⁹ and mass spectrometry data from IgLON family members⁵⁰ (Fig. 3A and Supplementary Table 3) were synthesized with >90% purity at GenScript and dissolved in DMSO at a stock concentration of 10 mM. Competition binding assays were conducted using the same plate-binding assay used for the identification of strong binders. Specifically, each reaction contained 25 nM of thrombin-cleaved HLA-monomer, 400 μM of IgLON5 peptide and 10 μM of bio-PLXC1 or bio-HTSF1 (Fig. 3B and C) for DQ0101/DQ0104/DQ0105 and DR10 containing reactions, respectively. All reactions were tested both with and without the addition of 25 nM HLA-DM and in duplicate. Controls were run for each plate by substituting both IgLON5 and biotin-conjugated peptides for DMSO (negative control) or just IgLON5 peptides for DMSO (positive control). Peptides with Eu TRF intensity between 25% and 50% of that of bio-PLXC1/HTSF1 alone were considered weak binders, whereas any with intensity <25% were considered strong binders. Candidate binders were further tested in titration at competitor-to-peptide ratios of 1:40, 1:100 and 1:200, using 40, 100 and 200 μM of IgLON5 peptide and 1 μM of bio-PLXC1 or bio-HTSF1, for DQ0101/DQ0104/DQ0105 and DR10 containing reactions, respectively (Fig. 3D and E and Supplementary Fig. 7B and C).

Figure 3

Risk-associated HLA-DQ5 molecules preferentially bind IgLON5 in a deamidated form. (A) The entire length and sequence of the IgLON5 peptide is shown, highlighting sites of post-translational modification (PTM) as predicted by MusiteDeep, NetPhos3.1, NetNGlyc1.0 and mass spectrometry data obtained from Itoh et al.⁵⁰ N_glyc = N-linked glycosylation, S_phos = serine phosphorylation, T_phos = threonine phosphorylation and Y_phos = tyrosine phosphorylation. (B) Schematic shows how peptide-HLA binding is determined using competition binding assay. (C) Results from competition binding assay are shown for (i) HLA-DQA1*01:05-DQB1*05:01, (ii) HLA-DQA1*01:01-DQB1*05:01, (iii) HLA-DQA1*01:04-DQB1*05:03 and (iv) HLA-DRA1*01:01-DRB1*10:01. Light bars show IgLON5 15mer peptides in native, and dark bars in PTM form (see legend at bottom). The y-axis (log scale) shows mean fluorescence intensity (mfi) of IgLON5 peptide together with biotinylated-competitor in competition, divided by the mfi of biotinylated-competitor alone. For i–iii, bio-PLXC1, and for iv, bio-HTSF1 was used as biotinylated-competitor. Along the x-axis, enumerated 15-mer IgLON5 peptides, sequentially encompassing the entire length of the IgLON5 protein, are shown (see Supplementary Table 3 for a full list of enumerated peptides). The dotted lines show the thresholds for weak (0.5) and strong (0.25) binders. (D) Binding of HLA-DQA1*01:05-DQB1*05:01 (green bars, left in each group), HLA-DQA1*01:01-DQB1*05:01 (purple bars, middle in each group) and HLA-DQA1*01:04-DQB1*05:03 (yellow bars, right in each group) to the five peptides (i) ⁵³SCFIDEHVTRVAWLN⁶⁷, (ii) ¹²⁵VYLIVHVPARIVNIS¹³⁹, (iii) ¹⁴¹PVTVNEGGNVNLLCL¹⁵⁵, (iv) ¹⁴⁵NEGGNVNLLCLAVGR¹⁵⁹ and (v) ¹⁴⁹NVNLLCLAVGRPEPT¹⁶³ is shown, with asparagine residues (N) in bold in an unmodified (‘native’, left), glycosylated (‘GLcNac’, middle) or deamidated (‘deamidated’, right) form. Dotted lines as for C; y-axis is also as described for C but linear. (E) Dose-dependent change in binding of peptides described in D(i–v) when deamidated at bold N residues (i.e. mimicking aspartic acid, D). The y-axis shows dose-dependent change in binding as a ratio relative to baseline concentration used in D (40 μM of IgLON5 peptide), given the variation of IgLON5 peptide concentrations shown along the x-axis. Error bars in C and D, and shaded regions in E, represent the standard error of the mean (sem). All assays were repeated at least in duplicate.

Open in new tabDownload slide

Biotinylation of HLA-DQA1*01:05-DQB1*05:01 (DQ0105), peptide exchange and spheromer assembly

DQ0105 monomer was biotinylated (bio-DQ0105) using a BirA biotin-protein ligase standard reaction kit (Avidity, Cat. No. BirA500) and extra free biotin removed by injection into a Superdex^® 200 Increase 10/300 GL (GE Healthcare, Cat. No. GE17-5174-01). Bio-DQ0105 was next transferred to a plate coated with monoclonal anti-DQ antibody (Biotium, Cat. No. BNUM0200-50), incubated for 1 h at room temperature and processed as described for the identification of strong binders. Biotinylation efficiency of the monomer was determined by detecting TRF intensity using a Tecan Infinite^® M1000 system. Following this, bio-DQ0105 was cleaved and peptides exchanged as previously described⁴⁵ with ¹²⁵VYLIVHVPARIVDIS¹³⁹ (bio-DQ0105/IgLON5). HLA-DM was removed and 5 μl of bio-DQ0105/IgLON5 (10 μM) incubated with 9.47 μl phycoerythrin-labelled streptavidin (2.78 μM, BioLegend, Cat. No. 405203) and 64.6 μl PBS (pH 7.4) for 30 min at room temperature. Uncleaved, unloaded bio-DQ0105 (bio-DQ0105/CLIP) (2.64 μM) was also incubated with 0.75 μl A488-labelled streptavidin (33.3 μM, BioLegend, Cat. No. 405235) and 59.3 μl PBS (pH 7.4) for 30 min at room temperature. The spheromers assembled with bio-DQ0105/CLIP monomers were used as a control for non-specific staining. Spheromer scaffold (0.9 μl, 471 μM)⁵¹ and PBS (20 μl, pH 7.4) were added to each reaction, and following brief vortexing, incubated at room temperature for 30 min in the dark on a shaking incubator (Fig. 4A).

Figure 4

CD4⁺ T-cell reactivity and phenotypic profiling. (A) Schematic shows how peptide-HLA-spheromers are assembled and used to isolate antigen-specific CD4⁺ T cells. (B) Proportion of CD4+ T cells reactive to ¹²⁵VYLIVHVPARIVDIS¹³⁹ in one control and two patients. (C) Neighbourhood clustering of isolated single cells by delineation into individuals [control (green), Patient 1 (purple), Patient 2 (pink)]. (D) T-cell markers. (i) Dot plot displays the expression levels of T-cell subset markers related to distinct transcriptional states in the control versus patient (Patients 1 and 2) group and (ii) related to distinct activation states. (E) T-cell receptor sequencing from antigen-specific cells isolated from control (green), Patient 1 (purple) and Patient 2 (pink) display the frequency of gene segments (i) TRAV, (ii) TRAJ, (iii) TRBV and (iv) TRBJ. (F) Pairings of (i) TRAV-TRAJ and (ii) TRBV TRBJ are shown. For B, C, E and F, see legend for colour coding and descriptions. T_naive = naive; T_em = effector; T_emra = highly effector; T_reg = regulatory.

Open in new tabDownload slide

Peripheral blood mononuclear cells and sequencing of anti-IgLON5 antigen-specific CD4⁺ T cells

Frozen peripheral blood mononuclear cells (PBMCs) were available from two patients with anti-IgLON5 disease carrying HLA-DQA1*01:05-DQB1*05:01 and from one HLA-matched control. Cells were thawed, counted and washed twice before settling in complete medium (RPMI 1640; Sigma Aldrich, Cat. No. 61770-036) supplemented with 10% heat-inactivated FBS (Gibco, Cat. No. 26140079) and 1× penicillin-streptomycin (Gibco, Cat. No. 15140122) and filtered through a 0.2 μm polyethersulphone (PES) membrane (Nalgene, Cat. No. 566-0020) at a concentration of 700 000 cells/300 μl/well in a 96-well flat bottom microplate (Thermo Fisher Scientific, Cat. No. 167008). Each well was stimulated with 1.875 μl of IgLON5-¹²⁵VYLIVHVPARIVDIS¹³⁹ (1 mM) and incubated in a tissue culture incubator at 37°C, 5% CO₂ for 10 days. Medium was changed every 2–3 days and cells stimulated with IL-2 on Day 8. On Day 10, cells were washed and resuspended in 200 μl fluorescence-activated cell sorting (FACS) buffer (1× PBS, pH 7.4; Gibco, Cat. No. 10010023) supplemented with 2 mM EDTA (Sigma, Cat. No. E4884-500G) and 0.5% BSA (Sigma, Cat. No. A7888-50G), filtered through a 0.2 μm PES membrane (Nalgene, Cat. No. 566-0020). Cells were stained with 25 μl of each bio-DQ0105/IgLON5 and bio-DQ0105/CLIP spheromer and incubated at 37°C, 5% CO₂ for 90 min with gentle vortexing every 20 min. Cells were further stained with anti-CD3 (BioLegend, Cat. No. 317343), anti-CD4 (BioLegend, Cat. No. 300557), anti-CD8 (BioLegend, Cat. No. 301014) and Aqua live/dead stain (Invitrogen, Cat. No. L34965) prior to analysis with FACS. CD4⁺/bio-DQ0105/IgLON5⁺ cells were sorted into 96-well plates filled with capture buffer (Takara, Cat. No. 634439) and processed for single-cell sequencing and T-cell receptor sequencing as previously described.⁴⁵ Single-cell data were analysed using Scanpy.⁵²

Standard protocol approvals, registrations and patient consents

This study was reviewed and approved by the Stanford University Institutional Review Board (Protocol No. 65073) as well as by local ethics committees of collaborating institutions. Written informed consent was obtained from all patients for the storage and use of biological samples and clinical information for research purposes.

Results

A primary association with HLA-DQ rather than HLA-DR in anti-IgLON5 disease

The intent of this study was to refine and extend prior HLA association studies in anti-IgLON5 disease. To do so, genome-wide SNP typing was conducted in 75/87 patients with the disease (Table 1, Supplementary Figs 1 and2 and Supplementary Table 1), the largest sample ever studied. This was used to match cases to 232 controls by ethnicity using PCA (Fig. 1A) and to impute HLA. Any questionable samples and a random collection of samples were HLA-sequenced for verification.

Following HLA imputation and typing (Supplementary Figs 1 and 3), HLA association analysis revealed strong dominant associations with HLA-DQB1*05:01 (frequency = 0.747, 56/75 patients), HLA-DRB1*10:01 (frequency = 0.440, 33/75 patients) and HLA-DQA1*01:05 (frequency = 0.413, 31/75 patients) in patients versus controls [Fig. 1B(i) and Table 2]. As shown in previous studies,^1,2,16HLA-DRB1*10:01 and HLA-DQB1*05:01 showed the strongest association in terms of significance. Importantly, however, HLA-DRB1*10:01, HLA-DQA1*01:05 and HLA-DQB1*05:01 are in strong linkage disequilibrium (LD), forming a conserved haplotype⁵³ (Fig. 1C and Supplementary Fig. 4). As such, association is with the entire haplotype, and although risk is higher with HLA-DRB1*10:01 and HLA-DQA1*01:05, HLA-DQB1*05:01 is also present in other haplotypes and was more frequently associated with the disease, most notably in the context of HLA-DRB1*01∼DQA1*01:01∼DQB1*05:01 haplotypes (Supplementary Fig. 4A).

Illustrating this, after excluding all carriers of HLA-DRB1*10:01∼DQB1*05:01 (33 cases and 6 controls), strong signals were still observed with HLA-DQB1*05:01 (frequency = 0.548, 23/42 patients), followed by HLA-DQA1*01:01 (frequency = 0.548, 23/42 patients) and HLA-DRB1*01:01 (frequency = 0.452, 19/42 patients) [Fig. 1B(ii) and Table 2]. Once again, these alleles formed a common conserved haplotype (Fig. 1C and Supplementary Fig. 4), with only one subject carrying the HLA-DRB1*01:02∼DQA1*01:01∼DQB1*05:01 haplotype. Given the remaining 19 patients did not carry either HLA-DRB1*10:01∼DQA1*01:05-DQB1*05:01 or HLA-DRB1*01∼DQA1*01:01∼HLA-DQB1*05:01, we further conditioned the analysis by removing carriers of HLA-DRB1*01:01/*01:02 alleles as well, which thus removed any carriers of HLA-DQA1*01:01∼DQA1*05:01. This identified a remaining association with HLA-DQB1*05:03 (frequency = 0.368, 7/19 patients), HLA-DQA1*01:04 (frequency = 0.368, 7/19 patients), HLA-DRB1*14:54 (frequency = 0.211, 4/19 patients) and HLA-DRB1*14:01 (frequency = 0.105, 2/19 patients) [Fig. 1C(iii) and Table 2]. Further conditioning with removal of HLA-DQA1*01:04∼DQB1*05:03 patients did not reveal additional significance [Fig. 1C(iv)].

Using genome-wide data, the highest association was with rs77456373 [minor allele frequency (MAF) = 0.32, OR = 4.09, P = 1.26 × 10⁻⁸, near HLA-DQA1], rs73729117 (MAF = 0.27, OR = 3.98, P = 1.07 × 10⁻⁸, near HLA-DRB1) and rs9271510 (MAF = 0.36, OR = 3.64, P = 1.63 × 10⁻⁸, near HLA-DQA1), all in tight LD (r² > 0.8) (Supplementary Fig. 2A). FINEMAP revealed two possible causal variants: rs17843573 with a probability of 0.99 (MAF = 0.40, r² = 0.56 with rs77456373) and rs9273329 with a probability of 0.96 (MAF = 0.43, r² = 0.64 with rs77456373). To confirm that these signals on chromosome 6 were secondary to the HLA association itself, we conditioned the GWAS association plot with rs73729117 (top SNP) (Supplementary Fig. 2B), rs17845573 (top FINEMAP SNP) (Supplementary Fig. 2C) and both (rs73729117 and rs17845573) (Supplementary Fig. 2D), showing that in all cases a residual effect above 10⁻⁵ remained in the HLA region. Similarly, controlling for HLA-DQB1*05:01 only (which did not account for the strong effect of HLA-DQA1*01:05 in conjunction with it), did not abolish the signal (Supplementary Fig. 2E). Only when all HLA effects we identified, HLA-DQA1*01:05, HLA-DQA1*01:01 and HLA-DQA1*01:04, were taken into account, did all association signal effects in the DR and DQ regions disappear (Supplementary Fig. 2F and G). While confirming the identified HLA association, single SNP data should be treated with caution, subject to the small sample size.

Altogether, this analysis indicated a rank-wise association with HLA-DRB1*10:01∼DQA1*01:05∼DQB1*05:01, DRB1*01∼DQA1*01:01∼DQB1*05:01 and DRB1*14∼DQA1*01:04∼DQB1*05:03 haplotypes, in order of descending predisposing risk. Importantly, conservation of DQA1*01-DQB1*05 (HLA-DQ5) with one to two amino acid differences across all three of these haplotypes (Supplementary Fig. 4B and C) indicated a primary role for HLA-DQ over HLA-DR.

HLA-DQ5 haplotypes associate with age at disease onset

Clinical data, together with (at least) 4-digit resolution HLA data for all DR∼DQ loci, were available in 71 patients (Table 1, Supplementary Fig. 1 and Supplementary Table 1). Comparison of clinical characteristics identified a later mean age at disease onset, in line with carrier status for all three ranked HLA-DQ5 haplotypes (Supplementary Table 2), which was significant for the two most dominant associations, HLA-DQA1*01:05-DQB1*05:01 [Fig. 2A(i)] and HLA-DQA1*01:01-DQB1*05:01 [Fig. 2A(ii)]. Specifically, subjects with HLA-DQA1*01:05-DQB1*05:01, DQA1*01:01-DQB1*05:01 and DQA1*01:04-DQB1*05:03 had a 11.3-, 8.6- and 0.9-years younger age of onset [Fig. 2A(i–iii)], compared with non-risk HLA haplotype carriers [Fig. 2A(iv)], respectively.

As previously reported,¹⁶ we observed a higher proportion of male carriers with HLA-DQA1*01:05-DQB1*05:01 (77% male versus 49% female, P = 0.37; Fig. 2B). Stratifying this analysis by gender showed a slightly older age at disease onset in females versus males [Supplementary Fig. 5A and B(i–iv)]; however, when controlling for gender as a covariate in the aforementioned analyses, the significance in the age shift (Fig. 1A) persisted [Supplementary Fig. 5C(i)]. We did not, however, confirm the presence of a sleep-dominant phenotype in HLA-DRB1*10:01 carriers (43% versus 46% in non-DRB1*10:01, P = 0.12) as initially suggested.¹⁶ Rather, stratifying by HLA-DQ5 revealed a more diverse clinical picture in non-HLA-DQ5 versus HLA-DQ5 carriers, although the numbers of non-DQ5 patients were too low to be conclusive for this subgroup (Fig. 2C). Moreover, there was no distinctive ethnicity associated with HLA-DQ5 subtype, as determined by PCA (Supplementary Fig. 5D).

HLA-DQ5 dosage effects and allele competition stratify disease risk

Given that HLA-DQA1*01-encoded alpha chains can heterodimerize with either HLA-DQB1*05- or 06-encoded beta chains, giving rise to DQ molecules with distinct binding profiles⁵⁴ (Supplementary Fig. 6), we investigated whether alterations in disease risk could occur in double-DQ1 heterozygotes, as reported for narcolepsy, where allele competition of other DQ1 molecules occurs with the DQ0602 heterodimer.⁵⁵ As predicted, we observed reduced frequency of trans-located DQB1*06 (DQ6, non-IgLON5-susceptibility associated DQ1 haplotypes) in patients versus controls (Table 3). Furthermore, age at disease onset was lowest in patients homozygous for DQB1*05 (DQ5/DQ5), moderate for heterozygous carriers of DQ5 with non-competing haplotypes (DQ5/other) and oldest in DQ5/DQ6 heterozygotes [Fig. 2D and Supplementary Fig. 5A, B(v–viii) and C(ii)], suggesting that competition with non-susceptibility DQ1 alleles modulates age of onset. While onset occurred significantly earlier in DQ5/DQ5 [Fig. 2D(i) and DQ5/other [Fig. 2D(ii)] subjects compared with non-DQ5 carriers [Fig. 2D(iv)], this effect was abrogated when DQ5 was inherited together with DQ6 [Fig. 2D(iii)]. Altogether, these observations further support a functional role of HLA-DQ5 in the development of anti-IgLON5 disease and suggest that rank-wise and dosage-dependent effects stratify disease risk.

Deamidation of N-glycosylation sites in the Ig2 domain primes IgLON5 peptide immunogenicity

To identify potential immunogenic IgLON5 peptides acting as T-cell auto-antigens, we derived 15-mer peptides (with 11-mer overlaps between peptides) encompassing the entire length of the IgLON5 protein (Fig. 3A and Supplementary Table 3) and screened these for binding (Fig. 3B) to all risk-associated HLA-DQ5 molecules: HLA-DQA1*01:05-B1*05:01 [Fig. 3C(i)], DQA1*01:01-B1*05:01 [Fig. 3C(ii)] and DQA1*01:04-B1*05:03 [Fig. 3C(iii)], as well as to DRA1*01:01-DRB1*10:01 [Fig. 3C(iv)].

In line with observations from other autoimmune diseases that show preferential immunogenic binding of peptides in PTM form,³⁰ we investigated possible modifications of IgLON5 (Fig. 3A). Computational predictions suggested that several serine, threonine and tyrosine residues could be phosphorylated,^47,48 and several asparagine residues N-linked glycosylated⁴⁹ (Fig. 3A). Notably, mass spectrometry confirmed several N-glycosylation sites in homologous IgLON peptides,⁵⁰ consistent with previous work.⁵⁶ Similarly, asparagine and glutamine can also be chemically deamidated as a result of protein ageing, which is a possibility, considering that the half-life of surface-expressed IgLON5 is over 3 days in vitro and could be higher in neurons.^57-60 We computationally predicted binding of these modified forms of IgLON5 to HLA molecules and subsequently narrowed down 124 15-mer peptides that we further tested using in vitro competition binding assay.

In general, we observed higher binding of peptides in their modified versus native forms (Fig. 3C and Supplementary Fig. 7A). We identified a handful of promising binders (e.g. IgLON5_91, IgLON5_178; Fig. 3C) that we titrated at different concentrations (Supplementary Fig. 7B and C). Higher binding at increasing concentrations suggests that the candidate peptide has a sound affinity for the HLA molecule, indicating their potential as strong binders with favourable HLA molecule interactions and immunogenicity profiles.

Notably, we found several peptides that bound disease-associated HLA-DQ5 monomers preferably when asparagine residues (N) N₆₇ [Fig. 3D(i) and E(i)], N₁₃₇ [Fig. 3D(ii) and E(ii)] or N₁₄₉ [Fig. 3D(iii–v) and E(iii–v)] were substituted for aspartic acid (D). N to D substitution mimics deamidation of key N-glycosylation sites,⁶¹ a PTM that is a known contributor to reducing self-tolerance and priming peptides for autoimmune T-cell recognition.^62,63 While HLA binding affinity of several initial candidate peptides was not confirmed by further titration experiments, such as IgLON5_91 [Fig. 3E(i, iii and v) and Supplementary Fig. 7], the two deamidated peptides ¹²⁵VYLIVHVPARIVNIS¹³⁹ N > D₁₃₇ [Fig. 3E(ii)] and ¹⁴⁵NEGGNVNLLCLAVGR¹⁵⁹ N > D₁₄₉ [Fig. 3E(iv)] showed a dose-dependent increase in binding, whereas their native or glycosylated forms did not (Supplementary Fig. 7C), thus confirming their strong HLA affinity and relevance as strong binders.

Moreover, comparison of the binding motifs of risk-associated HLA-DQ5 molecules showed that all three HLA-DQs bind D with high-affinity in pocket six of the 9-mer binding core (Supplementary Fig. 4B and C), thus further substantiating that deamidation may be key to making these peptides immunogenic.

Taken together, these results suggest that deamidation of specific asparagine sites within the IgLON5-Ig2 domain might prime specific IgLON5 peptides for HLA class II presentation through risk-associated HLA-DQ5 molecules.

Increased CD4⁺ T cell reactivity to a deamidated IgLON5-Ig2-derived peptide

To test whether deamidation may be a key modification to evoke a functional autoimmune response, we stimulated PBMCs of two HLA-DQA1*01:05-DQB1*05:01-carrying patients and one HLA-matched control with an Ig2-derived 15-mer peptide, ¹²⁵VYLIVHVPARIVDIS¹³⁹, mimicking deamidation at N₁₃₇ for 10 days, prior to isolating antigen-specific CD4⁺ T cells using spheromer⁵¹ (Fig. 4A). Notably, the low frequency of HLA-DQA1*01:05-DQB1*05:01 in the general population (∼2%) substantially challenged the screening for suitable HLA-matched control PBMCs. Both patients showed elevated CD4⁺ T-cell reactivity compared with the control, although this was more pronounced in Patient 2 (Fig. 4B). Antigen-specific CD4⁺ T cells were sorted and sequenced to investigate cellular phenotypes. While there was some admixture, patient cells formed a distinct cluster relative to control cells (Fig. 4C), suggesting similarities in affinities and possible clonality. Delineation into distinct T-cell transcriptional states, namely naive (T_naive), effector (T_em), highly effector (T_emra) and regulatory (T_reg) T-cell subsets [Fig. 4D(i)], underscored the expression of T_emra markers in the control versus patient population, while elevated activation and proliferation [Fig. 4D(ii)] in the latter population further supported the suggestion that antigen-specific T cells are more highly differentiated in patients. Furthermore, T-cell receptors (TCR) were sequenced and TRAV, TRAJ, TRBV and TRBJ gene segments studied (Fig. 4E); these make up the TCR alpha and beta chains and play a crucial role in the diversity and specificity of HLA-peptide-TCR interactions. In doing so, we identified several segments, especially in the TCR alpha chain, that were unique to patients [Fig. 4E(i and ii)]. Looking at pairings of TRAV, TRAJ, TRBV and TRBJ (Fig. 3F), we corroborated the finding of individual pairings in patients; however, several combinations were also found to be unique to the control.

Taken together, these preliminary findings suggested that reactivity against deamidated Ig2-derived IgLON5 may be elevated in anti-IgLON5 disease subjects and that unique TCR signatures associated with a more effector T-cell phenotype can be found.

Discussion

This study is the largest reported HLA-association analysis in anti-IgLON5 disease. Notably, it benefits from the inclusion of patients from multiple different ethnicities and countries and rigorous matching of controls by PCA.³⁹ In this work, we were able to confirm and refine the association of anti-IgLON5 disease with HLA-DQ over HLA-DR.^2,16 Interestingly, HLA-DQ5-dosage effects modulated both disease predisposition (haplotype carrier frequency differences) and age at disease onset, with a protective effect of HLA-DQ6, possibly through allele competition with HLA-DQ5. Future studies investigating the effects of these combinations on HLA-DQ5 expression and how trans-dimerization may affect HLA-peptide binding^64,65 will be needed to further substantiate these observations.

A ranked association with three HLA-DQ5 haplotypes, HLA-DQA1*01:05-05:01, HLA-DQA1*01:01-05:01 and HLA-DQA1*01:04-05:03, in order of descending risk was found in 85% (74/87) of all patients and further reflected in a delayed age at disease onset. Altogether, this strongly supports HLA-DQ, and not HLA-DR, as the actual determinant for disease risk. Disease risk in the remaining patients could not be explained by HLA; however, the fact that non-HLA-DQ5 carriers have a significantly later age at disease onset suggests that there might be a distinct trigger. While a previous report suggested more dominant presentation of sleep or bulbar symptoms in HLA-DRB1*10:01-carriers,¹⁶ we were not able to confirm this observation, nor decipher any other significant associations between HLA-haplotype and major clinical phenotypes. Interestingly, a retrospective analysis of clinical, polysomnographic and HLA data⁶⁶ showed a high prevalence of HLA-DQB1*05:01 in different types of NREM parasomnias, thus suggesting that stratification of sleep phenotypes may occur at the level of HLA-DQB1*05, as opposed to distinct HLA-DRB1∼DQA1∼DQB1 haplotypes. However, a recent report⁶⁷ showed that a progressive supranuclear palsy-like phenotype was underrepresented in HLA-DRB1*10:01-carriers, while carrier status did not affect the development of movement disorders more generally, suggesting that stratification of clinical phenotypes by HLA-carrier status may only occur at the level of specific clinical features. Future clinical reports should thus focus on a more fine-grained analysis of HLA association with detailed clinical phenotypes.

We did not observe that native IgLON5 peptides strongly bound HLA-DQ5. Rather, in line with observations from other autoimmune diseases such as type-1 diabetes mellitus or rheumatoid arthritis,³⁰ HLA-DQ5 was found to bind PTM IgLON5 peptides containing aspartic acid residues in several positions. In line with this observation, it is worth noting that HLA-DQA1*01:05/DQB1*05:01, HLA-DQA1*01:01/DQB1*05:01 and HLA-DQA1*01:04/DB1*05:03 all bind D in pocket six with high affinity, although less pre-eminently so for HLA-DQA1*01:04/DQB1*05:03 (Supplementary Fig. 4B). Deamidation that substitutes N for D can occur as proteins age and are targeted for degradation, and it has thus been suggested that this modification may serve as an age-dependent molecular clock,^57,61 an intriguing concept considering the late age of onset of the disease and slow turnover of the IgLON5 protein as has been shown by Heo et al.⁵⁹ (refer to Supplementary Table 2 in the aforementioned work). While deamidation of asparagine was previously shown to alter both HLA class II antigen presentation,⁶³ as well as T-cell recognition,⁶⁸ the physiological relevance of this modification in anti-IgLON5 disease remains enigmatic. Our hypothesis is that deamidations occur in vivo as the result of simple chemical deamidation,⁶¹ enzymatic deamidation or following de-glycosylation of N-linked glycopeptides at the corresponding residues,⁶⁹ although this latter mechanism was primarily shown for HLA-class I presentation of intracellularly processed peptides.⁷⁰ Notably, deficiency of the enzyme NGLY1, which catalyses the removal of N-linked glycans and plays a key role in endoplasmic reticulum-associated degradation, was recently discovered to be the cause of a severe, multi-symptomatic disorder that also involves a complex movement disorder,^71,72 while abnormalities in the NGLY1–NRF1 pathway have been associated with immune dysregulation.⁷³ Whether abnormalities in enzymatic N-deglycosylation could play a role in this process may therefore also be considered. Overall, a considerable shortcoming in the field of anti-IgLON5 disease research is the concurrent lack of availability of mass spectrometric or comparable proteomic data needed to confirm PTM changes speculated to occur in this condition beyond computational predictions. The significance of the PTM modifications suggested here, and their relevance and importance in triggering the disease, therefore need to be confirmed in vivo.⁷⁴

Of note, it has previously been suggested that the IgLON5-derived peptides ¹⁰LRLLAAAAL¹⁹ and ¹²⁷IVHVPARIV¹³⁶ constitute potential causative immunogenic cores, due to their high computationally predicted binding affinity to HLA-DRA1*01:01-DRB1*10:01 and HLA-DRA1*01:01-DRB1*01:01.¹⁶ Testing these peptides in vitro, however, we could not confirm binding of either peptide to HLA-DRA1*01:01-DRB1*10:01. Altogether, the fact that HLA-DRA1*01:01-DRB1*10:01 neither accounts for genetic risk in most patients nor shows binding of the proposed peptides, does not favour a role for HLA-DR in the disease. Furthermore, these results highlight the limitations of in silico predictions⁷⁵ and stress the importance of in vitro binding assays to substantiate any autoimmune HLA-peptide interactions.

The deamidated peptide we identified as HLA-DQ5-binder lies within the Ig2 domain of IgLON5. Interestingly, the Ig2 domain was previously shown to be the key target site of patient autoantibody recognition⁵⁶ and treatment with bacterial-derived PNGase F (N-glycosidase F, an endoglycosidic enzyme that deamidates N-glycosylated residues to aspartic acid through removal of oligosaccharides from glycoproteins) maintained autoantibody binding to IgLON5-expressing cells.⁷⁶ The binding of deamidated peptides to the risk-associated HLA-molecules we have shown here therefore corresponds with previously reported autoantibody target sites both in terms of location along the IgLON5 protein and the state of PTM. This suggests that B-cell and T-cell epitopes involved in the diseases may overlap or be in close proximity as is sometimes observed following epitope spreading.⁷⁷ T–B-cell interactions are a critical component of the pathophysiological trajectory in most autoimmune diseases^21,78 and the study at hand lays an important foundation for further investigation of this important interaction (Supplementary Fig. 8).

In the current study, we showed elevated T-cell reactivity against a single modified epitope, ¹²⁵VYLIVHVPARIVDIS¹³⁹ in two patients, compared with a single HLA-matched control. We used spheromers, state-of-the-art multimers that allow more efficient T-cell staining and improved TCR-binding properties,⁵¹ enabling the detection of low-affinity T cells that may be more disease relevant in autoimmunity.^79,80 While the T-cell data we present suggest functional relevance of ¹²⁵VYLIVHVPARIVDIS¹³⁹ in shaping an early autoimmune response in anti-IgLON5 disease driven by T cells in a late effectorness state,⁸¹ these data should be considered preliminary, subject to the small sample size. Randomized screening for the identification of matched HLA-control PBMCs proved challenging in the concurrent study, given the low frequency (∼2%) of HLA-DQA1*01:05-DQB1*05:01 in the general population. Future studies testing additional subjects and epitopes will be needed to confirm the relevance of ¹²⁵VYLIVHVPARIVDIS¹³⁹ and other epitopes in shaping anti-IgLON5 autoimmunity, as well as to define phenotypic signatures of disease-relevant, antigen-specific T cells. Similarly, the TCR sequencing data we have shown here is suggestive of unique TCRs in patients versus controls and motivates further assessment of TCR diversity, both in more patient and control subjects but also using different autoantigen candidate peptides and carriers of different HLA-DQ5 subtypes. Whether reactivity against the same epitope is shared across carriers of distinct HLA-DQ5 molecules, as suggested by the shared HLA-peptide binding properties shown in our current study, will require further examination. We hypothesize that future studies focusing on T-cell reactivity against multiple IgLON5-derived epitopes in distinct HLA-DQ5 carriers may explain the ranked-HLA-DQ5 disease risk, which the current study promotes on the basis of genetic, demographic and molecular evidence.

Taken together, this study has important pathophysiological and clinical implications for our understanding of anti-IgLON5 disease. First, it shows for the first time that anti-IgLON5 disease is primarily HLA-DQ associated, and that T-cell autoimmunity directed towards deamidated IgLON5 sequences presented by these molecules may be involved. Second, a delay in the age at disease onset between different HLA-DQ5 subtypes implies clinical differences secondary to peptide presentation by slightly different HLA-DQ heterodimers; thus, future studies should further investigate symptomatic and disease outcome distinctions between different HLA-DQ5 subtypes. Future studies incorporating a larger number of patients, a difficult endeavour in a rare disease, will be needed to further elucidate the genetic risk of this condition also at a genome-wide level. Concurrent to that, investigations of immune cell reactivity and phenotypes will be required to elucidate key pathophysiological events, aimed at disentangling the intricate interplay of autoimmunity and neurodegeneration within, and beyond, this condition. Ultimately, this understanding will pave the way to developing more effective therapeutics (e.g. targeting autoantigen-specific CD4⁺ T cells), promising to halt irreversible neurodegenerative mechanisms upstream in the pathology, thus instituting vital treatment strategies that currently remain amiss for this condition.

Data availability

Raw data are available on request to the corresponding author.

Acknowledgements

We thank Christopher Garcia (Stanford University) for providing insect cell lines, Gonzalo Montero-Martin and Marcelo Fernandez-Viña for HLA-validation typing, Daniel Fernandez for technical support during AKTA-FPLC and both the Stanford FACS and HIMC facilities for support with T cell isolation and sequencing, respectively. We also would like to thank all collaborators and doctors, who referred patients to this study, beyond those listed as co-authors: Olivier Flabeau (Bayonne, France), Eve Chanson (Clermont Ferrand, France), Mathilde Compoint (Dieppe, France), Tifanie Alberto (Lille, France), Julie Boucher (Lille, France), Isabelle Francillard (Lorient, France), Emilien Bernard (Lyon, France), Virginie Desestret (Lyon, France), Philippe Damier (Nantes, France), Nicolas Capet (Nice, France), Caroline Giordana (Nice, France), Dimitri Renard (Nîmes, France), Aurélie Meneret (Paris, France), Foucaud Du Boisgueheneuc (Poitiers, France), Maximilien Moulin (Reims, France), Flora Delamaire (Rennes, France), Marie Rafiq (Toulouse, France), Gilles Ryckewaert (Valenciennes, France).

Funding

This study was primarily funded by a seed grant awarded by The Stanford Autoimmune & Allergy Supergroup (SAAS) to E.M. and National Institutes of Health grant NIH-5U01NS120885-02. S.M.Y. received funding from the Einstein Center for Neurosciences Berlin (Einstein Foundation Berlin) and the German-Academic Exchange Service (DAAD). The French collection of samples is developed within the BETPSY project, which is supported by a public grant overseen by the French National Research Agency (ANR), as part of the second ‘Investissements d’Avenir’ program (reference ANR-18-RHUS-0012) and J.H. is supported by the European Reference Networks, the Research and Innovative Technology Administration and LabEx CORTEX (ANR-11-LABX-0042). C.F. is supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), grant numbers FI 2309/1-1 and FI 2309/2-1; and the Federal Ministry of Education and Research (Bundesministerium für Bildung und Forschung), grant number 01GM1908D (CONNECT-GENERATE). S.I. is supported by a senior clinical fellowship from the Medical Research Council (MR/V007173/1), Wellcome Trust Fellowship (104079/Z/14/Z), British Medical Association Research Grants- Vera Down grant (2013) and Margaret Temple (2017), Epilepsy Research UK (P1201), the US-UK Fulbright Commission (MS-Society research award) and by the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC). For the purpose of Open Access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript (AAM) version arising from this submission. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. L.S. is funded in part by Instituto de Salud Carlos III through the project FIS21/00165 (L.S., C.G.) (co-funded by the European Regional Development Fund ‘Investing in your future’). We thank the European Joint Programme on Rare Diseases (EJPRD) for funding a Networking event (contract number 463001015) on IgLON5 disease from the European Union’s Horizon 2020 research and innovation programme under the EJP RD COFUND-EJP N°825575. B.S. received funding through the Betty and David Koetser Foundation, Neuromuscular Research Association Basel. T.S. (22K07492) and A.K. (JP21K07457) received Grants-in-Aid for Scientific Research (C) from the Japan Society for the Promotion of Science. H.P. is supported by the German Research Foundation (DFG; grants FOR3004, PR1274/3-1, PR1274/5-1, and PR1274/9-1), Helmholtz Association (HIL-A03 BaoBab), and Bundesministerium für Bildung und Forschung (German Federal Ministry of Education and Research) (Connect-Generate 01GM1908D). A.M. is supported by National Institutes of Health (NIH): RO1NS126227, U01NS12090. J.D. is supported by Caixa Research Health 2022 (HR22-00221). This study was funded also in part by Instituto de Salud Carlos III (ISCIII)—Fondo de Investigaciones Sanitaria (FIS) through the project ‘PI21/00165’ (L.S. and C.G.) and the Networking Support Scheme funded from the European Union’s Horizon 2020 research and innovation programme under the EJP RD COFUND-EJP No. 825575.

Competing interests

S.R.I. has received honoraria/research support from UCB, Immunovant, MedImmun, Roche, Janssen, Cerebral therapeutics, ADC therapeutics, Brain, CSL Behring and ONO Pharma; and receives licensed royalties on patent application WO/2010/046716 entitled ‘Neurological Autoimmune Disorders’; and has filed two other patents entitled ‘Diagnostic method and therapy’ (WO2019211633 and US-2021-0071249-A1; PCT application WO202189788A1) and ‘Biomarkers’ (PCT/GB2022/050614 and WO202189788A1). S.M. received speaker honoraria from Biogen, Sanofy and Novartis. C.J.H. has been serving as a consultant for Univar and has received honoraria for lecturing and travel expenses/speaking honoraria from Abbott and Alexion and research support from Abbott. A.M. patents issued for GFAP and MAP1B-IgGs and patents pending for PDE10A, Septins-5 and -7 and KLCHL11-IgGs, and has consulted for Janssen and Roche pharmaceuticals, without personal compensation. J.D. holds a patent for the use of IgLON5 antibody testing and both he and F.G. receive royalties from Euroimmun for the clinical use of this test. The other authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.

References