Distinct Epitopes Are Associated With Clinical Phenotypes in Autoimmune Nodopathies With Anti-Contactin1 Autoantibodies

Abstract

Background and Objectives

Autoimmune nodopathy with anti-contactin1 (CNTN1) autoantibodies is a rare sensory-motor neuropathy characterized by subacute-onset sensory ataxia and variable disease courses. Comorbidities such as glomerulonephritis and diabetes mellitus are observed in some patients. The diversity in clinical presentation may reflect differences in the targeted CNTN1 epitopes.

Methods

To investigate the relationship between clinical features and the underlying epitopes, we analyzed serum samples from 16 anti-CNTN1–positive patients. Binding epitopes were assessed using cell-based assays with recombinant CNTN1 variants and peptide microarrays.

Results

Epitope mapping in 16 patients revealed that 13 harbored antibodies targeting CNTN1 immunoglobulin (Ig) domains while 3 recognized fibronectin (Fn) domains. Glomerulonephritis was exclusively observed in patients with autoantibodies binding to the Ig domain, whereas diabetes mellitus and a chronic course of disease were more prevalent in patients with binding to the fibronectin domain.

Discussion

Our findings highlight the heterogeneity of anti-contactin1 nodopathies, suggesting that distinct epitope-binding patterns may underlie different clinical manifestations and may be associated with different pathogenic mechanisms and comorbidities.

Introduction

Contactin1 (CNTN1) is a major component of the paranodal protein complex that constitutes the axoglial junction at the nodes of Ranvier.¹ Autoantibodies directed against CNTN1 are found in a subgroup of patients with autoimmune nodopathies and are associated with distinct clinical symptoms: severe sensorimotor peripheral neuropathy with (sub)acute onset and sensory ataxia can be found in all patients,^2,3 and other typical symptoms such as tremor, neuropathic pain, and glomerulonephritis only occur in some of the patients.^4-6 Diabetes mellitus has been reported to be associated with anti-CNTN1–mediated autoimmune nodopathy but is only found in a subgroup of patients.⁷ The course of disease also varies: chronic seropositivity and persistent symptoms over years have been described, as well as monophasic forms with complete clinical and serologic remission.^8,9 The detailed description of larger cohorts of anti-CNTN1–positive patients in the past few years raises the suspicion that the pathogenicity of anti-CNTN1 autoantibodies may differ among individual patients. Most anti-CNTN1 autoantibodies belong to the IgG4 subclass, but a predominance of IgG3 and variable levels of additional autoantibodies of the IgG1-3 subclasses have been reported and may affect the clinical phenotype.^9,10 Recent studies have reported variable capabilities of anti-CNTN1 autoantibodies to access their target, most probably because of different capabilities to pass through the myelin barrier.^10,11 Because interindividual differences in anti-CNTN1 autoantibodies cannot be explained solely by IgG subclasses, binding to distinct domains of the CNTN1 protein may also account for the phenotypes. CNTN1 consists of 6 immunoglobulin (Ig) domains and 4 fibronectin-III (FnIII) domains.^12,13 Indeed, several binding regions have been described: Labasque et al. reported epitopes at N-linked glycans of the Ig domain, thus requiring glycosylation as a prerequisite for autoantibody binding,¹⁴ whereas others¹⁵ could not confirm this finding in their cohort but found binding to the Ig domain of the protein core.

We, therefore, hypothesized that anti-CNTN1–mediated neuropathy represents a heterogeneous disease, in which distinct epitopes are linked to specific clinical phenotypes. Understanding these epitope-phenotype relationships may help predict disease course and guide treatment decisions.

Methods

Patients

Serum samples or plasma exchange material from 16 patients who had been positively tested for anti-CNTN1 autoantibodies as previously described^6,9 and who presented with a typical clinical phenotype of autoimmune nodopathy were included in the study. Serum samples from 5 healthy individuals were used as controls in the assays.

Standard Protocol Approvals, Registrations, and Patient Consents

All participants provided informed consent to use their samples in the study, and the study was approved by the Ethics Committee of the University of Würzburg. Demographic data of the patients are summarized in Table 1.

Table 1.

Demographic Data, Epitopes, and Clinical Data of Individual Patients

Patient ID	Epitope CBA/PM	Titer, IgG subclass	Diabetes m	Glomerulonephritis, proteinuria	Cranial nerve involvement	Mechanical ventilation	Course of disease	Follow-up serum	Treatment
1	FnIII/-	1:19,000, IgG4>IgG2	Yes	No	Yes	Yes (during pneumonia)	Chronic	1.5 mo, positive	IA, IVIG, rtx
2	FnIII/-	1:1,000, IgG4>IgG1	Yes	No	No	No	Monophasic	4 mo, negative	Steroids
3	FnIII/-	1:1,000, IgG4>IgG2	Yes	No	At onset	No	Chronic	15 y, positive	PE, steroids, cyclophosphamide, rtx
4	Ig1/-	1:7,500, IgG3>IgG4	No	Proteinuria	Yes	Yes	Died	n/a	PE, IVIG, steroids
5	Ig2/Ig6	1:2,000; IgG4>IgG3	Prediabetes	Glomerulonephritis	Yes	Yes	Chronic	5 y, negative	PE, IVIG, steroids, rtx
6	Ig1/Ig1, Ig1, Ig3	1:5,000; IgG4>IgG2>IgG1	No	Proteinuria	Yes	Yes	Monophasic	3 y, negative	IA, IVIG, steroids, rtx
7	Ig5/Ig1, Ig1, Ig3	1:15,000; IgG3>IgG4>IgG1	No	Proteinuria	Yes	Yes	Monophasic	2 y, negative	PE, IVIG, steroids, rtx
8	Ig1/Ig2	1:12,000; IgG4	No	Proteinuria	Yes	No	Monophasic	n/a	PE, IVIG, rtx
9	Ig1/-	1:500, IgG4	No	No	No	No	Monophasic	2 y, negative	PE, IVIG, rtx
10	Ig1/-	1:1,000, IgG4>IgG2>IgG1	No	Glomerulonephritis	No	No	Died	n/a	PE, IVIG, rtx
11	Ig1/-	1:4,000, IgG4>IgG2	No	No	No	No	Monophasic	9 mo, negative	PE, IVIG, rtx
12	Ig1/-	1:30,000, IgG4>IgG3	No	Glomerulonephritis	At onset	No	Monophasic	10 mo, negative	PE, IVIG, rtx
13	Ig1a/-	1:200, IgG2	No	Glomerulonephritis	At onset	No	Monophasic	1 y, negative	PE, IVIG, rtx
14	Ig1/-	1:300,000, IgG4>IgG3>IgG1>IgG2	No	Proteinuria	No	No	Monophasic	n/a	PE, IVIG, rtx
15	Ig1/-	1:50,000, IgG4>IgG3>IgG1	No	Proteinuria	Yes	Yes	Improvement	7 mo, positive	PE, IVIG, rtx
16	Ig1/-	1:50,000, IgG4>IgG1	No	Proteinuria	No	No	Chronic (5 mo of follow-up)	5 mo, positive	PE, IVIG, steroids, rtx, bortezomib

Abbreviations: IA = immune adsorption, IVIG = IV immunoglobulin, PE = plasma exchange, rtx = rituximab

^aNot to deglycosylated CNTN1.

Plasmids

Rat CNTN1 DNA¹⁶ was cloned and inserted into a pRK5 plasmid (provided by P. Seeburg, MPI for Medical Research, Heidelberg, Germany) for mutagenesis of deglycosylated variants. For mutagenesis of deletion constructs, rat CNTN1 DNA was cloned and inserted into a pEGFP-N1 vector (Clontech Laboratories, Mountain View, CA) to verify the expression of the protein by coexpression of GFP.

PCR Mutagenesis

All primers (Invitrogen, Karlsruhe, Germany) are listed in eTable 1. To disrupt specific N-linked glycosylation sites within the Ig5 domain of CNTN1, PCR mutagenesis was used. Given that N-linked glycosylation typically occurs at asparagine residues (Asn-X-Ser/Thr), we strategically altered the DNA sequence to substitute asparagine with glutamine in the polypeptide chain. Glutamine, being chemically akin to asparagine, is unlikely to affect the overall protein structure. Site-directed mutagenesis was executed using custom-designed forward and reverse primers in a two-step PCR approach. Initially, mutant primers were used in separate PCRs to generate DNA fragments with overlapping ends. Subsequently, an overlap extension PCR was conducted, where the products from both PCRs were annealed. In this way, we introduced mutations for eliminating the N-linked glycosylation sites at positions 457 (N457Q), 473 (N473Q), and 494 (N494Q). Double mutations N457,473Q; N457,494Q; and N473,494Q as well as triple mutation N457, 473,494Q were generated sequentially. Mutagenesis of deletion constructs (CNTN1ΔIg1, CNTN1ΔIg1,2, CNTN1ΔIg1-3, CNTN1ΔIg1-4, CNTN1ΔIg1-5, and CNTN1ΔIg1-6) was conducted in the same way as for deglycosylation constructs. Sanger sequencing (Eurofins Genomics, Ebersberg, Germany) was performed using appropriate primers to ensure the accuracy of the modified constructs.

Plasmid DNA Preparation

Plasmid DNA was transformed into competent NEB5a E.coli cells (New England Biolabs, Frankfurt, Germany) and isolated after overnight growth at 37°C using the NucleoBond Xtra Maxi kit (Machery & Nagel, Düren, Germany) according to the manufacturer's instructions.

Cell Lines and Transfection of HEK293 Cells

HEK293 cells (human embryonic kidney cells; CRL-1573; ATCC—Global Bioresource Center, Manassas, VA) were grown in minimum essential medium (Life Technologies) supplemented with 10% fetal bovine serum, 2 mM l-glutamine, 100 U/mL penicillin, and 100 µg/mL streptomycin and were transiently transfected using a modified calcium-phosphate precipitation method. A total of 200,000 cells were seeded on glass coverslips in 35-mm culture dishes or 1.5 × 10⁶ cells per 10-cm dish. Transfection was performed 24 hours after seeding. For 35-mm and 10-cm culture dishes, CNTN1 wild-type, glycosylation, or deletion variants were transfected with 1 µg of each plasmid for 35-mm dishes and 10 µg for 10-cm dishes. The plasmid DNA was supplied with 2.5 mM CaCl₂, 2× HBS buffer, and 0.1× TE buffer. It was mixed and then incubated for 20 minutes at room temperature.

Immunocytochemical Staining

Transiently transfected HEK293 cells were used for binding assays with patients' serum samples or PE material. Initially, nonspecific binding sites were blocked for 20 minutes using 10% BSA/PBS at 37°C and 5% CO₂. Subsequently, cells were incubated with patients' material, control serum samples, or anti-CNTN1 (R&D systems, Minneapolis, MN) at 1:250 dilution in 2% BSA/PBS for 1 hour at 37°C and 5% CO₂. After incubation, cells were fixed with 4% PFA for 20 minutes. Appropriate secondary antibodies (donkey anti-goat Cy3, donkey anti-human Cy3; Jackson ImmunoResearch, West Grove, PA) were then applied at 1:250 dilution in 2% BSA/PBS.

Peptide-N-Glycosidase F Digestion

Lysates of HEK293 cells transiently transfected with CNTN1 were prepared using the CytoBuster™ Protein Extraction Reagent (Merck, Darmstadt, Germany). N-linked glycans were cleaved from the protein by peptide-N-glycosidase F (PNGase F) (New England Biolabs, Frankfurt, Germany) digestion according to the manufacturer's instruction. Binding of patients' serum samples to the deglycosylated protein was analyzed using Western blot.

Biotinylation Assay

Transiently transfected HEK293 cells coexpressing CNTN1 wild-type and deletion variants fused to GFP were used. The surface proteins were labeled by incubating the cells (10-cm dish) with 1 mg/mL of EZ-Link Sulfo-NHS-LC-Biotin (sulfosuccinimidyl-6-(biotin-amido)-hexanoate) (Pierce Biotechnologies, Rockford, IL) for 20 minutes, followed by incubation with quenching buffer (192 mM glycine, 25 mM Tris in PBS, pH 8.0) for 10 minutes. Cells were detached using cold PBS buffer, followed by a centrifugation step of 10 minutes at 1,000g. The cell pellet was lysed in TBS (Tris-buffered saline) with 1% Triton X-100 and a protease inhibitor mixture tablet (Roche Diagnostics, Mannheim, Germany) and centrifuged for 1 minute at 13,000 g. Next, the whole protein fraction (supernatant) was incubated with streptavidin-agarose beads (Pierce Biotechnologies, Rockford, IL) for 2 hours at 4°C under constant rotation. The supernatant was then removed, and the beads (surface fraction) were washed in TBS buffer. For the elution of biotinylated proteins, they were boiled with 50 μL of 2× SDS buffer for 5 minutes at 95°C. 20 μg of the whole-cell protein fraction and 20 µL of the surface protein fraction were analyzed by Western blot.

SDS-PAGE and Western Blot

Protein samples were separated by SDS-PAGE using 9% (w/v) gels. Samples were then transferred onto a nitrocellulose membrane (GE Healthcare, Little Chalfont, UK), followed by blocking for 1 hour with 5% milk powder (for CNTN1) or 5% BSA (for pan-cad and β-actin) in TBS-T (TBS with 1% v/v Tween 20); membranes were incubated overnight at 4°C with patient or control serum (1:250 or 1:500) or primary antibodies [anti-CNTN1 (R&D Systems, Minneapolis, MN, AF904, 1:400 or 1:500), pan-cadherin (Cell Signaling Technology, Danvers, MA, 4068S, 1:1,000), or β-actin (GeneTex - Biozol, Eching, Germany, GTX26276, 1:5,000)]. Proteins were visualized using horseradish peroxidase (Dianova, Hamburg, Germany) and detected through chemiluminescence using the Clarity Western ECL substrate (Clarity Western Peroxide Reagent, BioRad 170-5061, California, US).

Image Analysis

Immunofluorescence images were taken using an Olympus Fluoview ix1000 microscope with an UPLSAPO 60× oil objective and diode lasers at 405 nm, 495 nm, and 550 nm, with 1,024 × 1,024 pixels. Binding assays were analyzed using a fluorescence microscope (Axio ImagerM.2 with Apotome) equipped with Colibri 7 LED and an Axiocam 506 mono CCD camera (all: Zeiss, Oberkochen, Germany).

Peptide Microarray Synthesis

The human CNTN1 sequence (UniProtKB: Q12860) was displayed in a microarray format as a 20-mer peptide (5-aa offset), 15-mer overlapping peptide (1-aa offset), and 20-mer triple nonoverlapping alanine mutants. Peptide arrays were synthesized as previously described¹⁷ via a MultiPep RSi robot (CEM GmbH, Germany) on cellulose discs containing 9-fluorenylmethyloxycarbonyl-β-alanine (Fmoc-β-Ala) linkers.

LC-MS was performed using amyloid-beta quality controls that were cleaved from the solid support following a recently described protocol.¹⁸ 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 room temperature. For analysis, the amyloid-beta quality controls were diluted 1:3 and sent to LC-MS (Agilent Technologies, Santa Clara, CA).¹⁸

Peptide Microarray Immunoassay

PCCs were mixed 2:1 with saline–sodium citrate buffer after transfer to a 384-well plate. A SlideSpotter (CEM GmbH) was used to transfer the PCC solutions onto white-coated CelluSpot blank slides (76 × 26 mm, Intavis AG Peptide Services GmbH and Co. KG). After the printing procedure, the slides were left to dry for a minimum of 3 hours. The CNTN1 slides were blocked for 60 minutes in 5% (w/v) skim milk powder (Carl Roth) and 0.05% Tween 20 phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, and 1.8 mM KH2PO4; pH 7.4). After blocking, the slides were incubated for 30 minutes with blocking buffer and human CNTN1 (15 patients) or control (5) serum. With the exception of patient 8 and their follow-up material, which was tested at 1:250 dilution (with 0.1% Tween 20), all other positive samples presented optimal signals over noise at a ratio of 1:500. The samples were then washed 3× with 0.05% Tween 20 in PBS for 1 minute. IgG antibodies were detected via a goat anti-human IgG (H + L) secondary antibody, HRP (TF: 31,410). For the detection of chemiluminescence, an Azure c400 imaging system (lowest sensitivity) and SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific GmbH, Germany) were used. The raw grayscale intensities were measured via MARTin software.¹⁹ The data were analyzed via Origin Pro version 2023b (OriginLab). The signal intensities were averaged (n = 3) and normalized to the flag peptide background using z-scores, and related standard deviations were calculated. Hits with z-scores ≥4 in the 20-mer library were considered of interest and included in follow-up validation via alanine mutants or 15-mer peptides.

Statistical Analysis

For Western blot quantification, Fiji/ImageJ software was used.²⁰ Data were further analyzed using GraphPad Prism Software (version 10.5.0 for Windows, GraphPad Software, Boston, MA) and are presented as mean ± SEM (standard error of the mean). The number of experiments is displayed in the figure legend. Statistical significance was calculated using the one-way ANOVA, followed by the Dunnett test for multiple comparisons. Significance is indicated as *p < 0.05, **p < 0.01, and ***p < 0.001. Differences in the number of positive patients between groups were compared using the Fisher exact test.

Data Availability

Anonymized data not published within this article will be made available by request from any qualified investigator.

Results

Glycosylation of CNTN1 Is Not Necessary for Autoantibody Binding In Vitro

Binding of anti-CNTN1–positive serum samples to HEK293 cells transfected with the unmodified CNTN1 wild-type DNA was confirmed for all patient samples, whereas all healthy controls were negative (Figure 1A). Glycosylation of the CNTN1 Ig5 domain has been proposed as a prerequisite for autoantibody binding.¹⁴ To assess the effect of glycosylation on autoantibody binding, additional experiments were performed using deglycosylated mutants of the CNTN1 Ig5 domain with all serum samples. One serum sample did not bind to any of the deglycosylated mutants, whereas all other serum samples showed clear binding to all tested deglycosylated mutants (Figure 1B). For 5 patient serum samples (patients 1, 3, 4, 5, and 9) and one healthy control, an additional Western blot was performed. Two serum samples (patients 1 and 3) were positive by Western blot, and PNGase treatment was conducted and confirmed binding to the deglycosylated protein core (Figure 1C). Thus, CNTN1 glycosylation was not found to be necessary for autoantibody binding in most patients.

Figure 1. Binding of Serum Samples to Deglycosylated CNTN1 Mutants.

All patient serum samples bind to transfected HEK293 cells expressing full-length CNTN1 wild-type mutants (A). One serum (Pat 13) did not bind to deglycosylated CNTN1 mutants, and all other serum samples did as exemplarily shown for patient 5 (B). Western blot with and without PNGase treatment confirmed binding to glycosylated and deglycosylated CNTN1 (C). Scale bars = 50 µm.

Anti-CNTN1–Positive Serum Samples Bind to Epitopes Within the FnIII or Ig Domain

To gain further insight into the CNTN1 autoantibody–binding sites, we analyzed deletions of the Ig domains at the N-terminus of CNTN1, leaving the Fn domains unaffected. Deletion of domains bears the disadvantage of potentially impaired trafficking to the cellular surface. Therefore, all constructs were investigated for membrane expression. Using live-cell imaging, we found surface-expressed CNTN1 for all deletion mutants analyzed, as seen by the red ring around transfected HEK293 cells (Figure 2A). The surface expression was further confirmed by Western blot and by a biotinylation approach, labeling surface proteins and thus discriminating them from total expressed protein (Figure 2, B and C). Again, our quantitative analysis revealed cellular surface expression with no significant differences in the CNTN1 wild-type construct compared with all 6 deletion variants (Figure 2D).

Figure 2. CNTN1 Deletion Mutants Are Expressed at the Cellular Surface.

(A) Transfected HEK293 cells expressing full-length CNTN1 wild-type and deletion mutants of the extracellular Ig domains 1–6 (deletion of Ig1 domain corresponds to CNTN1ΔIg1; deletion of Ig domains 1 and 2 CNTN1ΔIg1,2; and further deletions CNTN1ΔIg1-3, CNTN1ΔIg1-4, CNTN1ΔIg1-5, and CNTN1ΔIg1-6; red). CNTN1 wild-type and deletion mutants were fused to GFP (cyan). (B) Representative Western blot of CNTN1 wild-type and deletion variants. Note, the decrease in molecular weight on increase of the deletion (white open arrowheads between 125 and 80 kDa). # marks the lower protein bands representing most likely the nonglycosylated isoforms. β-Actin was used as housekeeping protein (black arrowhead; 48 kDa). Dotted line marks cut of the nitrocellulose. (C) Quantification of the whole-cell protein fractions of CNTN1 variants normalized to β-actin after a biotinylation assay (n = 5); significance value **p < 0.01. (D) Quantification of the surface protein fractions of CNTN1 variants normalized to pan-cadherin (pan-cad, membrane marker) (n = 3). Dotted lines in C and D point to the expression level of CNTN1 wild-type mutant. Error bars represent SEM.

Further binding assays with anti-CNTN1–positive serum samples on HEK293 cells transfected using the truncated CNTN1 constructs revealed different binding sites: serum samples of 3 patients showed clear binding to all deletion mutants, even after deletion of the whole Ig domain, indicating epitopes within the FnIII domain (patients 1–3, Figure 3A). In one patient (patient 7), clear binding after deletion of Ig1 was found but only weak binding occurred in the deletion variants Ig1-2, Ig1-3, and Ig1-4, indicating an epitope in the Ig2 domain and possible further epitopes in the domains Ig3-5. In one patient (patient 5), clear binding was detectable after deletion of Ig1, but not with the other deletion constructs, suggesting an epitope in the Ig2 domain (Figure 3B, Table 1). Samples from eleven patients did not bind to any of the deletion mutants, indicating epitopes in the Ig1 domain or conformational or modified epitopes that are not constituted by the truncated CNTN1 variants (Figure 3C). Binding assays with untransfected HEK293 cells and control serum samples were performed in parallel and showed no specific reactivity (data not shown).

Figure 3. Binding of CNTN1-Positive and Control Serum Samples to the Fn or Ig Domain.

Serum of Pat 3 equally binds to HEK293 cells transfected with CNTN1 wild-type mutants as well as Ig1-3 or Ig1-6 deleted mutants indicating binding to the Fn domain (A). Binding of serum of Pat5 is lost after deletion of Ig1-2, but not after deletion of Ig1 (B), whereas serum of Pat 9 does not bind to HEK293 cells that are transfected with any deletion mutants (C). Scale bars = 50 µm.

Peptide Microarrays Identify Linear Epitopes in the Ig Domain of 4 Patients

To investigate the epitopes identified in the cell culture experiments in detail, autoantibody binding was analyzed using serum samples from 15 patients and 5 controls, with an initial CNTN1 20-mer library (Figure 4) in the microarray format. Clear binding to multiple linear epitopes in several Ig domains was observed in 4 patients (Table 1). Following stringent cutoffs (z-score ≥4), 8 epitopes were selected for further validation by alanine scanning of 15-mer sequences (eFigure 1). For the remaining 11 serum samples, no binding was detected in the peptide microarray, possibly because of conformation or modification requirements of the epitopes. Five control serum samples excluded nonspecific binding within this experimental setup (eFigures 1 and 2).

Figure 4. Binding of CNTN1-Positive Serum Samples to a 20-Mer Peptide Microarray.

Peptide microarray results showing unique epitope binding patterns for patient 5, patient 6, patient 7, and patient 8 at baseline and follow-up (A). Neighboring and partially overlapping epitopes are detected in patients 6 and 7 across different time points. (B) The structure of CNTN1 using AlphaFold2²¹ and rendered with PyMOL (Schrödinger Inc.), highlighting different Ig1, Ig2, Ig3, and Ig6 epitopes. Patient epitopes are assigned to their respective Ig domains. The data points are visualized as a heatmap by z-score normalization (min–max scale bar), with each individual sample normalized internally (n = 3 replicates). Close-up view of epitope in patient 8 at the interphase between CNTN1 in gray and NF155 in pink. Chataigner et al. identified residues F177 and F180 as critical for the heteromeric interaction, which notably overlaps with the epitope mapped in our study.

Patient 5 exhibited stable, unique binding to the epitope LDSNGE (562–567) in the Ig6 domain (Figure 4A). Patient 6 exhibited unique binding to the epitope STEATL (125–130) in the Ig1 domain (Figure 4B). Patient 7 displayed consistent binding to the epitope DIRWRKVLEPMP (272–283) in the Ig3 domain (Figure 4C). Patient 8 showed binding to EFPVFITMDK (176–185) in the Ig2 domain (Figure 4D).

None of these determined epitopes was localized in or close to the protein domain involved in CNTN1 glycosylation. The specificity of all identified epitopes was confirmed via either 15-mer or alanine scanning of serum samples against the corresponding target peptides. It is important to note that mutation of distinct core regions within the initial hits entirely abolished binding, thereby validating specific epitope binding and further identifying the amino acids that are critical mediators of autoantibody binding (eFigures 1 and 2).

Taking advantage of longitudinal serum samples collected at different stages of disease, we assessed how the presence and persistence of epitope-specific autoantibodies relate to the clinical course.

Of the 3 patients with binding to the Fn domain, follow-up serum samples were available for 2. In one patient (patient 3), reactivity to the Fn domain persisted over more than 6 years and was associated with a chronic disease course. The other patient (patient 2) showed clinical improvement after 4 months, and follow-up testing revealed complete loss of antibody binding by ELISA and HEK293 cell-based assay. The patient for whom no follow-up sample was available also experienced a chronic course of disease.

Among 13 patients with Ig domain–binding antibodies, follow-up samples were available for testing at our laboratory from 9; one patient's follow-up sample tested positive at a commercial diagnostic laboratory. Two patients died during the acute phase, and one improved after rituximab administration but lacked follow-up sampling. Of the 9 patients with available follow-up samples, only one remained positive by ELISA and cell-based assays after treatment, albeit at a lower titer and with partial clinical improvement. The remaining 8 showed no detectable antibodies in follow-up testing and had experienced a monophasic disease course, with marked improvement in 6 patients and persistent severe symptoms in 2.

Notably, 4 patients showed Ig domain epitopes readily detectable in the microarray format. In this study, follow-up samples collected 8 months to 5 years after baseline remained positive and revealed largely consistent epitope profiles in 3 patients (patients 5, 7, 8), as well as one additional epitope (“EGIYEC”, AA305–310, Ig3 domain) in patient 6. Binding was slightly reduced in 2 patients. Clinically, 2 patients had largely recovered except for residual symptoms while 2 continued to suffer from severe symptoms, possibly reflecting a pathogenic effect of persistent low-level autoantibodies detectable by short linear epitopes.

Different Epitopes Are Associated With Distinct Clinical Features

Next, we compared clinical symptoms of patients with autoantibodies against epitopes in different domains of CNTN1 with those of patients with positive peptide microarrays, suggesting linear epitopes and suspected conformational epitopes (not detectable by peptide microarrays) (Table 1). While all patients suffered from the key symptoms of an autoimmune nodopathy, namely severe (sub)acute-onset sensorimotor neuropathy and sensory ataxia, other symptoms seemed to be associated with certain types of epitopes.

All patients with an epitope in the FnIII domain (n = 3) suffered from diabetes mellitus, which has been reported to be a risk factor of anti-CNTN1 autoantibodies,⁷ whereas in the patients with an Ig epitope (n = 13), only a single patient had prediabetes (p = 0.007). Comparison of outcomes tended to differ but failed to reach significance within the limited cohort size: Two of the 3 patients with binding to the FnIII domain showed a chronic course of disease, whereas in the patients with an Ig epitope, autoantibodies detected by routine diagnostic testing only persisted in a single patient (analyzed 7 months after onset) and in 8 patients, no positive signals were detectable during the course of disease. Eight of 10 patients with available clinical follow-up data experienced remission of disease with residual symptoms.

Glomerulonephritis was observed exclusively in patients with autoantibody epitopes in the Ig domain (4 of 10). An additional 7 patients with Ig epitopes showed proteinuria without a formal diagnosis of glomerulonephritis. By contrast, neither glomerulonephritis nor proteinuria was detected in any patient with an epitope in the FnIII domain. Thus, glomerulonephritis/proteinuria was more frequently found in patients with epitopes in the Ig domain (p = 0.018).

Patients with positive signals in the peptide microarrays appeared to have a more severe disease course, independent of the autoantibody titer or subclass. Three of the 4 patients with detectable linear epitopes required mechanical ventilation during the acute phase, compared with only one patient among those without detectable peptide binding and none of the patients with epitopes restricted to the Fn domain (except for one case of transient ventilation due to pneumonia). This association reached statistical significance (p = 0.027), suggesting that linear epitope targeting may be linked to more severe clinical presentations. By contrast, age, elevated CSF protein, neuropathic pain, and tremor did not differ among patients with different epitope profiles (Table 1). Of note, the only patient who required glycosylation for antibody binding was a female with a typical phenotype of anti-CNTN1–associated autoimmune nodopathy and glomerulonephritis. While her clinical course during the active phase did not differ from other patients, she relapsed one year after full remission with a subacute sensorimotor neuropathy. Anti-CNTN1 autoantibodies remained undetectable, but new reactivity to Caspr1 was observed.²²

Discussion

By analyzing the epitopes of anti-CNTN1 autoantibodies of 16 patients, we showed that epitopes can be localized in the FnIII domain as well as Ig domain and can partly be confirmed by peptide microarrays, indicating linear epitopes in some patients. Comparison with clinical data suggests that certain concomitant disorders such as glomerulonephritis and diabetes as well as the severity and chronicity of the disease are associated with epitopes in different protein domains.

The existence of epitopes in different protein domains suggests that there might be variability in the underlying pathogenic mechanisms even in patients with the same autoantigen. Recent studies have described diverging properties of anti-paranodal autoantibodies, including differences in accessibility to the nodes and in effects in passive transfer experiments when comparing anti-NF155 and anti-CNTN1 autoantibodies.^11,23 Manso et al. suggested impaired regeneration of NF155 as the major pathogenic mechanism in anti-NF155–associated neuropathy,²³ whereas in anti-CNTN1–associated neuropathy, direct axoglial disjunction is supposed to induce conduction blocks.^10,24 NF155 autoantibodies are directed against epitopes in the Fn domain of the protein,^25,26 i.e., epitopes are not located within the direct binding site of NF155 to CNTN1—in contrast to anti-CNTN1 autoantibodies that, according to our data and other studies, are mostly directed against epitopes of the Ig domain that comprises the binding site to NF155 ^13-15. Autoantibodies that are directed against epitopes at the binding site may directly induce detachment of the axoglial junction and, therefore, induce a more severe onset of disease as found in our patients with suspected linear epitopes within the Ig domain who often even needed mechanical ventilation during the acute onset. Binding to the FnIII domain may indirectly lead to steric changes, disturbing axoglial attachment or affecting the protein turnover, as suggested for anti-NF155 autoantibodies²³ and, therefore, may lead to a less severe onset but a more chronic course of disease as seen in 2 of our patients with epitopes within the FnIII domain.

A recent study provided evidence that the formation of immune complexes is associated with glomerulonephritis in patients with anti-CNTN1 autoantibodies,⁵ and proteinuria was suggested to be a possible marker of autoimmune nodopathies.²⁷ Based on our observation that glomerulonephritis is only found in patients with an epitope in the Ig domain, one may hypothesize that these epitopes may facilitate the formation of immune complexes.

Furthermore, different epitopes may not only induce different effects of autoantibodies but also indicate diverse triggers of autoantibody production: Ectopic expression of proteins, molecular mimicry by viral or bacterial infections, or cell debris may induce autoantibody production.²⁸ Our observation that diabetes mellitus, which is a risk factor of anti-CNTN1 autoantibodies,⁷ is preferably found in patients with an epitope in the FnIII domain may reflect different initiators of autoantibody production in patients with different epitopes. Diabetes-induced damage of nerve fibers may trigger the secretion of autoantibodies against the FnIII domain. However, further studies are needed to elucidate potential triggers.

In a previous Japanese study, binding to the Ig5 and Ig6 domains was reported in 8 of 10 patients.¹⁵ In our cohort, binding to the Ig domain was also found in most of the patients but they did not seem to be restricted to the Ig5/Ig6 region. Reasons may be differences in prevalence of certain epitopes in European and Asian patients. In another previous study, only binding to N-glycosylated CNTN1 not to the protein core was found in 3 of 4 patients with anti-CNTN1 autoantibodies, binding to mannose-rich N-glycans was found in another patient.¹⁴ By contrast, in our study, binding to the protein core was detectable in all patients except for one patient. In summary, these differences between studies and the detection of different epitopes even within the same study argue in favor of different possible epitopes in patients with anti-CNTN1 autoantibodies, mostly localized within the Ig domain, and demonstrate that intraindividual differences of autoantibody effects need to be taken into account when studying the pathogenicity of paranodal autoantibodies.

Remarkably, the only patient whose autoantibodies did not bind to deglycosylated CNTN1 mutants was a female with a monophasic course of anti-CNTN1–positive autoimmune nodopathy who developed anti-Caspr1–positive autoimmune nodopathy one year after remission of anti-CNTN1 autoantibodies.²² Thus, binding to glycosylated paranodal proteins might be associated with a certain susceptibility (acquired or hereditary) to developing autoimmunity against paranodal proteins and spreading of epitopes. Regarding patients with a chronic course of anti-CNTN1–associated neuropathy, epitopes were consistent during the course of disease. Linear epitopes detected via microarrays remained stable in follow-up samples, even after resolution of disease, indicating possible persistence of low-level autoantibodies or the presence of nonpathogenic antibodies below conventional detection thresholds. While this underscores the high sensitivity of the peptide array method, it also raises questions about the role of these antibodies in disease maintenance vs recovery.

A limitation of our study is the size of the cohort, specifically the relatively small number of patients with FnIII-targeted epitopes, which restricts statistical power for subgroup comparisons. Future studies should systematically explore the prognostic and therapeutic implications of epitope-specific subclassification.

In summary, our data highlight the existence of clinically meaningful subgroups within anti-CNTN1 nodopathy, defined by distinct antibody epitopes. This epitope-level heterogeneity has to be taken into account when studying the pathophysiology as well as treatment response and clinical phenotype of patients with anti-CNTN1 autoantibodies.

Glossary

CNTN1

contactin1

Fn

fibronectin

Ig

immunoglobulin

PNGase F

peptide-N-glycosidase F

Author Contributions

J. Grüner: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data. I. Talucci: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data. C. Kurth: major role in the acquisition of data. M. Bayer: major role in the acquisition of data. L. Appeltshauser: major role in the acquisition of data. A. Steinbrecher: major role in the acquisition of data. L. Väli: major role in the acquisition of data. S.U. Vay: major role in the acquisition of data. A. Grimm: major role in the acquisition of data. M. Fuchs: major role in the acquisition of data. Albrecht Günther: major role in the acquisition of data. C. Geis: major role in the acquisition of data. C. Sommer: major role in the acquisition of data. H.M. Maric: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data. C. Villmann: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. K. Doppler: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data.

Study Funding

The study was funded by a research grant of the Interdisciplinary Center for Clinical Research (IZKF) of the University Hospital Würzburg to KD and CV (N-356) and KD and HM (A-F-N-419).

Disclosure

The authors report no relevant disclosures. Go to Neurology.org/NN for full disclosures.