IgA autoantibodies demonstrate a novel mechanism of MuSK myasthenia gravis pathology

Abstract

Patients with muscle-specific tyrosine kinase (MuSK) myasthenia gravis (MG) develop muscle weakness due to functionally monovalent immunoglobulin (Ig) G4 autoantibodies (Abs) that target MuSK and disrupt acetylcholine receptor (AChR) clustering. Emerging evidence from other autoimmune conditions suggests that Abs of the IgA class — the archetypal immunoglobulin isotype at mucosal sites —may synergize with IgG Abs, contributing to pathology. In MuSK MG, however, the presence of disease-specific IgA Abs has not yet been recognized, limiting a broader understanding of IgG4-driven autoimmunity. To address this knowledge gap, we leveraged cell-based binding assays, patient-derived recombinant monoclonal autoantibodies (mAbs), high-throughput B-cell receptor sequencing, C2C12 mouse myotube cultures, and passive immunization experiments.

Among 112 sera collected from 25 patients with MuSK MG (discovery cohort), MuSK-specific IgA Abs were detected in eight samples, corresponding to 3/25 (12%) patients. This finding was validated in a second, independent cohort (validation cohort; n=14 individuals), wherein one additional patient (1/14; 7.1%) harbored MuSK IgA Abs. Across all cases, MuSK IgA Abs coexisted with MuSK IgG Abs, while domain-mapping demonstrated a polyclonal IgA response in 2/4 (50%) patients. Notably, in a paradigmatic case with longitudinal samples spanning over a decade, MuSK IgA seropositivity was persistent, and clonally expanded MuSK IgA B cells showed remarkable resistance to therapeutic B-cell depletion.

We generated three patient-derived MuSK-specific IgA mAbs for in-depth molecular profiling. These mAbs were highly hypermutated, bound to distinct MuSK domains, and shared target epitopes with MuSK IgG4 mAbs. Two IgA mAbs cross-reacted with murine MuSK, enabling functional studies with C2C12 myotubes. Mechanistically, individual IgA mAbs promoted AChR clustering (indicative of MuSK agonism) and partially antagonized the pathogenic effects of MuSK IgG4 monovalency. When modeling a polyclonal response, however, the combination of the same IgA mAbs significantly impaired cluster formation, demonstrating cooperative pathogenic potential. Consistent with this, passive transfer of both IgA clones into mice induced a myasthenic phenotype characterized by progressive weight loss, muscle weakness, and structural disruption of the neuromuscular junction (effects similar to those elicited by a functionally monovalent MuSK IgG4 mAb).

These findings uncover a previously unrecognized pathogenic Ab isotype in a subset of patients with MuSK MG and show a mechanism through which bivalent MuSK Abs may synergize to induce pathology. The identification of MuSK-specific IgA B cells could signal a role of mucosal or environmental factors in the pathogenesis of MuSK MG and other IgG4-mediated diseases, offering a worthy avenue for future research.

Introduction

Autoimmune myasthenia gravis (MG) is a rare autoantibody (Ab)-mediated neurologic condition that manifests with fluctuating muscle weakness and fatigability.^1,2 Immunoglobulin (Ig) G Abs that recognize the muscle-specific tyrosine kinase (MuSK) are found in 5–10% of patients with MG.^2–4 Compared to the most prevalent MG subtype — wherein Abs target the muscle nicotinic acetylcholine receptor (AChR) — MuSK MG is more commonly characterized by prominent bulbar weakness and shows favorable response to anti-CD20 B-cell depletion.^5,6 Indeed, following therapeutic administration of rituximab (RTX), an anti-CD20 chimeric monoclonal antibody (mAb), most patients achieve clinical remission and show a marked reduction in their MuSK IgG titers.^7–12

From an immunological perspective, and in contrast with the IgG1–3-driven AChR MG pathology, the autoimmune response occurring in MuSK MG is dominated by IgG4, an antibody isotype with unique molecular properties.^13–15 Because of specific amino acid residues in the heavy chain constant region, IgG4 is less effective at activating complement or engaging with effector cells compared to other IgG subclasses. Furthermore, circulating IgG4 antibodies stochastically exchange half-molecules with other unrelated IgG4, becoming bispecific and unable to cross-link target antigens, a process known as fragment antigen binding (fab)-arm exchange (FAE).^16,17 These “anti-inflammatory” features notwithstanding, in vitro and in vivo studies support a direct, prominent role of IgG4 Abs in MuSK MG pathogenesis.^3,18–21

At the neuromuscular junction (NMJ), MuSK interacts with the agrin/low-density lipoprotein receptor-related protein 4 (LRP4) complex and orchestrates a signaling cascade that promotes aggregation of AChR on the muscle endplate, contributing to NMJ development/maintenance and effective neuromuscular transmission.^22–24 In MuSK MG, however, IgG4 Abs (which comprise more than 90% of circulating MuSK IgG) sterically block the interaction between MuSK and LRP4, thus hindering MuSK activation and its intracellular signaling cascade.^13,25 This event ultimately leads to the impairment of AChR clustering and the failure of neuromuscular transmission.

The role of the less abundant MuSK-specific IgG1, 2, and 3 Abs is still under investigation, but cumulative evidence suggests that their mechanism of action differs from that of IgG4 Abs. Unlike IgG4 Abs that are functionally monovalent due to FAE, IgG1–3 MuSK Abs are bivalent and have the potential to force-dimerize MuSK and activate its signaling cascade. Indeed, studies on MuSK-specific mAbs have demonstrated the capacity of bivalent clones to induce MuSK phosphorylation and agrin-independent activation of the AChR clustering pathway.^21,26,27 Differences in Ab valency seem to dictate divergent functional effects in vivo as demonstrated by a passive transfer experiment in which bispecific, monovalent IgG4 mAbs induced severe and rapid myasthenic symptoms in mice, while injection of the same clones as bivalent antibodies led to absent or milder pathogenic effects.²⁸ In a recent study, however, IgG4-depleted pooled sera induced MuSK, DOK7, and βAChR phosphorylation in vitro, but cluster formation was impaired, mirroring the detrimental effects induced by IgG4 Abs.²⁹ A crucial difference between these investigations lies in the experimental approach of testing single mAbs versus polyclonal sera.

Beyond IgG4 Abs and the smaller population of IgG1–3 Abs, whether immunoglobulins of other isotypes play a role in the autoimmune response against MuSK remains unclear. Growing evidence from other autoimmune conditions suggests that IgA, an antibody isotype that dominates mucosal immune responses and the second-most prevalent isotype in serum, may synergize with IgG and contribute to pathology.^30,31 In MuSK MG, however, data regarding the presence and role of IgA Abs are lacking, limiting a broader understanding of the mechanisms that govern IgG4-mediated autoimmunity.³²

In this study we show that a subset of patients with MuSK MG harbors IgA Abs coexisting with IgG Abs. Functional profiling of patient-derived MuSK-specific IgA mAbs revealed divergent mechanistic properties depending on the experimental context. Individual IgA mAbs activated MuSK signaling and partially rescued the dispersal of AChR clusters induced by pathogenic IgG4 Abs. In sharp contrast, combining two IgA clones (an experiment designed to mimic serum polyclonality) led to striking impairment of AChR clustering and triggered myasthenic symptoms in passively immunized mice. Collectively, this study identifies a novel MuSK MG Ab isotype with demonstrated pathogenic capacity and highlights a previously unappreciated potential link between mucosal triggers and MuSK autoimmunity.

Material and methods

Detailed methods are described in the Supplementary material.

Results

MuSK IgA Abs are found in a subset of patients with MuSK MG

To measure MuSK-specific IgA Abs in serum, we optimized a live, MuSK-expressing cell-based assay (CBA) using flow cytometry.²⁰ First, we validated the specificity of a commercial anti-IgA secondary antibody using MuSK1B,²⁰ a patient-derived MuSK-specific IgG3 mAb that was subcloned into expression vectors encoding the heavy chain constant regions of the four IgG subclasses (IgG1, 2, 3 and 4), IgA1 (the predominant IgA subclass in serum), and IgM. As expected, the anti-IgA secondary antibody recognized MuSK1B subcloned into the IgA1 backbone but did not show IgG or IgM cross-reactivity (Fig. 1A). Serological assays that measure IgA (and IgM) can show non-specific background signal especially when certain antigens are expressed at high density on cells.³² Thus, we conducted a series of preliminary experiments in which we tested representative sera from patients with MuSK MG and controls (healthy individuals and AChR MG patients) using increasing serum dilutions. Based on these tests, a 1:100 dilution factor yielded an optimal signal-to-noise ratio that increased the assay specificity without reducing its sensitivity.

Figure 1. MuSK-specific IgA Abs are detected in a subset of patients with MuSK MG.

(A) Validation of the specificity of a commercial anti-IgA secondary antibody using MuSK1B, a MuSK-specific human mAb subcloned into an IgG1, IgG2, IgG3, IgG4, IgA1 and IgM backbone. Binding of isotype- and subclass-specific monoclonal antibodies (mAbs) was detected using flow cytometry. (B) MuSK IgA screening of 112 MuSK MG, 30 HD and 32 AChR MG sera (discovery cohort). The cutoff for positivity (dashed line) was set at the median MFI ratio of the HD group + four standard deviations (SDs). Each data point represents the mean of experimental duplicates. CBA-positive sera are color-coded, with each color corresponding to a distinct patient. (C) MuSK IgA screening of a validation cohort including 23 MuSK MG and 25 HD sera. (D) Cell-based assays using confocal fluorescence microscopy. Representative images include a healthy donor (HD) and three MuSK MG patients, two of whom (MuSK-1 and MuSK-4), belonging to independent cohorts, tested positive for both MuSK IgG and IgA Abs, while MuSK-5 only harbored MuSK IgG Abs. First row: nuclei counterstained with DAPI. Second row: eGFP signal showing MuSK-expressing cells. Third row: IgG (cyan) and IgA (red) detection is shown for each representative case. Fourth row: merge, confirming co-localization of IgG and/or IgA (when detected) with MuSK-expressing cells. All images were captured at 63× magnification. Scale bar: 20μm. (E) Correlation between MuSK IgA and MuSK IgG antibody levels, assessed using Pearson correlation coefficient. (F) Longitudinal tracking of serum MuSK IgG and IgA levels in a patient (MuSK-1) with fourteen serum samples (timepoints) collected over a decade of disease activity. Full circles represent serum samples above the positivity cutoff, while empty circles indicate those below the cutoff. (G) Panels showing superimposed changes of MuSK IgG and IgA levels between two timepoints. The upper panel corresponds to timepoints 11–12 (months 69–78), while the lower panel corresponds to timepoints 13–14 (months 127–138).

In the initial MuSK IgA screening, we included a total of 112 longitudinally collected sera derived from 25 patients (discovery cohort) with a clinical and laboratory-confirmed diagnosis of MuSK MG, as well as 30 sera from healthy donors (HD) and 32 sera from patients with AChR MG. Based on the positivity cutoff determined by the HD group, eight out of 112 MuSK MG sera, corresponding to 3/25 unique patients (12%), were positive for MuSK IgA Abs (Fig. 1B). To validate our findings in an independent cohort, we screened an additional 23 sera from 14 patients with MuSK MG (validation cohort) and identified one patient with four longitudinally collected sera harboring MuSK IgA Abs (Fig. 1C). Collectively, MuSK IgA Abs were detected in 12 out 135 (8.9%) tested sera and 4/39 (10.3%) patients with MuSK MG. Flow cytometry testing of an additional 112 control sera from patients with other neurologic disorders — AChR MG, neuromyelitis optica spectrum disorder, and myelin oligodendrocyte glycoprotein antibody-associated disorder (MOGAD) — identified one sample from a MOGAD patient exhibiting low positivity (1/112, 0.9%). In this case, IgA binding was restricted to cells expressing high membrane densities of MuSK (MuSK-GFP high) (Supplementary Fig.1A–C). Conversely, across all MuSK MG sera, IgA Abs recognized MuSK irrespective of its expression level. Thus, a threshold of MuSK cell density discriminated IgA Abs in MuSK MG patients from a disease control, probably reflecting qualitative differences in the binding capacity of the detected Abs (Supplementary Fig.1D).

In the two MuSK MG cohorts, MuSK IgA Abs were always co-detected with MuSK IgG Abs (representative fluorescence microscopy images are shown in Fig. 1D), but no association was found between IgA and IgG Abs’ levels measured by flow cytometry (Fig. 1E). Previous studies have demonstrated a good correlation between disease status and MuSK IgG (and IgG4) levels.^12,33,34 In our cohort, MuSK IgG —but not IgA—levels moderately correlated with disease severity at sampling (Supplementary Fig. 2A). Estimation of absolute MuSK IgG4 and IgA titers by flow cytometry showed an IgG4 predominance in most cases, confirming the primary role of IgG4 Abs in the disease (Supplementary Fig. 2B). Demographic and clinical data of the four MuSK IgA seropositive patients (MuSK-1, MuSK-2, MuSK-3, and MuSK-4) are summarized in Supplementary Table 1. No significant differences were identified when comparing the clinical features of these patients with those who were seronegative for MuSK IgA Abs (Supplementary Table 2). In addition, we reviewed corresponding clinical records to investigate factors (including documented infections and immunizations) potentially influencing the production of MuSK IgA Abs, but the clinical histories of the four patients were unrevealing in this regard.

For three MuSK IgA seropositive patients, longitudinal samples were available, which allowed us to track their Abs levels over time. In the case of MuSK-2, a patient with three consecutive sampling timepoints, MuSK IgA Abs were detected only in the first sample collection. In MuSK-4, instead, four longitudinal samples collected within the first year of symptom onset were positive for MuSK IgA Abs (Supplementary Fig. 3). In the third patient (MuSK-1), fourteen serum samples were available, spanning almost twelve years (140 months) of disease, and within this timeframe RTX was administered six times to treat or prevent clinical relapses (Fig. 1F). Notably, while MuSK IgG Abs were detected at all timepoints, MuSK IgA Abs were first detected at month 61 of the serial collection (9^th sampling timepoint) without evidence of preceding infections or vaccinations, and this IgA response persisted for nearly 7 years, up to the last available sample. Both MuSK IgG and IgA levels diminished following each RTX cycle, but the magnitude of reduction was greater for IgG than for IgA (11^th-12^th timepoint: −66.4% vs. −13.2%; 13^th-14^th timepoint: −88.8% vs. −43.86%) (Fig. 1G). Taken together, these data show that (1) a subset of patients with MuSK MG harbors MuSK IgA Abs coexisting with MuSK IgG Abs; (2) MuSK IgA Abs can be detected in both early and advanced stages of the disease; (3) both MuSK IgA and IgG Ab levels decline following RTX administration, albeit with different proportions, suggesting that MuSK-specific IgA and IgG-producing B cells may differ in their susceptibility to anti-CD20 depletion.

Patient-derived MuSK-specific IgA mAbs have heterogenous properties

To examine the molecular and functional features of MuSK IgA Abs and compare them to those of MuSK IgG4 Abs, we leveraged a B-cell culturing approach through which we previously generated human-derived IgG mAbs from patients with MuSK and AChR MG.^20,35,36 Patient CD27⁺ B cells were cultured in 96 well-plates and differentiated in antibody-secreting cells. Following a high-throughput screening of B cell supernatants, the B cells from those wells containing MuSK IgA Abs were cloned to obtain recombinant mAbs. Based on the number of CD27+ B cells isolated from each patient, the frequency of circulating MuSK IgA B cells was 0.01% (3 cells) for patient MuSK-1, 0.001% (1 cell) for patient MuSK-2, 0.012% (2 cells) for patient MuSK-3. Similar frequencies were observed in a parallel MuSK IgG screening (Supplementary Table 3), further confirming the exceptional rarity of blood-derived MuSK-specific clones.^20,21,35,37 Using this culturing strategy, we generated three MuSK-specific IgA mAbs (referred to as aMu1, aMu2, and aMu3), each derived from a distinct MuSK IgA seropositive patient (MuSK-1, MuSK-2, MuSK-3). In this nomenclature, “a” denotes the Ab isotype (IgA), and “Mu” refers to MuSK. Immunoglobulin subclass PCR showed that all three mAbs derived from IgA1-expressing B cells; accordingly, an IgA1 recombinant expression vector was used for the expression of these three MuSK IgA mAbs unless otherwise indicated. In humans, serum IgA is chiefly monomeric, whereas secretory IgA is mainly dimeric.³⁸ In line with this, no evidence of dimeric MuSK IgA Abs was found by testing patients’ sera (Supplementary Fig. 4), suggesting that the majority of circulating MuSK IgA Abs exist as monomers. Therefore, the recombinant IgA mAbs were expressed in a monomeric form.

First, we confirmed the specificity of the newly generated IgA mAbs by live CBA, using both flow cytometry and fluorescence microscopy. The three IgA mAbs bound to full-length human MuSK over a wide range of concentrations, and their relative binding strength was comparable to that of MuSK1B, a control mAb with subnanomolar affinity for MuSK (Fig. 2A).³⁹ Confocal fluorescence microscopy demonstrated a clear membrane staining and selective labeling of MuSK-expressing cells (Fig. 2B), whereas cross-reactivity with two clinically relevant neurologic autoantigens, AChR and myelin oligodendrocyte glycoprotein (MOG), was not observed (Supplementary Fig. 5). V(D)J sequencing analysis showed that the three mAbs were heavily somatically hypermutated, with an average V gene amino acid replacement mutation frequency of 19.5 ± 5.2% and 13.9 ± 3.4% (mean ± SD) for the heavy and light chains, respectively (Table 1). Two N-linked glycosylation sites—a feature of many MuSK IgG4 mAbs^40,41— were found in aMu1 (both encoded by germline-derived nucleotides in the VH4–34 gene), but not in the other two IgA mAbs.

Figure 2. Binding properties of patient-derived MuSK-specific IgA mAbs.

(A) Binding curves of three MuSK-specific IgA mAbs tested in a MuSK cell-based assay over a range of serial dilutions. MuSK1B (specific to MuSK) and mAb01a (specific to AChR), both expressed as IgA isotype antibodies, were used as experimental controls. The y-axis shows the calculated MFI ratio of the anti-IgA secondary antibody, representing binding of the mAbs to MuSK. All data are presented as a mean of experimental triplicates and SDs. (B) Cell-based assays using confocal fluorescence microscopy, confirming the specificity of the IgA mAbs. First row: nuclei counterstained with DAPI. Second row: eGFP signal showing MuSK-expressing cells. Third row: anti-IgA secondary antibody signal. Fourth row (merge): for aMu1, aMu2, aMu3 and MuSK1B, the IgA binding signal (red fluorescence on cell membrane) colocalizes with MuSK-expressing cells (green fluorescence), indicating MuSK recognition of individual mAbs. All images were captured at 63X magnification. Scale bar: 20μm. (C) Bar graph displaying the domain mapping results. All mAbs were subcloned into IgG1 vectors and tested for binding to HEK293T cells expressing different MuSK domains. MuSK mAbs binding to three different MuSK domains were used as controls. A negative control mAb (mAb09) was also included. The y-axis represents the MFI ratio of an anti-IgG-AF647 antibody. Experimental triplicates were run for each test. The means of triplicate values and SDs are shown. (D) Heatmap showing binding competition between MuSK-specific IgA and IgG4 mAbs. The x-axis shows three IgG4 mAbs (tested in a bivalent format), while the y-axis lists three MuSK-specific IgA mAbs and mAb01a-IgA (AChR-specific), used as a control. For each pairing, IgG binding (expressed as MFI ratio) was normalized to the non-competing condition. The color intensity correlates with the competition level shown by each IgA-IgG4 combination. Black represents the highest level of competition. (E) Schematic of the MuSK extracellular domain and the binding specificities of the IgA and IgG4 mAbs included in the study (mAbs with the same color compete for identical or partially overlapping epitopes).

Table 1.

Molecular characteristics of MuSK-specific human recombinant IgA mAbs

mAb ID	Isotype and subclass	V-(D)-J genes and alleles	AA replacements in variable region gene segment	AA replacements in CDR3	N-linked glycosylation site counts
aMul	IgAl	IGHV4-34*01, IGHJ4*02, IGHD1-7*01	24	2	2 (germline-encoded)
	IgK	IGKV3-20*01, IGKJ1*01	17	1	0
aMu2	IgAl	IGHV1-69*09, IGHJ1*01, IGHD5-12*01	19	4	0
	IgK	IGKV3-11*01, IGKJ2*01 (or IGKJ2*02)	12	4	0
aMu3	IgAl	IGHV3-23*04, IGHJ5*02, IGHD3-16*02	14	5	0
	Igx	IGLV2-18*02, IGLJ1*01	11	2	0

The IMGT database top-predicted V(D)J genes, V gene and CDR3 amino acid (AA) replacements, and number of V gene N-linked glycosylation (N-Glyc) sites are shown for both heavy and light chain pairs. V gene AA mutations were counted from the first codon of framework 1 to the invariable cysteine (C) at position 104. CDR3 AA mutations were counted between cysteine 104 and the invariable tryptophan (W) or phenylalanine (F) at position 118 in the heavy and light chain respectively, considering only mutations identified through alignment to V(D)J genes and excluding untemplated nucleotides. N-linked glycosylation motif: N-X-S/T, where X is any AA except proline.

MuSK is a tyrosine kinase receptor comprised of three immunoglobulin (Ig)-like domains, one Frizzled-like (Fz-) domain, and an intracellular kinase domain.^42–44 Most (if not all) patients with MuSK MG harbor IgG Abs that recognize the Ig-like 1 domain.¹⁹ Conversely, IgG Abs targeting other MuSK domains are found less frequently.⁴⁵ To determine the domain specificity of the three IgA mAbs, we performed an experiment in which—as we previously demonstrated—each mAb is tested for binding to single MuSK domains.²⁰ Using this strategy, we determined that aMu1, aMu2, aMu3 recognized the Ig-like 1, the Fz-, and Ig-like 2 domain, respectively (Fig. 2C). Because MuSK IgA and IgG Abs coexist in sera, and IgG4 is the predominant IgG subclass in MuSK MG, we then asked whether the IgA mAbs could compete with IgG4 Abs for MuSK binding. To this end, we employed three IgG4 mAbs that were characterized in previous studies: MuSK1A which recognizes the Ig-like 2 domain, and 6C6 and 2E6 both targeting the Ig-like 1 domain.^20,35 After incubating MuSK-expressing cells with an IgA mAb in excess, an IgG4 mAb was added and detected by flow cytometry using an experimentally validated anti-IgG Fcγ secondary antibody (Fig. 1A). In these experiments, pre-incubation with aMu1 IgA reduced the detection of 2E6 IgG4 by 93.0 %, while aMu3 IgA hindered binding of MuSK1A IgG4 by 92.6% (Fig. 2D).

In summary, we successfully generated three patient-derived MuSK-specific IgA1 mAbs from the CD27⁺ memory B-cell compartment. These mAbs show the hallmarks of affinity maturation, target three different MuSK domains, and exhibit competitive binding with patient-derived IgG4 Abs, suggesting identical or partially overlapping target epitopes (Fig. 2E). Given the competitive binding, we asked whether high titers of MuSK IgG4 Abs could have hindered the detection of MuSK IgA Abs in the screening CBA. To address this, we depleted IgG from MuSK IgG-positive sera in the discovery cohort and tested the IgG-negative fractions by CBA. No additional MuSK-IgA positive samples were identified following IgG removal, indicating that the CBA provided sufficient antigen density to allow parallel detection of both IgA and IgG Abs (Supplementary Fig. 6).

High-throughput B-cell receptor sequencing identifies RTX-resistant MuSK-specific IgA clones

We previously demonstrated that MuSK-specific IgG4 B cells can resist RTX and reemerge into circulation prior to relapses.³⁵ We next sought to verify whether MuSK-specific IgA B cells exhibit a similar behavior. We queried B-cell receptor (BCR) repertoire libraries previously generated through either bulk or single-cell sequencing of samples derived from patient MuSK-1, from whom aMu1 IgA mAb was cloned (Supplementary Table 1).³⁵ Additionally, we performed bulk BCR sequencing on newly collected samples belonging to patients MuSK-1 and MuSK-2 (the latter patient was the source of aMu2 IgA mAb). B-cell clones and clonal variants —cells that originate from a common V(D)J rearrangement—were identified by clustering BCR heavy chain sequences by sequence similarity. Based on this search, we found 24 distinct clonal variants of aMu1, belonging to five different timepoints that spanned over six years of disease (Fig. 3A–B). In addition, one clonal variant of aMu2 was found in the same timepoint from which aMu2 was generated (Supplementary Fig. 7). Clonal variants of aMu1 were detected before and after multiple RTX cycles. Nine clonal variants (7–11; 21–24) were found in peripheral blood mononuclear cell (PBMC) samples collected within three months after RTX administration, indicating remarkable resistance to anti-CD20 B-cell depletion (Fig. 3A). All variants were highly mutated, with between 12.6 and 24.2% nucleotide difference from their predicted unmutated V gene (Fig. 3B). Further, the mean of somatic hypermutations in this lineage was significantly positively correlated with sample time (slope=0.79 mutations/month, date randomization p=4×10⁻⁵), indicating that this lineage is measurably evolving and likely undergoing additional affinity maturation during its 77 month sampled period.⁴⁶ Five representative clonal variants, corresponding to the earliest detection of MuSK IgA in serum, were expressed, and their specificity was confirmed by CBA (Fig. 3C). Notably, all clonal variants were IgA, and thus evidence of MuSK IgA-IgG class-switching events was not observed.

Figure 3. Longitudinal durability of a clonally expanded MuSK-specific IgA clone.

(A) Clonal variants (CVs) of the MuSK-specific mAb aMu1 were identified in longitudinally collected samples from patient MuSK-1. The graph shows the identification of 24 distinct CVs (indicated by arrows) spanning 77 months of disease activity. The detection of MuSK IgA in serum (red circles) is also shown. (B) Clonal lineage tree containing BCR sequences from bulk RNA sequencing of serial samples from patient MuSK-1 from whom aMu1 was cloned. A maximum likelihood tree of aMu1 is shown, with 24 IgA CVs. Edge lengths represent the expected number of intervening somatic mutations between nodes (see scale bar). Colors correspond to the collection time point (months) at which each sequence was identified in relation to the first available collection timepoint. Duplicate sequences or those differing only by ambiguous nucleotides were collapsed, as indicated by the size of each tip. (C) Schematic diagram illustrating five representative clonal variants (CV1–5) of aMu1. Lollipop-shaped symbols indicate nonsynonymous nucleotide mutations as compared with the germline IGHV4–34 gene. Amino acid replacement residues are indicated under each symbol. The amino acids replaced in the germline gene segment (IGHV4–34) from which aMu1 derived are indicated in red. Right: heatmap summarizing the cell-based assay results that confirmed MuSK specificity for the five representative CVs. An AChR-specific mAb (mAb01a), subcloned into an IgA1 vector, was used as a negative control.

Individual IgA mAbs can behave as MuSK agonists and functionally antagonize IgG4 pathogenic monovalency in vitro

Because bivalent MuSK IgG mAbs have previously shown agonistic potential by inducing varying degrees of MuSK phosphorylation and AChR clustering,²⁷ we reasoned that the IgA mAbs could exhibit a similar behavior. To test this hypothesis, we cultured C2C12 myotubes which have the key signaling machinery, except neuronal agrin, to form AChR clusters. Treatment with agrin activates the LRP4-MuSK-DOK7 pathway and leads to the formation of AChR clusters (modelling nascent NMJs), which can be readily visualized and quantified using fluorescently labeled alpha bungarotoxin. Crucially, the addition of functionally monovalent MuSK IgG4 Abs blocks MuSK-LRP4 interaction and hampers cluster formation.

To assess whether the MuSK IgA mAbs could elicit AChR clustering independent of agrin, we incubated fully differentiated myotubes with the IgA mAbs in the absence of agrin and counted the number of “mature” (≥ 15 μm²) AChR clusters.²⁷ 8–18C5 IgA, a mAb that binds to MOG,⁴⁷ showed no conspicuous effect on clustering level (Fig. 4A). In sharp contrast, both aMu1 and aMu3 increased the number of AChR clusters, achieving 57.9% and 40.3% of the agrin effect, respectively (p<0.0001 compared to 8–18C5). A modest increase in clusters, albeit not statistically significant, was also observed with aMu2, the IgA mAb specific for the Fz-like domain. Further corroborating the agonistic potency of aMu1 and aMu3, immunoblotting showed that both mAbs induced MuSK phosphorylation, whereas aMu2 and 8–18C5 did not (Fig. 4B). Despite their agonistic activity, aMu1 and aMu3 moderately reduced agrin-dependent cluster formation (Fig. 4C), suggesting functional interference with the agrin-LRP4-MuSK pathway.

Figure 4. Individual IgA mAbs demonstrate a dual role, acting both as MuSK agonists and as antagonists to pathogenic monovalent IgG4 autoantibodies.

The addition of neural agrin to C2C12 myotubes leads to activation of MuSK and the formation of AChR clusters that can be readily visualized with fluorescent α-bungarotoxin and then quantified. (A) Assessment of the agonistic capacity of three MuSK IgA mAbs (aMu1 aMu2 and aMu3) in the absence of agrin. 8–18C5, a mAb that binds to MOG, was subcloned into an IgA1 backbone and used as a control. The number of AChR clusters is normalized to the agrin-only effect of each individual experiment (red line). Each data point represents the mean value obtained from two wells (six image fields) with fully differentiated myotubes, and for each condition at least three independent experiments were performed. Bars represent the mean of means and error bars SDs. Multiple comparisons ANOVA (against 8–18C5 mAb), Dunnett’s test; * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, only shown when significant. Representative microscopy images for each condition are shown below the graph. The panels in the bottom left corner of each image show magnifications of representative areas to highlight AChR clustering (red arrow). Scale bar: 100 μm. DM: differentiation media. (B) Normalized densitometry analysis results from the MuSK phosphorylation immunoblots are plotted. Each data point represents an independent immunoblot experiment. Bars represent the mean and error bars SDs. Phosphorylation of MuSK (detected by a commercial anti-phosphotyrosine antibody) was normalized to MuSK expression (detected by a commercial anti-MuSK antibody after stripping the blot). The ratio of phosphorylated MuSK/total MuSK is plotted. Multiple-comparisons ANOVA (against 8–185C IgA), Dunnett’s test; * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, only shown when significant. Below the graph is a representative immunoblot showing phosphotyrosine bands and relative MuSK expression of C2C12 myotubes following incubation with differentiation media (DM), agrin, the three MuSK IgA mAbs (aMu1, aMu2, aMu3) or a mAb control (8–18C5 IgA). (C) The three IgA mAbs were tested as described in (A), but with the addition of agrin. (D-F) Three MuSK IgG4-derived Fabs (MuSK1A, 6C6 and 2E6) were tested on C2C12 myotubes with agrin alone or with the addition of a MuSK-specific IgA mAb at equimolar concentrations. AChR clusters were quantified as described above. Bars represent the mean of means and error bars SDs. Multiple comparisons ANOVA (against each Fab), Dunnett’s test; * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, only shown when significant.

The IgA mAb aMu2, by contrast, was the only clone lacking significant agonistic activity on mouse C2C12 myotubes, which prompted us to investigate the underlying cause. Notably, sequence alignment of human and mouse MuSK demonstrates high interspecies identity (96–99%) in the Ig-like 1 and 2 domains respectively, but reduced homology (approximately 86%) in the Fz-like domain (Supplementary Fig. 8A). Based on this observation, we asked whether aMu2 lacked cross-reactivity with mouse MuSK. To test this hypothesis, Ab binding to mouse MuSK was assessed by CBA. Whereas aMu1 and aMu3 bound strongly to both human and murine MuSK, aMu2 displayed only minimal reactivity toward the murine isoform (Supplementary Fig. 8B). We therefore concluded that the C2C12 system could not provide a reliable assessment of aMu2’s functional activity and excluded this mAb from subsequent experiments.

Because the other two mAbs, aMu1 and aMu3, exhibited agonistic potency in the C2C12 clustering experiments, we then asked whether they could antagonize the pathogenic effects of FAE MuSK IgG4 Abs and rescue cluster formation in vitro. To test this, we leveraged an established experimental approach^21,39, and employed three MuSK IgG4-derived Fabs to mimic the pathogenic properties of MuSK IgG4 functional monovalency (Fig. 4D–F). As expected, agrin-dependent AChR clustering was diminished by MuSK1A Fab (−98.5%), 6C6 Fab (−86.6 %), and 2E6 Fab (−46.34%).^35,39 In the presence of MuSK1A Fab, both aMu1 and aMu3 partially rescued AChR clustering (respective increases of +44.5%; p< 0.0001, and +45.1%; p<0.0001, compared to the “Fab plus agrin” condition). In a similar pattern, when co-incubated with 6C6 Fab, aMu1 and aMu3 increased the number of AChR clusters (respective increases of +32.6%, p< 0.0001; +31.6%, p<0.001). No differences, however, were observed upon treatment with Fab 2E6, the Fab with the least pathogenic capacity among the three, as the addition of the IgA mAbs did not significantly change the number of clusters. These findings confirmed that, similar to bivalent IgG, individual MuSK IgA Abs can exhibit both agrin-independent agonistic activity and the potential to antagonize functionally monovalent IgG4 Abs.

The combination of two IgA mAbs potently inhibits AChR cluster formation despite persistent MuSK phosphorylation

Multiple lines of evidence suggest that the production of MuSK IgG is polyclonal: clonally-unrelated IgG mAbs have been generated from single individuals, and circulating IgG Abs in a given patient can target different MuSK domains.^20,21,45 To investigate whether polyclonality is also a feature of circulating MuSK IgA Abs, we tested representative sera from the MuSK IgA-seropositive patients, leveraging the same domain-mapping strategy that was previously employed for the mAbs. In two patients (MuSK-1 and MuSK-3) we detected IgA Abs against a single domain (the Ig-like 1 and Ig-like 2 domains, respectively), whereas in sera from MuSK-2 and MuSK-4, the IgA Abs recognized two or three MuSK domains, indicative of a polyclonal response (Fig. 5A).

Figure 5. MuSK IgA mAbs act synergistically when combined, resulting in enhanced pathogenicity and potent inhibition of AChR clustering.

(A) Serum IgA domain-mapping demonstrating targeting of multiple MuSK domains in two patients (MuSK-2 and MuSK-4). A conservative IgA binding cutoff (>1.5) was used. (B) Representative microscopy images showing the effect of combined aMu1 and aMu3 on AChR cluster formation. Scale bar: 100 μm. (C) Quantification of clusters in the absence agrin. (D) Quantification of clusters in the presence of agrin. In both experiments 8–18C5, a MOG-specific mAb, was included as an additional control. 8–18C5, aMu1 and aMu3 were combined in various permutations for analysis. The number of AChR clusters was normalized to the agrin-only effect of each individual experiment (red line). Each data point represents the mean value obtained from two wells (six image fields in total) with fully differentiated myotubes, and for each condition at least three independent experiments were performed. Bars represent the mean of means and error bars SDs. Multiple comparisons ANOVA (against aMu1+aMu3), Dunnett’s test; * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, only shown when significant.

This finding prompted us to assess how the simultaneous action of multiple IgA mAbs can affect AChR clustering. We therefore tested the combination of aMu1 and aMu3 on C2C12 myotubes and counted the number of clusters (Fig. 5B). Contrary to expectations, in the absence of agrin the two mAbs did not exert an additive or synergistic effect on cluster formation but, rather, we found very few clusters, with a clustering level comparable to that of a non-binding control mAb (p>0.99 compared to 8–18C5) (Fig. 5C). Moreover, this combination potently inhibited agrin-dependent AChR clustering to a greater extent than either mAb alone, achieving ~90% reduction in clusters compared to agrin (Fig. 5D). This inhibitory effect required the simultaneous action of two mAbs exhibiting MuSK specificity, because pairing either clone with an irrelevant IgA mAb (such as the MOG-specific 8–18C5) did not replicate the result (Fig. 5C–D). Despite this impairment of clusters, immunoblotting confirmed that the combination of aMu1 and aMu3 retained the capacity to induce MuSK phosphorylation (Supplementary Fig. 9). Collectively, these data indicate a complex interplay between Ab clonality and MuSK signaling. Indeed, single IgA mAbs activated MuSK (thereby promoting cluster formation) while partially inhibiting agrin activity. Instead, the combination of two IgA mAbs failed to sustain cluster formation and, at the same time, strongly antagonized agrin-mediated clustering. Thus, the simultaneous engagement of MuSK by two IgA mAbs was responsible for enhanced pathogenic effects in vitro.

Passive transfer of patient-derived MuSK IgA mAbs induces a myasthenic phenotype in mice

Finally, to investigate the pathogenic potential of MuSK IgA Abs in vivo, we established a passive immunization MG model. In a previous study, intraperitoneal (IP) injections of functionally monovalent IgG4 MuSK-specific mAbs binding to the Ig-like 1 domain induced progressive muscle weakness in immune-deficient NOD-SCID mice.²⁸ Thus, as proof of concept, we first tested MuSK1A, a patient-derived MuSK IgG4 mAb binding to the Ig-like 2 domain (Supplementary Figure 10A).²⁰ To emulate the monovalency of IgG4 Abs, we generated a bispecific mAb using a controlled FAE process.⁴⁸ The resulting mAb binds to MuSK with one arm, while the second antibody arm recognizes the keyhole limpet hemocyanin (KLH). In alignment with the findings achieved with two other MuSK IgG4 clones (13–3B5 and 11–3F6),²⁸ mice receiving three IP injections of MuSK1A-KLH (1 mg/kg on day 0, 3 and 6) showed overt signs of muscle weakness within one week of treatment, whereas two control groups (untreated mice and mice receiving a KLH IgG mAb) remained unaffected (Supplementary Fig. 10B–E). Notably, a decline in body weight was observed in mice treated with MuSK1A-KLH, likely due to the decreased ability to reach water and food and the prominent vulnerability of bulbar muscles to MuSK Abs.^5,49

After confirming the pathogenic capacity of a human-derived IgG4 mAb in our model, we next sought to evaluate the effects of MuSK IgA Abs. Because in immune-deficient mice the half-life of adoptively transferred IgA is relatively shorter than that of IgG (15–22 hours vs. 2–3 days),⁵⁰ we implemented a daily passive transfer protocol to ensure continuous in vivo IgA exposure. Additionally, to emulate the polyclonal IgA profile observed in patients, we treated mice with a 1:1 mixture of two MuSK IgA mAbs (aMu1 and aMu3, 5 mg/kg total) (Fig. 6A). This regimen caused progressive weight loss, with a 20% reduction in body weight on day 9 compared to baseline (Fig. 6B). Signs of muscle weakness emerged after one week of treatment, as shown by a significant decrease in grip strength on day 9 (−31.7%, p=0.0079, compared to mice receiving an equivalent daily dose of KLH IgA) (Fig. 6C), and shorter hanging times — assessed using two approaches (−71% in both tests, p=0.0079) (Fig. 6D–E). At the end of the experiment, sera collected from mice exhibiting muscle weakness confirmed the presence of circulating MuSK-specific IgA Abs (Fig. 6F), and whole-mount immunostaining of the epitrochleoanconeus (ETA) — a fast-twitch muscle in the upper forelimb, distal from the injection site — demonstrated IgA deposition at the NMJs (Fig. 6G). In sharp contrast to the intact, pretzel-like morphology of healthy NMJs observed in the KLH IgA group, the NMJs of MuSK IgA-treated mice lost their complex architecture and appeared fragmented (Fig. 6H). In sum, passive transfer of two patient-derived MuSK IgA mAbs induced a myasthenic phenotype in mice, demonstrating the pathogenic capacity of MuSK IgA in vivo.

Figure 6. Passive transfer of two MuSK-specific IgA mAbs causes myasthenic symptoms in NOD-SCID mice.

(A) Experimental design of passive immunization with recombinant IgA mAbs. NOD-SCID mice (n=5 per group) received a daily IP injection of either KLH IgA (control mAb) or a mixture (1:1) of two IgA MuSK mAbs (aMu1 and aMu3); final mAb concentration: 5 mg/Kg. (B) Body weight measurement. (C) Grip strength assessment. (D) Rolling loop and (E) inverted mesh hanging tests. Data represent mean and error bars SDs. Two-tailed unpaired t test. ** p<0.01. (F) Cell-based assay testing of MuSK IgA Abs in blood samples collected on day 9. (G) Detection of bound IgA mAbs, colocalizing with NMJs, in whole-mount preparations of epitrochleoanconeus. Top panel: alpha-bungarotoxin (green signal), labelling NMJs. Mid panel: anti-IgA signal (red signal). Bottom panel (merge): colocalization of bungarotoxin and anti-IgA signals, confirming the deposition of IgA Abs on NMJs. Scale bar: 20 μm. (H) Confocal microscopy images at low (20X) and high magnification (63X) showing the morphology of representative NMJs on diaphragm muscle preparations.

Discussion

The detection of MuSK-specific IgG Abs represents the diagnostic hallmark of MuSK MG. Here we showed that, beyond Abs of the IgG isotype, approximately 10% of MuSK MG patients harbor MuSK IgA Abs and that these Abs have pathogenic potential in vivo. Our data contribute to the otherwise limited evidence suggesting a role for IgA Abs in neurologic autoimmunity. In N-methyl-D-aspartate receptor (NMDAR) encephalitis, for example, pathogenic IgG1 Abs are found in most patients but NMDAR-IgA seropositivity is enriched in those with a paraneoplastic cause, primary ovarian teratomas.⁵¹ Histological examination of such tumors revealed germinal center-like architectures in close proximity with neuroglial tissues structures, which are potential sites for B-cell autoimmunization, affinity maturation, and class-switching.⁵¹ NMDAR IgA Abs have also been detected in a subset of patients with slowly progressive cognitive impairment and showed pathogenic potential in vitro, wherein patients’ purified IgA (but not that of controls) decreased membrane NMDAR expression in cultured neurons.⁵² More recently, IgA Abs recognizing MOG have been found in patients with central nervous system demyelination, presenting with a distinct clinical phenotype, and with MOG- and aquaporin-4 IgG seronegativity.⁵³ Given the detection of MuSK IgA Abs in a limited number of patients, our study was not positioned to conclusively establish whether MuSK IgA Ab seropositivity correlates with specific clinical patterns; larger cohorts will be required to identify such associations. This limitation notwithstanding, we found MuSK IgA Abs in patients from two independent MG cohorts, further supporting the generalizability of our findings. Cell-based screening of individuals with other neurologic diseases (OND) identified a MOGAD sample harboring low levels of MuSK IgA Abs. Notably, two different IgA binding thresholds were observed when comparing the OND and the MuSK MG sera. This finding could signal qualitative differences (such as variations in affinity) among MuSK IgA Abs,⁵⁴ which warrants further confirmation. It is conceivable, nonetheless, that, by overexpressing MuSK, our screening method enabled detection of both low- and high-affinity IgA Abs. Low-affinity Abs, for instance, can include naturally occurring Abs (NAbs), germline-encoded immunoglobulins that are polyreactive and target conserved epitopes.^55,56 Predominantly of the IgG and IgM isotype, NAbs also include IgA and are produced without explicit antigenic stimulation.⁵⁷ In sharp contrast to NAbs, the IgA mAbs generated from three MuSK MG patients showed a high frequency of somatic hypermutation, clear specificity for MuSK, and relative binding curves that were comparable to that of MuSK1B, an IgG4 mAb with known, high affinity for MuSK.³⁹ Although direct affinity measurements lie beyond the scope of this study, the molecular features of patient-derived IgA mAbs confirm the contribution of a previously unrecognized Ab isotype in the MuSK-specific repertoire.

In humans, IgA has two subclasses: IgA1 and IgA2. The three MuSK IgA mAbs were cloned from IgA1 B cells. IgA1 is the predominant IgA subclass in serum where it is mostly found in monomeric form, whereas at mucosal surfaces, secretory IgA1 and IgA2 antibodies exist as dimers, and their relative proportions vary based on the mucosal site.^58,59 While secretory IgA antibodies represent the first line of defense against toxins and pathogens invading the mucosa, the function of serum IgA is poorly understood. Serum IgA can bind to multiple receptors including the myeloid-cell-specific type I Fc receptor for IgA (FcαRI), which is expressed by neutrophils, eosinophils, monocytes, and macrophages, and it is believed to induce inhibitory signals and contribute to immune homeostasis. Of note, FcαRI has been shown to facilitate the internalization of IgA1-opsonized bacteria by hepatic Kupffer cells in a non-inflammatory context, suggesting that IgA1 antibodies represent a second mechanism of defense against intestinal bacteria that invade the portal venous system.⁶⁰ Whether MuSK IgA1 Abs can recognize bacteria or other pathogens remains undetermined; this possibility would imply molecular mimicry, a mechanism through which certain pathogens trigger the production of antibodies that, in turn, cross-react with self-proteins by virtue of structural homology.⁶¹ Some of the strongest evidence supporting this concept comes from recent investigations on Abs in rheumatoid arthritis (RA).^62,63 Clonally related IgG and IgA Abs binding to RA-relevant autoantigens have been shown to cross-react against a gut-derived bacterial strain of Subdoligranulum, and mice colonized with the bacterial isolate developed arthritis with pathology similar to human RA.⁶² The role (or lack thereof) of gut commensals (or other pathogens) in MuSK MG initiation and/or progression remains undetermined, but the generation of MuSK-specific IgA mAbs provides both a rationale and an unprecedented opportunity for analogous, mechanistic investigations.

Molecular and functional profiling of individual IgA mAbs showed binding to different MuSK domains, varied agonistic potencies on MuSK signaling, and functional antagonism of two IgG4-derived Fabs in vitro. The latter effect was not simply due to competitive binding for the same (or proximal) MuSK epitopes, because a partial rescue of clustering was also observed when an IgA mAb and Fab targeted different MuSK epitopes. This observation aligns with a MuSK MG mouse model in which an agonist mAb targeting the Fz-domain activated MuSK and functionally antagonized Ig-like 1 domain-specific FAE IgG4 mAbs.⁶⁴. In addition, both aMu1 and aMu3 partially reduced AChR clustering in the presence of agrin, suggesting pathogenic potential, and their combination significantly amplified this effect. These findings share commonalities with a study in which bivalent IgG1–3 Abs from patients’ pooled sera impaired AChR clustering while still activating MuSK signaling.²⁹ IgG1–3-mediated pathogenic effects were also observed with DOK7-overexpressing myotubes, wherein cluster formation is spontaneous and does not require agrin.²⁵ Because DOK7 acts downstream of MuSK, it was hypothesized that the effect of MuSK IgG1–3 Abs may not be limited to the disruption of the canonical clustering pathway.

Further details of the mechanism underpinning the detrimental effect of bivalent clones remain to be found, but our experiments indicate that polyclonality may be a key factor that amplifies the pathogenicity of bivalent Abs. In line with this hypothesis, a recent investigation demonstrated induction of myasthenic symptoms in mice using a polyclonal mixture of Abs specific for the Fz-like domain.⁶⁵ It is plausible that while the pathogenic potential of single bivalent clones is counterbalanced by their agonistic capacity, the combination of two (or more) clones can override MuSK agonism, leading to overt pathogenicity. Consistent with this, in our passive immunization experiment, injection of two bivalent IgA mAbs — emulating patient polyclonal serum —induced a myasthenic phenotype in mice. A speculative working model is graphically depicted in Fig.7. One attractive mechanistic hypothesis is that the binding of multiple Abs could induce conformational changes that “lock” the extracellular portion of MuSK in a non-physiological configuration, hampering the subsequent formation and/or stability of AChR clusters. Indeed, a role of the MuSK ectodomain in mediating the association of MuSK with other synaptic components has been proposed.^66,67 An alternative possibility is that the activation of the MuSK signaling cascade induced by multiple Abs stimulates a negative feedback loop, resulting in clustering impairment. In dose-response and time-course experiments performed on myotubes, however, agrin and an agonist Fz-specific mAb elicited remarkably similar phosphorylation cascades.²⁶ In aggregate, our findings strongly suggest that the cooperation among bivalent clones represents a novel pathogenic determinant in MuSK MG. We and others have described a similar mechanism in the context of AChR MG immunopathology wherein multiple AChR-specific clones can synergize to enhance pathogenic effects.^36,68 Future studies are warranted to elucidate the molecular and functional interactions of IgA and IgG1–3 clones in serum and to assess how these interactions affect MuSK function.

Figure 7. Speculative working model for MuSK IgA autoantibody-mediated pathogenic mechanisms.

(A) In the healthy neuromuscular junction (NMJ), neural agrin is secreted from motor nerve terminals and binds to LRP4 on the postsynaptic membrane, enhancing MuSK-LRP4 interaction and promoting MuSK dimerization and (auto)phosphorylation of its cytoplasmic domain. Activated MuSK initiates a complex intracellular cascade that culminates in AChR clustering, which is essential for efficient neuromuscular transmission (levels of clustering are denoted as +++/++/+ throughout the whole schematic). (B) MuSK IgG4 Abs act through their functional monovalency (which prevents cross-linking of MuSK) and sterically block MuSK-LRP4 binding, thereby inhibiting agrin-dependent signaling. This results in a marked impairment of AChR clustering and the failure of neuromuscular transmission. Compellingly, passive transfer of MuSK1A, a patient-derived IgG4 mAb — expressed as a bispecific mAb (MuSK1A-KLH) to emulate IgG4 monovalency — induced myasthenic symptoms in passively immunized mice. (C) MuSK IgA Abs, unlike IgG4 Abs, are structurally bivalent and thus capable of cross-linking and activating MuSK. As single monoclonal species, they behave as MuSK agonists and promote clustering (+/++). Furthermore, it is conceivable that each Ab can interfere with agrin signaling (similar to pathogenic IgG4 Abs), thereby hampering agrin-dependent clustering activity. Due to the concurrent agonistic potency, however, the net effect of single IgA clones is only a partial reduction in clusters compared to the agrin-only effect. This reduction may lead to no or modest pathogenicity. (D) When multiple MuSK IgA Abs of different epitope specificities are present (e.g., in a polyclonal response), their combined bivalency can result in dysfunctional cross-linking. This multivalent binding may hamper signaling pathways associated with the MuSK ectodomain or it kinase activity, leading to impaired cluster formation. In this case, the inhibitory mechanisms of multiple IgA clones outweigh the agonism of single clones, and agrin-mediated clustering is severely diminished, resulting in conspicuous pathogenic effects. In keeping with this, co-administration of two IgA mAbs (aMu1 and aMu3) induced muscle weakness in vivo. In aggregate, this mechanistic model proposes that the pathogenic capacity of MuSK IgA and potentially of other bivalent Abs is enhanced by the synergistic action of multiple clones. Single bivalent clones may partially counteract MuSK IgG4 Abs’ effect due to their agonistic potency (E), but multiple clones can override MuSK agonism and demonstrate cooperative pathogenicity.

The identification of aMu1 and multiple, historic clonal variants resisting consecutive cycles of RTX generates questions regarding the phenotype of MuSK IgA B cells and the specific mechanisms that confer resistance to anti-CD20 depletion. We previously showed that short-lived plasmablasts are key IgG4 Ab-producing cells in MuSK MG and that MuSK-specific IgG4 clones can also escape depletion, reemerging into circulation prior to clinical relapse.^35,37 In the current study, appreciable differences between the kinetics of MuSK IgG and IgA levels in response to RTX suggest that IgA B cells may be less susceptible to CD20 B-cell depletion than IgG4 B cells. A “resilient” pool of circulating IgA B cells has also been described in patients with multiple sclerosis, treated with ocrelizumab.⁶⁹ Among the numerous hypotheses that can explain resistance to CD20 depletion, the most plausible explanations are the limited efficacy of the drug in targeting tissue-resident B cells or the lower expression of CD20 in residual memory B cells.^70,71 Because in humans IgA production primarily occurs in the intestinal Peyer’s patches, an attractive explanation is that a subset of mucosal resident IgA B cells is self-sufficient and not replenished by CD20⁺ B cells. Compellingly, in a study on patients with RA, the majority of circulating plasmablasts — observed during B-cell depletion —expressed IgA as well as β7 integrin and CCR10, which are receptors for mucosal homing.⁷² In the same study, proliferating IgA plasmablasts, but not CD20⁺ cells, were detected in mucosal biopsies from patients with diffuse large B-cell lymphoma treated with RTX. These data support the hypothesis that mucosal B cells can resist RTX treatment and continuously differentiate into circulating IgA-producing plasmablasts. It is tempting to speculate that RTX-resilient MuSK IgA B cells could represent an unrecognized cellular reservoir poised to perpetuate MuSK autoimmunity. It is also plausible that MuSK IgA and IgG4 B cells respond similarly to RTX but differ in the kinetics that govern post-RTX repopulation. Future investigations will focus on the origin, location, and phenotype of such cells.

The detection of IgA Abs in an IgG4-driven disease opens new questions regarding the pathways leading to IgG4 class-switching. These are of substantial interest not only for MuSK MG but also for a better understanding of the autoimmune mechanisms underlying other IgG4-mediated conditions. Prompted by the isolation of MuSK IgA1 B cells, we analyzed bulk BCR sequencing libraries derived from two MuSK IgA-seropositive patients, but we did not find clones sharing MuSK IgA and IgG isotypes which would provide evidence for IgA-IgG class-switching events in vivo. These data align with recent findings in pemphigus vulgaris, an IgG4-mediated blistering disorder in which Abs target the adhesion proteins desmoglein (DSG) 3 and DSG1. In this condition, phylogenetic analysis combined with subclass-specific B cell deep sequencing showed that the majority of anti-DSG IgG4 B-cell clones lack relatives in other subclasses and evolve independently of IgA1.⁷³ Conversely, most anti-DSG IgA1 B cells demonstrated evidence of class switch from IgG1 or to IgA2.

Finally, our study has translational implications. Recent therapeutic advances for Ab-mediated autoimmune conditions include inhibitors of the neonatal crystallizable fragment receptors — such as efgartigimod and rozanolixizumab, recently approved for MG^74,75 — as well as engineered IgG-specific cleaving enzymes which are emerging as promising interventions.⁷⁶ These treatments are effective at reducing pathogenic IgG levels but do not affect other immunoglobulin isotypes. Given that MuSK IgG4 Abs were detected at higher titers compared to MuSK IgA Abs, it is reasonable to expect that most patients will show clinical improvement with IgG-targeting therapies. Nevertheless, future studies are needed to evaluate the long-term efficacy of such interventions and clarify the extent to which they will benefit patients harboring MuSK IgA Abs.

Additionally, in our cohort, two patients with MuSK IgA Abs were treated with RTX and followed longitudinally for several years. One patient (MuSK-2) exhibited a transient MuSK IgA response and — following two RTX cycles — achieved complete stable remission, which persisted at the last follow-up, eight years after immunosuppression withdrawal. In contrast, the other patient (MuSK-1) experienced a protracted disease course, characterized by durable MuSK IgA positivity, clonally expanded MuSK IgA B cells, and recurrent relapses requiring chronic RTX administration. These observations suggest that sustained MuSK IgA responses may define a subset of patients who remain dependent on long-term B-cell depletion. Collectively, our findings generate new hypotheses regarding mechanisms of clinical relapse, resistance to treatment, and disease perpetuation, warranting future research.

Limitations of the study

Immunosuppressive treatments during sampling may have affected Ab levels, potentially leading to an underestimation of the true prevalence of MuSK IgA Abs. Second, given the total number of patients in our cohorts and the limited number of MuSK IgA seropositive cases, our results may not be broadly representative of the whole MuSK MG population. For several patients, however, serially collected samples were available, which allowed longitudinal tracking of their MuSK Ab levels. Third, the epitope mapping strategy employed to evaluate polyclonality may have overlooked unique clones targeting different epitopes within the same domain. In addition, while our BCR sequencing analysis identified IgA clonal variants, no evidence of MuSK IgA-IgG classs-witching events was found, possibly due to limited sampling of the circulating B-cell repertoire.

Conclusions

By combining serologic profiling, generation of patient-derived mAbs, and in vitro and in vivo experiments, we characterized a novel Ab isotype in MuSK MG that offered provocative insights into disease mechanisms. Our findings provide pathways for further research investigating the role of IgA in MuSK autoimmunity and in other IgG4-driven autoimmune conditions.

Supplementary Material

Supplementary material is available at Brain online.

Data availability

Anonymized data will be shared at request by qualified investigators following the execution of appropriate materials transfer agreements.