Autoantibody screening in Guillain–Barré syndrome

Guillain–Barré syndrome (GBS) is an acute inflammatory neuropathy with a heterogeneous presentation. Although some evidences support the role of autoantibodies in its pathogenesis, the target antigens remain unknown in a substantial proportion of GBS patients. The objective of this study is to screen for autoantibodies targeting peripheral nerve components in Guillain–Barré syndrome. Autoantibody screening was performed in serum samples from all GBS patients included in the International GBS Outcome study by 11 different Spanish centres. The screening included testing for anti-ganglioside antibodies, anti-nodo/paranodal antibodies, immunocytochemistry on neuroblastoma-derived human motor neurons and murine dorsal root ganglia (DRG) neurons, and immunohistochemistry on monkey peripheral nerve sections. We analysed the staining patterns of patients and controls. The prognostic value of anti-ganglioside antibodies was also analysed. None of the GBS patients (n = 100) reacted against the nodo/paranodal proteins tested, and 61 (61%) were positive for, at least, one anti-ganglioside antibody. GBS sera reacted strongly against DRG neurons more frequently than controls both with IgG (6% vs 0%; p = 0.03) and IgM (11% vs 2.2%; p = 0.02) immunodetection. No differences were observed in the proportion of patients reacting against neuroblastoma-derived human motor neurons. Reactivity against monkey nerve tissue was frequently detected both in patients and controls, but specific patterns were only detected in GBS patients: IgG from 13 (13%) patients reacted strongly against Schwann cells. Finally, we confirmed that IgG anti-GM1 antibodies are associated with poorer outcomes independently of other known prognostic factors. Our study confirms that (1) GBS patients display a heterogeneous repertoire of autoantibodies targeting nerve cells and structures; (2) gangliosides are the most frequent antigens in GBS patients and have a prognostic value; (3) further antigen-discovery experiments may elucidate other potential antigens in GBS.


Background
Guillain-Barré syndrome (GBS) is an acute inflammatory neuropathy with a heterogeneous presentation that includes diverse clinical variants [1][2][3]. Diagnosis is based on clinical criteria; diagnostic biomarkers are not available for most patients [4]. The exact immunopathogenic mechanisms of GBS are relatively unknown, but it is considered a paradigmatic post-infectious autoimmune disease [5]. Diverse mechanisms, including humoral and cellular immune responses, autoantibodies and complement, activated macrophages and lymphocytes, have been implicated in GBS pathogenesis [6,7].
Considering the broad clinical and epidemiological spectrum of GBS, the diverse infectious triggers and the T-cell independent nature of the immune reaction leading to the appearance of autoantibodies [20], we hypothesized that a broad repertoire of autoantibodies targeting diverse nerve components may be causing nerve pathology in GBS. This study aims to (1) screen for autoantibodies against known antigens; (2) screen for antibodies against human and rodent nerve cells and monkey nerve tissue; (3) describe the diversity of staining patterns and (4) perform clinical-immunological correlations, in a well-characterized GBS cohort.

Patients and controls
Serum samples from 100 GBS patients included in the Spanish cohort of the International GBS Outcome study (IGOS) [21] were used in this screening. The IGOS is a multicentre, prospective, observational cohort study that investigates factors that determine and predict the clinical course, subtype and outcome of GBS [22]; including patients fulfilling diagnostic criteria of the National Institute of Neurological Disorders and Stroke (NINDS) [1] or Miller Fisher syndrome (MFS) and other variants of GBS [3,23]. Patients from the Spanish cohort were enrolled between February 2013 and January 2020. All patients fulfilled diagnostic criteria for GBS and were included within 2 weeks from onset of weakness. Serum samples were aliquoted and stored at − 80 °C until needed. In this study, we used serum samples extracted at baseline. Sixty-two (62%) of the baseline samples analysed were collected before starting treatment.
Clinical variants were defined as sensorimotor, pure motor, pure sensory, Miller Fisher syndrome (MFS) and ataxic. Nerve conduction studies results were classified as acute inflammatory demyelinating polyneuropathy (AIDP), acute motor axonal neuropathy (AMAN), acute motor-sensory axonal neuropathy (AMSAN), equivocal or normal. The outcome of all patients with GBS at 6 months and 1 year from disease onset were assessed using the GBS disability score (GDS), a widely accepted system for evaluating the functional ability of patients [24]. Patients unable to walk independently (≥ 3) at 6 months were defined as having a poor outcome in this study.

Autoantibody screening protocol
Autoantibody screening experiments included antiganglioside antibody detection with ELISA, nodo/paranodal (NF155, NF140, NF186, CNTN1 and CNTN1/ CASPR1 complex) antibody detection by ELISA and cell-based assays, immunocytochemistry using patient sera on murine dorsal root ganglia (DRG) neurons and neuroblastoma-derived human motor neurons (IgG and IgM) and reactivity pattern assessment by immunohistochemistry on monkey sciatic nerve sections (IgG and IgM).
Cell-based assays were used as previously described [25] as a second confirmatory technique for questionable cases. Briefly, mammalian expression vectors encoding human NF140, NF186, NF155, CNTN1 or CASPR1 were transfected into HEK293 cells using Lipofectamine 2000 (Invitrogen). Cells were then fixed with 4% paraformaldehyde and blocked. ICC experiments were performed using patient's sera and appropriate primary and secondary antibodies.

Testing for anti-ganglioside antibodies
Patients' sera were screened for the presence of antiganglioside antibodies using a previously validated ELISA protocol [26] as the general detection method, and thin layer chromatography [27] for confirmatory experiments. Anti-ganglioside antibodies were considered positive at a 1/1000 titre.

Rat dorsal root ganglia neurons immunocytochemistry
DRG were dissected from E16 rat embryos, dissociated and plated in glass coverslips coated with laminin (Invitrogen) and poly-d-lysine (Sigma, MO, USA). Cells were grown in neurobasal medium (Gibco BRL, NY, USA) supplemented with B27 (Gibco), Glutamax (Gibco) and nerve growth factor (NGF) (Invitrogen). After 24 h, cytosine arabinoside (ARA-C) (Sigma) and fluorouracil (5-FU) (Sigma) were added to the medium to remove fibroblasts. Then, medium was replaced every other day until reaching complete growth and differentiation of DRG neurons.
Live DRG neurons were incubated for 1 h with patients' sera diluted 1/100 (for IgG experiments) or 1/40 (for IgM experiments) in culture medium at 37 °C. Cells were then fixed for 10 min with 4% PFA and incubated with secondary antibodies. Goat anti-human IgG or IgM AF488 (Molecular probes, Oregon, USA) were used as secondary antibodies at 1/1000 concentration.
Coverslips were mounted with Vectashield with DAPI and fluorescence signal intensity was scored in a 0-3 scale by two independent researchers. Images were obtained with an Olympus BX51 Fluorescence Microscope (Olympus Corporation, Tokyo, Japan).
Coverslips were mounted with Vectashield with DAPI and fluorescence signal intensity was scored in a 0-3 scale by two independent researchers. Images were obtained with an Olympus BX51 Fluorescence Microscope.

Peripheral nerve immunohistochemistry
Macaque peripheral nerve tissue slides (Inova Diagnostics, Inc., San Diego, CA) were blocked with 5% normal goat serum in PBS; followed by incubation with patients' sera at 1:10 (for IgM) or 1:20 (for IgG). Monkey-adsorbed goat anti-human IgG AF488 (Southern Biotech, Alabama, US) or goat anti-human IgM AF488 (Molecular Probes) were used as secondary antibodies at 1/500 concentration. Finally, slides were mounted with Fluoromount medium (Sigma) and examined by two independent observers. Immunostaining patterns were analysed scoring fluorescence signal intensity of each nerve structure in a 0-3 scale. The nerve structures analysed were: nodes or paranodes, myelin from small myelinated fibres, myelin from large myelinated fibres, Schwann cells from unmyelinated fibres, large-fibre axons, and small-fibre axons. Reactivity against other non-nerve structures (fibroblasts, connective tissue, vessels) was not considered in the analysis.
Images were acquired using Leica TSC SP5 confocal microscope.

Statistical analysis
Results were analysed by GraphPad Prism v8.0 (Graph-Pad Software). Statistical comparison of proportions among groups was performed using contingency analysis with the application of Chi-square and a two-tailed Fisher's exact test, accepting an alpha-level < 0.05 for statistical significance. To represent the results and perform hierarchical clustering of the results heatmap diagrams using the Clustvis web tool were performed [28].
To investigate the association between anti-ganglioside antibodies and prognosis a multivariable logistic regression analysis to predict the inability to walk at 6 months and at 1 year of follow-up (GDS ≥ 3) was performed using the STATA software. A stepwise backward regression modelling to select variables independently associated with the outcome was performed first. The variables introduced in our initial multivariable models were selected based on known prognostic factors: age, initial GDS, diarrhoea, AMAN, serum NfL levels (analysed in a previous study with the same cohort) [29], serum anti-GM1 IgG antibodies and serum anti-GD1a IgG antibodies [29][30][31][32]. To perform the multivariable analysis patients with MFS were excluded, because our aim was to predict GBS prognosis and MFS is considered a different disease, including different pathophysiology, clinical presentation (it does not present with weakness), treatment (often untreated) and outcome (considered self-limiting and benign). Finally, the ability of the variable "presence of anti-GM1 IgG antibodies" to predict the inability to run at 1 year of follow-up (GDS ≥ 2) was evaluated in our previously reported multivariable logistic regression analysis [29].
Odds-ratios (OR) for the logistic regression analysis were reported with 95% confidence intervals and p values.

Baseline characteristics
We included 100 participants from 11 Spanish centres participating in the IGOS study. GBS patients had an average of 57.4 years and were predominantly men (57%). 65% of patients presented with the sensorimotor variant, 19% presented with the pure motor GBS variant, 10% with MFS, 5% with the pure sensory variant and 1 patient with the ataxic variant. Regarding nerve conduction studies, 59% of patients were classified as AIDP, 12% as AMAN, 7% as AMSAN, 8% as normal, and 14% as equivocal. Detailed epidemiological features of the cohort were described elsewhere [29].

Screening for known autoantibodies
None of the GBS patients included in the study reacted against the paranodal and nodal proteins tested (NF155, NF140, NF186, CNTN1 and CASPR1).

Antibodies targeting peripheral nerve neurons
ICC experiments with primary cultures of rat DRG neurons and human motor neurons derived from a neuroblastoma cell line were used to identify novel IgG and IgM reactivities against peripheral nerve neurons. The screening was performed in 100 serum samples from GBS patients and 90 serum samples from a control group (including healthy controls and patients with other neuromuscular diseases). ICC results were grouped in three separate categories: moderate-to-strong positives (including scores 2 and 3), all positives (including scores 1, 2 and 3), and negatives (score 0). Detailed results are shown in Additional file 2, Fig. 1 and Table 1.
Antibodies against DRG neurons appeared significantly more frequently in GBS patients than in controls (32% vs 6.7%, p < 0.0001) taking all positive tests in account; the same happened if only moderate and strong positives were considered, both in IgG (6% vs 0%, p = 0.03) and IgM experiments (11% vs 2.2%, p = 0.02).
In neuroblastoma-derived neuron ICC experiments 28 (28%) samples from the GBS group showed IgM autoantibodies; of these 8 (8%) showed moderate or strong reactivity. These proportions were significantly higher than in the control group (12.2% and 2.2%, respectively; p = 0.011). Differences in autoantibody proportions between GBS patients and controls were not observed  in neuroblastoma-derived neuron experiments when assessing IgG antibodies.

Antibodies targeting peripheral nerve tissue
We analysed the full GBS cohort and 56 controls. We analysed the staining intensity of six different structures within the nerve, including nodes or paranodes, myelin from small myelinated fibres, myelin from large myelinated fibres, Schwann cells from unmyelinated fibres, large-fibre axons, and small-fibre axons. Staining patterns can be found in Additional file 3.
IgG and IgM reactivity against nerve tissue was frequently detected in GBS patients and controls. Overall, about 70% of GBS patients and controls sera bound to nerve structures. IgG and IgM from GBS patients reacted moderately in 17 (17%) and strongly in 10 (10%) against monkey nerve structures. In the control group IgG and IgM reacted moderately in 8 (14.3%) and strongly in 1 (1.8%) against monkey nerve structures. The difference between the amount of GBS patients and controls reacting moderately or strongly against monkey peripheral nerve was statistically significant (p = 0.0455) only for IgG autoantibodies (Table 1).
Differences in IHC patterns of reactivity from GBS patients and controls were not statistically significant for any of the structures analysed (Additional file 4). Nonetheless, some specific reactivity patterns were only detected in GBS patients and not in controls (Fig. 2). Eight (8%) GBS patients' IgG reacted strongly against myelin, whereas only 2 controls showed weak reactivity against this structure. Moreover, we observed that 13 (13%) GBS patients' IgG had a strong reactivity against Schwann cells (myelinating and non-myelinating) while only one of the controls (1.8%) showed strong reactivity against Schwann cells (this difference is statistically significant; p = 0.0192).

Combined autoantibody screening analysis
We also analysed if GBS patients with or without antiganglioside antibodies differed in the reactivity patterns in the peripheral nerve cell and tissue autoantibody screening experiments. No differences were found between those two groups (Additional file 4), suggesting that the heterogeneity of the autoantibody repertoire appears even when a specific antigen is found.
We used a heatmap graph to represent all the autoantibody screening results performed in our GBS cohort (Fig. 3) [28]. This graph provides visual representation of the heterogeneity of the autoantibody repertoire in GBS sera.

Clinical correlations
Among patients with Miller Fisher syndrome, 8/10 (80%) had IgG anti-GQ1b antibodies, whereas in the rest of GBS patients only 14.4% (13/90) had these antibodies, usually in combination with other reactivities. IgG anti-GM1 antibodies were more frequently detected in patients with the pure motor variant than in those with other clinical variants [13/19 (68.4%) vs 14/81 (17.3%)]; and in patients classified electrophysiologically as AMAN than in the rest of GBS patients (83.3% vs 19.3%). All these differences were statistically significant (p < 0.0001). When we analysed the general clinical characteristics of the subgroup of patients with strong IgG reactivity against Schwann cells (n = 13), we did not observe any specific pattern that could distinguish them from the rest of the cohort. Ten (76.9%) of these patients presented with the sensorimotor clinical variant, whereas in the general cohort 65% presented this variant; and the proportions of nerve conduction studies subgroups were similar to those found in the general cohort (53.8% vs 59% of AIDP electrophysiological variant). Regarding the outcome, the percentage of patients having a good outcome at 6 months and 1 year are similar in the two groups (about 75%).
In the subgroup of patients with IgG or IgM reactivity against DRG neurons (n = 14), we did not find clinical differences with the whole GBS cohort. Briefly, 71.4% of patients staining strongly DRG neurons were classified as AIDP, and 64.3% presented with the sensorimotor clinical variant.
We did not detect any difference in peripheral nerve cell and tissue reactivity patterns or frequencies between the samples collected before starting the treatment (62%) and those collected after the treatment (38%). We neither observed any correlation of

Prognostic value of anti-ganglioside antibodies
First, we conducted a univariate analysis to select variables that were associated with the outcome. Patients with serum IgG anti-GM1 antibodies presented poorer outcomes than patients without the antibodies at 6 months [38.1% vs 16.1% (p = 0.04)], and 1 year [35.3% vs 9.7% (p = 0.014)]. Anti-GD1a IgG antibodies were not associated with prognosis ( Table 2). For the multivariate analysis, we included GM1 IgG, serum NfL levels, diarrhoea, age, and initial GDS.
We observed that having anti-GM1 IgG antibodies at baseline was independently associated with the inability to walk at 1 year of follow-up, after a backward stepwise selection modelling (OR 6.98, 95% CI 1.6-30.36; p = 0.01). However, the presence of anti-GM1 IgG antibodies was not independently associated with having a poor outcome at 6 months ( Table 2).
To analyse if anti-GM1 titres were associated with the GBS disability score, we performed a linear regression. We did not observe a positive correlation between antibody titres and disability at 6 months and at 1 year.
Finally, when we included the presence of anti-GM1 antibodies in our previously reported prognostic study [29], we observed that having anti-GM1 IgG antibodies at baseline was associated with the inability to run at 1 year, but this association was not independent from the other known prognostic factors and sNfL, age and AMAN remained in the model as independent factors associated with residual disability at 1 year.

Discussion
Our work describes a comprehensive autoantibody screening that provides experimental evidence of the heterogeneity of the autoantibody repertoire in patients fulfilling GBS diagnostic criteria.
Our study shows that GBS patients have a heterogeneous repertoire of autoantibodies targeting nerve cells and structures. Except for patients with anti-ganglioside antibodies and a minor subset of patients with antibodies targeting Schwann cells and the myelin sheath, this repertoire varies in frequency and intensity of staining, but it is not qualitatively different from controls. Antibodies targeting peripheral nerve cells of both IgG and IgM isotypes are significantly more frequent in patients than in controls, but no clear differences are seen when antibodies are tested using immunohistochemistry on monkey nerve preparations. Considering that whole nerve monkey preparations likely display protein antigens in a conformation that is phylogenetically closer to that of human nerves, this may imply that autoantibodies targeting nerve structures are present in normal human repertoire at lower titers, that they arise as a natural epiphenomenon of a T-cell mediated damage and are not pathogenic, or that other autoantibodies, targeting different types of molecules (such as lipids or glycans), for which our techniques are not optimized, are yet to be discovered.
Whether these autoantibodies arise from a process of molecular mimicry, or from an unspecific and polyclonal activation of pre-existing B cells, remains unclear. The general absence of common patterns suggests the latter, but the well-established molecular mimicry process described in anti-ganglioside-associated GBS supports the former. In anti-GM1-associated GBS the sequence of pathogenic events includes an immune response to an infection leading to the appearance of antibodies cross-reacting with peripheral nerve and nerve root gangliosides and triggering post-infectious inflammation [33]. Interestingly, in this screening we did not find clear differences in the reactivity patterns between GBS patients with or without anti-ganglioside antibodies, but we observed in both groups a higher amount of patients staining nerve structures than in controls. These findings suggest that the immune response in GBS is not restricted to the production of anti-ganglioside antibodies, but it is also targeting other peripheral nerve structures. This observation may either reflect the presence of a polyclonal, not antigen-driven, reactivation of a pre-existing repertoire that, in some patients, includes gangliosides, or the concomitant activation (by epitope spreading or bystander activation) of unspecific B cells in addition to the ganglioside-driven antigen-specific response.
Previous studies in other inflammatory neuropathies such as chronic inflammatory demyelinating polyneuropathy (CIDP), showed that frequencies of reactivity against DRG neurons in CIDP patients did not differ from healthy controls [34], in contrast with our results (shown in Table 2). GBS and CIDP are similar diseases both clinically and electrophysiologically, so this difference supports the idea that a heterogeneous autoantibody response against multiple nerve antigens arises in GBS while this does not happen in chronic inflammatory and demyelinating neuropathies in which a specific, antigendriven autoantibody response arises, as the recent discovery of the nodo-paranodal antibodies supports [35]. Differences in severity of these two diseases may also account for this observation.
We observed that 13% of GBS patients showed strong IgG reactivity against Schwann cells of monkey peripheral nerve. This observation is in agreement with previous findings: Kwa et al. observed that 24% of GBS patients had IgG antibodies against  non-myelinating human Schwann cells [36], and Vallat et al. also detected that a significant percentage of CIDP and GBS patients (about 25%) presented with IgG or IgM reactivity against myelin and that the staining patterns on Schwann cells were diverse, suggesting that diverse myelin antigens are being recognized by the autoantibodies [37].
Our study also confirms, in a well-characterized GBS cohort, that gangliosides are the most frequent specific antigens in GBS patients and that they associate to specific disease variants. The value of testing antiganglioside antibodies in the GBS routine clinical care is controversial, but it is clear that some antibodies are associated with specific clinical phenotypes [38]. IgG anti-GQ1b antibody is a diagnostic marker and a pathogenic antibody in MFS, and is often cross-reactive with GT1a [9]. Moreover, IgG anti-GM1 antibodies associate with the pure motor (clinical) and AMAN (electrophysiological) variants. Our results, with 80% of MFS patients having anti-GQ1b antibodies and 68.4% of pure motor patients having anti-GM1 antibodies at baseline, confirm these associations. However, our study lacks power to find other potential associations previously described (anti-GD1b with acute ataxic neuropathy, anti-GT1a and pharyngo-cervico-brachial variant) [39,40] that will need to be confirmed in even larger cohorts. Likewise, the clinical relevance of antibodies targeting other structures (neurons, peripheral nerve tissue…) is unclear, since their association with GBS is not completely specific or the number of patients with each particular reactivity is too low to draw any conclusions.
Some studies have reported a correlation between IgG anti-GM1 and anti-GD1a antibodies with a poor outcome in GBS patients [13,14,33,41]. In our cohort, IgG anti-GD1a antibodies did not associate to a poor outcome of the disease [13]. However, our data confirm that IgG anti-GM1 antibody is an independent prognostic factor that associates with poor prognosis at 1 year, supporting that it may be a marker for long-term axonal damage. Whether the presence of complement-fixing anti-GM1 antibodies is the driver of this long-term disability, an important therapeutic question (that would enable the use of complement inhibitors in these patients), remains to be elucidated.
Although in this study we analysed the prognostic value of anti-ganglioside antibodies using the traditional outcome measures: inability to walk (GDS ≥ 3) at 6 months and at 1 year, we have recently used the inability to run (GDS ≥ 2) as a measure of the presence of longterm residual disability. In this recent study we showed that high baseline sNfL were independently associated with inability to run at 1 year [29]. In agreement with these findings, we observed that including in the model the variable "presence of serum IgG anti-GM1 antibodies", sNfL levels remained as an independent prognostic factor, whereas anti-GM1 antibodies did not. These results confirm that sNfL levels are a prognostic factor that informs better on axon status and, consequently, on long-lasting disability.
It is interesting to note that we did not find any patient with anti-nodal/paranodal antibodies (CNTN1, NF140, NF186, NF155 and CASPR1) in our GBS cohort. Although previous studies from other authors have found some GBS positive patients in their cohorts [14][15][16][17][18], and case-reports and series describe the association of antinodal/paranodal antibodies with aggressive inflammatory neuropathies frequently misdiagnosed as GBS, these antibodies are rare and we cannot rule out the possibility that they are present in other selected patients that our cohort failed to capture.
One of the limitations of our study is the number of patients and controls included. We have small groups of patients with similar staining patterns in which it is difficult to establish clear clinical-immunological correlations. Nevertheless, this is the first large prospective study assessing the autoantibody repertoire against peripheral nerve structures in GBS patients and antigenidentification experiments will follow in those patients showing specific staining patterns that are absent in controls.
The existence of clear subgroups associated with antiganglioside antibodies, in contrast with the diversity in the new reactivities analysed, suggests that this apparent heterogeneity may be also due to technical caveats, because our study protocol is optimized for proteins and not for lipids or glycans. Moreover, other, not properly controlled factors, could have influenced heterogeneity in staining patterns (treatment, comorbidities…), and will need to be assessed in larger cohorts.

Conclusions
In conclusion, our study highlights the heterogeneity of the profile of autoantibodies targeting peripheral nerve structures, confirms gangliosides as the most frequent target antigens in the GBS autoantibody repertoire and their prognostic value in long-term GBS prognosis, and identifies small subsets of GBS patients with specific staining patterns in which further antigen-identification experiments could demonstrate novel and clinically relevant autoantibody reactivities in the future.