1 INTRODUCTION

Guillain-Barré syndrome (GBS) is an immune-mediated polyradiculoneuropathy and the most common cause of acute flaccid paralysis worldwide.1 GBS is usually preceded by an infection, and several pathogens have been associated with GBS in case-control studies, including Campylobacter jejuni, hepatitis E virus (HEV), cytomegalovirus (CMV), Epstein-Barr virus (EBV), and Mycoplasma pneumoniae.2-5 During the Zika virus (ZIKV) epidemic in 2015-2016 in Latin America, an increased incidence of GBS patients was observed and an association between ZIKV and GBS has later been confirmed.6-8

ZIKV is a flavivirus that is transmitted by the Aedes aegypti mosquito. Other arthropod-borne viruses (arboviruses) transmitted by the same mosquito, including dengue virus (DENV) and chikungunya virus (CHIKV), have also been associated with GBS, although evidence of an association is limited in comparison with ZIKV.6, 9-18 Most studies on DENV and GBS are limited to case series,10, 18-22 although two surveillances studies17, 18 showed a temporal association between the incidence of GBS and DENV, and one case-control study provided evidence of an association between GBS and DENV.23 Several studies have linked clusters of GBS cases with outbreaks of CHIKV,15, 24-26 and a case–control study9 demonstrated that CHIKV is a risk factor for GBS. Arboviruses have been increasingly recognized as a global health threat, as their geographic distribution has spread dramatically over the past decades.12, 27, 28 Roughly half of the world's population is currently living in areas at risk for transmission of these viruses, and especially countries in Latin America and Southeast Asia are at risk.29

Previous studies that demonstrated a link between GBS and ZIKV or other arboviruses were carried out during epidemic phases of viral transmission, and it is unknown whether these viruses also play a role in the occurrence of GBS in endemic phases. Another aspect of arbovirus-related GBS that has not been illuminated is the possible role of coinfections with other known triggers of GBS, as most previous studies only tested for arbovirus infections. Furthermore, the underlying pathophysiology and the role of antibodies to specific gangliosides and other glycolipids on the nerve axon has not been uniformly demonstrated for GBS related to arboviruses.26, 30-33

The International Guillain-Barré syndrome Outcome Study (IGOS) is an international observational prospective cohort study on the disease course and outcome of GBS patients.34 The protocol and infrastructure of this study were used and adapted to develop a case–control study (“IGOS-Zika study”) to investigate the association between GBS and arboviruses, and specifically whether these infections drive the occurrence of GBS beyond the peaks of epidemics. Samples were tested for a broad range of infections that are known to trigger GBS and for antibodies against glycolipids to investigate the role of coinfections and anti-glycolipid antibodies in arbovirus-related GBS.

2 METHODS 2.1 Study design

The study protocol of IGOS has been published elsewhere.34 This protocol was adapted to investigate the association between arbovirus infections and GBS. Additional questions regarding immunization history and preceding symptoms and signs of arbovirus infections were collected. Where possible, two hospital-based controls were collected for every case. Controls were sex- and age-matched (age difference <10 years) and were treated in the same hospital and collected within 10 days of the included case. Controls were excluded if they had been diagnosed with GBS 1 year prior or if they were admitted for a (post-)infectious disorder. The same questions on arbovirus history and a serum sample were collected from the controls. Otherwise, the protocol was identical to the original IGOS protocol. Patients were enrolled in two study sites in Brazil, four sites in Argentina, and one site in Malaysia. The IGOS study (MEC-2011-477) and the amendment of the study protocol (NL38706.078.11) were approved by the review boards of Erasmus MC University Medical Center, Rotterdam, The Netherlands. The study protocol was also approved by the local institutional review boards of all participating hospitals or universities. Written informed consent was obtained from all patients or their legal representatives.

2.2 Data collection

Data were collected on demography, antecedent events, and neurological symptoms and signs of GBS at study entry and at 1, 4, and 26 weeks.34 Additional collection of data at weeks 2, 8, 13, and 52 was optional. Muscle strength was recorded by the Medical Research Council (MRC) score and disability by the GBS disability score.35, 36 Disease nadir was defined as the first visit that the lowest MRC sum score was found during the first 4 weeks from study entry. When there was no muscle weakness, the GBS disability score was used instead. The results of routine cerebrospinal fluid (CSF) examination and nerve conduction studies were collected. To determine the electrophysiological subtype, raw data of the first nerve conduction study, local reference values, and an algorithm were used to classify each nerve conduction study according to the criteria of Hadden et al by two independent clinical neurophysiologists (SA, JD).37 Patients were categorized according to the Brighton Collaboration criteria based on the available data.38 Insufficient data were available to categorize the Miller Fisher syndrome (MFS) patients according to the published criteria, and all patients with clinical variants of GBS without limb weakness were categorized as Level 4. The ability to walk at 6 months was used to determine the outcome. For patients with missing data at the 6-month visit, who were able to walk independently at the previous visit (week 13 or week 8), this visit was used to determine the outcome.

2.3 Diagnostic virology and bacteriology

All patients and controls with available serum samples were tested for a recent infection with C jejuni, HEV, M pneumoniae, CMV, EBV, DENV, ZIKV, and CHIKV. Serum samples collected at entry or week 1 were used where possible; otherwise, samples collected at week 2 or 4 were used. Antibodies against C jejuni were determined using an indirect enzyme-linked immunosorbent assay (ELISA) for IgG and antibody class capture ELISAs for IgM and IgA antibodies, as previously described.39 IgM and IgG antibodies against HEV and M pneumoniae were determined using commercially available ELISAs (Wantai, Beijing, PR China, respectively, Serion ELISA classic M pneumoniae, Serion GmbH, Würzburg, Germany). The presence of IgM and IgG antibodies and IgG avidity against CMV and of VCA IgM and viral capsid antigen (VCA) IgG and EBV nuclear antigen (EBNA) was determined by LIAISONXL (DiaSorin, Italy), a semi-automated system, which uses chemiluminescent immunoassay (CLIA) technology for detection of antibodies. The presence of IgM and IgG antibodies against ZIKV and DENV were determined using commercially available ELISA (EuroImmun, Lübeck, Germany). The presence of IgM and IgG antibodies against CHIKV was determined using a commercially available ELISA (Novatec), and immunofluorescence was performed to verify the presence of IgM. Immunofluorescence was leading in the interpretation of the results. In all patients that were IgM- or IgG-positive against ZIKV, a virus neutralization test (VNT) was performed to differentiate between a recent DENV and ZIKV infection.40 In general, IgM positivity is a good marker for a recent arbovirus infection, as studies have shown that ZIKV, CHIKV, and DENV IgM become positive starting the first week after onset of symptoms and usually persist for up to 2-3 months.41-43 Evidence of a recent infection was defined as IgM positivity for M pneumoniae and HEV, and IgM and/or IgA positivity for C jejuni. For CMV, IgM positivity with negative IgG or IgG with low avidity, and for EBV, VCA IgM, and VCA IgG positivity with negative EBNA IgG was considered indicative of a recent infection. For ZIKV, IgM positivity confirmed by VNT, and for CHIKV, IgM positivity in immunofluorescence was considered indicative of a recent infection. For DENV, NS1 positivity was considered indicative of a recent (re)infection as well as the combination of IgM and IgG positivity. Low-positive or borderline IgM with positive IgG was considered indicative of a previous infection (with possible reinfection with a different DENV strain) (Table S1).

2.4 Anti-glycolipid serology

Sera were tested with ELISA for IgG and IgM antibodies against GM1, GM2, GA1, GD1a, GD1b, GT1a, GQ1b, and GD3, and using combinatorial glycoarray for IgM and IgG anti-glycolipid antibodies against GM1, GM2, phosphatidylserine, GA1, GD1a, GD1b, GT1a, GQ1b, GD3, GalC, lactosylceramide, and sulfatide, plus their possible heterodimeric complexes.44, 45 Combinatorial glycoarray was performed using a thin-layer chromatography autosampler, which spotted glycolipids and glycolipid combinations onto in-house-made glass slides containing a polyvinylidene difluoride (PVDF) membrane.46 Antibodies were detected using AF647-conjugated goat anti-human IgM and Cy3-conjugated goat anti-human IgG (Jackson ImmunoResearch). Fluorescent intensity was measured using the appurtenant LuxScan software. The mean and SD were calculated for each glycolipid (−complex) using the fluorescent intensities of the control patients. Fluorescent intensities were considered positive if more than the mean plus three times the SD.

2.5 Statistical analysis

We used SPSS Statistics 21.0 for data analysis. Continuous data are presented as medians with interquartile ranges (IQRs) and dichotomized or categorical data as numbers and proportions. We used the Mann-Whitney U test and Kruskal-Wallis test to compare continuous data and the χ2-test or Fisher's exact test to compare proportions. A two-sided P-value of <0.05 was considered significant. For the case-control analysis, crude odds ratios were calculated (not matching for pairs) using contingency tables, and 95% confidence interval were calculated according to Altman, 1991.47, 48 The Cox proportional hazards model was used for the individually paired case-control analysis (SPSS COXREG function), adjusting for age and sex.49, 50 We used R version 3.6.1., packages dplyr 1.0.5 and ggplot2 3.3.2 for the development of the heatmaps. Raw data were clustered based on a distance matrix using Pearson's correlation and hierarchical cluster algorithm (Ward.2D) and clipped at a 10 000 upper limit.51

3 RESULTS

In total, 54 patients were included between July 2017 and December 2019. Five patients were excluded, four because of insufficient clinical data and one because of an alternative diagnosis (chronic inflammatory demyelinating polyradiculoneuropathy). For 22 of the remaining 49 patients, paired controls were collected, and they were included in the case-control analysis part of the study (Figure 1). Demographic and clinical features, ancillary investigations and outcome of the full cohort are described in Table 1.

Flowchart of inclusions in cohort and case-control part of the analysis. *Family control (brother) instead of hospital control (n = 1), hospital control admitted with Alzheimer's and chikungunya fever (n = 1)

TABLE 1. Demography, clinical features at entry, and outcome of the full cohort of patients with GBS All cases (n = 49) Sex (male) 32 (65) Age (years) 42 (23-57) <18 y old 7 (14) Country of inclusion Brazil 31 (63) Argentina 7 (14) Malaysia 11 (22) Antecedent event—onset weakness (days) 7 (4-15) Antecedent symptom (any) 36 (74) Fever 20/36 (56) Respiratory tract infectiona 15/36 (42) Gastro-intestinal infectionb 18/36 (50) Rash 4/36 (11) Cranial nerve deficits 29/48 (60) Oculomotor 10/48 (21) Facial 18/48 (38) Bulbar 10/48 (21) Limb weakness 37/48 (77) MRC sum score 45 (32-58) Hypo-/areflexia 42/48 (88) Sensory deficitsc 23/47 (49) Sensory symptoms 27/41 (66) Ataxiac 13/41 (32) Onset weakness—nadir (days) 10 (5-15) GBS clinical variant Sensorimotor 19/48 (40) Pure motor 14/48 (29) MFS (overlap) 10/48 (20) Other 5/48 (10) Nerve conduction studiesd 48/49 (98) Demyelinating 28/48 (58) Axonal 6/48 (13) Equivocal 13/48 (27) Immunomodulatory treatment 44/49 (90) IVIg 43/49 (88) Plasmapheresis 1/49 (2) ICU admission 20 (41) Mechanical ventilation 12 (25) Able to walk unaided at 6 moe 28/33 (85) Note: Data are presented as n/N reported (%) or median (IQR). Clinical features presented are at study entry. a Sore throat, nasal cold and/or cough. b Diarrhea or nausea/vomiting. c If “unable to examine” coded as missing. d One patient tested negative had an inexcitable EMG. e Patients able to walk at 8 or 13 wk and missing data at week 26 were included in this category. 3.1 Viral and bacterial serology

Evidence of a recent infection was found in 27/49 (55%) of patients and included arbovirus infections in four patients (8%), including CHIKV in two (4%), DENV and CMV in one (2%), and DENV and C jejuni in one patient (2%). In addition, in one patient, a low-positive IgM, positive IgG, and negative NS1 indicated a possible reinfection with a different DENV strain, and in one patient, a borderline-positive IgM, and positive IgG and VNT for ZIKV indicated a possible recent ZIKV infection. For the purpose of this study, these patients were not considered positive for a recent infection with these viruses. Details of serological test results for arbovirus infection-positive cases are shown in Table S2. The patients with a recent CHIKV infection and the patient with a recent DENV and C jejuni infection were included in Northeast Brazil between May and July 2019. The patient with a DENV and CMV infection was included in Malaysia in August 2019 (Table 2). C jejuni was the most common preceding infection in 15 patients (31%), followed by M pneumoniae in five (10%), and one additional patient had evidence of a recent infection with both these pathogens. Evidence of a recent EBV infection was found in one patient (2%), and none of the patients had evidence of a recent HEV infection. Samples were collected at a median of 11 days (IQR 7-19) after the onset of weakness.

TABLE 2. Demographic and clinical features of GBS patients with evidence of a recent arbovirus infection Sex, age, country Antecedent event Clinical features (entry) GBS clinical variant EMG subtype Treatment, ICU, and ventilation Disease nadir Outcome last follow-up CHIKVa male, 72 y/o, Brazil Nasal cold (20 d prior)

Brighton Level 1.

Bulbar and oculomotor palsy, limb weakness, sensory deficits, blood pressure dysfunction

MFS-GBS overlap Demyelinating IVIg (5 d), admitted to ICU (7 d) and MV (3 d)

MRC-SS =32, GBS-DS = 5.

Onset-nadir 9 d

MRC-SS = 60, GBS- DS w8 = 0 (w8 last follow-up) CHIKVb female, 37 y/o, Brazil Fever, joint pain, rash (4 d prior)

Brighton Level 1.

Bulbar and facial palsy, limb weakness, sensory deficits, blood pressure dysfunction

Sensorimotor Demyelinating IVIg (5d), admitted to ICU (21 d) and MV (17 d)

MRC-SS = 28, GBS-DS = 5

Onset-nadir 11 d

MRC-SS = 58, GBS- DS = 4 (w8 last follow-up) DENV and CMV male, 30 y/o, Malaysia Fever, myalgia, arthralgia, headache, retro-ocular pain (13 d prior)

Brighton Level 4.

Facial palsy, sensory deficits, ataxia

Ataxic form Demyelinating IVIg (5 d), no ICU or MV

MRC-SS = 60, GBS-DS = 3

Onset-nadir 7 d

MRC-SS = 60, GBS- DS = 0 (w26 last follow-up) DENV and Campylobacter jejuni male, 19 y/o, Brazil Fever, diarrhea (5 d prior)

Brighton Level 2.

Limb weakness

Pure motor Axonal IVIg (5 d), no ICU or MV

MRC-SS = 40, GBS-DS = 3

Onset-nadir 5 d

MRC-SS = 54, GBS- DS = 2 (w26 last follow-up) Abbreviations: C. jejuni, Campylobacter jejuni; CHIKV, chikungunya virus; CMV, cytomegalovirus; DENV, dengue virus; GBS-DS, GBS disability score; ICU, intensive care unit; IVIg, intravenous immunoglobulins; MFS, Miller Fisher syndrome; MRC-SS, MRC sum score; MV, mechanical ventilation; y/o, years old. a P40 in Figure 2. b P39 in Figure 2. 3.2 Clinical features, ancillary investigations, and outcome of the full cohort

The median time between onset of neurological symptoms and hospital admission was 6 days (IQR 3-10). Lumbar puncture was done in 46/48 reported patients (96%). In 73%, an increased protein level (>0.45 g/L) and a cell count below 50 cells/μL was found (albuminocytological dissociation).34, 52 The median cell count was 1.0 (1.0-3.5), and none of the patients had a cell count above 50 cells/μL. Nerve conduction studies were performed in 48 (98%) patients. To exclude differential diagnoses, MRI of the spinal cord was performed in eight patients and was normal in six and showed enhancement of the cauda equina in two. According to the Brighton Collaboration Criteria, 25 (51%) had Level 1, 7 (14%) Level 2, and 17 (35%) Level 4. Patients were categorized as Brighton Level 4 because of: time to nadir >28 days (n = 2), normal (n = 4) or increased tendon reflexes (n = 1), clinical variant of GBS without limb weakness (n = 9), and missing data on time to nadir (n = 1). Four of five patients with normal or increased tendon reflexes had evidence of a recent C jejuni infection. Nerve conduction studies showed signs typical of a poly(radiculo)neuropathy in 16/17 patients (96%), and 9/17 (53%) had an albuminocytological dissociation in the CSF.

At disease nadir, 79% of patients were unable to walk unaided (GBS disability score ≥ 3), and the median MRC sum score was 43 (IQR 31-46). When including the patients with missing data at 6 months, but who were able to walk at week 8 or week 13 after study inclusion, 28/33 (85%) were able to walk unaided at 6 months. Eighteen of 19 patients (95%) who were followed up to 1 year or more were able to walk unaided at 1 year. One patient died due to complications of pulmonary tuberculosis 5 months after the onset of GBS.

3.3 Comparison of infection groups

Preceding symptoms of an infection were reported in 36 (74%) of the patients and included fever, and gastro-intestinal and respiratory tract infection. Of the patients with preceding symptoms of an infection, 16 (44%) had no serological evidence of a recent infection. In contrast, of the 27 patients with serological evidence of a recent infection, 7 (26%) did not have preceding infectious symptoms. Antecedent events other than infectious symptoms included vaccination (n = 4) and surgery (n = 1). The types of vaccination were influenza, polio, and tetanus. All patients that reported a recent vaccination had serological evidence of a recent infection, with C jejuni (n = 1), M pneumoniae (n = 1), EBV (n = 1), and CHIKV (n = 1). The patient with surgery also had preceding infectious symptoms, including fever, gastro-intestinal complaints, and joint pain. She was negative for the tested infections.

The clinical features of the patients with evidence of a recent arbovirus infection are shown in Table 2. The two patients with a recent CHIKV infection had different clinical variants (MFS-overlap and sensorimotor), the same electrophysiological subtype (demyelinating), and a similar clinical progression; both were admitted to ICU and ventilated, had a low MRC sum score at nadir, but near-complete recovery of strength at 8-week follow-up. One of these patients had typical antecedent symptoms of CHIKV infection, including fever, joint pain, and rash; the other reported a nasal cold 20 days prior. The patient with a recent DENV and CMV infection reported preceding symptoms of fever, myalgia, arthralgia, headache, and retro-ocular pain and had an ataxic variant and demyelinating subtype of GBS. The patient with a recent DENV and C jejuni infection had preceding symptoms of a gastro-enteritis and a pure motor variant and axonal subtype of GBS.

In patients with a recent C jejuni infection, gastro-enteritis was the most common reported antecedent event (78%). The pure motor variant of GBS was most frequently reported (12/15, 80%), cranial nerve involvement was infrequent (5/15, 33%), and the MRC sum score at entry was relatively low (41 [IQR 30-46]). Nine of 12 reported patients (75%) were able to walk unaided at 6 months. The five patients with a M pneumoniae infection were frequently <18 years old (2/5, 40%), and had a relatively long time between antecedent event and onset of weakness (18 days [IQR 11-21]), and a high MRC sum score at entry (59 [IQR 56-60]), and 2/2 reported patients had fully recovered at 8 weeks. The patient with a recent EBV infection was 9 years old, and had preceding symptoms of headache and nausea, a sensorimotor demyelinating variant, and full recovery of disability at 13-week follow-up. Details on the clinical features per infection group are displayed in Table S3.

3.4 Anti-ganglioside antibodies

The presence of serum anti-ganglioside antibodies (IgM and IgG) against 12 commonly studied glycolipids in GBS was tested in ELISA and combinatorial glycoarray.

In ELISA, 21 patients (43%) were positive for one or more of these antibodies (IgM or IgG), vs none of the 32 tested controls (Table 3). In patients with a CHIKV or EBV infection, no anti-ganglioside antibodies were found in ELISA. In patients with a C jejuni infection, antibodies against GM1, GM2, and GD1a were most frequently reported. In the patient with a C jejuni and DENV infection, IgM antibodies against GM1, GM2, and IgM and IgG against GD1a were found, and in the patient with a CMV and DENV infection, IgM antibodies against GM2 were found. The presence of anti-ganglioside antibodies (IgM or IgG) was found in patients with an axonal (4/6, 67%) as well as in patients with a demyelinating electrophysiological subtype of GBS (14/28, 50%).

TABLE 3. Anti-ganglioside antibodies in serum (ELISA) Controls (n = 32)a All cases (n = 49) Campylobacter jejuni (n=15) Mycoplasma pneumoniae (n = 5) IgM IgG IgM IgG IgM IgG IgM IgG Any 0 (0) 0 (0) 11 (22) 15 (31) 6 (40) 9 (60) 1 (20) 1 (20) GM1 0 (0) 0 (0) 6 (12) 5 (10) 4 (27) 4 (27) 1 (20) 0 (0) GM2 0 (0) 0 (0) 6 (12) 1 (2) 4 (27) 1 (7) 0 (0) 0 (0) GD1a 0 (0) 0 (0) 4 (8) 5 (10) 4 (27) 4 (27) 0 (0) 0 (0) GD1b 0 (0) 0 (0) 1 (2) 6 (12) 0 (0) 3 (20) 0 (0) 1 (20) GD3 0 (0) 0 (0) 0 (0) 1 (2) 0 (0) 0 (0) 0 (0) 0 (0) GQ1b 0 (0) 0 (0) 0 (0) 3 (6) 0 (0) 0 (0) 0 (0) 0 (0) a In 32/35 controls, sufficient serum sample was available for anti-ganglioside antibody testing.

In glycoarray, 19 patients (39%) were positive for IgM and 25 (51%) for IgG antibodies against single glycolipids, and 26 patients (53%) were positive for IgM and 36 (74%) for IgG antibodies against glycolipids in complexes. In contrast, of the 32 controls, 2 (6%) were positive IgM and 6 (19%) for IgG antibodies against single glycolipids, and 11 (34%) for IgM and 10 (31%) for IgG antibodies against glycolipids in complexes. In Figure 2, glycoarray findings are visualized in a heatmap. Binding of IgG antibodies to glycolipids is clearly lower in cases vs controls although some reactivity against GalC, lactosylceramide, and sulfatide is seen in both cases and controls. Similar to the ELISA results, no or only low reactivity was found in patients with (arbo)virus infections. The patient with a recent CHIKV infection that had a MFS-GBS overlap variant (P40) was positive for IgG and IgM antibody binding to GD3 in complex with several other glycolipids, including GQ1b, but binding was low and not visible in the heatmap (Figure 2). In the other patients with a recent CHIKV infection (P39), no antibody binding to glycolipids was found. In one patient with M pneumoniae infection and a sensorimotor variant of GBS (P49), reactivity was found against complexes with GD1a, GD1b, GD3, and GQ1b. In patients with a C jejuni infection, a large variety of reactivities were found, but clusters were mostly seen in complexes with GM1, GD1a, and GD1b. The patient with a C jejuni and DENV infection (P41) showed complex reactivity similar to that of other patients with a C jejuni infection.

Heatmap of IgG antibody binding to glycolipids as assessed by glycoarray. Each row presents one patient (P1-P49) or control (C1-C23); each column presents one of the tested glycolipid antibodies (single or in complex). Raw data were was clustered based on a distance matrix using Pearson's correlation and hierarchical cluster algorithm, and clipped at a 10 000 upper limit

3.5 Case-control study

In total, 35 paired controls were collected of 23 cases. One of these cases was excluded because of an alternative diagnosis, leaving 22 patients with 33 paired controls for the paired case-control analysis (Table S4). None of the cases or controls included in this analysis had evidence of a recent infection with ZIKV, CHIKV, or EBV. Calculated crude odds ratio and adjusted odds ratio of recent infections were not significant.

We also performed an unpaired case-control analysis, comparing all 49 cases to all 35 controls (Table 4). Although all infections occurred more frequently in cases vs controls, calculated crude odds ratio were not significant. Evidence of a recent infection with DENV, CHIKV, CMV, or EBV was only found in cases. Furthermore, two cases had a possible recent arbovirus infection (one ZIKV infection and one DENV reinfection), and in none of the controls were such borderline results found.

TABLE 4. Unpaired case-control analysis Evidence of recent infectiona Controls (n = 35) Cases (n = 49) Crude odds ratio (CI)b P-value Dengue virus 0/35 (0%)