Antecedent infections in Guillain‐Barré syndrome in endemic areas of arbovirus transmission: A multinational case‐control study

Abstract Half of the world's population is at risk of arthropod‐borne virus (arbovirus) infections. Several arbovirus infections have been associated with Guillain‐Barré syndrome (GBS). We investigated whether arboviruses are driving GBS beyond epidemic phases of transmission and studied the antibody response to glycolipids. The protocol of the International Guillain‐Barré syndrome Outcome Study (IGOS), an observational prospective cohort study, was adapted to a case‐control design. Serum samples were tested for a recent infection with Zika virus (ZIKV), dengue virus (DENV), chikungunya (CHIKV) virus, hepatitis E virus, Epstein‐Barr virus (EBV), cytomegalovirus (CMV), Campylobacter jejuni, and Mycoplasma pneumoniae, and for antibodies to glycolipids. Forty‐nine patients were included from Brazil (63%), Argentina (14%), and Malaysia (22%). Evidence of a recent infection was found in 27/49 (55%) patients: C jejuni (n = 15, 31%), M pneumoniae (n = 5, 10%), CHIKV (n = 2, 4%), EBV (n = 1, 2%), C jejuni and M pneumoniae (n = 2, 4%), CMV and DENV (n = 1, 2%), and C jejuni and DENV (n = 1, 2%). In 22 patients, 35 paired controls were collected. Odds ratio for recent infections did not significantly differ between cases and controls. No typical anti‐ganglioside antibody binding was associated with recent arbovirus infection. We conclude that arbovirus infections occur in GBS patients outside of epidemic viral transmission, although not significantly more than in controls. Broad infection and anti‐ganglioside antibody serology are important to establish the most likely pathogenic trigger in GBS patients. Larger studies are necessary to determine the association between arboviruses and GBS.

Odds ratio for recent infections did not significantly differ between cases and controls. No typical anti-ganglioside antibody binding was associated with recent arbovirus infection. We conclude that arbovirus infections occur in GBS patients outside of epidemic viral transmission, although not significantly more than in controls. Broad increased incidence of GBS patients was observed and an association between ZIKV and GBS has later been confirmed. [6][7][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][10][11][12][13][14][15][16][17][18] Most studies on DENV and GBS are limited to case series, 10,[18][19][20][21][22] although two surveillances studies 17,18 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][25][26] and a case-control study 9 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][31][32][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.

| 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.

| 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 (Table S1). Case-control study F I G U R E 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) 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 antihuman IgG (Jackson ImmunoResearch). Fluorescent intensity was measured using the appurtenant LuxScan software. The mean and SD were calculated for each glycolipid (Àcomplex) using the fluorescent intensities of the control patients. Fluorescent intensities were considered positive if more than the mean plus three times the SD.

| 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 χ 2 -test or Fisher's exact test to compare proportions. A two-sided P-value of <0.05 was considered significant. For the casecontrol 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

| 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.
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.  Table S3.

| 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 antiganglioside 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,

| 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)   IgM seroprevalence of 5% for CHIKV and 2% for DENV and ZIKV. 55 One study from Malaysia performed between 2012 and 2017 showed 0.6%-2.2% seropositivity for ZIKV neutralizing antibodies, 56 and another study performed in 2015 in a rural area showed ±11% IgM seroprevalence of DENV. 57 We were not able to find reliable data on CHIKV IgM seroprevalence in Malaysia or of any of the three arboviruses for Argentina. Although the proportion of positive cases found in this study is higher than most of these seroprevalence studies, we are unable to draw any conclusions due to the differences in the study population.
We also tested our cohort for other infections that have previously been associated with GBS and found evidence of a recent C   23,26,71 This may indicate that arbovirus-related GBS is associated with a more severe initial disease course and/or respiratory insufficiency, but patient numbers are too small to draw conclusions.
Serology of anti-ganglioside antibodies clearly showed higher reactivities in patients compared to controls, both in ELISA and in glycoarray, confirming the role of these antibodies in the pathophysiology of GBS. 2,30 The patients with a recent C jejuni infection mainly displayed binding of GM1, GM2, GD1a, and GT1a, as has been reported previously. [72][73][74] In one of the patients with a recent CHIKV infection, low binding of GD3 antibodies in complex was found, and in the other CHIKV-positive case, no binding was found. The patient with a recent DENV and C jejuni infection had an anti-glycolipid complex reactivity similar to that of the patients with a C jejuni monoinfection, and in the patient with a recent DENV and CMV infection, IgM antibodies against GM2 were found, similar to previously published cases of CMV-related GBS. 75,76 This is in line with a previous study from Northeast Brazil where we did not find a specific antiganglioside antibody profile related to arbovirus infections. 26 Although a study on ZIKV-related GBS in French-Polynesia demonstrated antibody activity against GD1a, this has not been replicated in any study on GBS conducted during the Latin American ZIKV epidemic. 31 In general, anti-ganglioside antibodies have rarely been demonstrated in GBS patients with preceding virus infections, indicating that the underlying pathophysiology may be different from bacterium-related GBS.
The fact that the anti-ganglioside antibody profile of the patients with a DENV was more typical for their coinfection suggests that the CMV or C jejuni infection were the actual trigger of GBS in these cases, and the DENV infection was a coincidental finding. The patient with a C jejuni and DENV infection also had a clinical profile most compatible with a C jejuni infection, with preceding diarrhea and a pure motor axonal variant of GBS. This is similar to findings of a study from Bangladesh conducted during an endemic phase of ZIKV transmission, where 9/18 ZIKV-positive GBS cases also had evidence of a recent C jejuni infection, and a clinical phenotype typical for that infection. 77 Our study has several limitations. Most importantly, the casecontrol study was underpowered. It was not always possible in clinical practice to collect two paired controls for every case, as per the original protocol, an unfortunate but unavoidable feat in a multinational study. Second, participating centers were mostly academic or teaching hospitals, and inclusion of patients may have been biased towards more complicated or severe cases. Third, although we used sophisticated serological testing to identify the presence of arbovirus and other preceding infections, we were not able to perform PCR (for the viruses) or culture (for the bacteria) due to sample and cost limitations, and may have missed patients that did not (yet) mount a serological response.
In conclusion, we found that preceding infections with CHIKV and DENV occur in GBS patients outside of epidemics, although not significantly more often than in controls. Broad serological testing and anti-ganglioside antibody diagnostics, as well as clinical and electrophysiological findings, may be helpful in determining the actual trigger in GBS patients with coinfections. Larger studies on arbovirus-related GBS are necessary to further study the association with GBS in endemic phases of transmission.