The spectrum of antecedent infections in Guillain-Barré syndrome

The Guillain-Barré syndrome (GBS) is an acute polyneuropathy that may lead to a variety of motor and sensory deficits. It is considered a postinfectious disease as approximately two-thirds of patients report some form of preceding infectious illness. These infections may effect an immune response against peripheral nerve antigens, as suggested by the interval of 1 to 4 weeks between antecedents infection and onset of weakness. A remarkable diversity of infectious agents has been reported in patients with GBS.¹ The association between GBS and infections with Campylobacter jejuni and cytomegalovirus (CMV) has been demonstrated in case-control studies.^2-4 Furthermore, C. jejuni infections are associated with antibodies against the ganglioside GM1, and a severe, pure motor variant of GBS with a poor prognosis after plasma exchange.^5,6 GBS patients with a CMV infection more frequently have antibodies against the ganglioside GM2,^7,8 and have severe sensory deficits and cranial nerve involvement.⁹ These findings strongly indicate the C. jejuni and CMV infections determine the specificity of the immune response against peripheral nerves leading to distinct clinical variants of GBS. The variety of reported infectious agents in GBS may therefore underlie the immunologic and clinical heterogeneity in GBS.

Controversy exists over whether infections with other frequently reported agents, such as Epstein-Barr virus (EBV) and Mycoplasma pneumoniae, are more often present in GBS patients compared with controls^10,11 or not.² The majority of these "GBS-related infections" were reported only in small and selected groups of GBS patients, without age-, sex-, and season-matched controls, and may therefore represent coincidental findings. In addition, many of these infections were clinically defined without serologic confirmation, or with serology performed on serum samples obtained long after neurologic onset or after treatment. Moreover, in most studies, the possible association among these infections, antiganglioside antibodies, and clinical presentation was not investigated. Therefore, the role of infectious agents in GBS, other than C. jejuni and CMV, is not established.

The aim of the current study was to determine the spectrum of antecedent infections in a large, unselected group of GBS patients and to investigate whether these infections are more common in GBS than in a matched control group. In addition, we investigated whether these infections in GBS patients are associated with specific antiganglioside antibodies and distinct clinical presentations.

Patients and methods. Patients. The GBS patients in this study participated in the Dutch Guillain-Barré trial comparing the therapeutic effect of plasma exchange (PE) and intravenous immunoglobulins (IVIg),¹² or in the pilot study evaluating the therapeutic effect of methyl-prednisolone in addition to IVIg (MP-IVIg).¹³ The 147 patients in the PE/IVIg trial entered the study from June 1986 to December 1989, and the 25 patients in the MP-IVIg pilot study from September 1990 to September 1992. The GBS patients were referred to academic or other large hospitals in The Netherlands. All patients fulfilled the diagnostic criteria for GBS,¹⁴ were unable to walk 10 meters independently, and were admitted within 2 weeks of onset of weakness. All patients were evaluated with respect to the presence of paresthesias, sensory deficits (two-point discrimination, position sense, and tactile function in hands and feet), cranial nerve involvement, functional score, and Medical Research Council (MRC) sumscore, at study entry and subsequently at 16 time points according to a previously established protocol during a follow-up period of 6 months. At entry, all patients were asked if they had infectious illness in the preceding 4 weeks. Pretreatment serum samples obtained within 2 weeks of neurologic onset were available from 154 (90%) of the 172 patients included in these therapeutic studies. The 18 excluded cases did not differ from the other patients regarding their antecedent illnesses, neurologic manifestations, and course of disease.

Controls. As a control group for the serology studies, we used serum samples from 154 sex- and age (±4 year)-matched control patients with other neurologic diseases (OND) than GBS. The controls resided in the southwest area of The Netherlands and were referred to the University Hospital Rotterdam in the period from 1990 to 1996. The population in this area reflects that of other parts in The Netherlands with respect to relevant demographic characteristics, such as urbanization ratio. The time of serum sampling in the control group and the group of GBS patients had the same seasonal distribution. The OND controls had chronic inflammatory demyelinating polyneuropathy (CIDP) (n = 21), other forms of neuropathy (n = 60), MS (n = 41), cerebrovascular accidents (n = 14), motor neuron disease (n = 6), MG (n = 6), or other neurologic disorders (n = 6). All samples were obtained at admission to the hospital and before treatment. In addition, serum samples from 50 healthy subjects were tested.

Infection serology. Serum samples from GBS patients were tested to determine recent infections with 16 viruses or bacteria frequently reported in GBS. The assays were performed according to routine techniques with previously established criteria for positivity. C. jejuni infections were defined as the presence of IgM, IgA, or high titer IgG in ELISA.⁶ CMV,⁸ M. pneumoniae, and hepatitis A virus infections were defined as the presence of IgM, and Haemophilus influenzae infections as the presence of IgM, IgA, and IgG antibodies in ELISA. Hepatitis B virus infection was defined as the presence of hepatitis B virus surface antigen in ELISA. Infections with herpes simplex virus, varicella zoster virus, measles virus, influenza A and B virus, parainfluenza 1 and 2 virus, adenovirus, and respiratory syncytial virus were determined by complement fixation and were defined as the presence of an antibody titer ≥64. EBV infections were determined as the presence of IgM against viral capsid antigen in immunofluoresence. To determine IgM antibodies to EBV and CMV, sera were preabsorbed with Gull-sorb (Gull Laboratories, Salt Lake City, UT) to remove the IgG antibodies to these agents. To determine IgM antibodies to C. jejuni, H. influenzae, M. pneumoniae, and hepatitis A virus, μ-specific class-capture ELISAs were used. These precautions to determine the IgM serology increased the sensitivity of the assays and prevented false positivity by binding of IgM rheumatoid factors with IgG antibodies. Because this study focused on the infections that were more frequent in GBS patients than controls, we only determined the serology for specific infections in the controls if more than five GBS patients were positive for this infection, the minimal number to be significant in the McNemar's test. The frequency of CMV and C. jejuni infections in the patients included in the PE/IVIg trial were reported previously,^6,8,9 but without comparison with the matched control group.

Antiganglioside antibodies. Pretreatment serum samples from all GBS patients were tested for IgM, IgG, and IgA antibodies against the peripheral nerve gangliosides LM1 (sialosyl paragloboside), GM1, GM2, GM3, GD1a, GD1b, GT1b, and GQ1b by ELISA and confirmed in thin-layer chromatography overlay, according to methods described previously.⁶

Statistical analysis. Differences in frequency of infections between GBS patients and OND controls were tested with the McNemar's test,¹⁵ and between GBS patients and healthy controls with the chi-square test without continuity correction or Fisher's exact test. The odds ratio for each infectious agent was also estimated using a multivariate conditional logit model for matched case-control data, with disease status (GBS or control) as dependent variable and the infectious agents as independent variables.¹⁶ With respect to the clinical features of GBS patients, differences in proportions were tested with the chi-square test without continuity correction or Fisher's exact test, and differences in medians with the Wilcoxon-Mann-Whitney U test. The time for patients to reach independent locomotion was analyzed by the Kaplan-Meier method and the log-rank test. All calculations were performed using STATA 5.0 for Windows 95 (Stata Statistical Software, Release 5.0, College Station, TX). A p value < 0.05 was considered to be significant.

Results. Infections and antecedent illness in GBS patients. In the 4 weeks preceding GBS, 105 (68%) of the 154 patients reported some form of infectious illness, as indicated by their clinical symptoms. The most common identified cause of recent infections in patients with GBS was C. jejuni (32%) (table 1). Other recent infections in GBS patients were CMV (13%), EBV (10%), M. pneumoniae (5%), and, less often, H. influenzae (1%), parainfluenza 1 virus (1%), influenza A virus (1%), influenza B virus (1%), adenovirus (1%), herpes simplex virus (1%), and varicella zoster virus (1%) (see table 1). The majority of GBS patients with positive serology had clinical symptoms of preceding infectious illness (see table 1). Patients with C. jejuni infections had diarrhea more often compared with patients without this infection (p < 0.001); symptoms of upper respiratory tract infection (URTI) were less frequently found (p = 0.001). C. jejuni infections were also associated with other manifestations of infectious illness than diarrhea and URTI, such as nausea, vomiting, and abdominal pain. Patients with antecedent CMV infection frequently had URTI and other manifestations of infection such as nausea, fever, and exanthema, but none had diarrhea. Other manifestations of infection were fever and sore throat in EBV-infected patients and pneumonia and pancreatitis in M. pneumoniae-infected patients. In the 49 patients who did not report symptoms of antecedent infectious illness, a substantial proportion had positive serology for infections with C. jejuni (29%), CMV (18%), EBV (16%), and M. pneumoniae (4%) (see table 1). The high proportion of C. jejuni infections in these patients is in accordance with the frequent subclinical course of this infection. Positive serology for recent infection with parainfluenza 2 virus, hepatitis A and B virus, respiratory syncytial virus, or measles virus could not be demonstrated. None of the infections in GBS patients showed a clear seasonal predominance.

Positive serology for more than one infection was found in 13 (8%) of the GBS patients. Most of these GBS patients had positive EBV serology, and were additionally positive for C. jejuni and CMV (n = 2), C. jejuni (n = 2), M. pneumoniae (n = 3), CMV (n = 2), and H. influenzae (n = 1). The other patients were positive for CMV and additionally for C. jejuni (n = 1), M. pneumoniae (n = 1), and influenza A virus (n = 1).

Antecedent infections in GBS patients compared with controls. Serology for infections with C. jejuni, CMV, EBV, and M. pneumoniae was also performed in the two control groups. Univariate analysis showed that these infections were all significantly more frequent in GBS patients compared with matched OND controls (table 2). The attributable proportion of these infections, i.e., the proportion of patients that can be attributed to these infections provided they play a causal role in GBS, was calculated as 0.44 (see table 2). Using a multivariate conditional logit model for matched case-control data, only infections with C. jejuni, CMV, and EBV appeared to be independently associated with GBS. Acute infections in the OND controls were not significantly associated with any of the diagnostic categories. The OND patients with positive serology had MS (n = 8), CIDP (n = 4), other polyneuropathies (n = 7), cerebrovascular accident (n = 1), MG (n = 1), Friedreich's ataxia (n = 1), or concussion (n = 1). Differences in the frequency of C. jejuni, CMV, EBV, and M. pneumoniae infections between GBS patients and controls could not be adjusted to differences in their inclusion period because retrospective analysis revealed that the annual incidence of these infections did not change significantly in the time period from 1986 to 1996. In addition, we found no significant association between the hospital or year of admission and the frequency of specific infections in the groups of GBS patients or controls.

Antecedent infections, antiganglioside antibodies, and clinical subgroups. C. jejuni infection was significantly associated with antibodies against GM1 and GD1b, and CMV infections with antibodies against GM2 (table 3). Interestingly, the only GBS patient with herpes simplex virus had high titers of IgG antibodies against LM1, GM3, GD1a, GD1b, GT1b, and GQ1b. This patient, who had a pure motor variant of GBS with ophthalmoplegia and hypoglossal palsy, needed artificial ventilation and recovered to independent locomotion 69 days after IVIg. Infections with EBV or M. pneumoniae, or any of the other agents, were not associated with the tested antiganglioside antibodies. Antibody titers to GM1 and GM2 were also determined in a subgroup of the OND and healthy controls, but were not found to be associated with the presence of C. jejuni or CMV infections.^6,8

Infections with C. jejuni and CMV were each associated with a distinct clinical presentation in GBS, as we previously described.^6,9 In addition, the case-control study enabled us to investigate the association between these infections and sex and age. The interaction between sex and C. jejuni infection was almost significant; the odds ratio (95% confidence interval) in male GBS patients was 14.4 (3.1 to 68.1) and in female GBS patients 2.9 (3.1 to 7.1) (p = 0.08). The median age of the GBS patients with EBV infection was 29 years, and with M. pneumoniae infection 28 years, indicating that they were significantly younger than the other GBS patients (median 50 years) (p = 0.002 and p = 0.001, respectively). Eleven (69%) of the 16 EBV infected patients had respiratory insufficiency, which was a significantly higher proportion compared with patients with other or no infections (37%) (p = 0.01). Infections with these or other micro-organisms were not significantly associated with time to reach nadir, MRC sumscore at entry and nadir, distribution of weakness, cranial nerve involvement, sensory loss, paresthesias, or time to recovery. Also, the GBS patients with multiple positive serology did not significantly differ from the other patients with respect to their neurologic presentation.

Discussion. In the current study, C. jejuni, CMV, EBV, and M. pneumoniae were identified as the most common causes of antecedent infections in GBS. These infections were found more frequently in GBS patients than in age- and sex-matched controls with OND. Infections with herpes simplex virus, varicella zoster virus, influenza A and B virus, parainfluenza 1 virus, adenovirus, and H. influenzae were also identified, but their frequency was not significantly higher than in controls. These findings confirm the concept that certain infections are specifically related to GBS, although a spectrum of infections may precede the disorder.

Previous case-control studies demonstrated C. jejuni infections in 14% to 36% of GBS patients, and in 1% to 10% of controls,^2-4 which is in accordance with our findings. The current study justifies the conclusion that C. jejuni is the predominant cause of antecedent infection in GBS. However, C. jejuni is also the predominant cause of acute bacterial enteritis in The Netherlands, and our study shows that in patients with OND and healthy controls, C. jejuni occurs more frequently than infections with CMV, EBV, and M. pneumoniae. Therefore, one may argue that GBS is a postinfectious disease not related to specific infections, in which the predominance of C. jejuni infections simply reflects the high frequency of this infection in the community. However, there is a strong association between C. jejuni infections and a distinct severe and pure motor variant of GBS.^4,6 This argues for a specific role of these bacteria in inducing immune responses to motor nerve antigens in a subgroup of GBS patients. GM1 may be one of the targets because anti-GM1 antibodies are associated with C. jejuni infections in GBS,^5,6 and higher concentrations of GM1 are present in myelin of motor fibers compared with sensory fibers.¹⁷

CMV was the most common viral cause of antecedent infection, present in 13% of GBS patients. This finding is in accordance with the 15% reported by Dowling and Cook,¹¹ and the 11% and 22% reported by others.^2,18 The presence of IgM antibodies to CMV suggests a recent CMV infection, although reactivation of a latent infection cannot be excluded. According to the report of Dowling and Cook, CMV infections in GBS patients are clustered in time periods of 10 to 16 weeks,¹¹ but this could not be confirmed in our study. The high frequency of sensory and cranial nerve involvement in CMV-associated GBS patients may indicate that CMV infections induce specific antibodies against antigens present in these nerves. Anti-GM2 antibodies are associated with CMV infections in GBS,^7,8 but the distribution of GM2 in the human peripheral nerve system is unknown. Others reported that CMV infections induce antibodies against sulfatide and other sulfated glycosphingolipids from human peripheral nerves.¹⁹ Interestingly, antisulfatide antibodies were reported to be related to sensory nerve involvement in GBS and other neuropathies.²⁰ The association between CMV infections and antisulfatide antibodies in GBS patients, however, has not been established.

The current study is the first to demonstrate a significantly higher frequency of EBV and M. pneumoniae infections in GBS patients compared with case-matched controls. These frequencies in GBS patients are similar to those found by the group of Dowling and Cook, i.e., EBV 10% versus 8% and M. pneumoniae 5% versus 5%.^11,10 Provided these infections play a role in the pathogenesis of GBS, 9% of GBS cases can be attributed to EBV infections, and 4% to infections with M. pneumoniae. These significant but low attributable proportions may explain why Winer et al.,² in a smaller case-control study, were not able to demonstrate an association with one of these infections. Accordingly, M. pneumoniae infections in our study were significantly associated with GBS in univariate analysis only. We found no relation between EBV and M. pneumoniae infections and antibodies against the major peripheral nerve gangliosides. However, M. pneumoniae infections in GBS were reported to be associated with antibodies against galactocerebroside, the predominant glycolipid in human peripheral nerves.²¹ The ubiquitous distribution of this antigen in the peripheral nervous system may explain why we did not find an association with involvement of specific nerve fibers.

Thirteen (8%) of the GBS patients in our study had a positive serology for more than one infectious agent, as was also found in other serologic studies.¹⁸ This may indicate that dual-antigen induced immune responses play a role in a subgroup of GBS patients, as was also suggested for other immune-mediated neuropathies.²² Because we did not culture the organisms, the possibility that some of these patients had a single infection inducing polyclonal B cell activation or antibodies against common or cross-reactive structures on different micro-organisms resulting in a false double positive serology cannot be excluded. However, studies based on culture alone will underestimate the frequency of recent infections, as most infections precede GBS for several weeks, and the infectious agents may be cleared at the time of neurologic onset. Our findings should be confirmed in prospective large case-control studies using culture, serologic, and molecular biology techniques to determine recent infections.

C. jejuni, CMV, EBV, and M. pneumoniae may have a general feature that explains their specific association with GBS. These agents may share the capacity to induce immune responses to peripheral nerve glycoconjugates, as indicated by their coexistence with antibodies to gangliosides and other glycolipids in GBS patients. Moreover, infections with M. pneumoniae, CMV, and EBV are associated with cold agglutinins, which preferentially bind to carbohydrate antigens at 4 °C, a characteristic shared with antiganglioside antibodies.²³ Monoclonal cold agglutinins to disialylated gangliosides from patients with chronic ataxic neuropathy were found to reduce nerve excitability and neurotransmitter release in the mouse hemidiaphragm preparation.²⁴ Antiganglioside antibodies in GBS patients with C. jejuni infection may be induced by molecular mimicry, as ganglioside-like structures were identified in C. jejuni lipopolysaccharides.²⁵ In addition, it was demonstrated that anti-GM2 antibodies cross-react with CMV-infected cells.⁷ The four GBS-related micro-organisms also express receptors or toxins that bind to carbohydrate antigens and may act as carrier proteins in the production of of antiglycoconjugate antibodies. C. jejuni produces a toxin that specifically binds to gangliosides.²⁶ M. pneumoniae binds with sialylated carbohydrates from the blood group I antigen and induces antibodies against this structure.²⁷ Further research is needed to investigate if molecular mimicry with gangliosides or production of glycolipid-reactive carrier proteins explains the predominance of these infections in GBS.