We determined mean values, standard deviations, odds ratios, p-values, paired T-tests, 95% confidence limits, and interquartile ranges. Software included Lotus 123 and EpiInfo (a statistical software package designed by the Centers for Disease Control and Prevention [CDC]).

Seven GBS-associated strains were serotyped with the O (formerly heat-stable) serotyping scheme of Penner and Hennessy (17) and the heat-labile serotyping scheme of Lior (18). Serotyping was conducted (19) at the CDC Campylobacter Reference Laboratory. A GBS-associated strain was defined as one isolated from a patient with GBS (n = 6) or from a patient infected during an outbreak in which at least one infected person contracted GBS (n = 1). Four strains were from patients in the United States, and three were from patients in Germany; four of these were isolated from women.

Serotype O:19 was present in one of four U.S. GBS-associated C. jejuni isolates and one of three German isolates. Thus, 2 (29%) of 7 GBS-associated C. jejuni isolates from patients in the United States and Germany were O:19. The O and heat-labile serotypes of the seven GBS-associated strains are shown in Table 1.

We studied 44 GBS- and enteritis-associated strains from all over the world. We divided C. jejuni isolates into two major groups. Group 1 (17 GBS-associated strains) was divided into two subgroups. Subgroup 1A (GBS-associated/type O:19) consisted of seven C. jejuni O:19 isolates: three from the United States, three from Japan, and one from Germany. Subgroup 1B (GBS-associated/non-O:19 serotype) consisted of 10 C. jejuni isolates with serotypes other than O:19: three from the United States, two from Germany, and five from England. Group 2 (27 enteritis-associated strains from patients with uncomplicated enteritis but no known GBS association) was also divided into two subgroups. Subgroup 2A (not GBS-associated/type O:19) consisted of 11 C. jejuni O:19 isolates: 10 from the United States and one from Japan. Subgroup 2B (not GBS-associated/non-O:19 serotype) consisted of 16 C. jejuni isolates from U.S. patients with serotypes other than O:19. Serum-resistant C. fetus strain 23D and its spontaneous serum-susceptible mutant, 23B, were used as controls (20).

Strains were grown on trypticase soy blood agar plates (BBL Cockeysville, MD, USA) at 37^oC for 48 hours in an incubator containing 10% CO₂, 5% O₂, and 85% N₂. The bacteria were harvested in 0.15M saline and centrifuged at 8,000 g for 15 minutes. The supernatant was discarded, the pellet was resuspended in 300 µl of saline, and 10-fold serial dilutions were performed in saline. From the 10^-4, 10^-5, and 10^-6 dilutions, 400 µl aliquots of cells were added to 1.2 ml of Hanks balanced salt solution (HBSS).

Two sources of NHS were used in these assays. Pool 1 consisted of serum from five healthy adults that was pooled, aliquotted, and frozen at -70°C. The level of C. jejuni antibodies in pool 1 serum was determined by enzyme-linked immunosorbent assay (ELISA) (4). The optical density ratio (ODR) for immunoglobulin G (IgG) was 5.14; the ODR for IgM was 0.66. Documentation of previous C. jejuni infection was not required to be included in this pool. In contrast, pool 2 consisted of serum from two healthy adults with only low-titer antibody. The ODR for pool 2 serum in the IgG assay was 0.37; and the ODR in the IgM assay was 0.30. As a control, the NHS was heated to 56^oC for 30 minutes (heat inactivated NHS [HINHS]) to ablate all complement activity.

The assay to determine susceptibility to NHS was performed in sterile, disposable 96-well microtiter U-bottom plates (Falcon MICROTEST III, Becton Dickinson & Co., Franklin Lakes, NJ, USA). From each of the bacterial suspensions, a 150 µl aliquot was added to duplicate wells. In addition, 50 µl of NHS or HINHS (diluted to 40% with HBSS) was added; the final serum concentration in the suspension was 10%. After the assay plate was incubated at 37^oC for 1 hour, 50 µl of the suspension from each well was poured onto blood agar plates and incubated for 48 hours; the number of CFUs was then calculated. The difference between the counts for cells incubated with NHS and HINHS was expressed as median log₁₀ kill for each strain; greater than 1 log₁₀ kill was considered a serum-susceptible strain (21). If less than 1 log₁₀ kill occurred, the strain was considered serum-resistant. The identical procedure was performed using pool 1 and pool 2 serum.

The resistance of the 44 C. jejuni strains studied to killing by NHS is shown in Table 2. Of these strains, 12 (27%) were resistant to killing by C. jejuni antibody-positive pool 1 serum and 10 (23%) were resistant to killing by C. jejuni antibody-negative pool 2 serum; thus, as described previously (22), antibodies to C. jejuni had little impact on serum-killing. The GBS-associated strains were no more likely to be resistant to serum-killing than were the strains from patients with uncomplicated enteritis. However, O:19 strains were significantly more likely than other C. jejuni serotypes to resist serum-killing, regardless of GBS-association and serum pool used (Table 2). In serum pool 1, only 1 of 18 O:19 strains was serum-susceptible compared with 11 (42%) of 26 non-O:19 strains (odds ratio = 12.5, p = 0.008). Similarly, in serum pool 2, no O:19 strain was serum-susceptible compared with 10 (38%) of the non-O:19 isolates.

Eight strains from Group 1 (four randomly selected from each subgroup) and eight from Group 2 (four randomly selected from each subgroup) were grown on blood agar plates as described above. C. fetus strains 23D and 23B again served as controls and the assays were conducted (21). In brief, bacteria from each plate were harvested in 1.5 ml of HBSS and were centrifuged at 8,000 g for 15 minutes; the pellet was resuspended in 0.5 ml HBSS and adjusted to OD₄₅₀ = 3.0. For each strain studied, a suspension of 4 µl ¹²⁵I-C3 (with 20,000 cpm to 65,000 cpm) and 2.5 µl of either NHS or HINHS from pool 2 and 100 µl of the bacteria-HBSS suspension were incubated at 37^oC for 15 minutes. ¹²⁵I-C3 was prepared in the laboratory of one of the authors (RGW) (21). The suspensions were then centrifuged twice at 175xg for 5 minutes, the pellet was resuspended in HBSS, and the supernatant was discarded. The bottom 5 mm (containing the pellet) of each tube was clipped, and emissions were determined in a gamma counter. The counts in the negative control mixtures containing HINHS were subtracted from the NHS counts. An assay was considered valid only if the net pellet counts (NHS minus HINHS) for control strain 23B were at least four times higher than for 23D. To control for nonspecific binding, the counts for each strain studied were expressed as the ratio of net counts relative to the serum susceptible control (23B). Each strain was assayed two to four times.

Mean binding of serum-susceptible control strain 23B was 497 cpm, whereas mean binding of serum-resistant control strain 23D was 54. The mean ratio of C3-binding for strain 23D to 23B of 0.114 was as expected (21). In contrast, the mean ratio for C3 binding to C. jejuni strains in comparison with strain 23B was 0.022 to 0.464 (mean 0.216). The eight O:19 strains were not significantly different from the eight non-O:19 strains in their ability to bind ¹²⁵I-C3 (Table 3).

This study of GBS patients in the United States and Germany confirms the observation made in Japan that serotype O:19 is overrepresented among patients with C. jejuni-induced GBS. Of 298 randomly collected Campylobacter isolates from patients with enteritis in the United States, only 3% were serotype O:19 (19). A similarly low prevalence of O:19 strains is found in all parts of the world, including North and South America, Asia, and Europe (5,6,19,23). Although specific serotyping surveys have not been done in Germany, it is unlikely that serotype O:19 is more frequent among German C. jejuni isolates. Thus, the GBS-associated strains in our study were more than 11 times as likely to belong to this serotype (p = 0.03). Although the association among GBS and C. jejuni serotype O:19 was not as marked as in Japan (where more than 80% of GBS-associated isolates are O:19), this serotype clearly is overrepresented among GBS-associated strains in other countries. We conclude that the association of O:19 strains with GBS is not just a local phenomenon in Japan but likely reflects a fundamental characteristic of O:19 strains. O-serotype 2 and heat-labile serotype 4, which were common among the GBS strains, are commonly represented among infected persons in the United States (23).

Despite the frequency of Campylobacter infections in GBS patients, such strains are difficult to obtain for several reasons. First, in most C. jejuni-infected patients, stools are clear before neurologic symptoms begin. Second, most neurologists do not culture stool samples when GBS is first diagnosed. And finally, even if a stool culture is ordered and Campylobacter is present, few microbiology laboratories save their isolates; by the time the case is reported, the strain has been discarded. Thus, these seven strains represent one of the largest collections of GBS-associated strains described. Additionally, these are likely to be representative of the population of GBS-associated C. jejuni strains. Unless serotype O:19 strains persist in stools longer, are more easily cultured, or are less likely to be discarded by microbiology laboratories (and no data support any of these possibilities), these strains probably are not different in any systematic way from other GBS-associated isolates.

Most C. jejuni strains are susceptible to killing by human serum (22), but because we studied highly selected strains (most either serotype O:19 or from GBS patients), a high proportion of strains in this investigation were resistant. In this context, the finding that C. jejuni strains from GBS patients were no more likely than the strains from patients with uncomplicated enteritis to be serum resistant is not surprising. Even when the analysis was limited only to serotype O:19 strains, no differences were found between GBS- and enteritis-associated strains. However, serotype O:19 strains were substantially less serum-susceptible than strains of other serotypes and were 12 times as likely to be serum-resistant. Although serotype O:19 represents only a small percentage of C. jejuni strains from patients with uncomplicated enteritis in the United States or Japan, it is the most common serotype identified in GBS patients in both locales. The relatively small variation in serum susceptibility of the O:19 strains is consistent with the close genetic relationship observed among these strains (24).

To better understand the basis for the relative serum-resistance among the O:19 strains, we compared their ability to bind C3 in relation to non-O:19 strains. Because C3 binding occurs after activation by either the alternative or classical complement pathways, it is a screen for differences in these early steps. The serum-resistance of C. fetus is explained by the inability of C3 to bind to the cell surface (25). In contrast, C3 binding is normal in serum-resistant Salmonella, but the C5-9 membrane attack complex does not insert properly (26). In the present study, the lack of substantial C3 binding differences among O:19 and non-O:19 C. jejuni strains suggests that the early steps in complement activation are similar among strains. However, clear differences in serum-susceptibility bespeak either rapid inactivation of C3 or reduced assembly or function of the membrane attack complex in the more resistant strains. Future studies should address this point.

The increasing awareness of the importance of C. jejuni infection in triggering GBS is another example of how previously well-described diseases have emerged as sequelae of acute infectious illnesses. This study attempts to begin to characterize the nature of this association; however, there is much to learn about how an acute gastrointestinal infection results in ascending paralysis. One fact is quite clear: many more people are infected with C. jejuni than contract GBS subsequently. Perhaps some persons are predisposed to contracting GBS after infection with campylobacters that might cause only uncomplicated enteritis in another patient. Conversely, as we have suggested in this paper, some strains may be more likely than others to trigger GBS. No associations between human leukocyte antigen (HLA) types and GBS have been found (27,28). However, in Great Britain and Japan, an association between HLA type and C. jejuni-associated GBS has been suggested (9,29). Perhaps some combination of familial susceptibility, HLA type, strain serotype, or other host or strain characteristics together play a role in the pathogenesis of C. jejuni-induced GBS.

The relative serum-resistance of O:19 strains correlates with mechanism. Furthermore, the relevance of these in vitro assays to the susceptibility of organisms in vivo cannot be known with certainty. We speculate that the relative insensitivity of these strains to the lytic effects of complement allows them to trigger a heightened specific immunologic response. We further speculate that this heightened immunologic response leads to injury of peripheral nerve structures. Since only a small fraction of infections caused by C. jejuni O:19 lead to GBS (estimated incidence 1 in 158) (30), additional factors also must be involved in vivo.