Bacterial shedding and serologic responses following an outbreak of Salmonella Typhi in an endemic cohort

Our study shows that, even in a population where there had been a recent outbreak of typhoid fever, sero-surveillance could not identify culture-confirmed shedding of Salmonella Typhi. Sero-surveillance combined with PCR may have identified one participant shedding Salmonella Typhi. Anti-Vi IgG titres fell amongst participants reporting persistent fever during the outbreak. Anti-H:d IgG titres fell amongst participants with and without persistent fever. There was no change in anti-H:d IgM, irrespective of fever status. We did not find an environmental reservoir of Salmonella Typhi, but non-typhoidal salmonellae were present in water samples from the reservoir supplying the nursing school, and a tap in the kitchen. The presence of these bacteria indicate contamination and under-treatment of the water supply, and cast suspicion that this could have been a route through which the outbreak was sustained. Feedback from this investigation has resulted in more robust chlorination procedures.

Our study population was young (mean 23.5 years). Although we did not formally ascertain this, biliary pathology is rare in sub-Saharan Africa [25]. A similar study design might have detected Salmonella Typhi shedding amongst an older cohort, or in an endemic setting where rates of biliary disease are high. 140 of 156 participants meeting the case definition for ‘suspected typhoid’ received antibiotics, compared with 16 of 212 not meeting the case definition – it is likely that antibiotic administration reduced the amount of Salmonella Typhi that could be recovered post-outbreak. Our study design was contingent upon how swiftly ethical approval could be obtained after the outbreak. Thus, we selected three months for the first study visit, given that the immediate serologic signal caused by recent enteric fever would have started to settle, enabling us to identify participants with persistently elevated antibody titres attributable to ongoing shedding. It may be that we would have identified acute temporary shedding (shedding occurring within three months of illness)[10] if we had taken stool samples at an earlier time-point, but this would not have allowed us to test our hypothesis that presently elevated antibody titre was associated with ongoing shedding. Unfortunately, the serotypes of the non-typhoidal salmonellae we recovered from stool and from water were not recorded, and we can only state that they were non-typhoidal salmonellae (Salmonella enterica subspecies enterica).

Participants from the Nursing College were on clinical placement in multiple institutions across Southern Malawi. This led to participants being lost to follow up between 3 and six months. Given the intermittent nature of shedding, several samples over time would have increased our chances of detecting Salmonella Typhi by culture. Another approach (had resource allowed) would have been to take a stool sample from all participants, regardless of antibody titre, and thence investigate the difference in titres between culture confirmed shedders with that of the rest of the cohort. One participant was identified as having PCR amplification of the fimbriae target. This suggests that they may have been carrying / shedding Salmonella Typhi, but this could not be verified through culture confirmation. This participant had the highest anti-Vi IgG and anti-H:d IgG titres of all those who submitted stool.

Although we have no pre-outbreak serologic data to base this on, antibody titres to our selected antigens are likely to have risen and peaked around six weeks after the outbreak [26]. As described by Voysey and Pollard, decreases in antibody concentrations provide serologic evidence of infection [27]. This is supported by the fall in anti-Vi IgG (amongst those with suspected typhoid) and the cohort-wide fall in anti-H:d IgG between three and six months. Anti-H:d IgG has been shown to fall swiftly after controlled human infection, and this may explain why the signal was observed across the cohort for this antigen but only in those with suspected typhoid for the Vi antigen [19]. The static anti-H:d IgM concentration suggests that any change in titre precipitated by the outbreak has been and gone at three months.

The Vi (virulence) antigen comprises a polysaccharide expressed on the surface of Salmonella Typhi, as well as Salmonella Paratyphi C, Citrobacter freundii and Salmonella Dublin [28, 29]. IgG antibodies to the Vi antigen have been detected in high titres among microbiologically confirmed shedders of Salmonella Typhi [13, 30, 31]. Salmonella Typhi shedders have been identified through their high anti-Vi titre in the context of two outbreaks in the United States [11, 12]. Yet the role of anti-Vi IgG for population screening for Typhoid shedding remains unclear: a 2004 study obtained sera from 3209 adults in a Typhoid-endemic area (Vietnam). Multiple rectal swabs were taken from 103 participants with the highest anti-Vi IgG titres at different time points: no Salmonella Typhi was isolated [14]. Although we obtained fewer faecal specimens, our study population had known recent exposure to typhoid. The paucity of Salmonella Typhi detection in both settings suggests that anti-Vi IgG might be elevated for reasons other than biliary carriage, such as cross-reactivity with other bacteria that express the Vi antigen, or convalescence from acute typhoid fever. It is plausible that both factors would be more significant in areas of the world where typhoid fever is endemic, explaining the geographical discrepancy. A Chilean study comparing anti-Vi IgG amongst bacteriologically proven Salmonella Typhi carriers and healthy control subjects and those with acute typhoid fever supports our finding that anti-Vi IgG is a non-specific marker in an endemic setting: although the authors demonstrated higher antibody titre among the shedders, one quarter of chronic carriers did not have a high anti-Vi IgG titre [13]. There was no licensed typhoid vaccine in Malawi at the time that our study was conducted. We are therefore confident that our cohort’s serologic responses have not been impacted by prior immunisation.

Flagellin proteins comprise the bacterial flagellum, which is used for motility – predominantly by Gram-negative bacteria [32]. Flagellin epitopes are distinct among Salmonella serovars, which is why they may be used in serotyping [33]. Elevated anti-H IgG titres have been found amongst people infected by other Salmonella spp., other Enterobacteriaceae, and even Staphylococcus aureus [34, 35]. In our study, we show elevated titres of anti-H:d IgG amongst participants with and without persistent fever (although the fall between visit one and two was statistically greater amongst those with persistent fever). In our study setting, it is likely that participants will have been exposed to other Enterobacteraciae and Salmonella spp., which may explain why we observed elevated anti-H:d titres across the cohort. This illustrates the difficulty of using serology to identify shedding in endemic settings – a combination of antigens may be more specific.

A recent study utilising culture confirmed carriers who underwent cholecystectomy in Nepal found 13 putative antigens in sera that were expressed highly in the carrier state, and which might bear closer scrutiny in future studies looking for shedding [36]. Future studies should focus on robust microbiological confirmation of Salmonella Typhi shedding status to provide clear comparators for prospective serologic markers. The detection of carriers will require repeated stool sampling over time, and newer PCR techniques may increase sensitivity over culture alone [22, 37].

Despite the success of conjugate typhoid vaccines in preventing typhoid cases, indirect protection is yet to be demonstrated [38]. Wider public health measures, including identifying those who continue to shed Salmonella Typhi after infection, remain key to reducing transmission and to ultimately eradicating typhoid.