The Impact of Vaccination and Prior Exposure on Stool Shedding of Salmonella Typhi and Salmonella Paratyphi in 6 Controlled Human Infection Studies

Six Salmonella Typhi or Paratyphi human challenge studies were conducted, and daily stool cultures performed. Vi-containing vaccines reduced bacterial shedding, Ty21a or an experimental vaccine did not. Higher Vi immunoglobulin G titers were associated with reduced shedding.

Infection with Salmonella enterica subspecies enterica serovars Typhi or Paratyphi (S. Typhi or S. Paratyphi) is estimated to be responsible for between 11.6 and 26.9 million cases of enteric fever and 75 000-216 510 deaths annually [1][2][3][4][5]. Transmission of S. Typhi and S. Paratyphi occurs primarily through consumption of contaminated food or water, via short-cycle or longcycle transmission. Short-cycle transmission is defined as the contamination of food and water in the immediate environment, whereas long-cycle transmission is defined as contamination of the broader environment, such as pollution of water supplies by sewage or inadequate treatment of piped water [6]. The relative contribution of each transmission mode may vary depending on the epidemiological context and may differ between S. Typhi and S. Paratyphi [7]. As these serovars are human restricted, all modes of transmission involve shedding of the organism by infected individuals during incubation, acute disease, or convalescence or by chronic long-term carriers, ultimately resulting in contamination of food or water consumed by susceptible individuals. Disease control is therefore likely to require the integration of initiatives to improve water quality, sanitation, and hygiene, coupled with the deployment of effective vaccines [8].
Vaccines that both protect against clinical disease as well as reduce shedding would likely have enhanced effectiveness by interrupting transmission and providing indirect protection to unvaccinated individuals. Live attenuated oral typhoid vaccine, Ty21a, reduced stool shedding in the early Maryland challenge studies [9] and appears to induce herd immunity in field trials [10]. However, there are conflicting data on the indirect protection conferred by Vi-polysaccharide vaccines (Vi-PS) [11,12] and limited data on the impact of new Vi-tetanus-toxoid conjugate vaccines (Vi-TT) on stool shedding [13]. Finally, the impact of previous exposure to typhoidal Salmonella on stool shedding after subsequent exposure has not been previously explored.
Experimental challenge studies of closely monitored volunteers can be used to describe microbial dynamics in clinical and subclinical typhoid and paratyphoid infections, including the timing and pattern of stool shedding after challenge in naive or vaccinated individuals [14]. Early experimental human challenge studies in Maryland indicate that shedding is more common in individuals who develop typhoid disease. However, individuals who fail to develop disease following challenge can continue to shed bacteria for several weeks [15]. Differences in stool shedding patterns between S. Typhi and S. Paratyphi A are poorly understood. Improved estimates of stool shedding dynamics in enteric fever are needed, particularly in the context of different immune states, as these form important variables in models of typhoid transmission dynamics and estimates of vaccine impact [16,17].
We performed an analysis of stool shedding dynamics in healthy volunteers enrolled into closely monitored S. Typhi and S. Paratyphi human challenge studies. Our aims in this study were to model stool shedding after experimental challenge and to compare those challenged with S. Typhi vs S. Paratyphi, participants who did or did not develop enteric fever after challenge, those who received typhoid vaccines vs those who were unvaccinated, rechallenged individuals previously exposed to S. Typhi or S. Paratyphi vs previously unexposed individuals, differences according to demographic variables, and the relationship between antibody levels and stool shedding.

Typhoid and Paratyphoid Human Challenge Studies
Data were available from 6 enteric fever human challenge studies conducted in Oxford between 2011 and 2017. A list of included studies is provided in Table 1.
All challenge studies followed comparable protocols, detailed elsewhere (Supplementary Materials) [13,18,19,20,21,22]. Briefly, healthy adults drank 120 mL of sodium bicarbonate solution prior to challenge. After challenge, daily blood and stool cultures were collected for 14 days. Participants were diagnosed with enteric fever if they had fever of 38 o C for ≥12 hours and/or S. (Para)Typhi bacteremia detected ≥72 hours from challenge. Antibiotics were initiated at the time of diagnosis or at day 14 for those not diagnosed. All participants were effectively treated, and no chronic carriers were identified.
Typhoid challenge was performed using the Qualies strain (genotype 3.1.0) [15,23]. Paratyphoid challenge was performed using the S. Paratyphi A NVGH308 strain [14] Stool Culture Stool cultures were performed according to local procedures based on national guidance at Oxford University Hospital National Health Service (NHS) Foundation Trust (Supplementary Materials) [24].

Statistical Analyses
Stool culture data were combined in mixed effects logistic regression models, which included participant-specific random intercepts to account for the multiple samples per person. Models were adjusted for the vaccine received, study, and, where applicable, challenge dose (high or low). "Day" was included in the model as a categorical factor to allow the odds of stool shedding to vary by day. An overall interaction term (day-by-vaccine) was tested to determine if vaccination altered the pattern of shedding over time. An interaction term for day-by-diagnosis status was used to compare the pattern of shedding in those who were diagnosed with enteric fever and those who remained undiagnosed at day 14.
The linear predictor from the model was exported for each participant for each day and converted into a probability by taking the inverse logit. These probabilities are presented in figures with a loess smooth to illustrate the findings from the logistic regression models. Odds ratios (ORs) presented from logistic regression models represent the ratio of the odds of shedding in comparative groups on average across all 14 days.
All models were fitted in SAS version 9.4, and code is displayed in the Supplementary Materials.

RESULTS
In total, 4934 stool samples from 430 participants were analyzed; 3698 samples were from S. Typhi challenge participants and 1236 samples were from S. Paratyphi challenge participants (Table 2). In total, 14.5% of stool samples from S. Typhi-challenged participants were positive compared with 7.5% of samples from those challenged with S. Paratyphi A.

Differences in S. Typhi Stool Shedding According to Diagnosis Status
There were 331 participants challenged with S. Typhi, of whom 186 (56.2%) met the diagnostic criteria for typhoid fever [18]. The highest incidence of shedding was observed on day 1 and day 2 after challenge. The pattern of shedding over time was significantly different between those diagnosed and those who did not develop enteric fever (P < .0001 day-by-diagnosis interaction). The odds of shedding 1 day after challenge were 2.5 times greater in those who were later diagnosed than in those who remained undiagnosed (OR, 2.49; 95% confidence interval [CI], 1.32-4.69; P = .0049). On day 2 odds were 9 times higher in those later diagnosed (OR, 8.93; 95% CI, 3.86-20.61; P < .0001) and on day 3 were 23 times higher (OR, 22.59; 95% CI, 6.83-74.7; P < .0001). In the second week after challenge, many diagnosed participants had commenced antibiotics, and shedding ceased. In the undiagnosed participants, increased shedding was observed from day 10 onward, until those participants also received treatment on day 14 ( Figure 1A). A similar pattern was observed in historic challenge studies (Supplementary Figure S1).

Differences in Stool Shedding After S. Typhi or S. Paratyphi Challenge in Unvaccinated Participants
There were 197 unvaccinated participants exposed to S. Typhi and 109 unvaccinated participants exposed to S. Paratyphi. The odds of shedding in participants exposed to S. Typhi were twice as high as in those exposed to S. Paratyphi (OR, 1.97; 95% CI, 1.00-3.88; P = .049). A sensitivity analysis excluding low-dose challenge gave a very similar estimate (OR, 1.97; P = .044). The dose received was nonsignificant in the model ( Figure 1B).

Effect of Vaccination on Stool Shedding After S. Typhi Challenge
Data from 5 S. Typhi studies were analyzed. Participants challenged with a low dose or a genetically modified S. Typhi were excluded [21]. Participants who received Vi-PS or Vi-TT vaccine had lower rates of shedding than unvaccinated participants (OR, 0.34; 95% CI, 0.15-0.77; P = .010 and OR, 0.41; 95% CI, 0.19-0.91; P = .029 for Vi-PS and Vi-TT, respectively). In Ty21a vaccine recipients, there was a nonsignificant reduction in shedding compared with unvaccinated controls (OR, 0.57; 95% CI, 0.27-1.20; P = .14; Figure 1C). There were no differences between unvaccinated participants and those who received the M01ZH09 vaccine or between any other groups after adjusting for study-specific variation. The vaccine-by-day interaction term was nonsignificant, showing that the pattern of shedding over time was similar for all vaccines even though the amount of shedding differed.

Effect of Previous Exposures
Participants previously exposed to S. Typhi had less stool shedding than previously unexposed individuals (T-T vs T, OR, 0.33; 95% CI, 0.1-0.8; P = .016; Figure 2). For those participants who received S. Typhi challenge, previous exposure to S. Paratyphi significantly increased stool shedding compared with previous S. Typhi exposure (P-T vs T-T: OR, 7.5; 95% CI, 2.0-28.4; P = .003) and nonsignificantly increased the rate of bacterial shedding compared with previously unexposed individuals (P-T vs T: OR, 2.5; 95% CI, 0.8-8.0; P = .125). For those receiving S. Paratyphi challenge, there were no differences between groups, and a lower rate of shedding in general was observed in comparison with those exposed to S. Typhi challenge.

Antibody Levels Prior to Challenge
Higher anti-Vi IgG antibody titers prior to challenge were associated with less bacterial shedding after challenge with S. Typhi (P < .0001). There was no relationship between anti-LPS IgG or anti-Hd IgG and shedding after challenge with either S. Typhi or S. Paratyphi (Figure 3). IgA and IgM responses to S. Typhi LPS and Hd antigens were nonsignificant (Supplementary Figure S2).

DISCUSSION
Shedding of S. Typhi and S. Paratyphi bacilli is a prerequisite for onward transmission in these human-restricted infections. This is the first comprehensive analysis of bacterial shedding from almost 5000 stool samples taken after deliberate challenge of healthy volunteers with S. Typhi or S. Paratyphi. We demonstrate that stool shedding is more common in individuals who meet the case definition of enteric fever (fever 38 o C for ≥12 hours and/or S. (Para)Typhi bacteremia), but shedding can also occur in the absence of bacteremia or clinical symptoms of disease. Vaccination with Vi-PS or Vi-TT significantly reduced stool shedding of S. Typhi following controlled human infection, suggesting that these vaccines are likely to reduce onward transmission of disease. The decreased shedding following vaccination with the live-attenuated Ty21a vaccine was not significant, possibly due to the small sample sizes available for these comparisons and moderate protective efficacy of Ty21a in the challenge model. In earlier challenge studies with Ty21a involving larger numbers of participants who received a freshly harvested formulation of vaccine, shedding of S. Typhi was significantly reduced in recipients of the live oral vaccine [9].
The effects of Vi-PS or Vi-TT vaccination on indirect protection and stool shedding are poorly understood. A cluster randomized control trial in Kolkata, India, demonstrated that Vi-PS vaccination can result in indirect protection against typhoid fever in unvaccinated individuals resident in population clusters randomized to Vi-PS vaccine [11]. However, indirect protection of Vi-PS was not observed in another cluster randomized trial conducted in Karachi, Pakistan. One difference is that the trial in Kolkata vaccinated adults as well as children [12].
The reduction in stool shedding observed in individuals vaccinated with a Vi-TT conjugate vaccine is an important finding. To date, there are no completed cluster randomized trials of typhoid conjugate vaccines. However, 1 trial is ongoing in Bangladesh, and individually randomized trials are ongoing in Nepal and Malawi [26]. A previous Vi-conjugate vaccine with Pseudomonas aeruginosa exotoxin A as a carrier protein was shown to have a high vaccine efficacy (VE) in Vietnam (VE, 91.1%; 95% CI, 78.6%-96.5%) [27]. Vi-TT vaccine is highly immunogenic in children [28], with demonstrated efficacy of 54.6%-87.1% in a controlled human infection model (depending on the diagnostic criteria used) [13] and vaccine efficacy of 85% estimated from serological data [29]. In October 2017 the World Health Organization recommended the introduction of typhoid conjugate vaccines for children aged >6 months in typhoid-endemic countries [8]. It remains to be determined to what degree the reduction in shedding associated with Vi-TT vaccination will translate to indirect protection of unvaccinated persons in field studies. If the reduction in shedding translates to indirect protection of nonvaccinees in field settings, the overall effectiveness of Vi-TT conjugate vaccines could be significantly higher than is estimated in challenge studies.
Current models that predict the potential impact of Vi-conjugate vaccines account for this potential indirect protection by assuming that transmission of S. Typhi is reduced in a manner that is proportional to the vaccine efficacy (ie, by preventing infection and hence shedding among a proportion  of vaccinated individuals). The results presented here support such assumptions. The protection afforded by Vi-TT against typhoid infection is similar to its effect against stool shedding. When both are expressed as ORs, the effects of Vi-TT against typhoid diagnosis and shedding have largely overlapping CIs (typhoid diagnosis: OR, 0.16; 95% CI, 0.05-0.46 and shedding: OR, 0.41; 95% CI, 0.19-0.91). However, more analysis is needed to examine the relationships between vaccination, stool shedding, and the development of clinical typhoid and how this may vary between the human challenge model and field settings. Additionally, the mechanism by which anti-Vi antibodies prevent stool shedding requires further investigation.
The nonsignificant reductions in shedding with Ty21a vaccine contrast with data from early challenge studies, where Ty21a was associated with a reduction in any stool shedding of S. Typhi from between day 4 and day 30 post-challenge (OR, 0.08; 95% CI, 0.02-0.29) [9]. The differences may result from methodological differences between the Oxford and Maryland challenge studies, including mode of administration, pretreatment with sodium bicarbonate instead of milk, differences in challenge dose (Maryland, 10 5 organisms; Oxford, 10 4 ), and criteria for initiating antibiotics. Of note, Ty21a appeared to provide herd immunity in field studies conducted in Chile [10].
As was observed in early typhoid challenge studies, early shedding increased the likelihood of subsequent development of typhoid fever [15]. Interestingly, we observed an increase in shedding in the undiagnosed group from day 10 onward. It is possible that late shedding represents a harbinger of subsequent bacteremia or fever, such that a proportion of the "undiagnosed" group may have developed enteric fever had the infection not been halted by commencing antibiotics at day 14. Conversely, late, asymptomatic shedding has been described in undiagnosed participants from early challenge studies, where shedding peaked in the second week post-challenge before clearing spontaneously by 6 weeks (in the absence of clinical disease or antibiotic treatment) [15]. These data emphasize that a proportion of individuals exposed to typhoidal Salmonella will act as asymptomatic short-term carriers who transiently shed in the absence of overt clinical disease; this is a factor that should be considered when determining the target population for vaccination campaigns.
In individuals previously challenged with S. Typhi, we detected a significant reduction in the rate of stool shedding compared with naive controls. A single episode of typhoid exposure is thought to confer moderate protection against subsequent clinical disease (estimated at approximately 23% from historical challenge studies [30]), and modeling studies assume that multiple episodes of typhoid exposure are required to induce functional immunity [31]. These are the first data to suggest that prior typhoid exposure significantly affects the pattern of shedding following rechallenge, albeit from a small sample size (n = 27 rechallenged). Interestingly, we observed that individuals previously exposed to S. Paratyphi A had substantially higher rates of shedding when rechallenged with S. Typhi. The reasons for the increased risk of typhoid shedding on heterotypic rechallenge are unclear and are the subject of ongoing studies focusing on the role of secretory IgA [32].
When shedding rates were compared between typhoid and paratyphoid challenge studies, rates of shedding following S. Typhi challenge were twice as high as those following S. Paratyphi A challenge. The higher rate of shedding following typhoid challenge may reflect the higher challenge dose administered in the typhoid model (10 4 vs 10 3 colony-forming units), or differences may exist between these serovars in host-pathogen interactions and transmission mechanisms. Several studies in areas coendemic for S. Typhi and S. Paratyphi A have suggested that transmission dynamics and risk factors may differ between the 2 serovars [7,33,34]. For example, S. Paratyphi A cases appear to be more spatially dispersed than S. Typhi cases in an urban area [33] and may be particularly associated with foodborne transmission [7,34]. Improved understanding of paratyphoid shedding and transmission dynamics will be an important consideration in the development of vaccines for paratyphoid fever [35].
There are several limitations to assessing stool shedding using data from human challenge studies. The population enrolled comprises adults from a nonendemic country; shedding dynamics may differ in individuals from endemic countries with prior immune priming or in children, who represent the majority of enteric fever cases (but not necessarily the majority of shedders). In this analysis, data were pooled across 6 studies conducted over 6 years. While all samples were processed in the same laboratory using consistent protocols, study-to-study variation may still exist in the sensitivity of stool testing; thus, "study" was adjusted for in all models. Furthermore, challenge was conducted using only a single strain of S. Typhi (Quailes) or S. Paratyphi A (NVGH308) at a single dose, which may not mirror the shedding dynamics of contemporary circulating strains in Asia and Africa, such as the multidrug-resistant associated H58 (genotype 4.3.1) strain of S. Typhi [36]. These limitations primarily reflect safety considerations required for controlled human challenge, and these data should be interpreted alongside emerging data from surveillance studies [37].
In summary, we have performed the first detailed model of shedding dynamics in the context of controlled typhoid and paratyphoid challenge and provide evidence for efficacy of new and existing typhoid vaccines (Vi-PS and Vi-TT) in reducing rates of shedding. These studies illustrate the value of closely monitored experimental human challenge studies in obtaining novel insights into hostpathogen interactions and microbial dynamics, which can directly inform the disease control efforts for priority pathogens.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. A. J. P., M. A. G., A. S., and V. C. designed, oversaw, and obtained funding for study OVG2014/01. A. J. P. supervised the work and obtained funding for the 6 studies. All authors reviewed and edited the manuscript and approved the final version.