An Outpatient, Ambulant-Design, Controlled Human Infection Model Using Escalating Doses of Salmonella Typhi Challenge Delivered in Sodium Bicarbonate Solution

Oral delivery of escalating-dose Salmonella Typhi (Quailes strain) using sodium bicarbonate buffer solution in an outpatient, ambulant-design human infection study demonstrates safety, requires a lower challenge inoculum than that used in historical studies, and offers a unique insight into host–pathogen interactions.

Typhoid infection is a major global health problem [1]. Salmonella enterica serovar Typhi (S. Typhi) is a human-restricted pathogen; in the absence of suitable animal models, many details of the human response to infection remain unclear [2,3]. From 1952 to 1974, human typhoid challenge studies were performed at the University of Maryland to address this critical issue [2,[4][5][6]. A successful model was developed in which volunteers ingested S. Typhi suspended in 45 mL of milk; without buffer, 50% of participants developed clinical infection following ingestion of 10 7 colony-forming units (CFU) of wild-type S. Typhi [2,5]. Data from this model were directly applied to improve understanding of antibiotic mechanisms [7] and used in vaccine (a first step toward eventual licensure of Ty21a vaccine) [8][9][10] and diagnostic development [11,12].
Here we describe the development of a new controlled human infection model of S. Typhi challenge using outpatient, ambulant participants. The primary objective of this study was to ascertain the challenge inoculum ("dose") of S. Typhi (Quailes strain) required to produce an attack rate of 60%-75% in typhoid-naive volunteers when ingested with sodium bicarbonate buffer solution.

Study Design
An observational, dose-escalation study of controlled human infection using S. Typhi (Quailes strain) was performed. The challenge agent was delivered by oral ingestion of bacteria suspended in sodium bicarbonate solution (NaHCO 3 [aq]) using a predetermined dose-escalation strategy ( Figure 1). Attack rate was defined as the proportion of participants diagnosed with infection by day 14 after challenge meeting clinical (temperature ≥38°C sustained for ≥12 hours) and/or microbiological (blood culture-confirmed S. Typhi bacteremia) endpoints (per protocol population; "typhoid diagnosis"). An independent data and safety monitoring committee reviewed participant safety and attack rate data throughout the study, in particular cumulative data gathered following the first, fifth, 10th, and 20th challenges performed. Secondary objectives included description of the human response to and the microbiological dynamics of infection.
The study was approved by Oxfordshire Research Ethics Committee A (10/H0604/53) and conducted in accordance with the principles of the International Conference of Harmonisation Good Clinical Practice guidelines.

Setting and Participants
The study was performed at the Centre for Clinical Vaccinology and Tropical Medicine, Oxford, United Kingdom. The United Kingdom is nonendemic for typhoid fever; the rate of typhoid fever notification in Oxfordshire is 0.5 per 100 000, the majority of which is travel related [13].
Healthy adults aged 18-60 years who had not previously received typhoid vaccination or been resident >6 months in typhoid-endemic areas were eligible to participate. With written informed consent, screening was performed to assess participant health status. Exclusion criteria were extensive and included any significant medical, surgical, or psychiatric illness; evidence of gallbladder disease (by ultrasound examination); antibiotic allergy; food-handling occupation; or contact with susceptible individuals (including healthcare workers and carers of young children or elderly or immunocompromised persons). Consent was not obtained from close household contacts, but participants were required to provide these contacts with written information providing details regarding the study and measures to reduce the risk of infection. After completion of antibiotics, participant contacts were also offered the opportunity to be screened for infection.

Challenge Strain
Salmonella Typhi Quailes strain was isolated in 1958 from the gallbladder of a known chronic carrier. Quailes strain can express the Vi antigen and is fully antibiotic susceptible [2,5]. A fresh working cell bank was manufactured under Good Manufacturing Practices (GMP) guidelines prior to storage at −80°C. Whole-genome sequencing was used to determine the phylogenetic relationship of the Quailes strain to other S. Typhi (Supplementary Figure 1). These data also confirmed that the challenge strain encoded the expected repertoire of virulenceassociated determinants.

Inocula and Challenge
Inocula were freshly prepared prior to each challenge by defrosting and suspending the required number of bacteria in 30 mL/0.53 g NaHCO 3 (aq). Participants fasted for 90 minutes before ingesting 120 mL/2.1 g NaHCO 3 (aq). Two minutes later, participants ingested the prepared challenge suspension and were monitored for 15 minutes. To calculate the actual challenge dose given, direct plating of aliquots from the challenge inoculum and remaining working cell bank vials was performed in triplicate using tryptone soya agar (Oxoid) and culture for 24 hours before colony counting.

Clinical Evaluation of Participants
Participants were reviewed daily for at least 14 days, recording the duration and severity of all solicited (Supplementary Table 1) and unsolicited symptoms experienced, and twicedaily oral temperature readings, using supplied diary cards. Participants had 24-hour access to a study physician; additional reviews were performed if infection was suspected, at medical discretion, and at participant request.
Indications for antibiotic treatment (ciprofloxacin, 500 mg twice daily, 14 days) included reaching typhoid diagnosis, unmanageable symptoms, or clinical necessity. All remaining participants received treatment at day 14. Treatment compliance was ensured by direct observation of antibiotic ingestion at study visits and daily reminders by telephone contact and/or text message. Subsequent visits were performed at 21, 28, and 60 days after challenge. Clearance of infection after treatment was confirmed by bacterial culture of ≥2 stool specimens obtained at least 1 week apart, at least 3 weeks after completion of antibiotics.

Assessment of Hematological, Biochemical, and Serological Response to Challenge
Routine blood hematology and biochemistry were assessed on alternate days after challenge and at typhoid diagnosis; a maximum of 1110 mL blood was collected from each study participant by day 28. Serological response to challenge was performed using blood collected at baseline (day 0) and 14, 28, and 60 days later. Specific immunoglobulin G (IgG), immunoglobulin M (IgM), and immunoglobulin A (IgA) responses to Vi polysaccharide (Sanofi Pasteur MSD, Maidenhead), lipopolysaccharide (LPS) (L2387; Sigma-Aldrich, Dorset), and flagellin ( prepared in-house by shearing and centrifugation of a whole-cell preparation) were measured in serum by enzyme-linked immunosorbent assay, using goat antihuman IgG, IgM, and IgA conjugated to specific horseradish peroxidase (AbD Serotec, UK).

Assessment of Microbiological Dynamics After Challenge
Salmonella Typhi shedding in stool was assessed using daily selfcollected samples, cultured according to national standard operating procedures [14]. Blood (10 mL or 5 mL at typhoid diagnosis) was cultured daily by direct inoculation into broth (BACTEC Plus Aerobic vials, BD) and subsequent automated growth detection (BACTEC FX System, BD), in accordance with standard methods [15]. Salmonella Typhi growth and serotype were confirmed by biochemical profile (API-10S, bio-Mérieux, France) or slide agglutination according to the Kauffman-White classification, respectively [16].
Quantitative blood culture was performed at typhoid diagnosis by inoculation of 10 mL blood into an ISOLATOR 10 tube (a commercial lysis-centrifugation system; Alere, UK). After centrifugation, the resulting pellet was directly plated onto XLD agar (Oxoid, UK). Quantitative stool culture was performed by suspending 1 g of stool in sodium selenite, followed by subculturing onto XLD agar (Oxoid). Blood-or stool-inoculated plates were then incubated (37°C for 24 hours) prior to identification and counting.

Statistical Analyses
An initial challenge dose of 10 3 CFU was chosen as the highest dose from the historical studies that did not lead to clinical infection [5,17,18]. With a final sample size of 20 participants at each dose level, 95% confidence intervals (CIs) of 36%-81% and 46%-88% for measured attack rates of 60% or 70%, respectively, were anticipated. Analyses are descriptive and comprise proportions or means with associated 95% CIs. Diagnostic odds ratios (DORs) (a single indicator describing the ratio of the odds of positivity in infection relative to the odds of positivity in the noninfected; 1 indicating that the test does not discriminate between infected and uninfected) [19] and 95% CIs were calculated to assess performance of individual symptoms reported for typhoid diagnosis to be made.

Participants
From February to October 2011, 41 participants were enrolled and challenged with 10 3 or 10 4 CFU of S. Typhi (Table 1). Forty participants were included in the per protocol analysis ( Figure 2); 1 participant challenged with 10 3 CFU was treated for symptoms related to infection before day 14 without meeting diagnostic criteria and was therefore excluded.

Clinical Safety
Challenge was well tolerated; no participants required hospital admission, intravenous antibiotics, or fluids. Four participants fulfilled criteria for severe infection ( Table 2). No episodes of shedding or carriage following the initiation of antibiotics were identified.
Symptoms consistent with typhoid infection were reported from day 4. The clinical profile of infection varied considerably between participants, although headache was universally re- Clinical observations made are summarized in Table 3. Relative bradycardia was seen at typhoid diagnosis in 4 of 7 participants with temperatures ≥38.3°C (ie, recorded heart rate <110 bpm) or 2 of 2 participants with temperatures ≥38.9°C (heart rate <120 bpm) [20].

Hematological and Biochemical Response
In participants developing typhoid infection, laboratory abnormalities were evident prior to the onset of clinical typhoid symptoms (Supplementary Figure 3).
Hemoglobin concentrations fell progressively in all groups after challenge (Supplementary Figure 3E ). An initial fall in   participants diagnosed with typhoid infection; platelet counts fell to <150 × 10 3 cells/µL in 11 of 24 participants (Supplementary Figure 3F ).

Microbiological Dynamics
Salmonella Typhi was cultured from blood in 10 of 11 (90.9%) and 11 of 13 (84.6%) diagnosed participants given 10 3 or 10 4 CFU, respectively (Table 4); however, those participants ingesting 10 4 CFU had a greater proportion of positive blood cultures prior to antibiotics being initiated (10 3 CFU, 6.2%; 10 4 CFU, 9.6%). Quantitative blood culture performed at typhoid diagnosis demonstrated median bacterial loads of 0.47 CFU/mL and 1.10 CFU/mL in those challenged with 10 3 CFU and 10 4 CFU, respectively (Table 4). Predefined criteria for typhoid diagnosis were either clinical (reaching and sustaining an oral temperature of ≥38°C for ≥12 hours) or microbiological (having 1 or more confirmed positive blood cultures). Criteria for severe typhoid included any of the following: oral temperature measurement ≥40°C, lethargy or confusion, gastrointestinal bleeding or perforation, or any grade 3 or higher laboratory abnormality (these are described for the 2 individual cases in footnotes a and b below).    Stool quantification data for participants challenged with the 10 3 CFU dose demonstrated that the median number of bacteria excreted was 32 (IQR, 18-41) CFU/g feces (n = 13).

Serological Response
Low levels of anti-H and anti-Vi antibody were measured in all participants at baseline (Supplementary Figure 4 and Supplementary Table 2), and did not correlate with subsequent risk of infection using the study diagnosis definition (data not shown). In participants developing typhoid infection, increases in IgG, IgM, and IgA to LPS and H antigens were seen, whereas anti-Vi levels remained unchanged throughout (Supplementary Figure 4). Little change was seen in participants not succumbing to infection after challenge.

DISCUSSION
This is the first investigation of the human pathophysiological response to S. Typhi challenge in nearly 40 years [2]. We have demonstrated that human challenge with S. Typhi (Quailes strain) using an outpatient model is safe and well tolerated, even in participants developing blood culture-confirmed symptomatic infection. Using coadministered sodium bicarbonate solution, an almost 4-log 10 lower dose of S. Typhi was required to produce the target attack rate of 65% in comparison to historical challenge studies [5,17]. A target attack rate of 60%-75% was selected based on historical typhoid challenge data from Maryland, providing a balance between using a smaller and therefore more physiological exposure dose while achieving a moderately high attack rate. A relatively high attack rate is advantageous in performing vaccine and treatment studies as smaller sample sizes are required, thereby reducing the number of individuals exposed and making studies more logistically feasible [18]. For some human challenge models of bacterial enteropathogens, eg, Vibrio cholerae O1, inocula require coadministration of a buffer agent for clinical infection to ensue. For others, such as Shigella dysenteriae or enterotoxigenic Escherichia coli, administration with sodium bicarbonate buffer both reduces the inoculum dose required and results in a more consistent pattern of clinical infection [21][22][23][24]. From this study, Salmonella Typhi challenge appears to fall into the latter category, the features of which are seen as desirable prerequisites for using a human challenge approach to assess vaccine and therapeutic efficacy. In particular, lower levels of pathogen exposure are thought to more closely replicate natural infection and there is less chance of overwhelming any infection-or vaccinederived protective responses seen, an issue noted previously in typhoid challenge/vaccine studies [10,22,23,25].
Diagnostic endpoint definitions are critical to determining the attack rate in human challenge studies, for drawing conclusions regarding dose response, and for their practical application to assessment of diagnostics, vaccines, and therapeutics. The effect of altering the stringency of the diagnostic criteria to make a diagnosis of typhoid infection was highlighted retrospectively in a reanalysis of the Maryland studies; the stringent criteria used potentially led to missed cases of milder disease, thus revealing a dose response when less stringent criteria were used [17]. We provide some confirmation of this dose response with use of bicarbonate, in which a higher challenge dose produces a higher attack rate (with a variety of definitions) and a shorter duration to the development of bacteremia. Interestingly, there was no association between dose and severity of clinical disease or time to onset of symptoms, which was also noted in the Maryland studies [26]. The reasons for this disparity remain unclear and are under further investigation. Sensitivity analysis around the diagnostic definition we used demonstrates a wide range of resulting attack rates, from 32% to 68% (Table 2). We suggest that reporting a range of outcomes in this way is useful for extrapolation to a variety of different settings; that is, a more relaxed endpoint definition may be most useful in calculating likely vaccine efficacy in field settings, whereas strict criteria may be more applicable to the development of novel diagnostic assays.
This study has also provided further data regarding the "milder" end of the typhoid infection spectrum, confirming that subclinical and asymptomatic infections may be relatively more common than previously anticipated or measured. Although cases of asymptomatic bacteremia were occasionally seen in the Maryland studies, these were often complicated by prior receipt of endotoxin challenge or may have been explained by previous vaccination or exposure [27]. Here, we demonstrate that asymptomatic bacteremia is a distinct pattern of infection after exposure (challenge), which may have implications for assessment of patients in the field, and for transmission modeling and assessment of vaccine mechanisms.
Whereas the range of symptoms reported by participants in our study is in keeping with those from previous Maryland studies and community settings, in general, our participants were more symptomatic throughout the challenge period [28][29][30][31][32][33]. Reasons for this may include heightened symptom reporting (and solicitation) due to perceived study risk profile, lower levels of preexisting immunity, or the relatively high exposure dose used in this artificial setting. This range and severity of symptoms may also be more in keeping with the true early, untreated (both antibiotic, analgesic, and anti-inflammatory) clinical presentation of typhoid infection, however [34,35], and are more compatible with those noted among ill returning travellers, rather than patients in endemic settings or children [30][31][32][33].
Important laboratory findings in this study include further detail regarding the development of hematological changes prior to and following the development of clinical typhoid infection. A gradual fall in hemoglobin and hematocrit was seen in all participants, likely as a result of the venesections performed, but for which there may be additional explanations including bacterial suppression of bone marrow function (a fall in platelets was also seen in typhoid-diagnosed participants). Early and almost complete loss of eosinophils from peripheral blood was observed in all participants subsequently developing infection. There are few previous reports regarding this phenomenon, but these include observations made in returning travelers [36]; although not specific for typhoid fever diagnosis, this finding may be useful in predicting likely outcome in the highly controlled challenge setting.
Although the number of bacteria required to cause natural typhoid infection remains to be elucidated, microbiological blood quantification data provide indirect support for the doses used here. We found very low numbers of bacteria to be present in the peripheral blood at the time of clinical diagnosis, a finding supported by field data, albeit in an endemic setting [37]. These very low numbers reiterate the complexity of developing useful sensitive diagnostics for typhoid. Many participants were stool culture positive in the days preceding the development of bacteremia and fever, a period that did coincide with the initial onset of symptoms (headache, abdominal discomfort, etc). This prepatent period may represent a period of heightened transmission risk, for example, to household contacts caring for a sick relative, and supports renewed national recommendations for contact screening [38].
Longitudinal sample collection during the course of bacterial challenge provides unique insight into the kinetics of antibody response following pathogen exposure. In agreement with previous studies, we found relatively high baseline levels of S. Typhi-specific LPS antibody in nonexposed/nonvaccinated study participants [39]. Contrary to the findings from Maryland, these baseline titers did not correlate with subsequent infection risk. Similarly, those participants not developing typhoid fever consistently demonstrated scant response to the surface-expressed antigens studied [28]. Although the use of different antigen preparations in various studies may be partially responsible, the reasons for this discrepancy remain unclear. In participants diagnosed with typhoid, most developed measurable LPS and H antibody responses by day 14 or day 28; among those who did not were several individuals with blood culture-confirmed infection. This finding underscores a key central limitation in using serology to sensitively detect acute infection. Interestingly, no change in Vi antibody titer was seen in study participants, supporting previous findings from Maryland and those from studies of engineered Vi + Ty21a vaccine strains [28,40]. One reason for initially choosing the Quailes strain for challenge was its consistent Vi expression; indeed, all S. Typhi isolates retrieved from blood in this study demonstrated Vi positivity. Failure to mount an antibody response to the Vi antigen in acute infection has not been adequately explained and requires further investigation; these data do, however, suggest that natural infection-derived immunity is unlikely to be Vi mediated.
Limitations to the interpretation of data derived from this typhoid challenge model include logistic, ethical, and financial factors restricting the feasible challenge observation period to 2 weeks. The usual incubation period following exposure to natural infection is 7-14 days (range, 3-60 days) [41]; hence, treatment of all participants with antibiotics at day 14 may have curtailed or prevented infections that would otherwise have occurred had observation been extended. Furthermore, it is possible that the Quailes strain is not truly representative of currently circulating virulent strains; however, we demonstrate close phylogenetic relatedness to known disease-causing strains (Supplementary Figure 1).
In summary, we report the successful development of a new controlled human infection model of S. Typhi using healthy community volunteers. In addition to enabling detailed investigation of host-pathogen interactions and providing data to inform transmission modeling, use of this new challenge model will facilitate and expedite the development of new diagnostics, vaccines, and therapeutics, which are much needed for improved global control and ultimate eradication of this human blight.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.