Environmental surveillance for Salmonella Typhi as a tool to estimate the incidence of typhoid fever in low-income populations

Background: The World Health Organisation recommends prioritised use of recently prequalified typhoid conjugate vaccines in countries with the highest incidence of typhoid fever. However, representative typhoid surveillance data are lacking in many low-income countries because of the costs and challenges of diagnostic clinical microbiology. Environmental surveillance (ES) of Salmonella Typhi in sewage and wastewater using molecular methods may offer a low-cost alternative, but its performance in comparison with clinical surveillance has not been assessed. Methodology/Principal Findings: We developed a harmonised protocol for typhoid ES and its implementation in communities in India and Malawi where it will be compared with findings from hospital-based surveillance for typhoid fever. The protocol includes methods for ES site selection based on geospatial analysis, grab and trap sample collection at sewage and wastewater sites, and laboratory methods for sample processing, concentration and quantitative PCR to detect Salmonella Typhi. The optimal locations for ES sites based on digital elevation models and mapping of sewage and river networks are described for each community and their suitability confirmed through field investigation. We will compare the prevalence and abundance of Salmonella Typhi in ES samples collected each month over a 12-month period to the incidence of blood culture confirmed typhoid estimated from cases recorded at referral hospitals serving the study areas and community surveys of healthcare seeking for individuals with fever. Significance: If environmental detection of Salmonella Typhi correlates with the incidence of typhoid fever estimated through clinical surveillance, typhoid ES may be a powerful and low-cost tool to estimate the local burden of typhoid fever and support the introduction of typhoid conjugate vaccines. Typhoid ES could also allow the impact of vaccination to be assessed and rapidly identify circulation of drug resistant strains.

6.4 billion people living in low-and middle-income countries. The WHO recently prequalified highly 47 immunogenic typhoid conjugate vaccines (TCVs) and recommended that the introduction of public 48 vaccination programmes should be based on an understanding of the local epidemiology of 49 infection. However, representative data on the community level describing incidence of 50 microbiologically confirmed typhoid cases are limited, reflecting the costs and challenges of 51 implementing blood culture for Salmonella Typhi (S. Typhi). Therefore, cost-effective alternative 52 means of community surveillance are required to assess overall burden of disease. 53 We developed a protocol for typhoid environmental surveillance, including methods for the 54 identification of suitable sampling sites, sample collection and laboratory processing. The detection 55 of S. Typhi in wastewater and sewage samples will be examined against the outcomes of hospital-56 based surveillance for typhoid fever with healthcare utilisation surveys to evaluate and validate the 57 use of ES as an effective means of assessing community level incidence of typhoid. The findings of 58 this study will inform potential wider implementation of ES for typhoid to support the introduction 59 of TCVs and monitor their impact. 60 conditions are common [4,5]. Incidence is typically lower in rural areas, although recent findings in 68 rural or peri-urban sites describe a higher burden than previously appreciated [6,7]. Whilst 69 increasing access to safe drinking water, improved sanitation and antibiotics have helped reduce the 70 incidence of typhoid fever, its persistence in low-income countries and the emergence of antibiotic 71 resistance, including extensively resistant (XDR) strains, are a major concern [8,9]. 72 The isolation of S. Typhi by culture remains the gold standard method of detecting typhoid 73 both in clinical practice and for the purposes of typhoid surveillance [10]. However, microbiological 74 culture is time-consuming and resource intensive, and therefore has been difficult to sustain in low-75 Study design 120 The locations for this study will be the central area of Vellore, India and the city of Blantyre, 121 Malawi. Study boundaries are defined as those areas covered by hospital-based blood culture 122 surveillance and planned healthcare utilisation surveys. In Vellore, this corresponds to the study area 123 originally identified by the Surveillance of Enteric Fever in India (SEFI) program [28] and represents 124 an area of 16 km 2 . In Blantyre, this includes the entire municipality, consisting of the central urban 125 districts and surrounding areas, representing 214 km 2 . We will compare the prevalence and 126 abundance of S. Typhi in ES samples between these study locations and within locations over time 127 and space during a 12-month period. We will also assess the association between the prevalence 128 and abundance of S. Typhi in ES samples and the incidence of culture-confirmed typhoid fever cases 129 in ES site catchment populations as reported at local hospitals. 130 The estimated incidence of typhoid fever among children in Vellore during 2017-2020 was 131 about twice that for Blantyre during a similar time period 2016-2018 (approximately 2% vs 1% per 132 year of observation for children <15 years old) [29]. Pilot ES surveillance data from Vellore based on 133 direct molecular detection methods using either grab samples (bag-mediated filtration) or trap 134 samples (Moore swabs), detected S. Typhi in 14% of all samples. Similar data are not yet available 135 from Blantyre. If typhoid ES is to give a reasonable indication of the incidence of typhoid fever, we 136 might expect the prevalence of S. Typhi in ES samples in Blantyre to be about half that observed in 137 Vellore (i.e. 5-7% vs 14%). To detect this difference in proportions with 80% power we would need 138 to sample between 26 and 47 ES sites with monthly sample collection in each location, using a 139 mixed-effects logistic regression model to allow for clustering of detection by site (the intraclass 140 correlation coefficient in Vellore was 0.08 for the pilot data) [30]. We therefore decided to select 40 141 ES sites at each location and to sample every month. We will also compare the detection (yes/no) 142 and abundance (genome copies) of S. Typhi at each site over time with the incidence of typhoid 143 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 22, 2021. ; https://doi.org/10.1101/2021.05.21.21257547 doi: medRxiv preprint population density estimates for both the full catchments and those spatially clipped to the study 170 areas. Digital land use and human population datasets indicated key differences in the 171 environmental composition between the two study areas. Specifically, Vellore consists exclusively of 172 medium to high population density urban environments, whilst the larger Blantyre study site 173 exhibits a variety of land use types, including agricultural lands, natural grassland and forested areas 174 in addition to both peri-urban and urban areas characterised by a wide range of residential settings 175 (e.g. informal, traditional and permanent dwellings). To harmonise the study design and improve 176 consistency and comparability between the 2 locations, we only retained those candidate ES sites 177 throughout the Blantyre study area that produced catchments consisting mostly of medium and high 178 population density urban areas. All other ES sites which did not meet this requirement were deleted. 179 Each study area was divided into an array of equal size rectangular cells (i.e. fishnet) used to 180 generate a sampling grid. Due to the differences in areal extents between the two study areas, 181 different size grid networks were selected to ensure appropriate resolution for each location: 500 m 182 and 75 m for Blantyre and Vellore, respectively. The number of candidate ES sites per grid cell was 183 calculated and for each sample cell that contained multiple potential ES locations the downstream 184 site was retained whilst all others were deleted. However, if two or more ES sites in the same grid 185 cell were situated on different river streams or drainage network channels, therefore generating 186 independent catchments, then both/all were retained. 187 Finally, to ensure that the finalised list of ES sites for each location represented a wide range 188 of catchment sizes, candidate ES sites were stratified by catchment population estimates and 189 categorised into 3 approximately equal size groups (terciles) representing small, medium and large 190 catchment populations. In order to maximise detection sensitivity, ES sites with the greatest 191 population density estimates for each catchment size category were identified and retained as 192 priority ES sites. Once finalised, the full lists of priority candidate ES sites were visited by field-based 193 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 22, 2021. ; https://doi.org/10.1101/2021.05.21.21257547 doi: medRxiv preprint teams to undertake detailed site descriptions and assess their overall suitability for inclusion in the 194 study based on accessibility, perceived quality of sampling location (e.g. adequate depth and flow of wastewater) and personnel safety concerns. Sites that were deemed unsuitable where removed. Of 196 the suitable locations, a total of 40 sites for each study area were confirmed as the project ES sites. 197 Where available, previously obtained data during initial pilot studies were used to help guide the 198 decision-making process. During this iterative review, it was considered important to retain broad 199 spatial coverage throughout both study areas and maintain a wide distribution of catchment sizes as 200 determined by population estimates. 201

202
Following the recommendations of the BMGF Typhoid Environmental Surveillance Expert 203 Advisory Committee (EAC), wastewater/sewage samples will be collected using two separate 204 collection procedures: trap sampling using Moore swabs (i.e. gauze pad) and grab samples with 205 membrane filtration. Moore swabs made of hospital gauze will be prepared following the protocol 206 described by Sikorski and Levine [35]. At field ES sites, swabs will be fixed in place using wire or 207 twine tied to a fixed object or stone and suspended in flowing wastewater or sewage for 48-72 208 hours. Upon retrieval, Moore swabs, including their attachment line, will be placed in containers 209 containing 450 ml of Universal Pre-Enrichment (UPE) broth (BD Difco, Fisher Scientific) [36]. For the 210 grab samples, wastewater/sewage samples will be collected using standard sample collection 211 equipment (e.g. bucket and rope, long handled dipper or similar) used to fill a 1 litre container. Once 212 collected, all sample containers (Moore swab and grab samples) will be disinfected externally and 213 placed in an ice box for immediate return to the laboratory within 4 hours of retrieval. Sample 214 containers will be labelled with sample ID, date, sample type (e.g. sewage or wastewater) and initials 215 of personnel collecting the sample. All other sampling equipment will be wiped down with 216 disinfectant. Suitable personal protection equipment (PPE) (including gloves, lab coat, glasses, and 217 face mask) will be donned upon arrival at the sampling site and GPS coordinates will be recorded 218 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 22, 2021.
using handheld GPS device. Each quarter, wastewater physicochemical properties will also be 219 measured at the site (including temperature, pH, oxidative reduction potential (ORP), dissolved 220 oxygen, total suspended solids (TDS), salinity and turbidity) using a water quality probe following a 221 published protocol [17]. At the first visit field staff will also collect information on local infrastructure 222 and ES site characteristics using an electronic questionnaire. 223 Blood culture surveillance 224 Information on the number, age, and place of residence for blood culture confirmed typhoid 225 cases recorded at hospitals serving the study communities during the study period will be extracted 226 from the electronic records. In Vellore, the Christian Medical College (CMC) hospital, a 1,721 bed 227 tertiary care centre, and its two satellite facilities, the Low Cost Effective Care Unit and the CHAD 228 hospital, serve the study area along with the government-run Pentland Hospital. We will use health 229 seeking survey instruments to determine the probability that a typhoid case seeks care at one of 230 these participating hospitals. Age-appropriate blood volumes (3 ml in <1 year; 5 ml in 1 to 15 years; 8 231 to 10 ml >15 years) will be obtained from those hospitalised with an acute febrile illness and culture 232 for S. Typhi done using the automated BacT/ALERT system (bioMérieux, France) at the CMC Vellore 233 microbiology department. 234 In Blantyre, the Queen Elizabeth Central Hospital (QECH) serves the study area and 235 surrounding districts. Blood will be sampled from paediatric patients presenting with non-specific 236 febrile illness, who test negative for malaria, are severely ill with suspected sepsis, or fail initial 237 malaria treatment and remain febrile, and from all febrile adult patients admitted. 2-4 mL of blood 238 will be taken for culture from children (aged younger than 16 years) and 7-10 mls from adults under 239 aseptic conditions. All blood samples will be cultured using the automated BacT/ALERT system 240 (bioMérieux, France). 241 Bottles that flag as positive after incubation will have Gram stain done and Gram-negative 242 bacilli will be identified using API biochemical testing (bioMérieux, France). Salmonellae, including S. 243 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

246
On arrival at the laboratory, Moore swabs will be incubated at 37°C for 24 hours in the 247 containers of UPE to enrich bacteria trapped in the swab. 40 ml will be filtered in two aliquots of 20 248 ml under a vacuum and the filter discs cut into strips and stored at -20°C until DNA extraction. Grab 249 samples for membrane filtration will be pre-filtered using a coffee filter, then filtered under a 250 vacuum using up to five filter discs (0.45 µm) which will then be eluted using 10 ml Ringers lactate 251 solution. 1 ml of the Ringers lactate solution will then be centrifuged, and the resulting pellet stored  Table 1. 263 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted CTT CCT CTC AGA ACC CCT ATC C HF183_P [

FAM]-CT AAT GGA ACG CAT CCC-[BHQ1] 266
Reactions can be carried out in a multiplex or as single reactions and will be carried out in a 267 final volume of 25 µl using a qPCR mastermix with a ROX reference dye for normalisation (multiple 268 mixes will be evaluated). Primers and probes will be included at a final concentration of 0.5µM and 269 0.1µM respectively for tviB, staG and HF183, and ttr primers and probe at 0.25µM and 0.5µM. 270 Cycling conditions for all qPCR reactions will be as follows: 50°C for 2 minutes then 95°C for 15 271 minutes followed by 40 cycles of 95°C for 30s, 60°C for 30s and 72°C for 30s. The Ct values for each 272 gene target will be assessed to decide whether a sample is positive for S. Typhi. A standard curve for 273 each target will be produced by running a qPCR on a dilution series of gBlocks Gene Fragment DNA 274 standards (IDT) in water. 275 276 We will compare the prevalence of S. Typhi in ES samples between Vellore and Blantyre 277 using a mixed-effects logistic regression model with a random effect on the intercept by ES site to 278 allow for repeated measures. In each location, mixed-effects logistic regression and binomial 279 regression will be used to investigate the association of S. Typhi detection with ES site characteristics 280 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 22, 2021. ; https://doi.org/10.1101/2021.05.21.21257547 doi: medRxiv preprint including the physicochemical properties of the wastewater/sewage, local infrastructure, climate 281 variables, catchment population size and characteristics, including the number of blood culture 282 confirmed typhoid cases among residents within the catchment area. We will also conduct 283 exploratory analyses of the spatial distribution of S. Typhi detection, looking for hotspots for 284 transmission and their association with local population characteristics. 285 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

287
The geographic distribution of both candidate and field-based confirmed ES sites for the 288 Blantyre and Vellore study areas is shown in Fig 1A and Fig 1B, respectively, and summary statistics  289 for both the full and spatially intersected catchments are presented in Table 2. Full details can be 290 found in the supplementary material (S1 Table) For Blantyre, an initial list of 93 unique priority 291 candidate locations were identified using the GIS-based site selection approach. Candidate ES sites 292 were relatively well distributed throughout the study area; however, the majority were located 293 throughout the northern and eastern quarters of Blantyre, consistent with the distribution of 294 densely populated areas served by the Mudi and Lukhubula rivers which drain to the southwest and 295 northwest, respectively. Hydrological modelling indicates that catchments for candidate ES site 296 ranged from approx. 0.1 km 2 to 21.2 km 2 (mean: 5.1 km 2 ). Moreover, catchment population 297 estimates varied between approx. 400 to over 107,000, stratified into 3 equal number groups 298 (terciles; n31): small <9,000; medium 9,000 -35,000; and large >35,000. Mean population density 299 for all 93 catchments was 8,161 inhabitants/km 2 (range: 3,550 -30,268 inhabitants/km 2 ), remaining 300 relatively consistent between the population stratified groups with estimates of 9,086 301 inhabitants/km 2 , 8,056 inhabitants/km 2 and 7,340 inhabitants/km 2 for the small, medium and large 302 population terciles, respectively. Most of the ES sites catchments overlapped with the study area, 303 such that the mean extent and population of the full catchments and those clipped to the study area 304 were similar. 305  For Vellore, a total of 98 candidate ES site locations were initially identified. The bulk of the 317 sites were located within the study boundary and spatially consistent with those areas better served 318 by the artificial open drainage sewer network. Overall, the mean area for the full unprocessed 319 catchments was 22.5 km 2 , however, this masks significant variability and a highly skewed 320 distribution of catchment sizes ranging from <1 km 2 to 175 km 2 (median 0.81 km 2 ) with many 321 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 22, 2021. ; https://doi.org/10.1101/2021.05.21.21257547 doi: medRxiv preprint catchments extending well beyond the study boundary. Similarly, catchment population estimates 322 varied considerably from <50 to over 280,000. Using the tercile stratification, small medium and 323 large class catchments were based on approximate population thresholds of <1,300, 1,300 -22,000 324 and >22,000, respectively. Catchment extent and population metrics were significantly altered when 325 using the processed catchments that spatially intersected with the study boundary. Using these 326 data, the maximum catchment size was 15.6 km 2 (approx. 94% of the study area), representing an 327 estimated population of approximately 111,750. Moreover, mean population density estimates 328 using the cropped catchment extents remained relatively consistent, ranging from 6,863 329 inhabitants/km 2 to 5,534 inhabitants/km 2 for small and large category catchments, respectively. 330 These refinements were not required for Blantyre as the entire urban area was selected as described 331 above. 332 Candidate ES site locations underwent a significant review process and following field-based 333 site investigations and detailed discussions between study partners in each study location, the 334 proposed lists of candidate sites were used to identity and confirm the 40 ES sites to be retained for 335 the study. ES sites confirmed for inclusion in the study are presented in Fig 1a and Fig 1b.  336 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. To address these gaps, we describe the design of a study to detect S. Typhi in environmental 350 samples and correlate these findings with contemporaneous hospital-based blood culture 351 surveillance serving the same communities over a 12-month period. This will provide timely 352 information to help ascertain the technical feasibility and reproducibility of typhoid ES and its utility 353 in determining the local incidence of typhoid fever. It will also inform our understanding of the 354 spatiotemporal patterns of S. Typhi transmission within the community and the optimal ES site 355 characteristics for S. Typhi detection. 356 The relatively low cost and scalability of ES means it can also be expanded to cover large 357 geographic areas that include a range of settings (i.e. rural, urban and peri-urban) therefore allowing 358 decision-makers to better understand the epidemiology of infection and develop and deploy an 359 optimum region-specific vaccination programme. Specifically, such information could allow for the 360 more appropriate distribution of TCV relative to the spatial burden of typhoid fever. ES could also be 361 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 22, 2021. ; https://doi.org/10.1101/2021.05.21.21257547 doi: medRxiv preprint deployed in the event of a suspected outbreak of typhoid fever and used to locate high risk areas for 362 targeted vaccination. Finally, ES may also be valuable in evaluating the impact of TCV following its 363 roll out by monitoring potential reductions in S. Typhi prevalence or abundance in environmental 364 samples. 365 In preparing the current study protocol, several important considerations were made. 366 Firstly, we develop a representative site selection strategy that ensures the findings of the ES can be 367 effectively compared with the incidence of typhoid fever estimated from healthcare utilisation 368 surveys and the number of cases reported by local hospitals serving the community. Secondly, the 369 study design sought to maximise the comparability of results between countries by standardising the 370 methodological approach. Thirdly, we aimed to generate an easily reproducible approach to be 371 adopted as a standardised procedure for future studies. This was particularly important as we plan 372 to expand the proposed study to include a third study site in the Asante Akim North district, Ghana. 373 We believe that the design of this prospective study has several strengths. First, we 374 anticipate that the inclusion of two distinct, yet comparable study sites characterised by differing 375 environmental conditions will provide an opportunity to elucidate potential differences or 376 similarities in the distribution and abundance of S. Typhi and performance of ES between these 377 settings. Second, in ensuring the broad geographic distribution of ES sites at each study area, 378 therefore capturing a wide range of microenvironments, we will be able to investigate how different 379 ES site and catchment characteristics might correlate with the detection of S. Typhi. We also 380 acknowledge that the proposed study may be affected by several limitations. First, upon the initial 381 commencement of the study, the program of ES will be geographically restricted to two locations 382 and temporally limited to just one year of active surveillance. Moreover, due to the ongoing global 383 COVID-19 pandemic, we recognise that healthcare utilisation patterns within the study areas are 384 likely to be impacted during the study period. Regarding the Vellore study location, due to the 385 relatively coarse spatial resolution of the DEM used during the hydrological modelling process, 386 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 22, 2021. ; https://doi.org/10.1101/2021.05.21.21257547 doi: medRxiv preprint coupled with the high level of detail in the wastewater network, there is some uncertainty in the 387 drainage catchments, and therefore population metric estimates, for the proposed ES site locations. 388 However, we anticipate that the collection of ancillary data (such as local hydrologically) during the 389 initial phase of the study, will allow us to refine the catchment modelling process and improve 390 predictive accuracy of catchment wide population estimates. Finally, we recognise that despite 391 significant advancements in molecular detection methods, there remain uncertainties regarding its 392 sensitivity to detect S. Typhi in ES samples [26]. For example, molecular detection from complex 393 environmental samples is often hampered by relatively low sensitivity and specificity [35] and 394 therefore may yield false-negative results, thus prompting an underestimation of S. Typhi in the 395 environmental samples [43]. 396 Whilst active acute febrile illness surveillance with microbiological confirmation by culture 397 remains the gold standard for estimation of the burden of typhoid fever in a community, such 398 approaches are hugely time and resource demanding. Well designed and sensitive ES of S. Typhi may 399 offer a practical and relatively low-cost alternative that could rapidly generate vital information 400 regarding the prevalence of infection. This information could be used by national technical advisory 401 groups and others to determine the likely benefits of TCV introduction and to monitor the impact of 402 vaccination or WASH interventions to tackle typhoid fever. 403 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 22, 2021. ; https://doi.org/10.1101/2021.05.21.21257547 doi: medRxiv preprint