Background
Food-borne diseases are responsible for considerable morbidity, mortality and economic cost [
1‐
3]. The global human health impact of non-typhoidal
Salmonella infection can be as high as 93.8 million illnesses and 155,000 deaths each year [
1]. Many cases of salmonellosis could be prevented if common outbreak sources could be identified rapidly, thus enabling earlier public health interventions. However, increased mobility of people and complexity of food production, processing and distribution systems have complicated the recognition and investigation of outbreaks [
4]. Salmonella-related outbreaks are increasingly associated with a diverse range of sources, yet the mechanisms of contamination often remain poorly understood [
4,
5]. Purely epidemiological approaches to distinguishing outbreaks from sporadic cases significantly underestimate the proportion of outbreak cases. Furthermore, linking laboratory results to public health actions and increasing the timeliness of case follow-up are critical to reducing delays in the investigation of outbreaks [
6]. Recent evidence suggests that the geospatial clustering of food-borne isolates supports a more timely detection of otherwise unrecognized outbreaks [
7].
Salmonella enterica subsp.
enterica serovar Typhimurium (
S. Typhimurium/STM) has been predominant for decades in Australia and elsewhere in the world [
8]. It represents approximately one third of human and one quarter of bovine and chicken isolates, albeit with considerable geographic and temporal variation. Multilocus sequence typing of STM isolates from sub-Saharan Africa suggests the continuous evolution of STM towards a more human-adapted invasive life-style, similar to that seen in
S. Typhi [
9]. STM is a diverse serovar and additional subtyping is needed for outbreak detection and investigation. Traditionally, the method of choice has been phage typing, which is based on the susceptibility of isolates to a panel of 34 bacteriophages. The current phage typing scheme recognises 207 definitive and many more provisional phage types [
10]. Over the past 5 years, phage types 9, 170, 135 and 135a accounted for 70% of the STM isolates received by the Enteric Reference Laboratory in New South Wales. Clusters are difficult to recognise among common phage types and a more discriminatory subtyping method is needed. Numerous subtyping methods for STM have been used previously, including pulsed-field gel electrophoresis (PFGE), IS
200 restriction fragment length polymorphism (RFLP) and amplified fragment length polymorphism (AFLP), but all have disadvantages [
11‐
13]. Recently, multiple locus variable-number tandem-repeat (VNTR) analysis (MLVA) has been suggested as a rapid alternative to RFLP [
14,
15], capable of discriminating strains within the most common phage types [
14].
While typing methods have usually been used to confirm epidemiological links, the rapid turn-around time of PCR-based methods makes them suitable for the prospective detection of clusters for further investigations [
16]. Other advantages of MLVA are the relative ease of implementation and harmonisation of the method between laboratories [
17]. The emerging evidence suggests that MLVA is superior to RFLP and PFGE for both surveillance and outbreak investigations [
15] and that prospective monitoring of the local epidemiology of STM using MLVA can be beneficial for tracing possible sources of community cases of salmonellosis [
18‐
21]. However, these studies were based on relatively small numbers of isolates and did not detail specific insights that such high resolution typing could provide.
In our previous study [
22], we described the importance of scalable cluster definitions adjustable to reflect changes in local STM disease prevalence and the availability of public health resources. The objectives of this project were to explore the MLVA dynamics of human STM infection in a 3-year prospective study of STM MLVA typing for public health surveillance. We also tested the sensitivity of STM MLVA cluster definition in a relatively low prevalence country, using New South Wales (NSW), the most populous state of Australia, as an example.
Discussion
Our findings extend previous studies by providing additional evidence on the utility of the prospective MLVA typing of STM. Prospective MLVA typing of STM improves the resolution of outbreak detection, breaks down each phage type into several MLVA profiles and offers new insights into the diversity of genotypes. At the same time, our results emphasize important variations in the discriminatory power of MLVA typing. We confirmed the previous observation [
20], based on a set with different distribution of phage types (DT101, 104 and 160) [
20], that the locus STTR10pl has the greatest number of different alleles, followed by STTR5 and STTR6.
The role of appropriate cluster definition was one of the most important lessons learnt from our study. Our regression analysis confirmed the validity of the initial cluster definition that was suggested for low prevalence settings [
15] and quantified the impact of increasing either the total number of cases or duration of a cluster (or both) on the sensitivity of its definition (Figure
4). The choice of cluster definition is heavily influenced by the local prevalence of infection and public health capacity to respond to surveillance alerts [
29] so the definition should be optimised for the local prevalence and distribution of STM MLVA patterns.
Our findings add further insights into the epidemiology of sporadic STM infection and offer a new way to document baseline STM disease burden. The relative stability of the ratio of novel MLVA types to all MLVA types observed at any given time suggests constant seeding from new or established STM infections and environmental sources. The monitoring of this ratio together with established measures, such as McIntosh's dominance index of diversity, could provide a useful benchmark for STM surveillance. Unfortunately, we did not observe any measurable impact of prospective MLVA typing on the trends of STM infections, perhaps due to the relatively short duration of the study and to the limited number of clusters that were investigated by public health units. A significant proportion of STM MLVA clusters had no obvious temporal and spatial clustering and was associated with endemic MLVA types. This is not surprising, as
Salmonella infections have been linked to a range of food vehicles including eggs, chicken, beef, pork, dairy products and salad vegetables [
30,
31] with an attack rate of 7-10% [
32], making public health control of human salmonellosis particularly challenging. The continuous introduction of new STM variants and re-introduction of endemic variants from different food outlets and farms offered an additional insight into the inherent complexity of controlling the STM endemicity and the importance of addressing the whole food chain [
31]. As in previous reports [
31], our STM MLVA typing identified identical patterns in isolates from the same source and from different sources at different points in time, demonstrated co-circulation and persistence of endemic STM types and occasionally found similar MLVA profiles in epidemiologically unlinked cases.
To our knowledge, this is the first report of the association between STM MLVA-guided public health interventions and the diversity of STM infections in a community. We did not detect any significant impact of MLVA-based surveillance on the size, number or the duration of human STM infections. The apparent delay in the re-occurrence of MLVA types linked to public health interventions is of particular interest. However, these observations and their possible cause-effect relationships require further confirmation in a larger longitudinal study with more detailed epidemiological data collection to enable the development of quantitative rules to exclude random effects and to differentiate primary clusters from secondary re-occurrences in low-endemicity settings.
Several potential limitations of this study should be acknowledged. First, the study was limited to one state of a large country with a disproportionate concentration of STM cases in one metropolis (i.e. Sydney) and lacked sufficient numbers of isolates with diverse MLVA types required to perform statistical analyses of causal associations. A study with statistical power for these analyses would require a decade of data collection, as multi-centre assessment may not be achievable due to potential disparities in local STM epidemiology and public health capacity. In addition, a limited number of endemic MLVA patterns constituted almost a third of all STM isolates. However, the epidemiology of STM infection and public health responses in New South Wales are similar to those in other regions of Australia and Western Europe [
33,
34], supporting the generalisability of our findings. Second, we relied on a single method of typing due to the logistics of our rapid prospective public health surveillance and restricted our cluster definition to STM isolates with identical MLVA patterns and excluded isolates differing by one repeat in one of five loci [
20]. Occasionally, such closely related isolates might represent epidemiologically linked cases of infection and alternative typing methods might be useful to clarify the relationships between those strains. However, it is unlikely that our more restrictive case definition would significantly affect the clustering rates or our overall conclusions. Third, we assumed the stability of the MLVA loci over the period of the study. While existing evidence supports this assumption [
35], the dynamics of MLVA loci over an extended period of time deserve further evaluation. Furthermore, minor variations in laboratory protocols may complicate the inter-laboratory comparison of MLVA results, therefore efforts to harmonise STM MLVA testing and to develop a unified nomenclature have received well-deserved attention [
25,
17]. Lastly, the trends observed in the study apply to a scenario of relatively low incidence of food-borne STM infections with access to laboratory testing and mandatory STM notification. They may not be readily generalizable to high STM endemicity settings with larger population densities and different diagnostic or public health practices.
Acknowledgements
This study was funded, in part, by the NSW Department of Health by its Capacity Building Grant. VS was funded by National Health & Medical Research (Grant 457122). Support and advice from public health epidemiologists from the Health Protection Branch, NSW Ministry of Health and OzFoodNet is gratefully acknowledged. The authors also thank colleagues from Microbiological Diagnostic Unit, University of Melbourne and the Division of Analytical Laboratories, ICPMR for providing phage typing results and initial testing of environmental isolates, respectively, and Dr Blanca Gallego of the Centre for Health Informatics, University of New South Wales for providing spatial cluster visualisation software.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
VS conceived the study, undertook analysis and drafted the manuscript. QW and GLG initiated the study and made major contribution to the study design. PH, CWYH, JM and KK worked on the collection of data used in the manuscript and drafted the manuscript All authors contributed to the writing of the manuscript and read and approved the final manuscript.