Ascertaining the burden of invasive Salmonella disease in hospitalised febrile children aged under four years in Blantyre, Malawi

Typhoid fever is endemic across sub-Saharan Africa. However, estimates of the burden of typhoid are undermined by insufficient blood volumes and lack of sensitivity of blood culture. Here, we aimed to address this limitation by exploiting pre-enrichment culture followed by PCR, alongside routine blood culture to improve typhoid case detection. We carried out a prospective diagnostic cohort study and enrolled children (aged 0–4 years) with non-specific febrile disease admitted to a tertiary hospital in Blantyre, Malawi from August 2014 to July 2016. Blood was collected for culture (BC) and real-time PCR after a pre-enrichment culture in tryptone soy broth and ox-bile. DNA was subjected to PCR for invA (Pan-Salmonella), staG (S. Typhi), and fliC (S. Typhimurium) genes. A positive PCR was defined as invA plus either staG or fliC (CT<29). IgM and IgG ELISA against four S. Typhi antigens was also performed. In total, 643 children (median age 1.3 years) with nonspecific febrile disease were enrolled; 31 (4.8%) were BC positive for Salmonella (n = 13 S. Typhi, n = 16 S. Typhimurium, and n = 2 S. Enteritidis). Pre-enrichment culture of blood followed by PCR identified a further 8 S. Typhi and 15 S. Typhimurium positive children. IgM and IgG titres to the S. Typhi antigen STY1498 (haemolysin) were significantly higher in children that were PCR positive but blood culture negative compared to febrile children with all other non-typhoid illnesses. The addition of pre-enrichment culture and PCR increased the case ascertainment of invasive Salmonella disease in children by 62–94%. These data support recent burden estimates that highlight the insensitivity of blood cultures and support the targeting of pre-school children for typhoid vaccine prevention in Africa. Blood culture with real-time PCR following pre-enrichment should be used to further refine estimates of vaccine effectiveness in typhoid vaccine trials.


Introduction
Both Salmonella Typhi and nontyphoidal Salmonellae remain prominent contributors to the large burden of bloodstream infection (BSI) in sub-Saharan Africa (sSA) [1][2][3][4]. Until recently nontyphoidal serovars Salmonella Typhimurium and Salmonella Enteritidis were the most prevalent in sSA, mainly affecting young children and HIV-infected adults [5][6][7]. Outbreaks of typhoid fever are now being reported across sSA [1,[8][9][10][11], largely caused by a multidrug resistant S. Typhi genotype 4.3.1 [12][13][14]. In Malawi where surveillance for bloodstream infections has been conducted for over 20 years, S. Typhi has become one of the commonest blood culture isolates amongst hospitalized febrile adults and children [1,15]. In this context, ineffective, commonly available antimicrobials and inadequate diagnostic tools with poor sensitivity and results turn-around time, hamper the identification, management and control of both iNTS and typhoid fever.
Blood culture identifies 40 to 80% of cases of invasive Salmonella disease [16]. Limitations of culture arise due to low numbers of bacteria in blood, requiring large volume samples that are not feasible from young children [17]. Prior antimicrobial use may further compromise the diagnostic yield. Currently available, commercial serological tests detecting antibody against S. Typhi antigens have limited sensitivity and poor specificity [18][19][20]. A number of PCR methods for diagnosis of Salmonella have been developed. The inclusion of a culture step prior to nucleic acid amplification has been suggested to increase sensitivity [21,22], but this has not been evaluated in the field.
Studies from countries in Asia with endemic typhoid suggest a considerable disease burden in young children, including children aged under 4 years [3,[23][24][25]. If this is also the case in sSA, this could influence the age at which vaccination might be implemented within the National vaccination expanded programme on immunisation [26]. In this study, we adopted real-time PCR with a pre-enrichment culture step alongside standard blood culture to improve case ascertainment of invasive Salmonella disease in children aged under 4 years in Malawi.

Ethics statement
The College of Medicine Research Ethics Committee approved the study, approval number P.08/13/1445. Written informed consent was sought from the parents or guardians before enrolment of participants into the study.

Participant enrolment
Recruitment of participants into the study was nested within the pre-existing sentinel surveillance for bacterial infections at Queen Elizabeth Central Hospital (QECH) and the Malawi-Liverpool-Wellcome Trust Clinical Research Programme, Blantyre, Malawi. Children with non-specific febrile illness (temperature �38˚C or a history of fever) aged between 0 and 4 years old presenting to QECH were recruited. QECH is the largest government hospital in Malawi, providing free healthcare to the district of Blantyre with a population of approximately 1.3 million and referrals from the Southern region [27]. Recruitment was from August 2014 to July 2016 covering three dry seasons and two rainy seasons. 5 mL blood was drawn from each child; 2 mL for culture, 2 mL for real-time PCR and 1 mL for serology. At six weeks in convalescence further 1 mL of blood was drawn and plasma was extracted for serology. Unless invasive bacterial infection was strongly suspected, children with a positive RDT or malaria blood film were not enrolled into the study.

Blood culture processing
Blood cultures were processed using the BacT/Alert automated system (Biomerieux, France), at the Malawi-Liverpool-Wellcome Trust Clinical Research Laboratories [1]. Bacteria were isolated and identified using previously described standard microbiological procedures [1]. A minimum of four days, including sub-culturing, biochemical testing and latex agglutination testing, was necessary to generate a definitive report of the diagnosis.

Pre-enrichment
A modified pre-enrichment procedure for Salmonella was adopted [28]. A maximum of 2 mL of blood was added to 8 mL of tryptone soy broth mixed with 3% Ox-gall powder (TSB/Oxgall) and incubated overnight at 37˚C. Blood samples with <2 mL were added to 8 mL TSB/ Ox-gall medium, topped up to 10 mL with sterile distilled water.

DNA extraction and real-time PCR
Pre-enriched blood samples were centrifugated in a macrocentrifuge at 3000 x g and the supernatant pipetted out leaving approximately 200 μl to resuspend the pellets. DNA from blood, including bacterial genomic DNA, was extracted following the UltraClean BloodSpin Kit (MO BIO Laboratories CA USA) extraction protocol. DNA was eluted in a final volume of 100 μl and stored at -20˚C. A 40-cycle real-time PCR was performed using a mastermix of Platinum Quantitative UDG reagents on an ABI 7500. Three primer pairs (Table 1) designed using Primer Quest, were added in monoplex reactions (S1 Table) to detect the presence of either S. Typhi or S. Typhimurium. A pan-Salmonella primer pair targeting invA was designed using sequences from S. Typhimurium ST313 D23580 (Accession FN424405) and S. Typhi CT18 (Accession AL513382). A primer pair specifically identifying S. Typhimurium was designed against the flagellin gene fliC and S. Typhi was detected using a primer pair targeting the fimbrial gene staG.

Establishing a cycle threshold cut-off
A positive PCR result was assigned to a sample only if the pan-Salmonella (invA) amplification was simultaneously positive with either one of the serovar-specific amplifications (fliC or staG) within a specified cycle threshold (CT) cut-off. The CT-value cut-off was determined using a combination of three approaches. Firstly, the primer pairs were tested against DNA extracted from serially diluted cultures of S. Typhi, S. Typhimurium, Staphylococcus aureus, E. coli, Klebsiella pneumoniae, Micrococcus spp., and Bacillus spp (S2 Table). Five replicates were prepared for each organism. A cut-off was set to exclude all CT-values observed for non-specific target amplifications. Secondly, the limit of target detection was established for the three primer pairs by running the real-time PCR assay on five replicates of serially diluted cultures of S. Typhi and S. Typhimurium. Thirdly, a receiver operating characteristic (ROC) curve was constructed to assess the cut-off value that maximises sensitivity and specificity of the PCR amplification, considering the blood culture result as a true positive.

S. Typhi serology
Serology for S. Typhi (Table 2), with previously identified serodiagnostic antigens [29,30] and S. Typhi Vi polysaccharide antigen [31], was conducted to validate the PCR amplification data in blood culture negative samples. Sera from febrile children acutely ill and plasma at six weeks in convalescence were analysed. Archived serum samples from healthy Malawian children aged 0 to 4 years (median age 10 months) were included in the analysis as controls [32]. IgM and IgG titres in acute and convalescent plasma samples and in healthy control sera were measured using a previously described ELISA with 3 purified protein antigens STY1498 (Haemolysin), STY1479 (a possible ATP binding protein), STY1886 (cytolethal distending toxin subunit B homolog) and S. Typhi Vi polysaccharide antigen.

Statistical analysis
Fisher's exact test and Chi-squared test were used to determine whether clinical characteristics (fever, vomiting and diarrhoea), hospital admission and reported prior use of antibiotics were related to detection of Salmonella by both culture and PCR. Statistical significance was defined as P < 0.05, although exact p-values were reported. Pearson's correlation coefficient was used to analyse the association between blood sample volume and cycle threshold value on PCR. Antibody (IgG and IgM) titre differences between S. Typhi positive and negative groups were logtransformed and tested using two-tailed two-sample t-tests and we also looked for differential clustering between these 2 groups using multidimensional scaling and principal component analysis.

Study population
A total of 661 children aged less than 4 years (median age 15.9 months; range 1-48 months; 357 males), with non-specific febrile disease were recruited at QECH over the study period. Eighteen children were excluded from the analysis; 5 had no blood culture results, 4 had no PCR sample, and 9 had inconclusive PCR results on re-testing (Fig 1). The baseline characteristics of the 643 children in these analyses are presented in Table 3.

Real-time PCR and CT-value cut-off
Real-time PCR amplifications were generated in 323 pre-enriched blood samples for the pan-Salmonella primers, in 234 for the S. Typhimurium specific primers, and for 168 of the S. Typhi specific primers, and the remaining 320 samples were negative.  (Fig 2). A cut-off was set at CT = 29 because non-specific PCR amplifications generated in validation experiments clustered above a CT value of 29 (Fig 3). An ROC curve showed that a CT value of 29 cut-off maximised sensitivity and specificity of the respective PCR primer pairs to identify S.

Clinical characteristics, antimicrobial use and Salmonella detection
Non-specific clinical presentations including fever, vomiting and diarrhoea were common and there was no association between clinical signs or symptoms and detection of either S.

Real-time PCR and sample volume
Blood sample volumes were recorded in 619 of the 643 recruited children. The targeted volume of blood (2 mL) for real-time PCR was achieved in 100 /619 (16.2%) children. The minimum volume of blood sample collected for PCR after venesection for blood culture was 100μl (median = 1,200μl, IQR 1,000μl-1,725μl). There was no correlation between the CT value and      Ascertaining the burden of invasive Salmonella disease culture alone) (n = 10) against STY1498 were significantly elevated in comparison to responses in febrile children with all other non-typhoid illnesses (n = 428) (IgM, P = 0.0172; IgG, P = 0.0001) (Figs 4 and 5). The concentration of IgG against STY1498 in convalescent plasma from children who previously had a negative blood culture but were PCR amplification positive for S. Typhi were significantly higher than in children with non-typhoid illnesses (P< 0.0001) (Fig 4). There was no significant difference in IgM or IgG titres (P> = 0.05) against STY1479, STY1886 or Vi between children with typhoid infection and those with other illnesses (Figs 4 and 5). Further, neither multidimensional scaling nor principal component analysis, when run on both IgM and IgG titres for all four antigens, was able to clearly separate Salmonella positive from negative cases on blood culture and / or PCR (S4 Fig).

Discussion
The precise epidemiology of invasive Salmonella infections remains elusive in many resourcelimited settings [33,34]. We have utilised a combination of blood culture and real-time PCR with pre-enrichment to improve ascertainment of S. Typhi and S. Typhimurium bacteraemia in young children by 62% and 94% respectively. In sub-Saharan Africa, including Malawi, invasive nontyphoidal Salmonellae (iNTS) has been widely reported in children less than five years old [5,6], leading to preventive strategies being targeted towards this age group [35]. In the context of multiple reports of declining incidence of iNTS in the region [1,36], our findings suggest that the residual burden of this high mortality bloodstream infection may be greater than supposed. In contrast to iNTS, the conventional understanding has been that typhoid fever is a disease of older children and adults [25,37,38]. Our data suggests a considerable hidden burden in children 0-4 years who present with non-specific clinical features. We speculate that an even greater number of cases are presenting to community health centres and receiving only partially effective antibiotics.
The performance of both real-time PCR with culture pre-enrichment and blood culture is likely to have been limited by small volumes of blood available from young children and prior antibiotic use [39]. Nonetheless, real-time PCR with pre-enrichment did identify additional cases to blood culture with few false negatives. Reported prior use of antibiotics, which is expected to impact negatively both real-time PCR with culture pre-enrichment and blood culture, was found not to have affected detection of invasive salmonella infection. However, the unreliable nature of reported use of antibiotics [40,41] may mean that our findings may still be an underestimate of invasive Salmonellosis in young children.
To validate case ascertainment by pre-enrichment PCR in the context of a negative blood culture, we have used antibody-based diagnostics [42,43], which are not dependent on bacterial concentration in the blood but can vary between populations [44]. In this study IgM and IgG antibody responses to S. Typhi antigens Vi and CdtB (STY1886) in Malawian children were not serodiagnostic as previously reported in Vietnamese [30] and Bangladeshi [31] populations. Only IgG and IgM responses to STY1498 (haemolysin gene, hlyE) separated S. Typhi cases confirmed by blood culture and PCR with culture pre-enrichment, from those with  convalescence (B, D, F, H) Ascertaining the burden of invasive Salmonella disease illness due to other infections. The serodiagnostic capacity of STY1498 in Malawian children was evident in sera from active infection, as described for Vietnamese population [30], and also in convalescent plasma screening. STY1498 serology also validated the blood culture negative/ PCR negative results.
In conclusion, the combination of real-time PCR with culture pre-enrichment with blood culture improved case ascertainment among children aged between 0 and 4 years. These data highlight the hidden burden of invasive Salmonellosis in young children and support the targeting of pre-school children for typhoid vaccine prevention. The recent roll-out of a typhoid conjugate vaccine trial in Malawi and subsequent implementation will likely avert a greater burden of disease than previously reported. However, impact assessment of the vaccine may be affected by poor sensitivity of blood culture which is the primary endpoint for the vaccine trial. Future vaccine trial designs should consider using a combination of blood culture and real-time PCR with culture pre-enrichment as part of the evaluation of the full impact of the intervention.  convalescence (B, D, F, H) Ascertaining the burden of invasive Salmonella disease is IgG and IgM responses where both blood culture and PCR were negative for Salmonella. Ctrl are healthy controls. (TIF) S1 Table. Concentrations of primers and probes for the PCR reaction. (PDF) S2 Table. Bacterial colony forming units per mL. Concentrations of bacteria serially diluted and DNA extracted to assess PCR non-specific amplification. (PDF) S3 Table. Area under the curve for the receiver operating curve. The pan-primer and S. Typhi primer in (a) and the pan-primer and S. Typhimurium primer in (b). (PDF)