Diagnostics for invasive Salmonella infections: current challenges and future directions

Introduction

Invasive Salmonellosis is caused by Salmonella enterica serotype Typhi (S. Typhi) or Paratyphi A and B (S. Paratyphi), the causes of enteric fever, S. Paratyphi C, which causes septicemia and metastatic purulent infections, or invasive non-typhoidal Salmonella (iNTS) serotypes, including S. Enteriditis and S. Typhimurium. Invasive non-typhoidal Salmonellosis has its highest burden among immunocompromised or malnourished individuals, especially children infected with HIV in resource-limited areas, among whom case fatality is high [1]. Similarly, enteric fever remains among the leading causes of disability from an infectious disease in the developing world [2]. Recent estimates of typhoidal Salmonella incidence have varied substantially [3–6], and iNTS estimates are sparse [7–9], in large part due to poor access to reliable diagnostics, particularly in low-resource outpatient settings where patients with these illnesses typically present for medical care. Measured by its burden and influence on antibiotic use, invasive Salmonellosis is perhaps the most important infectious disease cluster for which rapid and reliable (>90% sensitivity and specificity) diagnostics do not exist.

This diagnostic gap leads to under-diagnosis as well as inaccurate, over-diagnosis of enteric fever especially, the latter of which may lead to inappropriate and excessive antibiotic use. This results in selective pressure for the emergence of resistant bacteria, at a time in which highly resistant Gram-negative infections, including Salmonella [10–13], threaten to undermine reductions in case fatality rates for bacterial infections [14]. Additionally, inappropriate targeting of antibiotics for Salmonellosis results in inadequate therapy for other treatable infections, such as leptospirosis, rickettsia, and brucellosis. It also poses a challenge to the targeted rollout and evaluation of more effective, conjugated enteric fever vaccines, which are on the horizon [15,16]. A recent review (2011) of diagnostics for enteric fever provided a detailed summary of the state of existing diagnostics, with an emphasis on serologic assays and nucleic acid amplification-based tests [17]. Here, we briefly review the literature on currently available diagnostic approaches for both enteric fever and iNTS, and then provide an overview of diagnostic strategies under development, desirable test characteristics according to their utilization goal, and the development and implementation challenges for scale-up of new Salmonella diagnostics.

Available diagnostic approaches for enteric fever

Essentially all enteric fever diagnosis begins with evaluation of clinical signs and symptoms. For perhaps the majority of patients with suspected enteric fever worldwide, who live in settings where diagnostic microbiology is unavailable [18], this is also the end of the diagnostic algorithm, and a decision concerning empiric treatment is made at this juncture. Unfortunately, clinical diagnosis of typhoid is not reliable, as it is difficult to distinguish typhoid from other co-endemic acute febrile illnesses including influenza, dengue, leptospirosis, malaria, brucellosis, rickettsial infections, and other systemic infections. Fever and headache occur in the majority of patients, and a myriad of non-specific symptoms include abdominal pain, myalgias, chills, cough, sore throat, anorexia and nausea [19–25]. Diarrhea and constipation are both regularly reported in case series. Hepatomegaly, splenomegaly, and cervical lymphadenopathy are present in a minority of patients. Faget’s sign (relative bradycardia in the presence of fever) occurs in less than half of patients and is not specific for enteric fever. Rose spots—a salmon-colored maculopapular eruption typically on the trunk—are seen in less than 30% of cases in most series [21], and are similarly not pathognomonic [26]. Laboratory abnormalities are also non-specific. Most patients have normal leukocyte counts, though leukopenia is present in a minority. Mild increases in hepatic transaminases, creatine kinase and lactic acid dehydrogenase have been reported but are also common to other infections in the differential diagnosis [19,21]. While S. Paratyphi A has been considered to cause a more mild illness than typhoid, a recent large study from a co-endemic setting found these infections to be clinically indistinguishable [27]. Studies aiming to develop and validate prediction rules for enteric fever from clinical features and laboratory results have had very limited success [28–31].

After clinical diagnosis, the two most common diagnostic procedures for typhoid in use today were developed in the 19^th century: bacterial culture and the Widal test. Since Eberth’s discovery of the etiological agent of enteric fever in 1880 [32], culture has been the gold standard of enteric fever diagnosis. Whereas early culture techniques had low yield, Coleman and Buxton developed an improved approach by using larger quantities of blood (10 ml) and broth, and adding Ox-bile which lyses blood cells and inhibits antibacterial activity [33]. This approach, while effective, prevents isolation of many other important bacteria. Tryptone soy broths are among the most commonly used blood culture media today, with automated systems used in settings with sufficient resources. The majority of positive cultures are evident within 48 hours, and nearly all are positive by five days (Table 1). Subculture, biochemical testing, and agglutination with specific antisera are typically performed to identify Salmonella serovars.

The sensitivity of culture varies substantially according to the specific fluid and volume assayed, age of the affected individual, prior antimicrobial use, and stage of the illness. Bone marrow cultures have the highest sensitivity (>80%) and are relatively unaffected by antibiotics [34,35]; however, this diagnostic procedure is not commonly performed in clinical settings where typhoid is endemic due to its invasive nature and the need for training and specialized, sterile equipment. The sensitivity of blood culture has been variably reported at 40–80%, with higher sensitivity in the first week of illness, when the bacterial concentration in blood is an order of magnitude higher than in subsequent weeks [34–36]. Unlike with bone marrow cultures, antibiotics substantially diminish the yield of blood cultures. Stool cultures and rectal swabs have lower sensitivity (<40%), though they can be enhanced by culturing three specimens or performing multiple cultures from a single stool specimen [37]. Duodenal string sampling and culture may provide a higher yield than stool or rectal swabs [38,39], but with all gastrointestinal site sampling, it should be recognized that positivity may reflect chronic carriage rather than invasive illness. Urine cultures have a low yield and are unlikely to provide an incremental benefit to diagnostic yield over blood culture. When present, rose spots can be cultured and provide some enhancement to diagnostic yield, particularly later in the course of illness [34].

The Widal agglutination test, in which killed Salmonella Typhi and Paratyphi A antigen is reacted with serum to measure agglutinating antibodies to the flagellar (H) and lipopolysaccharide (O) antigens, was developed in the 1890s [40], modified and standardized in the 1950s [41], and today remains in widespread use throughout typhoid-endemic settings. The simplicity and rapidity of the test enables its use in settings with minimal laboratory infrastructure, but misuse and misinterpretation of the results remains a critical problem. A single agglutination test has limited sensitivity and specificity, particularly early in the course of illness and in endemic settings [42,43]; comparison of acute and convalescent titers improves test accuracy but has limited utility in guiding clinical practice [44,45]. A number of ELISAs have been evaluated for typhoid diagnosis by targeting antibodies to the O, H and virulence polysaccharide (Vi) antigens; however, for diagnosis in the acute phase, these tests suffer from the same limitations as the Widal test [44,46,47].

Several serologic tests have been developed for point-of-care diagnosis of enteric fever. The two that have been most widely studied are TUBEX TF (IDL Biotech, Sweden) and Typhidot (Malaysian Biodiagnostic Research, Malaysia). TUBEX TF assays for antibodies to S. Typhi LPS (O9) by quantifying inhibition of binding between O9 monoclonal antibodies and LPS-coupled magnetic particles [48]. Typhidot is a miniaturized dot-blot ELISA that detects IgM and IgG antibodies to a 50 kD S. Typhi outer membrane protein (OMP). Typhidot-M uses the same approach to detect IgM to OMP after removal of total serum IgG, to improve specificity for recent infection [49]. A recent systematic review and meta-analysis of TUBEX TF and Typhidot found sensitivity of 56%–95% and 56%–84%, respectively, with specificity of 72%–95% and 31–97% [50]. Several assays based on antigen detection (O9, Vi, and Hd) in urine or plasma have been evaluated and have had moderate sensitivity and specificity [51–56]; despite adaptation into dot blot and immunochromatographic assays [57,58], none appear to have gained widespread use. While point-of-care serologic tests have the advantage of rapid time-to-result and minimal laboratory infrastructure requirements, their modest accuracy and moderate costs have limited their use in routine clinical settings in high-burden countries.

A number of studies have assessed the yield of nucleic acid amplification tests for enteric fever diagnosis; flagellin genes (fliC-d for S. Typhi, fliC-a for S. Paratyphi A) have been most commonly targeted, but an array of alternative PCR gene targets have been assessed [59–63]. Parry et al. reviewed published literature on PCR assays for enteric fever performed on blood, and found excellent specificity (100%) but variable sensitivity [17]. In general, the sensitivity is very good (most studies >90%) among blood culture-positive patients but lower in blood culture-negative patients (3–13% in the two largest studies) [62,63]. PCR assays targeting fliC have also been applied to urine, showing promising results, though further validation of accuracy is needed [60,64]. Thus, in comparison with culture and conventional identification methods, the primary gain of PCR is rapidity of result rather than improved sensitivity. The downsides of PCR are that it requires fairly sophisticated laboratory infrastructure, which is limited in settings where typhoid is endemic, and current configurations do not provide information on antimicrobial resistance, for which a diversity of genes are responsible.

Available diagnostic approaches for iNTS

As with enteric fever, most treatment for iNTS is based on clinical presentation alone given the dearth of diagnostic infrastructure in settings where iNTS is common. However, iNTS causes a similarly non-specific febrile syndrome. Splenomegaly and hepatomegaly may be suggestive of iNTS in children [65,66], but are also commonly seen with malaria and other infections that are co-endemic in Africa, where most iNTS cases arise. Respiratory symptoms and signs (cough, tachypnea) are not uncommon, and pneumonia is frequently present [67,68]. Diarrhea is present in a minority of iNTS infections among children [65,69]. In a large prospective study in Tanzania, Nadjm and colleagues found that application of the WHO criteria to target antimicrobial therapy failed to identify one-third of invasive bacterial infections, among which non-typhoidal Salmonella were the leading cause [68].

For culture-based and molecular diagnostics, similar challenges present with typhoidal Salmonella—low bacterial burden (1 CFU/ml) in blood—may confound sensitive diagnosis of iNTS [70]. PCR assays have been clinically validated for diagnosis of enterocolitis [71], or for use on cultured isolates [72], but not for direct diagnosis of invasive infections.

In contrast with typhoidal Salmonella, there is more limited experience with serologic assays for iNTS. Despite similarity between the O antigens of S. Typhi and S. Enteritidis, the Widal agglutination test performs poorly [73]. Kuhn et al. recently reviewed ELISAs for non-typhoidal Salmonella infections, which were based on S. Enteritidis and S. Typhimurium LPS, and found most sensitivities in the 70–95% range, with specificities >90%, though most studies were small and many did not report specificity [74]. The higher specificity, in comparison with ELISAs for enteric fever, may reflect their evaluation in low endemic settings. At present, there is a lack of standardization of ELISA assays for non-typhoidal Salmonelloses. A challenge with developing serologic tests for iNTS is the diversity of non-typhoidal serovars globally, which will require empirically informed, and perhaps locally targeted, selection of LPS antigens to achieve adequate sensitivity. Additionally, it will be difficult for serologic diagnostics using current technologies to distinguish between the two syndromes iNTS can cause: self-limiting gastroenteritis and invasive systemic infection.

Novel diagnostic approaches for enteric fever and iNTS

All of the currently available diagnostic approaches for invasive Salmonellosis have limitations in terms of speed, sensitivity, infrastructure requirements, and suitability to point-of-care application. To date, a rapid, accurate, point-of-care diagnostic has remained elusive. However, a number of promising approaches are under development, which may address many of the limitations of the existing diagnostics. Here, we briefly review some of these technologies (Table 2).

Microbiologic culture has excellent specificity; however, drawbacks include lower sensitivity, 24–72 hour incubation need, and—critical for resource-limited settings—requirement for laboratory capacity. To mitigate at least the need for laboratory infrastructure at the point-of-care, novel approaches have been explored, including the development of an electricity-free incubation system and colorimetric readout that permits sensitive culturing of blood in areas with extremely limited resources [75].

Similarly, much of the non-culture based work under development has focused on addressing the limitations of molecular diagnostics in terms of sensitivity, speed, and instrumentation requirements. Zhou and Pollard demonstrated increases to the sensitivity of PCR using a three-hour enrichment step in culture media [76], and separately using selective removal of human DNA from blood using ox-bile for cell lysis and enzymatic digestion [77]. Several isothermal nucleic acid amplification procedures have been developed and commercialized in recent years, eliminating the need for thermocycling instruments [78–80]. Among these, loop-mediated isothermal amplification (LAMP) has received the most attention. LAMP assays involve amplification at 60°C with interpretation through a visual or fluorometric readout, and have the advantage over PCR of less inhibition by biological fluids, which reduces the sample preparation requirements. LAMP assays for Salmonella Typhi have been developed and studied using simulated specimens and a human challenge model [81,82]. However, there remains a need for minimally instrumented, low-cost approaches to DNA extraction and concentration in order for molecular diagnostics to be widely adoptable in resource-limited settings.

Novel amplification-free molecular methods may provide even more rapid diagnosis of Salmonella. Tennant and colleagues have demonstrated detection of 1 CFU/ml of non-typhoidal Salmonella in under 30 seconds using Microwave-Accelerated Metal-Enhanced Fluorescence (“MAMEF”) [83]. Evaluation of this technology using clinical specimens, and toward typhoidal Salmonella, is underway.

Newer generation serologic approaches may yield more accurate immunodiagnosis. One approach involves detection of IgA antibodies produced by active mucosal lymphocytes that migrate to the peripheral blood after infection [84]. This assay, referred to as TPTest (typhoid and paratyphoid fever test), had 100% sensitivity and 78–97% specificity in an endemic setting, and has been adapted into an immunodot format for use in settings without ELISA capability [85]. The procedure does require a centrifuge and incubator, and the chief limitation is the 24–48 hour incubation period, which precludes use as a rapid point-of-care diagnostic. Immunoscreening techniques have led to the identification of a novel protein (YncE), the IgG response to which appears to be specific for asymptomatic carriage, and could lead to the development of improved screening for chronic carriers of S. Typhi [86].

One area of growing interest in diagnostic development has been transcriptional profiling of host mRNA in peripheral blood cells, which has been applied in diagnosis of influenza [87], tuberculosis [88], pertussis [89], and other infectious diseases [90]. Thompson and colleagues identified distinctive gene expression signatures in patients during acute disease, convalescence and recovery from typhoid [91]. While their study did not specifically evaluate these signatures from a diagnostic perspective, this transcriptomic approach could potentially be adapted for diagnostic purposes. Metabolomic approaches are among the newest class of diagnostic approaches in infectious diseases, and utilize a pathogen non-specific screening of metabolite signals in biological samples to identify infectious agents. Exploratory studies have examined specific metabolites to differentiate bacterial infections, including S. aureus and E. coli [92]. Näsström and colleagues used gas chromatography with time-of-flight mass spectrometry on plasma samples and found that a combination of six metabolites could accurately distinguish S. Typhi, S. Paratyphi A and control patients [93]. While the cost, equipment, and analytic requirements of these approaches are currently far too high for their use in routine diagnosis, particularly in resource-limited settings, it is conceivable that insights generated from this work, combined with decreasing costs of analytic tools, may lead to novel classes of tests for enteric fever that convey both diagnostic and prognostic information.

Target product profiles for enteric fever and iNTS diagnostics

There is an increasing recognition that diagnostic development should be guided by specific needs for their utilization—for example, sensitive, rapid screening tests for use in clinics versus comprehensive resistance assays to be implemented in reference laboratories. Such detailed “target product profiles” (TPPs) have been developed for a number of tropical infectious diseases, including malaria [94], tuberculosis [95], and helminthiases [96]. TPPs are typically constructed by convening experts from the scientific community, technical agencies, industry, implementation agencies, clinicians, and national programs to identify and then prioritize specific diagnostic gaps. For enteric fever and iNTS, a set of TPPs is not currently available to our knowledge. While we do not attempt to propose a comprehensive list of needs nor would we prioritize those needs in the absence of input of relevant stakeholders, we here highlight what we perceive as several key focus areas for enteric fever diagnostics.

Perhaps the most obvious acute need is for a rapid diagnostic test with improved accuracy over existing rapid tests. A test with greater sensitivity would give clinicians confidence in excluding Salmonellosis as a diagnosis, and since Salmonella is likely present in a minority of acute febrile illnesses, this would likely have the greatest impact on treatment, as long as specificity remains reasonably good. Such a test could also be used to focus limited resources for blood culture, which is more costly, in settings where Salmonella is the most common recoverable organism in blood cultures. Because most treatment is provided in outpatient settings, including medical shops, the ideal rapid diagnostic would not require a formal laboratory, sophisticated ancillary instrumentation, extensive manual procedures, or advanced laboratory training. The result readout should be easily interpretable by non-laboratory personnel. Containing costs will be essential to maximize use in high-burden settings.

For inpatient settings, where patients with more severe febrile illnesses are likely to receive empiric therapy to cover an array of potential bacterial pathogens, speed and cost may be less important than accuracy and compatibility with drug susceptibility testing. Patients who are hospitalized with suspected typhoid are more likely to have already received empiric oral therapy as outpatients, and assessing for drug resistance is therefore more important in guiding therapy. The current standard against which this diagnostic would be measured is blood culture, which has limited sensitivity, particularly for patients with recent antibiotic exposure, and is not available in many hospitals in high burden settings due to laboratory infrastructure and personnel requirements.

Screening for chronic carriage is not commonly performed in endemic settings, but may be valuable in reducing transmission, particularly in transitional countries that are moving toward elimination of endemic enteric fever [97]. Because this is likely to be pursued in more advanced economies where sanitation has reduced transmission, and the urgency of providing antibiotics in carriage is less, cost and rapidity of diagnostics for carriage may be less critical compared with diagnostics for acute infection.

Finally, there have been few examples of sustained surveillance for Salmonella infections in resource-limited settings, particularly in clinics or medical shops where most patients initially present with fever. As conjugate vaccines become available, reliable surveillance methods will become increasingly critical for guiding vaccine deployment and ongoing assessment of vaccine impact. Diagnostic speed may be sacrificed but accuracy will be critical. Blood culture, which has excellent specificity, is currently filling this niche, but is unavailable in settings without microbiology laboratories, limiting the breadth of surveillance.

Convening experts and relevant stakeholders to refine and prioritize enteric disease diagnostic needs will be critical to focusing limited resources on those tests that will have the appropriate combinations of accuracy, speed, cost, and infrastructure requirements to enable them to be adopted broadly for patient care or disease surveillance.

Challenges for development and rollout of enteric fever and iNTS diagnostics

Among the central challenges to establishing a robust diagnostics pipeline for Salmonellosis is stimulating development and commercialization in the absence of a clear, profitable market for diagnostics. The overwhelming majority of enteric fever and iNTS infections occur among the poor in low- and middle-income countries, whose limited purchasing power is typically directed toward empiric antibiotics. Purely market-based approaches driven by private industry have failed to generate new diagnostics for many tropical infectious diseases. Nonprofit product development partnerships (PDPs), which collaborate with academic, industry, and governmental partners to accelerate diagnostic development and scale up, have emerged as a successful alternative model. For diseases of poverty, the Foundation for Innovative New Diagnostics (FIND) is the largest and most successful PDP, having launched important diagnostic tests for tuberculosis, malaria, and human African trypanosomiasis. FIND has identified non-malaria acute febrile illnesses as an important focus for diagnostic development and has initiated early-phase activities including disease mapping and target product prioritization, but has not reported specific diagnostics under development for enteric fever.

Upon development of candidate diagnostics, there are several challenges in conducting high-quality, clinical validation studies. First is the absence of an accurate, non-invasive reference standard. Blood culture is the most commonly used reference standard in Salmonella studies given its excellent specificity, but as previously noted has limited and variable sensitivity. Many studies report sensitivity among blood culture-positive cases and specificity in healthy controls or individuals with confirmed alternative diagnoses; this approach, while a reasonable compromise, is suboptimal when the goal is to know the overall sensitivity (including culture-negative cases) and unbiased specificity. Paired serologic tests from the acute and convalescent period may improve diagnostic accuracy [44], but it can be challenging to obtain convalescent samples consistently in resource-limited settings. Use of multiple reference diagnostics, along with reliable diagnostics for alternative infections, combined with latent class analysis (which accounts for imperfect reference standards), can enable more accurate estimation of sensitivity and specificity and should be encouraged for studies of novel diagnostics [98].

Once successful next generation diagnostics for enteric fever and iNTS are validated and brought to market, there will be critical questions about how to spur their rollout in high-burden, resource-limited settings. Chief among these is how to incentivize the use of a diagnostic when the alternative—empiric treatment—is low cost and has limited “apparent” harms (e.g. adverse effects, toxicities) at the individual level. For example, point-of-care serologic tests such as Tubex and Typhidot are estimated to cost $2–4 in kit costs alone, not accounting for labor costs and other laboratory consumables [43,99], while a 7-day course of the most commonly used treatments (Ciprofloxacin/Ofloxacin, Cefixime, Azithromycin) frequently retails for $1.00–$3.00 in South Asia. Similar challenges have been found with other disease-specific diagnostics, such as leptospirosis, for which empirical therapy has been found to be more cost-effective than diagnostic-guided treatment across a wide range of assumptions [100].

The patient-level harm of empiric treatment for Salmonella infections that receives less consideration is the failure to consider and treat the alternative diagnosis. In Nepal, for example, more than 500,000 cases of typhoid are reported in the public sector each year, nearly all empiric diagnoses [101]. This contrasts to 6 cases of leptospirosis and 1 case of rickettsial infection for the most recent year available [101]. Surveillance studies, however, suggest that these alternative diagnoses—which require different antibiotic therapy—are combined at least as common as enteric fever [102]. Untreated morbidity and mortality from these other infections may be considerable [103,104]. Additionally, the societal costs of indiscriminate antibiotic use may be substantial, as bacterial resistance rises amid selective pressure of antimicrobials [105].

The fundamental challenge is one of aligning incentives for the use of diagnostics by providers and patients with those of societal goals of improving health by appropriate therapy for infectious diseases and limiting unnecessary antibiotic use to contain drug resistance. An analogous, informative situation has been the challenge of scaling up rapid diagnostic tests (RDTs) for malaria, which is similarly over-treated and for which increasing resistance has also been problematic. Subsidizing RDTs for malaria at private drug shops led to increased uptake, improved targeting of treatment and less unnecessary, empiric therapy [106,107]. Similar strategies are being examined for tuberculosis diagnostics in India [108], though attention will need to be given to ensure that wealth disparities in access to diagnostics do not arise, as may occur with market-based subsidies.

Finally, it is important that the invasive Salmonella diagnostics development and rollout proceed in parallel with 1) enhanced surveillance for Salmonella and other febrile illnesses; and 2) improved diagnostics for other acute febrile illnesses. There has been a dearth of sustained public health surveillance of enteric fever or iNTS, particularly in rural areas [5,7]. If the prevalence of invasive Salmonellosis is low among acute febrile illnesses, even a diagnostic with a specificity of 95% may generate more false positives than true positives. Informing the pre-test probability of disease is critical to appropriate use of and geographic rollout of imperfect diagnostics, particularly given the non-specific clinical presentation of invasive Salmonella infections.

In the near term, a case could be made that a critical component of a best diagnostic for Salmonellosis, in terms of limiting unnecessary treatment, may therefore be the corollary development and use of accurate diagnostics for other common, alternative infections such as malaria, influenza or rickettsia. Diagnostics for invasive Salmonellosis and acute febrile illnesses will need to be complementary, and should be deployed as part of integrated approaches that are informed by local epidemiology, and aimed at reducing inappropriate antibiotic usage and enhanced targeting of therapy.