Laboratory-acquired infections of Salmonella enterica serotype Typhi in South Africa: phenotypic and genotypic analysis of isolates

Workers in clinical microbiology laboratories are exposed to a variety of pathogenic microorganisms [1, 2]. A review of published literature in 2009, reported that bacteria account for the greatest number of reports of laboratory-acquired infections; Shigella species, Brucella species, Salmonella species, Mycobacterium tuberculosis and Neisseria meningitidis were the most commonly reported bacterial causes of laboratory-acquired infections [1].

There have been very few reports of laboratory-acquired infections involving Salmonella enterica serotype Typhi (Salmonella Typhi), however many may go unreported. A PubMed literature search on 1 February 2017, using the key words ‘Salmonella Typhi laboratory acquired infection’ found 12 publications (English language) reporting laboratory-acquired Salmonella Typhi infections [3‐14]. These 12 papers were published over the period 1961 to 1997, so over the last 10 years, to the best of our knowledge, there are no recent publications of laboratory-acquired Salmonella Typhi infections. However, more recently, there have been reports of nontyphoidal Salmonella (NTS) laboratory-acquired infections. In 2016, a Salmonella enterica serotype Typhimurium (Salmonella Typhimurium) laboratory-acquired infection was reported from Canada by Alexander and coworkers [15]; while in 2014 and 2012, the Centers for Disease Control and Prevention (CDC) reported on Salmonella Typhimurium outbreaks in the USA of which the source of the outbreaks were traced to university teaching microbiology laboratories [16, 17] In 2015, Barker and coworkers reported on a Salmonella Enteritidis laboratory-acquired infection [18].

The results for molecular subtyping of isolates are as follows.

For case one, molecular subtyping using PFGE analysis showed that the PFGE pattern of the patient’s isolate was indistinguishable (100% identical) to that of a PFGE pattern shown by a cluster of Salmonella Typhi isolates that the patient (laboratory technologist) had been working on in Laboratory-A (Fig. 1), suggesting that the source of the patient’s infection was within this cluster of isolates that the patient had been working on. This ‘cluster of isolates’ was sourced from a region of South Africa located ~1500 km from where the laboratory technologist lived and worked; also the laboratory technologist had no travel history or any history of contact with persons that lived/travelled to this region of South Africa. Therefore, this could not have been a case of community exposure to Salmonella Typhi. For case one, MLST presented a ST 1 subtype for the patient’s isolate; the isolate was also PCR-positive for a marker associated with haplotype H58, a haplotype of Salmonella Typhi which is being reported with increasing frequency from many countries in Africa and Asia [27, 28].

For case two, molecular subtyping of isolates included PFGE analysis, MLST and SNP profiling. We investigated all Salmonella Typhi isolates that the patient (resident) could have been exposed to in Laboratory-B. Compared to three possible isolates, PFGE analysis showed a PFGE pattern match (Fig. 2) to a single isolate (isolate A) from Laboratory-B. Phylogenetic analysis and SNP profiling showed that isolate A was related to case two; the isolates only differed by 41 SNPs (Fig. 2). Both isolates also presented the MLST ST 2 subtype and both were PCR-negative for the marker associated with haplotype H58. These data suggested that the source of the patient’s infection was isolate A. The patient later confirmed that she had indeed worked with isolate A.

For case three, molecular subtyping of isolates included PFGE analysis, MLST and SNP profiling. We investigated all Salmonella Typhi isolates that the patient (laboratory technologist) could have been exposed to in Laboratory-C. Compared to two possible isolates (isolates B and C) that the patient could have been exposed to in Laboratory-C; PFGE analysis showed a PFGE pattern match (Fig. 3) to one of the isolates (isolate B). This was supported by MLST; case three presented ST 1 which matched the ST 1 of isolate B; isolate C presented ST 2 and so did not match the case. The case isolate and isolate B were both PCR-positive for the marker associated with haplotype H58. Phylogenetic analysis and SNP profiling confirmed that isolate B was highly related to case three with the isolates only differing by 6 SNPs (Fig. 3). These data suggested that the source of the patient’s infection was isolate B.

Case one highlight’s issues concerning the misdiagnosis and misreporting of a Salmonella Typhi isolate as a NTS by a diagnostic laboratory; with associated public health implications, as typhoid fever is a notifiable disease in most countries including South Africa. Salmonella Typhi has a high potential to cause outbreaks of disease [29, 30], therefore rapid diagnosis and reporting to the health authorities is vital to ensure appropriate investigation including follow up of the patient and tracing the patient’s contacts, as well as to design interventions to avert any potential outbreak.

Cases one and three highlights issues concerning the importance of accurate fluoroquinolone susceptibility testing and interpretation of results according to updated guidelines [20]. For case one, the patient was treated initially with inappropriate therapy with ciprofloxacin due to the use of defunct interpretative criteria for fluoroquinolone resistance [31], which was later followed with appropriate therapy with azithromycin. For case three, the discrepancy between the pefloxacin screening result (25 mm or susceptible) and the ciprofloxacin MIC (0.25 μg/ml or intermediately-resistant) for the isolate highlights the potential shortcomings of a single disk susceptibility screening test for fluoroquinolone susceptibility and suggests that confirmatory MIC testing should always be performed in cases of invasive Salmonella infections. While appropriate antimicrobial therapy does not always result in clearance of Salmonella Typhi from the stool, in both cases one and three, treatment with ciprofloxacin to which the isolates were intermediately-resistant failed to clear the pathogen whilst subsequent therapy with azithromycin [32] was successful.

All cases highlight issues concerning Salmonella Typhi vaccination of laboratory workers who are at high risk of exposure to Salmonella Typhi. Neither of the two laboratory workers described in cases two or three had received a Salmonella Typhi vaccination, despite exposures to pure cultures. Case one had received vaccine based on Vi polysaccharide antigen, but no currently available vaccine against typhoid is 100% protective; vaccination decreases risk of infection but does not completely eliminate the risk of infection. Vaccine failures are reported [33, 34]. For case one, vaccination was ineffective; vaccine failure may have occurred secondary to a faulty vaccine. In developed countries such as the USA, Canada, United Kingdom and Denmark; vaccination against Salmonella Typhi is offered to and encouraged for laboratory workers who are at high risk of exposure to Salmonella Typhi, such as those working in reference laboratories. In South Africa, this is certainly the policy and procedure at the CED at the NICD; however not all laboratory networks in South Africa provide occupation-specific vaccination. For reference laboratories in most other African countries, vaccination of laboratory workers against Salmonella Typhi is not routine practice. In 1980, a review of Salmonella Typhi laboratory-acquired infections reported that of 24 cases, the vast majority of patients (19 cases; 79.2%) had not been vaccinated; so lack of vaccination is a definite risk factor [8].

All our cases of laboratory-acquired infection were most likely the result of lapses in good laboratory practice (GLP) and laboratory safety. Case one had six years of working experience in a clinical microbiology laboratory, yet laboratory-infection still occurred. Lapses in laboratory safety do occur, even among laboratory workers with many years of experience working in a clinical microbiology laboratory. A recent report in 2015 described an NTS infection involving a laboratory technician with 20 years of experience in a clinical microbiology laboratory [18]. Over recent years, the decreasing numbers of reports concerning Salmonella laboratory-acquired infections probably reflect the current increased awareness and practice of GLP and laboratory safety. Risky procedures such as mouth-pipetting are now no longer practiced. Today, clinical microbiology laboratories in most countries enforce uniform safety policies nationwide, which would include: safety training; restricted laboratory access; use of biological safety cabinets; personal protective equipment (PPE) (gloves, masks, coats, etc.); hand-washing facilities with automatic-operating taps or elbow operating taps and automatic-operating paper towel dispensers; prohibiting eating, drinking and smoking in the laboratory; prohibiting cell phones in the laboratory; proper decontamination of laboratory benches before and after working with clinical specimens and bacterial cultures; etc. Laboratory accreditation to International Organization for Standardization (ISO) requirements ensures compliance and audit safety policies; diagnostic laboratories in South Africa are all accredited against the ISO 15189 standard [35]. With regards to use of biological safety cabinets, in South Africa (and most African countries), these are not mandatory for working with Salmonella Typhi. However, for some laboratories in Europe and the USA, it is mandatory to work with Salmonella Typhi within a biological safety cabinet (cabinets with a minimum of level-2 safety) or within a biosafety level-2 laboratory.

Molecular subtyping of bacterial isolates is an effective methodology to investigate the relatedness of bacterial isolates and to investigate the source of the infections. Molecular subtyping, including PFGE analysis [3, 15‐18] and WGS [15], has previously been reported to investigate laboratory-acquired infections involving Salmonella species. In particular, Alexander and coworkers [15] used WGS to identify the source of a laboratory-acquired infection involving a laboratory technologist; interestingly, the technologist was not responsible for any testing conducted on the causative strain, so it was concluded that the infection was probably acquired while cleaning laboratory benches or discarding biohazardous waste.

For our current cases, all isolates were investigated with PFGE analysis. Although PFGE analysis is an older molecular subtyping methodology, it still provides valuable information and still remains the primary bacterial subtyping methodology employed by PulseNet International, a molecular subtyping network for foodborne and waterborne disease surveillance. PFGE analysis is limited in that it may not always provide sufficient resolution; for example, if you have a case isolate and you query a database of isolates to find a closest match by genetic relatedness, you may find that the case isolate will equally match (by PFGE pattern) a number of isolates. This was the situation for our case one, where the PFGE pattern for the case isolate was found to match a PFGE pattern represented by a cluster of isolates (Fig. 1), suggesting that the source of the laboratory worker’s infection was within this cluster of isolates that the laboratory worker had handled. To pinpoint the best possible epidemiological match, PFGE data analyzed together with epidemiological data (dates, geographic locations, patient contact information, etc.) is helpful. This was the situation for cases two and three, whereby PFGE data interpreted together with epidemiological data allowed us to pinpoint the most likely source of the laboratory worker’s infection; source isolate A for case two (Fig. 2) and source isolate B for case three (Fig. 3).

For any finer degree of resolution, alternative molecular subtyping methodologies must be employed. The ultimate methodology is WGS data analysis; this WGS methodology was used to investigate isolates for cases two and three, thereby providing a secondary analysis to the primary PFGE analysis. Our analysis of WGS data included a phylogenetic analysis based upon a comparison of SNP profiles. When the number of SNP differences is compared among isolates, the lowest number of SNP differences implies the closest genetic relationship. This allowed us to determine the exact genetic similarity between a case isolate and its probable source. For case two, the case isolate and the probable source isolate (isolate A) differed by only 41 SNPs and were located within close proximity on a phylogenetic tree (Fig. 2). For case three, the case isolate and the probable source isolate (isolate B) differed by only 6 SNPs and were located within close proximity on a phylogenetic tree (Fig. 3).

Lastly, a third molecular subtyping methodology, MLST, was employed to investigate our case isolates. MLST is generally more suitable for a long-term epidemiological comparison of isolates or a global comparison of isolates. However, in some cases of short-term epidemiological investigations (outbreak investigations), it can sometimes provide valuable supporting data. This was the situation for case three, whereby the laboratory worker had acquired his infection by exposure to two possible isolates; isolate B with MLST ST 1 or isolate C with MLST ST 2. The case three isolate presented MLST ST 1, providing evidence supporting isolate B as the cause of the infection.