Antibiogram patterns and chromosomal DNA typing were used to compare 151 non-typhoidal Salmonella spp. (NTS) isolated from patients and 78 from animals, environmental or food specimens obtained within or near the homes of patients with invasive salmonellosis. The majority of NTS from humans (137; 90.7%) were Salmonella enterica serotype Typhimurium (S. Typhimurium) and S. Enteritidis. Chicken specimens and feeds produced (24; 52.2%) S. Enteritidis, while S. Agona was the predominant (20; 77%) serovar among pigs and dairy cows. The majority (97; 64.2%) of NTS from humans were multidrug resistant, while NTS from cows, pigs, beef carcass swabs and sewers were fully susceptible to all antibiotics tested. Pulsed-field gel electrophoresis patterns of Xba I-digested genomic DNA of NTS from the humans and the chickens were different. However, S. Enteritidis from chickens, and S. Braenderup and S. Agona from cows and pigs were clustered together in one group. There was no significant relatedness between NTS isolates from humans and those from animals, food or the environment in close contact to humans.
Non-typhoidal Salmonella spp. (NTS) are an important cause of infection in both humans and animals. Large outbreaks of infection have been associated with food-borne transmission including that from contaminated poultry and poultry products, meat and milk and other dairy products [1–3]. Although NTS typically cause gastroenteritis they are becoming increasingly important bacterial pathogens in developing countries causing bacteraemia and other invasive disease , and there is an increasing prevalence of multidrug resistance [5,6]. In Kenya, among immunocompromised individuals and the very young, NTS frequently causes bacteraemic infections. Multidrug resistance, particularly to the commonly available antibiotics, poses a major health concern, as alternative therapeutic choices are either unavailable or too expensive to be affordable for most patients.
Several studies have documented that farm animals are the major reservoir for NTS in industrialised countries. For instance in the USA it is estimated that over 95% of NTS infections are related to food-borne transmission . In most industrialised countries and some less industrialised countries Salmonella enterica serotype Enteritidis (S. Enteritidis) is transmitted through consumption of foods containing raw or incompletely cooked eggs and home-cooked products containing eggs [1,8–10]. Certain S. Enteritidis clones are stable over long periods of time causing several outbreaks in different geographical areas . The multidrug resistant (MDR) S. Typhimurium phage type (DT) 104 strain, which has been responsible for epidemics particularly in Europe, the USA and Canada, has reservoirs in cattle and is transmitted mainly through consumption of contaminated meat, milk and milk products [3,12–14]. However, there are no data from developing countries including Kenya on the likely sources of NTS that cause human infections. In the present study we used antimicrobial susceptibility testing, and plasmid and genomic DNA typing to investigate if NTS from animals and environmental sources were related to NTS isolated from humans living in close contact.
Materials and methods
These were acute adult admissions to the medical wards at the Kenyatta National Hospital and the Aga Khan Hospital, Nairobi from 1998 to 2000. Blood cultures and stools were obtained from all febrile patients prior to antibiotic treatment. Stools were obtained from cases of persistent diarrhoea.
Specimens from animals and environmental specimens
Cases from whom NTS were isolated were followed to their home areas where more epidemiological data were obtained and the following specimens were obtained: faeces and intestinal contents from rodents trapped around the homes, water from nearby rivers and streams, and raw and cooked food from homes of patients and vendors in markets serving the areas. In addition, rectal swabs from farm animals including pigs, chickens, cows, and goats were taken from around the homes of index cases or from neighbouring homes. Five abattoirs for beef and one for camel meat were sampled by means of carcass swabs, and effluent was sampled from various points inside and outside the abattoirs. For epidemiological comparison, faecal and liver specimens were obtained from sick chickens identified from two large commercial chicken farms 20 km from Nairobi. Stools were also obtained from a total of 16 workers from the abattoirs (6), pig and chicken farms (10).
Blood cultures and stool specimens were processed using standard techniques. Briefly, blood cultures were incubated in 5% CO2 at 37°C for 18 h and if signs of bacterial growth were observed (air bubbles, turbidity or both) they were subcultured on sheep blood agar and chocolate agar. The remaining blood cultures were reincubated for a further 7 days or until positive. Stools were processed by direct plating onto selective media (XLD and brilliant green agar) (Oxoid, Basingstoke, UK) and by overnight enrichment in selective Selenite F broth (Oxoid) followed by plating onto XLD and brilliant green agar (Oxoid), and incubated in air at 37°C for 18 h. NTS were identified using agglutinating antisera (Murex Biotech, Dartford, UK) and their identification was confirmed biochemically using API 20E strips (API System, Montalieu Vercieu, France).
Environmental samples were initially cultured in Rappaport–Vassiliadis soya broth (Oxoid) for enrichment. The broth culture was then subcultured onto XLD and brilliant green agars (Oxoid). NTS were identified as for blood and stool cultures. All NTS isolates were stored at −70°C on Protect Beads (Technical Service Consultants, Heywood, UK) until analysed.
Antimicrobial susceptibility testing
Susceptibility tests with commonly used antimicrobial agents were performed on Mueller–Hinton (Oxoid) agar by the disk diffusion technique. The antibiotic disks (Oxoid) used were ampicillin 10 µg, co-amoxiclav 10:20 µg, cefuroxime 30 µg, ceftazidime 30 µg, co-trimoxazole 25 µg, chloramphenicol 30 µg, ciprofloxacin 5 µg, gentamicin 10 µg, nalidixic acid 10 µg, streptomycin 10 µg, and tetracycline 30 µg. E-test strips (AB BioDisk, Solna, Sweden) were used to determine the minimum inhibitory concentrations (MICs) of the same antimicrobial agents according to the manufacturer's instructions. Escherichia coli ATCC 25922 was used as control for bacterial growth and potency of antibiotics on disks and E-test strips. Disk zone sizes and E-test MICs were interpreted according to the National Committee for Clinical Laboratory Standards guidelines .
Pulsed-field gel electrophoresis (PFGE) of macrorestricted chromosomal DNA
Chromosomal DNA was prepared in agarose plugs as described by Kariuki et al.  from an overnight bacterial culture in Luria broth (Oxoid). Agarose plugs were then digested with Xba I (Life Technologies, Paisley, UK) according to the manufacturer's instructions. PFGE of agarose plug inserts was then performed in a CHEF-DR II system (Bio-Rad Laboratories, Hercules, CA, USA) on a horizontal 1% agarose gel for 24 h at 120 V, pulse time of 1–40 s, at 14°C. A λDNA digest consisting of ca. 22 fragments of increasing size from 48 kb to about 1000 kb was included as a DNA size standard. The gel was stained with 0.05% ethidium bromide and photographed on an UV transilluminator (UVP Inc., San Gabriel, CA, USA). The restriction endonuclease digest patterns were interpreted by considering migration distance and intensity of all visible bands, and by using guidelines described by Tenover et al. . Dendrograms of genetic similarity were then constructed using data obtained by the Dice coefficient method and clustered by the unweighted pair group arithmetic averaging method (Molecular Fingerprinting Program version 1.4.1, Bio-Rad, UK) putting the isolates into PFGE types. Isolates within each PFGE type produced indistinguishable PFGE fragment patterns in which case they were likely to be identical strains or they showed one to two band differences in their fragment banding patterns and were considered to be closely related. Isolates from different PFGE were sufficiently different in their fragment banding patterns as to render them different strain types.
Plasmid DNA extraction was performed using a Plasmid Mini Prep Kit (Qiagen, West Sussex, UK) according to the manufacturer's instructions. Plasmids were separated by electrophoresis on horizontal 0.8% agarose gels at 100 V for 2 h. Plasmid sizes were determined by co-electrophoresis with plasmids of known sizes from E. coli strains V517 (NCTC 50193) (53.7, 7.2, 5.6, 3.9, 3.0, 2.7, 2.1 kb) and 39R861 (NCTC 50192) (147, 63, 43.5, 6.9 kb). DNA bands were visualised with an ultraviolet transilluminator (UVP Inc.) after staining with 0.05% ethidium bromide.
Bacteria from humans
A total of 592 specimens from 402 inpatients were processed between January 1998 and April 2000, giving a total of 151 non-duplicate NTS. A total of 58 NTS each came from blood and stools of the same patients; 12 NTS were isolated from blood and 49 NTS were obtained from stools only. These isolates came from sporadic cases that came to seek medical attention at two tertiary hospitals in Nairobi, Kenya. It was therefore difficult to incriminate any particular common food source. The main Salmonella serovars were S. Typhimurium (79; 52.3%) and S. Enteritidis (58; 38.4%). Other serotypes including S. Agona, S. SaintPaul, S. Braenderup and S. Durban in small numbers (1–3) were isolated from blood, stool cerebrospinal fluid, pus or liver biopsies.
Bacteria from animals and environmental specimens
A total of 220 swabs from beef carcasses and 47 abattoir effluent samples taken at various points along the open sewer lines from five abattoirs were processed. From these samples only four NTS were isolated. In addition, faeces were processed from 210 dairy cows, 122 pigs and 228 rodents, and these gave a total of 28 NTS. A further 39 NTS came from chickens and soya feed supplements from two large commercial farms just outside Nairobi while faeces from chickens from small-scale farmers from the study area yielded seven NTS. S. Agona (20; 76.9%) was the main serovar from pigs and cows which came from the same farms, while S. Enteritidis (24; 52.2%) was the main serovar isolated from chicken specimens. Faeces from goats, swabs from camel carcasses, and food and water samples from the homes of index cases did not yield any NTS.
Epidemiological data from homes of patients
All 151 patients that were followed back to their home areas came from within 40 km of the two study hospitals. A total of 122 (80.8%) kept 10–15 chickens and 1–5 other animals (goats, pigs, cows or pigs) in their homesteads to provide eggs, meat and milk, respectively. The main source of meat was the local butcher who got his meat from one of five abattoirs within the city. Occasionally, the chickens and goats were also slaughtered in homesteads for meat. These small-scale farmers also grew vegetables for sale and household use. The other 29 patients rented houses and depended wholly on local markets and butchers for vegetables and meat, respectively. Treated tap water was available to all visited homesteads but on occasional water shortages during the drought period (January–March) nearby rivers or streams became sources of water for over 50% of the homes. A survey on use of antibiotics by farmers showed that tetracycline is widely available for sale ‘over the counter’ and is used extensively in poultry rearing. It is added to commercial poultry feeds and in drinking water for birds of all ages. Tetracycline, penicillin and sulfonamides were also used extensively in dairy animals for prophylaxis, but there was no indication of their use in beef animals.
Drug susceptibility testing
There were two main antibiotic susceptibility patterns exhibited by the NTS from humans in the current study: fully susceptible isolates constituting 34% and MDR (resistant to two or more antibiotics) isolates accounting for 64.2%. For all commonly available antibiotics including ampicillin, co-trimoxazole, streptomycin, chloramphenicol, and tetracycline MIC values were high for the MDR isolates (Table 1). Three of the MDR S. Typhimurium isolates showed resistance to ceftazidime (MIC=16 µg ml−1). Few isolates (2%) were resistant to only one antibiotic, mainly to ampicillin or co-trimoxazole. Although no NTS were resistant to ciprofloxacin, a number of them (4%) showed reduced susceptibility (MIC=0.125 µg ml−1) and 11% were resistant to nalidixic acid.
|Antimicrobial agent||MIC (µg ml−1)||% resistant|
|Antimicrobial agent||MIC (µg ml−1)||% resistant|
All NTS isolated from cows, pigs and sewers were fully susceptible to all 11 antimicrobials tested. In contrast 15 (38.5%) of the NTS from chickens from the two large commercial farms were multiply resistant to ampicillin, co-trimoxazole, tetracycline and streptomycin (all MIC >256 µg ml−1) and co-trimoxazole (MIC >32 µg ml−1). All five S. Enteritidis from small-scale farms were resistant to tetracycline only (MIC=16–32 µg ml−1).
PFGE patterns of NTS isolates
For NTS from humans that gave indistinguishable PFGE banding patterns only representative isolates were selected for the dendrogram analysis. Fragment sizes of 110 kb or less were omitted from analysis as they may represent plasmids. Fig. 1A,B are simplified dendrogram analyses of S. Typhimurium and S. Enteritidis, respectively, showing the relatedness of strains from both humans and animals. From the PFGE typing, S. Typhimurium isolates produced three main patterns, one common to most isolates (54; 68.4% in PFGE type A) and two other patterns (PFGE types C and D) that differed from each other and from the main pattern by four bands. Within PFGE type A, 42 (53.2%) of S. Typhimurium produced PFGE patterns that were indistinguishable.
For S. Enteritidis from humans there were two main PFGE patterns, the commonest one contained 41 (70.7%) isolates in PFGE type A of which 28 (48.3%) produced PFGE patterns that were indistinguishable. A total of 12 (15%) were contained in PFGE type B. For both S. Typhimurium and S. Enteritidis there was limited diversity in PFGE patterns for the strains from one year to another during the study period. PFGE analysis of S. SaintPaul (three isolates), S. Agona (two isolates) and S. Braenderup (four isolates) from the humans indicated that the isolates within each serovar were closely related.
The 20 S. Agona isolates from domestic animals (12 isolates from pigs and eight isolates from cows) produced indistinguishable PFGE banding patterns. However, there were two PFGE patterns in S. Choleraesuis, one for isolates from chickens and the second for isolates from cattle carcass swabs. All nine S. Braenderup isolates from chickens produced indistinguishable PFGE banding patterns. Similarly, both S. Typhimurium and S. Enteritidis isolates from chickens from the commercial farms gave indistinguishable PFGE patterns (Fig. 1), indicating that single common strain types caused infection on both farms.
The PFGE patterns of S. Enteritidis and S. Typhimurium from humans were distinct and different from corresponding serovars isolated from either animals or environmental sources. Only three S. Enteritidis from humans (PFGE type C) showed close relatedness to S. Enteritidis from chickens, while three S. Typhimurium isolates from chickens were only distantly related to isolates from humans. For S. Enteritidis Xba I digest fragments from human and animal isolates differed in four bands at 120–280 kb, while for S. Typhimurium banding patterns differed in four bands at 520–640 kb. However, two S. Agona isolates from pig farm workers produced PFGE patterns that were indistinguishable from the 20 S. Agona isolates from animals. PFGE fragment digests from S. Braenderup isolates from humans differed from those isolated from chickens and chicken soya feed supplements in four bands at 200–280 kb, clearly indicating that these strains were different. Similarly, S. Saintpaul isolates from humans were significantly different from those isolated from cows in four bands at 150–220 kb.
Plasmid DNA analysis
All MDR S. Typhimurium and S. Enteritidis from humans and from chickens had a common 100–110-kb plasmid that was found to encode multiple antibiotic resistance phenotype. In vitro conjugation tests  showed that the 100–110-kb plasmids transferred resistance to ampicillin, co-trimoxazole, streptomycin and tetracycline en bloc to E. coli K12. In addition, S. Typhimurium isolates contained smaller non-transferable plasmids, 3, 5 and 15 kb in size, while S. Enteritidis isolates from chickens contained only the 3-kb additional plasmid. Plasmids could not be isolated from sensitive strains of NTS from humans, animals or environmental sources.
Although in the USA and Europe domestic animals are the major reservoir and foods of animal origin are the major vehicles of NTS infection in humans [1,3,12], our findings indicate that this may not be the case for the NTS we studied. We found that the distribution of NTS serovars among animals from home environments of index cases was important. No significant numbers of NTS were isolated from rodents, carcass swabs or effluent from abattoirs, and the serovars S. Agona and S. Choleraesuis that were isolated from pigs and dairy cows did not appear among humans. Human isolates were predominantly S. Typhimurium and S. Enteritidis. In contrast, serovar S. Enteritidis was the predominant isolate from chicken specimens and feed from two large-scale commercial farms, in addition to other serovars.
Antimicrobial susceptibility testing showed that NTS from humans were multiply resistant to antibiotics commonly available in Kenya whereas the NTS from animals and environmental sources close to the homes of index cases were fully susceptible to all antibiotics tested. A common transferable 100–110-kb plasmid was found to encode the multidrug resistance phenotype among S. Typhimurium and S. Enteritidis from humans and from the chickens from the large-scale commercial farms. As has been previously documented , the 100–110-kb plasmids were common among MDR NTS serovars from different ecological sources in Kenya. However, it appears that this MDR-encoding plasmid has not been transmitted to NTS isolated from cattle, pigs, or chickens from the small-scale farms that we studied despite the fact that index cases from whom MDR NTS were isolated lived within the same homes or homes close to these animals. The MDR NTS from chickens were all from the two large-scale chicken farms that used soya feed supplements imported from a source in Europe. These feed supplements had the same NTS serovars isolated from them, thus giving credence to the assumption that the imported soya feed was the source of the MDR NTS. Indeed feeds have been the main source of MDR NTS infection outbreaks among poultry from other industrialised countries as well . In addition, several studies from industrialised countries have also shown that NTS from farm animals are multiply resistant to commonly available drugs and lately to both quinolones and the extended spectrum β-lactams [12,19–21].
Using dendrogram analysis of Xba I-digested genomic DNA separated by PFGE we observed that there were certain dominant PFGE groups among the NTS from humans: three types for S. Typhimurium and two types for S. Enteritidis, indicating that a few dominant strain types were largely responsible for infection in humans. However, even for NTS serovars such as S. Typhimurium and S. Enteritidis that were common to humans and animals, there were major differences in the PFGE patterns as to render them separate genotypes. These NTS serovars from humans were clearly unrelated to corresponding serovars from animal sources. We also observed that a single clone of S. Agona was in circulation among pig herds, chickens and zero-grazed cows. In addition, S. Braenderup isolates from both chickens and cows produced indistinguishable PFGE patterns, indicating that these strains were likely to be from common sources. Comparing S. Braenderup and S. Saintpaul strains from humans and those from animals and environmental sources, there was no significant relationship between them, again suggesting that the sources of infection for these strains were different. However, S. Agona isolated from two farm workers and S. Agona from animals had indistinguishable PFGE patterns, indicating that these two workers had possibly acquired the infection from handling the animals on the farm. Apart from this case of S. Agona, the majority of NTS, including S. Typhimurium and S. Enteritidis from animals, were clearly different from NTS strains isolated from humans.
In agreement with our findings, other recent studies that were based on a mathematical model for serovar distribution  and PFGE genomic DNA typing  also observed that there was no significant evidence that NTS from pigs and cattle, and raw animal products including meat and poultry were sources of infection for NTS found in humans in parts of the USA, suggesting that animal reservoirs may not always represent a source for human NTS infection outbreaks. In addition, using several genotyping methods including genomic DNA, plasmid profiling and ribotyping, a study of S. Enteritidis from eight poultry farms from across England found that a total of 54 different strain types may be in circulation on the farms. In contrast, using PFGE-RFLP typing, Murphy et al.  found in Ireland that S. Typhimurium DT104 isolates from both humans and animal sources were predominantly belonging to one strain.
From the results of antimicrobial susceptibility testing, plasmid and genomic DNA typing, it appears that NTS from animal and environmental sources are not closely related to NTS isolated from humans living close to these animals, and therefore are unlikely to originate from common sources. We further observed that NTS serovars from animals and environmental sources from homes of index cases were unique antibiotic-sensitive single strain types that remained stable over the study period. Only the two large commercial farms had the same strain types of NTS from chickens that were MDR and these may have originated from imported feed supplements. However, as imports of protein feed supplements for animals, particularly chickens, are becoming more common, surveillance studies need to be implemented so as to detect NTS that may be a source of infection and spread of antimicrobial resistance among the animals from small-scale farmers. In addition, the MDR phenotype in humans needs to be monitored with a view to finding ways of implementing a prudent antibiotic usage policy in order to reduce resistance, particularly to commonly available drugs. Further studies will be required to investigate whether patient-to-patient transmission of NTS plays any role in the epidemiology of NTS infections in Kenya.
We thank the Director of the Kenya Medical Research Institute for permission to publish this work. S.K. was supported by the Wellcome Trust Research Development Award in Tropical Medicine.