Havelaar Microbiological Laboratory for Health Protection, National Institute of Public Health and the Environment, PO Box 1, 3720 BA Bilthoven, the Netherlands.
Aims: To develop an animal model to study dose–response relationships of enteropathogenic bacteria.
Methods and Results: Adult, male Wistar Unilever rats were exposed orally to different doses of Salmonella enterica serovar Enteritidis after overnight starvation and neutralization of gastric acid by sodium bicarbonate. The spleen was the most sensitive and reproducible organ for detection of dose-dependent systemic infection. Illness was only observed in animals exposed to doses of 108 cfu or more. At lower doses, histopathological changes in the gastro-intestinal tract were observed, but these were not accompanied by illness. Marked changes in numbers and types of white blood cells, as well as delayed-type hyperresponsiveness, indicated a strong, dose-dependent cellular immune response to Salm. Enteritidis.
Conclusions: The rat model is a sensitive and reproducible tool for studying the effects of oral exposure to Salm. Enteritidis over a wide dose range.
Significance and Impact of the Study: The rat model allows controlled quantification of different factors related to the host, pathogen and food matrix on initial stages of infection by food-borne bacterial pathogens.
Microbiological risk assessment is an emerging tool for the evaluation of the microbiological safety of food and water, and is increasingly used by regulatory agencies, both national and supranational. In this process, hazardous micro-organisms are identified, and exposure of the consumer to these organisms is estimated by a combination of observational data and mathematical modelling. The health risk arising from exposure is then estimated by use of a dose–response model that gives a quantitative description of the relationship between the exposure to a certain number of pathogens (the dose) and the probability of an effect, such as infection or disease (Anon. 1999).
Dose–response models can be based on observational or experimental data. Observational data (usually from food- or water-borne outbreaks) have the advantage that they are based on actual situations, but they are extremely limited in availability for several reasons. The dose may be too low to be measured accurately, such as in drinking water, the contamination may be rare, such as with foodstuffs, or samples of the causal product may simply not be available. Furthermore, the size of the exposed population is often not known. Experimental data have the advantage that they are obtained under well controlled conditions and can therefore be subjected to detailed mathematical analysis (Teunis et al. 1996). Typically, healthy volunteers are exposed to different doses of enteric pathogens that have been maintained and cultured under laboratory conditions. The volunteers are then closely monitored for signs of infection (excretion, seroconversion) and symptoms of illness. The limitations of the experimental approach are determined by the experimental set-up; for ethical reasons the exposed population is limited to healthy, usually adult, volunteers and the range of micro-organisms that can be tested is limited to those that are known to induce mild, self-limiting disease only. Furthermore, the laboratory-adapted cultures may not be representative of the micro-organisms as they occur in nature.
To overcome some of the limitations of the experimental approach in man, animal models are used to assess the dose–response relationship for infection, and possibly disease, by enteric pathogens. Such models should enable an evaluation of the effect of single factors related to the host, pathogen and food matrix, and should allow inferences to be derived about dose–response relations in humans. Typical questions to be addressed are the effects of factors such as age, immunological status and non-specific barriers (e.g. gastric acid, innate immunity) on the susceptibility of the host, and the effects of factors such as bacterial adaptation, stress proteins, protection by fatty foods etc. on the infectivity of the pathogen. As a next step, it is planned to develop kinetic models of the infection process that describe the dynamics of the host–pathogen interaction in the alimentary tract, for which the animal models should provide insight into important mechanisms and parameter estimates (Takumi et al. 2000).
Experimental salmonellosis in rodents has been extensively studied, and detailed information is available on the host–pathogen interaction from in vivo as well as in vitro experiments. Whereas in humans, non-typhoid salmonellae induce self-limiting gastroenteritis with only occasional bacteremia, systemic disease develops in rats and mice. In mice, exposure to oral doses higher than 106 colony-forming units (cfu) is usually lethal, whereas rats develop clinical disease only at doses higher than 108 cfu, and rarely succumb to the infection (Sato 1965; Kampelmacher et al. 1969). The differences in clinical response imply that the rodent is particularly useful for investigating the initial stages of the pathological cascade, i.e. invasion of the epithelium and innate immune response.
After oral exposure of rats to salmonellae, the intestine is rapidly colonized, and salmonellae can be detected in the small intestine and caecum within 2 h. Intestinal colonization is concentrated in the distal ileum and caecum, and may be detected by faecal excretion. Invasion is quickly established via the M-cells in Peyers’ patches, causing salmonellae to be detectable in mesenteric lymph nodes within 8 h (Naughton et al. 1996). Systemic infection results in high numbers of pathogens being present in spleen and liver, and is associated with an increase in organ weights. A major line of defence against invasion of salmonellae is uptake by macrophages. Later, these macrophages play a crucial role in the induction of T-cell-dependent immunity that can be detected by the development of a delayed-type hypersensitivity reaction (Collins and Mackaness 1968). Virulent salmonellae have adapted to survive and grow inside macrophages, and may induce cell death by apoptosis (Jones and Falkow 1996). Polymorphonuclear leucocytes (mainly neutrophils) play a key role in the clearance of salmonella infections (Vassiloyanakopoulos et al. 1998). The number of circulating neutrophils increases considerably after exposure (Hatchigian et al. 1989; Hougen et al. 1989; Radoucheva et al. 1994). T-cells play a crucial role in immunity to Salmonella infection (Hougen et al. 1989).
Despite the available insight into the pathogenesis and host response, the published literature is of limited use for dose–response modelling of food- and water-borne pathogens. Usually, only high doses were administered and were often inoculated via the intraperitoneal or intravenous route instead of orally. Most papers only considered a few end-points to detect infection, and it is not known whether colonization of the intestine and invasion occurred in a dose-dependent manner. It is commonly accepted that T-cells respond to minute stimuli, but the possibility of low numbers of bacteria surviving the effects of innate or acquired immunity is unknown. A series of experiments is therefore reported in which rats were extensively studied by microbiological, haematological, immunological and pathological measurements following oral exposure to different doses of Salmonella enterica subsp. Enteritidis (Salm. Enteritidis).
MATERIALS AND METHODS
Specific pathogen-free (SPF) male Wistar-Unilever (WU) rats were obtained from the breeding colony at the National Institute of Public Health and the Environment, Bilthoven, the Netherlands. The animals, 6–9 weeks of age, were housed individually in macrolon cages, 1–2 weeks prior to inoculation. Drinking water and conventional diet (RMH-B, Hope Farms BV, Woerden, the Netherlands) were provided ad libitum. The breeding colony of the animals was pre-screened/monitored for endogenous pathogenic viruses and bacteria, and was negative.
Salmonella enterica serovar Enteritidis 97-198 was a patient isolate (origin RIVM); Escherichia coli WG5, a nalidixic acid-resistant derivative of E. coli C (Havelaar and Hogeboom 1983) was used as a negative control. From all strains, a stock collection was made by pure culturing on Brain Heart Infusion (BHI) agar (18–20 h at 37°C), inoculating a single colony in BHI and incubating for 18–20 h at 37°C. After incubation, 0·7 ml of the culture was added to cryotubes filled with glass beads and 0·1 ml glycerol (82% w/v). Directly after adding the cultures, the cryotubes were thoroughly mixed and placed in a −70°C freezer.
Both strains were inoculated by placing one glass bead from the stock collection in BHI and incubating at 37°C for 18 h. After incubation, 100 ml of each culture were centrifuged at 5000 g for 10 min at 4°C. The supernatant fluid was discarded and the pellet was re-suspended in 100 ml physiological saline (PS), followed by re-centrifugation. Again, the supernatant fluid was discarded and the pellet was re-suspended in a volume of 4 ml PS. The cell suspension and serial dilutions in PS were delivered to the animal department on melting ice. Directly before administration to the animals, 4 ml of each bacterial suspension were mixed with 4 ml of a solution of 6% (w/v) NaHCO3. After administration, the remainder of the inoculum cultures was transported to the microbiological laboratory on melting ice for plate counts on sheep-blood agar (incubated as above).
The Central Animal Laboratory of RIVM possesses a licence under the Dutch ‘Animal Experiments Act’. In accordance with Section 14 of this Act, an officer was appointed to supervise the welfare of laboratory animals. All experiments were discussed and approved by an independent ethical committee prior to the study.
Each animal was implanted with a temperature transponder (BioMedic Data Systems, Seaford, DE, USA) subcutaneously in the neck. The transponder was coded for the (unique) animal number; it was used to detect body temperature each day at about 0900 h and, in some experiments, also in the afternoon. Body temperatures and animal numbers were registered using a Biomed Pocket Scanner (Plexx, Elst, the Netherlands).
After 1–2 weeks of rest (i.e. acclimatization), the animals were starved overnight (water adlibitum). After 16 h of starvation, 1 ml of the bacterial suspensions was orally administered by gavage. Directly after gavage (day 0), food and water was provided ad libitum.
Blood samples were taken via orbita plexus puncture, using a capillary under light ether anaesthesia, 10–14 days before and 5 days after oral inoculation, just before administration of the DTH reagent (see below).
Daily clinical observations were made on the general health of the animals. Special attention was paid to the consistency of the faeces. The animals were weighed each day (early in the morning), starting one day prior to the oral inoculation. Each morning, faeces were obtained from each rat in each group. The faeces were macroscopically evaluated and microbiologically tested on the same day (at weekends, day 1 and day 2 samples were stored at 4°C and examined on the following Monday, day 3).
The animals were sacrificed on day 6 after oral inoculation by bleeding from the abdominal aorta under KRA anaesthesia (intramuscular injection of 100 μl of a cocktail consisting of 7 ml ketalar (50 mg ml−1, Parke Davis, Spain), 3 ml rompun (20 mg ml−1, Bayer, Leverkusen, Germany) and 1 ml of atropin (1 mg ml−1, OPG, Utrecht, the Netherlands). Mesenteric lymph nodes, the gastrointestinal tract, the spleen and in some experiments, other organs were removed aseptically. The caecum of each rat was weighed and caecum weights relative to bodyweight were calculated. All organs were divided into two parts for microbiology and pathology, respectively.
Five experiments were performed under the standard conditions described above. Some experiments included slight differences in experimental protocol, which will be indicated in the text. In one experiment, the time course of infection was evaluated by including additional groups of animals that were sacrificed on day 12. Blood collection and injection of the DTH reagent for these animals was on day 11. In another experiment, the effects of fasting and neutralization of gastric acid by sodium bicarbonate were evaluated.
Delayed-type hypersensitivity reaction (DTH)
Five or 11 days after oral infection with Salm. Enteritidis, the thickness of both ears of each animal in each group was assessed in duplicate using an engineering micrometer (Mitutoyo 193–10, Veenendaal, the Netherlands). For this purpose, the animals were anaesthetized by intramuscular injection of 100 μl of a KRA solution. Directly after the ear measurements, 25 μl of a heat-killed suspension of Salm. Enteritidis (approximately 5 × 108 cfu ml−1) were subcutaneously injected into the pinnae of each ear of each rat (also in the control WG5 group). The increase in ear thickness was assessed 24 h after challenge under ether anaesthesia. The difference between ear thickness prior to, and 24 h after injection was calculated and reflected the DTH response, a valid parameter for T-cell-dependent (in vivo) immunity to Salm. Enteritidis. The swelling in control animals (i.e. exposed to E. coli WG5) reflected the background swelling response induced by the ear injection of Salm. Enteritidis itself.
Faecal moisture content
As an indicator for diarrhoeal-like effects, the moisture content of the faeces was determined each day by calculation of the weight loss after drying. For this purpose, the samples were weighed in a glass container with an air-tight lid and dried in an open container for 1 h at 103–105°C. After cooling in an exsiccator with silica-gel, the container was closed and weighed. This procedure was repeated several times until a constant weight was obtained.
As an indicator for (systemic) infection, haematology for each rat was determined on blood samples, obtained on day – 1 and 5, or day 11, anticoagulated with K3EDTA. The haematological analyses were performed using the H1-E, a multi-species haematology analyser (Bayer B.V., Mijdrecht, the Netherlands) with multi-species software, version 3·0.
The following parameters were determined: haemoglobin concentration (Hgb); haematocrit (Hct); red blood cell concentration (RBC); mean corpuscular volume (MCV); mean cell haemoglobin (MCH); mean cell haemoglobin concentration (MCHC); red blood cell distribution width (RDW); haemoglobin distribution width (HDW); platelets (Plt); mean platelet volume (MPV); white blood cell concentration (WBC); differentiation of white blood cells (% and absolute numbers) into neutrophils (neut), lymphocytes (lymph), monocytes (mono), eosinophils (eos), basophils (baso) and large unstained cells (luc).
Faecal samples were homogenized and diluted 1 : 10 (w/v) in peptone-physiological saline (PPS); appropriate 10-fold dilutions were spread on Brilliant Green Agar (BGA) for Salm. Enteritidis and on Tryptone Yeast Extract Glucose agar with nalidixic acid 100 μg ml−1 (TYGnal) for E. coli WG5. The BGA and TYGnal were incubated at 37°C for 22–26 h and 18–20 h, respectively. Internal organs were homogenized in PPS using an Ultra Turrax (Janke und Kankel, Breisgau, Germany). Appropriate 10-fold dilutions were spread-plated on BGA and TYGnal and incubated as described above. Blood samples were analysed by spread-plating 0·1 ml volumes on the same media.
Immediately after exsanguination, the abdomen and thorax were inspected and the gastro-intestinal tract was removed and processed. Macroscopic abnormalities were recorded per intestinal segment (volume and outer colour, gut-associated lymphoid tissue, quantity and quality of content, thickness of wall, aspect of mucosa). The stomach, three segments of the small intestine (duodenum, jejunum and ileum, each with Peyers’ patch), caecum, proximal and distal colon, mesenteric lymph nodes, liver, spleen, and tissue of macroscopically-abnormal organs were sampled and fixed in 3·8% (w/v) phosphate-buffered formaldehyde. The gut segments were processed using the Swiss roll technique (Moolenbeek and Ruitenberg 1981). After fixation, the tissues were embedded in paraplast; sections 4–5 μm thick were prepared and routinely stained with haematoxylin and eosin (HE, all dose groups) and periodic acid-Schiff (PAS, control and highest dose group). Histopathological examination in most animals was confined to the gut. Data for each individual animal were tabulated manually. Only well-cut parts of sections were scored. If oedema was present in the outer winding of the Swiss roll only, this was considered an artefact. Histopathological examination was performed without knowledge of the treatment.
The haematological data were analysed with Mathematica version 4·0 by means of linear regression using the following model:
The variable DAY means either prior to inoculation (DAY=0) or post inoculation (DAY=1). For the Log10(DOSE) term, a value of 0 was used for the control group. In this model, a significant day-effect due to experimental conditions is observed if β≠ 0, and a dose-related effect if ≫≠ 0. P-values less than 5% were considered significant.
Dose–response modelling was based on the following assumptions (Haas 1983): (i) single hit, i.e. one single surviving organism is capable of initiating an infection (Rubin 1987); (ii) independent action, i.e. the probability of one organism initiating infection is independent of the presence of other organisms (Meynell and Stocker 1957); and (iii) random distribution of the organisms in the inoculum.
In this article, infection is detected by isolation of salmonellae from either the spleen or the mesenteric lymph nodes. If it is assumed that the probability of any organism in any host to survive and initiate infection has a constant value r, then the exponential model follows:
where Pinf is the probability of infection and D is the mean dose. If r is assumed to follow a Beta-distribution with parameters α and β, the hypergeometric model follows (Teunis and Havelaar 2000). If α ≪ β and β ≫ 1, this can be simplified to the Beta-Poisson model:
The models were fitted to pooled data on infection of spleen and MLN using the solver in Microsoft Excel (Haas 1994) and model fits were compared by the likelihood-ratio test. If there was not a significant difference, the more parsimonious exponential model was preferred. ID50 values were calculated as:
for the exponential model and
for the Beta-Poisson model.
Table 1 shows the plate counts of different dilutions of the inoculum cultures for each experiment and the number of animals per dose group.
Table 1. Microbiological analysis (colony count ml−1) of inoculum cultures
The animals infected with the highest doses (more than 108 per animal) showed severe illness from 2 days after infection onwards. The animals were skinny, weak, cold, had ruffled fur, low muscle tension, nasal discharge, and red crusts around eyes and nostrils. At these doses, weight loss was observed whereas at all other doses, body weight increased at a rate similar to the controls. For ethical reasons, animals already severely ill on day 3 after infection were sacrificed. None of the animals developed diarrhoeal illness.
There was a dose-dependent decrease in body temperature of the infected animals shown by measurements taken in the afternoon. Because of relatively high inter-individual variability, the difference only exceptionally reached statistical significance. Temperature measurements taken during the morning did not show any dose-dependent differences. Faecal moisture content did not show any consistent trend over time in relation to infection.
Typically, exposure to Salm. Enteritidis resulted in a sharp increase of faecal counts, followed by a transient decrease and subsequent increase. Figure 1 shows data from two different experiments, indicating relatively good reproducibility. Doses below 104 cfu did not result in detectable faecal excretion. Due to overgrowth of plates by indigenous faecal flora, the limit of detection in some experiments was up to 104 cfu g−1. For this reason, faecal excretion was not considered to be a reliable indicator of infection.
Blood cultures were always negative for Salmonella. The results of microbiological examination of mesenteric lymph nodes (MLN) and spleen are shown in Fig. 2. Infection occurred in all exposed animals at doses above 105 per animal (MLN) or 103 per animal (spleen). Bacterial counts in the spleen of infected animals at day 6 were usually between 103 and 104 cfu g−1, irrespective of dose. Counts in MLN were usually between 105 and 106 cfu g−1, but in some experiments, values below 103 cfu g−1 were observed. On day 12, no bacteria were detectable in MLN. Salmonellae were recovered from the spleen but in fewer animals than on day 6. Between days 6 and 12, counts in the spleen had decreased by one to two orders of magnitude (data not shown).
The data for the probability of infection in the spleen on day 6 showed a regularly increasing trend with the dose and could be fitted to the exponential model (r=1·2 × 10−3, see Fig. 2). Between doses of 102 and 105 cfu, the fraction of animals infected in the MLN was widely dispersed between 0 and 1. At the 5% but not at the 1% level, the Beta-Poisson model (α=0·261, β=769) fitted the data for infection of MLN better than the exponential model (r=3·9 × 10−5). From these models, the ID50 is 600 cfu for the spleen and 10 200 cfu (Beta-Poisson model) or 17 600 cfu (exponential model) for MLN. These data indicate that isolation of bacteria from the spleen is a highly sensitive and reproducible end-point to detect infection.
Delayed-type hypersensitivity reaction (DTH)
Injection of heat-killed Salmonella antigens induced ear swelling in all animals. Animals previously exposed orally to Salm. Enteritidis in doses of 500 cfu and above had significantly more swelling than control rats exposed to E. coli WG5 (see Fig. 3). Note that at doses above 500 cfu, Salm. Enteritidis was also detectable in the spleen. The degree of swelling clearly increased with dose and the response to similar doses in different experiments was at the same level, which indicates good in vivo reproducibility of this immune parameter. On day 11–12, the ear swelling was more pronounced than on day 5–6 in all dose groups.
Oral exposure to Salm. Enteritidis resulted in significant, reproducible, dose-dependent changes in the white blood cell (WBC) population of the rats. WBC concentration increased significantly in several, but not all experiments. Lack of significance was related to variability in haematological parameters between different animals, and also within one animal on different days. The composition of the WBC population changed dramatically. On day 5, large increases were seen particularly in absolute and relative numbers of neutrophils. Figure 4 shows that detectable changes in neutrophil counts were observed from doses of 104 cfu upwards. Numbers of monocytes, and sometimes also basophils, increased as well, whereas the numbers of lymphocytes decreased significantly. Figure 5 shows the changes in the different subclasses of the white blood cell population in relation to time since exposure and dose. All leucocyte types, except eosinophils, showed clear dose–response relationships on one or both days. Neutrophils and monocytes peaked on day 5. On day 11, monocyte counts returned to the background level but neutrophil counts were still somewhat elevated. In contrast, lymphocytes and basophils showed little increase on day 5, but were clearly elevated on day 11. Large unstained cells (usually large lymphocytes or monocytes) increased above the background level on day 5 at doses greater than 103 cfu, whereas at day 11, an increase was also clear at lower doses.
Macroscopic pathological analysis indicated systemic illness after exposure to high doses, notably (broncho-) pneumonia. Also, there were limited, dose-related gastro-intestinal abnormalities, i.e. little gastro-intestinal content reflecting reduced food intake, and opacity of the caecal wall. Peyers’ patches in the ileum and the mesenteric lymph nodes were enlarged in the highest dose groups. In experiment number 2, extensive histopathological examination was carried out (see Table 2 and Fig. 6). No abnormalities were seen in the duodenum and distal colon. In all but one animal in the two highest dose groups, predominantly mononuclear inflammatory infiltration of the villi in the ileum was present, while the jejunum was generally normal. In one animal, however, the opposite situation was observed. Slight to moderate infiltration of the small intestinal mucosa (‘villitis’) was accompanied by a decreased villus–crypt ratio. Microgranulomas were seen mainly in Peyers’ patches, in one animal associated with local peritonitis, and also in the lamina propria of animals in the two highest dose groups as well as in one ileal Peyers’ patch from an animal in the intermediate group. In the large intestine, microgranulomas were limited to the gut-associated lymphoid tissue. Inflammation reached a higher grade of severity in the large intestine, especially in the caecum of the three highest dose groups. In the two highest groups, diffuse flattening of superficial caecal epithelium, massive erosions and local ulcerations, and decreased PAS-positivity were seen. In association, almost diffusely, very marked, mixed inflammatory infiltration was present in the mucosa, near ulcers also in the submucosa, accompanied by marked oedema. Caecal lesions in the intermediate group locally reached equal severity, but were less extensive. Infiltration was also present in the proximal, but not in the distal part of the colon.
Table 2. Histopathological evaluation of enteritis (D: duodenum, J: jejunum, I: ileum, C: caecum, C1: proximal colon, C2: distal colon), of adult, male WU rats, 6 days after oral inoculation with different doses of Salmonella Enteritidis (experiment 2)
In later experiments, histopathological analysis was confined to the caecum, demonstrating lesions that were comparable with those seen previously. Variation in the severity of mucosal lesions was observed at doses between 103 and 105 cfu. At doses below 104 cfu, infiltration and erosion of the caecal wall was noted at day 12, but not at day 6. One animal exposed to a dose of only 10 cfu showed remarkable lesions on day 12. The superficial epithelium was almost diffusely flattened/basophilic, and cellularity of the lamina propria increased, with evidence of crypt regeneration. This was the only animal at day 12 with slightly watery caecal contents, indicating the relative importance of extent rather than severity of the lesions.
Effect of fasting and neutralization of gastric acid
To promote survival of salmonellae so that they could reach target sites for colonization, the standard model included overnight fasting of the animals and neutralization of gastric acid by suspension of the inoculum in a sodium bicarbonate solution. The protective effect of these factors was evaluated separately and in combination (experiment number 5). Overall, the greatest protective effect was found for bicarbonate, with an additional effect of fasting in the low dose region (104 cfu). Figure 7 shows the effects as detected by the DTH reaction. There is a clear dose–response relationship for all treatments. At any dose, the effect was largest with fasting and bicarbonate (FB). Animals that did not fast but received the inoculum in bicarbonate (– B) had a slightly lower response, whereas the response was clearly lower in the groups without bicarbonate (F– and – –). Due to small group sizes, none of the effects reached statistical significance. Similar effects were observed for haematological data and for invasion into the spleen and MLN. Table 3 shows the results for invasion into the spleen. With fasting and bicarbonate, all animals at all doses were culture-positive, whereas with only bicarbonate, only two animals were positive at the lowest dose and at this dose, no animals were positive in the group that only fasted. There was no clear pattern in the spleen counts of positive animals.
Table 3. Effect of fasting (F) and neutralization of gastric acid by sodium bicarbonate (B) on invasion of the spleen by different doses of Salmonella Enteritidis (6 days after oral inoculation)
The dose–response relationship of exposure by intragastric gavage of adult, male WU rats to Salm. Enteritidis is well reproducible. In all experiments, the animals were shedding the bacteria in their faeces after exposure to intermediate to high doses (> 104 cfu). At lower doses, no faecal excretion was detected within the 6 day period of observation, but salmonellae could be isolated from the spleen and mesenteric lymph nodes. These findings are in accordance with the fact that Salm. Enteritidis is highly invasive in rodents, and that the intestinal tract may not be the major site of multiplication. The limited recovery from faeces may also be related to problems with the selectivity of the isolation medium for salmonellae from faeces.
Pathological results partly confirm these observations, indicating that the gastro-intestinal tract (although portal of entry) shows relatively few abnormalities in animals that succumb to severe systemic illness after oral inoculation with very high doses of Salm. Enteritidis. However, at lower, non-lethal doses, lesions typical for gastro-enteritis were observed in the ileum, caecum and proximal colon. There was no evidence of clinical illness associated with these histological abnormalities. Animals appeared healthy, the faecal moisture content of infected rats was similar to background levels, and body temperatures were only marginally affected. Note that in the present study, a decrease in body temperature was observed in resting animals, in contrast to increases reported for Fischer 344 rats infected with Salm. Typhimurium (Bradley and Kauffman 1988, 1990).
In these experiments, semi-inbred rats were exposed to a pure culture of Salm. Enteritidis prepared under laboratory conditions. In such cases, it appears reasonable to assume that the probability of any single organism establishing itself in the animal and causing infection is constant. Indeed, data for infection of the spleen could be fitted with the exponential model and indicated that, on average, 1 per 860 (i.e. 1/r) cells in the inoculum infected the spleen. In contrast, the reproducibility of detecting infection by microbiological examination of lymph nodes was poor, both with respect to the fraction of positive animals and the mean counts in positive animals. The Beta-Poisson model fitted the dose–response data in MLN slightly better than the exponential model. Figure 2 shows that the estimated probability of infection for the Beta-Poisson model is substantially less than 1 at doses as high as 106–108 cfu. This is in contrast with experimental observations in this and many other published studies. The exponential model agrees better with experimental data in this dose range than the Beta-Poisson model, but in fact neither model fits well to the highly variable MLN data. This may be due to inhomogeneous infection of individual nodes at intermediate doses. Only a small number of MLN were sampled for microbiological analysis, and it may well be that by chance, positive nodes were sampled in one experiment and negative nodes in another. Examining a larger proportion of the MLN would reduce this problem.
Rats are able to mount a vigorous cellular immune response to salmonella infection. It has been demonstrated that there is a consistent dose–response relationship for most leucocyte subsets. The only exception was a lack of response by eosinophils, which was not unexpected because these cells are mainly related to parasitic infections. The strong increase in monocytes on day 5 can be interpreted as a consequence of the innate, non-specific immune response in which tissue macrophages are recruited from a pool of blood monocytes to engulf the invading salmonellae. Initial depletion of the monocyte pool in blood by migration into tissues is over-compensated by increased production in the bone marrow (Volkman and Collins 1974). Recent studies (Vassiloyanakopoulos et al. 1998) have shown that survival and growth in macrophages is a major virulence mechanism for Salmonella. Macrophage death by apoptosis is a host defence mechanism, which leaves the bacteria susceptible to subsequent phagocytosis and killing by neutrophils (which increased strongly in the blood of infected animals). In serum of animals challenged with the highest doses, detectable levels of IL-8, which attracts neutrophils, were observed. There was a significant, dose-related decrease in lymphocyte counts on day 6 which may be related to killing by NO produced by macrophages (al-Ramadi et al. 1992). Later, the non-specific immune response is succeeded by a specific response, which is manifest by a strong increase in lymphocytes on day 11. There was also an increase in basophils on day 11. Recent studies indicate that basophils can bind various bacteria in the presence or even in the absence of opsonizing antibodies. This interaction can induce the release of a number of inflammatory mediators and cytokines, such as tumor necrosis factor alpha (TNF), which is a crucial potentiator of neutrophil responses (Abraham and Arock 1998). These authors suggest that basophils, like mast cells, are important in the early interaction with invading pathogens. In the present studies, the strongest basophil response was seen on day 11, which suggests an involvement in specific immune response instead. Analysis of the dose–response relationship of this kind of continuous dataset requires development of specific models, which will be described in a separate paper (Takumi et al. submitted).
The DTH reaction was also found to be a sensitive and reproducible parameter for detecting an immune response to salmonella infection. Positive reactions were observed at challenge doses above 100 cfu, which is equivalent to doses that lead to colonization of the spleen.
In conclusion, the rat model appears to be a valid tool for the evaluation of the role of different factors in the host–pathogen–food matrix triangle on initial stages of infection by food-borne bacterial pathogens. Sensitive and reproducible end-points of infection include colonization of the spleen, DTH reaction and white blood cell differentiation, particularly neutrophils and monocytes. Animals fasted overnight and exposed to bacteria suspended in a sodium bicarbonate suspension are most susceptible to infection. According to model predictions, systemic infection may occur at doses of 100 cfu or below, and will occur in virtually all animals at doses above 103 cfu.
The experiments in this article involved the work of many persons in different laboratories. Hans Strootman, Mariska van Dijk, Dirk Elberts and Bert van Middelaar were responsible for animal experiments. Coen Moolenbeek performed section of the animals, Paul Roholl was responsible for histotechnique, Henny Loendersloot and Sandra de Waal performed the histotechnical work and Yvonne Wallbrink participated in haematological analyses. Ellen Delfgou-van Asch and Wilma Ritmeester were involved in the microbiological analyses. Wim Jansen and Nan van Leeuwen provided bacterial strains. Joke van der Giessen, Marion Koopmans, Joop Schellekens and Ron Boot provided helpful advice on biomedical aspects, and Wout Slob and Peter Teunis on statistical aspects of the work.