Development of an In Vivo Model for Study of Intestinal Invasion by Salmonella enterica in Chickens

Abstract

An in vivo loop test model for the investigation of the invasiveness of Salmonella enterica in chickens was developed. Ten jejunal loops were made in 10- to 12-week-old Lohman Brown chickens under isofluoran anaesthesia. Salmonella at 5.0 × 10⁷ CFU was inoculated into each loop and left for 2 h, followed by a 1-h incubation with gentamicin in order to kill noninvading bacteria. After euthanasia, Salmonella invasiveness was measured as tissue-associated counts relative to a reference strain. The ability of Salmonella invasion was 1 log₁₀ CFU higher per 42-mm² mucosal tissue in the anterior than in the posterior part of jejunum. A statistically significant (P < 0.001) sixfold difference in invasiveness was observed between a wild-type S. enterica serotype Typhimurium strain and the corresponding invH mutant. The model was shown to be able to show small differences in invasive capability and allows for comparison of strains tested in different animals, provided that the same reference strain is present in all animals.

Invasion of the gut epithelium is essential in the pathogenesis of Salmonella infections (4), and differences in virulence among serotypes might be related to differences in invasiveness. In order to study this, both in vivo and in vitro invasion models have been used (10, 13). As pointed out by Stone et al. (11), in vitro models may not incorporate all possible invasion pathways. In the case of the chicken, the lack of suitable epithelial cell lines is a further limitation in the use of in vitro models. In order to elucidate further Salmonella invasion in chickens, the present study was undertaken to develop and evaluate a chicken intestinal loop model.

Lohman Brown female chickens, 10 to 12 weeks of age, were used because they are immunologically mature and have an established gut flora. The birds were given feed and drinking water ad libitum and acclimatized for 5 to 6 days before the study commenced. All birds were tested for excretion of salmonellae by examining pools of 10 fecal swabs by standard culture procedures using Rappaport-Vassiliadis (Oxoid CM866) as the selective enrichment medium. In order to decrease intestinal content, the birds were starved 16 to 20 h prior to experiments.

Chickens were anesthetized by the administration of Isofluran (Pharmacia) through a mask with continuous gas flow at a level of 1.5 to 3% depending on the depth of the anesthesia and an oxygen flow of 1.5 to 2.0 liters/min. The bird was covered with a sterile surgical blanket and defeathered at the abdominal surface. After disinfection of the incision area with 99% alcohol, the abdomen was opened by a transverse incision approximately 2 cm caudal to the sternum, and the jejunum was carefully exposed from the cavity by hand. Nine to ten ligated loops were constructed using epidermal polyester green surgery suture (Kruuse, Martofte, Denmark). The loops were ca. 1.5 to 2.0 cm long and separated by smaller “spacer” loops of 0.5 to 1 cm in order to allow excision of the loops and to minimize the risk of leakage and contamination from adjacent dosed loops. Sufficient blood supply was ensured to both regular and spacer loops. The anatomic position of the loops is schematically shown in Fig. 1.

FIG. 1.

Schematic representation of the chicken intestinal tract. The jejunal section of the gut is shaded black, with black transverse bars to indicate the region for loop construction. Key: 1, proventriculus; 2, gizzard; 3, duodenum and pancreas; 4, jejunum; 5, diverticulum ventricularum; 6, ileum; 7, ceca; 8, rectum; 9, cloaca.

Loops were dosed with bacteria through an injection performed into the middle third of the loop opposite to the mesentery line using a 0.4-mm (27 gauge) needle. After dosing, the loops were reintroduced into the abdomen, and the abdominal wall was clamped. Two hours later, 0.5 ml of phosphate-buffered saline (PBS) containing gentamicin (300 μg/ml) was added to each loop with an estimated final concentration of 150 to 200 μg/ml in the loop depending on the degree of absorption of the dose volume. The jejunum was relocated into the abdomen, and the abdominal wall was clamped again. After a further 1 h the chicken, still under anesthesia, was euthanized by decapitation. Loops were immediately removed under sterile conditions and placed on ice. In the laboratory they were cut clean at each ligature, opened along the mesenteric line, put into 5 to 10 ml of ice-cold mucosal medium (1), and gently rinsed in order to remove excessive gentamicin and non-tissue-associated salmonellae. Three circular biopsies of 6-mm diameter, a total of approximately 84-mm² mucosa, were subsequently taken from the gut tissues of each loop. The biopsies were placed in a 5-ml tube and homogenized with 2 ml of ice-cold PBS with 1% Triton X-100 using a Diax 900 dispersing machine (Heidolph Electro GmbH, Kelheim, Germany) for ca. 30 s to ensure homogeneity. After a lysis period of 10 min, 10-fold serial dilutions of the homogenate were made in physiological saline (0.9%), and subsequently 100 μl from each dilution was spread on xylose-lysine-desoxycholate agar (XLD-Agar; Oxoid CM469). The plates were incubated aerobically overnight at 37°C. The detection limit was approximately 10 bacteria per ml of biopsy homogenate, which is equivalent to 20 bacteria per 84-mm² mucosa. The counts per milliliter of homogenate, corresponding to counts per 42-mm² mucosa, were log₁₀ transformed, and this value was used to express invasion throughout the study. The first and the last loops were normally inoculated with 0.5 ml of PBS as control loops. The remaining loops were inoculated with test strains, which consisted of Salmonella serotype Typhimurium 4/74 and its corresponding invH mutant, Salmonella serotype Typhimurium 4/74 invH201::TnphoA (13), in addition to Salmonella serotype Choleraesuis A57 and Salmonella serotype Dublin 2229 (field isolates). All strains harbored serotype-specific virulence plasmids. The MIC values for gentamicin were 0.125 μg/ml for all strains.

The loop model is an indirect invasion assay, which demonstrates Salmonella counts from representative tissue sections rather than an exact measure of total internalized bacteria. Salmonella invades the loop mucosa during the 2-h incubation period, and the intracellular location leads to protection against the subsequent treatment with gentamicin. The efficacy of gentamicin for killing the noninvading bacteria is thus crucial to the outcome. To investigate this, cultures were made by adding 12 μl of an overnight broth culture to 10 ml of Luria-Bertani (LB) broth (CM-Lab) overlaid with 3 ml of paraffin oil and then incubating this mixture at 37°C for 6.5 h. This resulted in bacteria in late log phase and with low oxygen tension, which is favorable for invasion (6, 8). After incubation, the culture was harvested by centrifugation at 4,000 rpm for 15 min and suspended in 5 ml of PBS. The optical density at 620 nm was measured, and the dose culture adjusted to 0.1 optical density units corresponding approximately to 1.0 × 10⁸ CFU/ml. For each loop a dose of 0.5 ml containing 0.5 × 10⁸ CFU (log₁₀ 7.83) was desired. The bacterial concentration was verified by plate counts on XLD in duplicate. The mean dose (log₁₀ CFU) for all inoculations in this study was 7.81 ± 0.16. Loops (n = 5) dosed with the wild-type strain Salmonella serotype Typhimurium 4/74 and treated with gentamicin showed 1.23-log₁₀ (17-fold)-lower mucosal counts than loops not treated with gentamicin. The corresponding difference for the invH mutant was 2.62 log₁₀ (417-fold) (n = 5) (Fig. 2), reflecting that this strain has a much lower invasion ability (see below) and hence is less protected against the effect of the antibiotic chase. The effect of gentamicin was more pronounced in this study than was seen in the cattle intestinal loop assay (13); however, the difference in effect between the wild type and its invH mutant was comparable for the two studies.

FIG. 2.

Mean Salmonella counts (log₁₀ CFU/42-mm² mucosal tissue) in loops dosed with the wild-type Salmonella serotype Typhimurium 4/74 and the isogenic invH mutant with (G+) or without (G−) gentamicin treatment. The standard error of mean is also indicated.

Preliminary investigations indicated that the ability of Salmonella to invade was not uniform throughout the jejunum (data not shown). Therefore, counts from 35 loops dosed with the invH mutant in eight loop positions using the two outermost loops (loop 1 and 10) as PBS controls, were analyzed. The inoculum was prepared as described above except that 12 μl of a 10⁻³ dilution of the primary culture was added to 10 ml of LB broth with a paraffin cover, followed by incubation for 8.5 h. The counts of the invH mutant were seen to decline from the anterior part to the posterior part of the jejunum as seen from Table 1. A statistical analysis of log₁₀ transformed bacterial counts was conducted as standard F tests in linear normal models, including effects of strain (s), loop position (l), and animal (h). The analysis, which also included loop counts for loops dosed with Salmonella serotype Typhimurium (wild type) (n = 10), Salmonella serotype Dublin (n = 4), and Salmonella serotype Choleraesuis (n = 4), supported an exponential decrease or a linear decrease after log₁₀ transformation of the bacterial counts (log linear decrease) in invasiveness in the craniocaudal direction. The decline in invasion was not invH specific, since the three other strains showed a similar trend. The description of the effect as a log linear decrease, depended on the exclusion of loop 10 (the most caudal loop) due to relative high counts of the Salmonella serotype Typhimurium 4/74 wild type at this position. Consequently, we decided to use loops 1 and 10 as blank PBS controls in the model. The analysis showed an average decline in Salmonella uptake over eight jejunal loops (loops 2 to 9) of 1.06 log units (10-fold) (95% confidence interval, 0.56 to 1.55) as a mere effect of the loop position. In order to be able to compare the invasion of strains which are not dosed at the same loop positions a calculated correction for the effect of loop position was chosen. The calculated correction fits the actual data, and as such the validity of the model does not depend on the exclusion of loop 10. This exclusion is only for the convenience of a more perceivable description of the effect.

TABLE 1.

Mean counts (log₁₀ CFU) of Salmonella serotype Typhimurium 4/74 invH201::TnphoA from loops 2 to 9

Loop	Mean count (log10 CFU)	SEM	n
1	–a
2	5.16	0.24	5
3	4.47	0.20	4
4	4.58	0.18	4
5	4.06	0.28	5
6	3.90	0.35	5
7	3.44	0.24	5
8	3.56	0.17	5
9	4.12	0.11	2
10	–

^a–, Blank control. 

The jejunum was chosen as the intestinal site for the loop construction for practical surgical reasons. It has been shown that invasion can take place in all parts of the jejunum in day-old chicks (12), but it was not known whether Salmonella is able to invade the jejunum in immunocompetent animals with a fully developed gut flora. To directly demonstrate this, loops for histological examination were prepared separately. Sections of entire cross-sections of the intestine were fixed in 10% neutral buffered formalin and subsequently processed according to standard procedures. Slices of 3 μm in thickness were mounted on Super Frost*/Plus slides (Menzel-Gläser). One slice of each sample was stained with Giemsa, hematoxylin, and eosin or stained by a peroxidase-antiperoxidase (PAP) immunohistochemical method according to standard procedures employing a polyclonal rabbit antiserum against serotype Typhimurium. Tissues from loops with both strains of serotype Typhimurium exhibited normal appearance of the mucosa with long slender villi. Minor early degenerative epithelial changes with vacuolization and loss of lateral cell adherence were only observed at the apex of villi. Immunohistochemical preparations demonstrated invasion of epithelial cells (enterocytes) of both wild-type serotype Typhimurium 4/74 and the corresponding invH mutant. Multiple intraepithelial salmonellae were demonstrated apically in several villi (Fig. 3). Some of the intraepithelial bacteria appeared to be within vacuoles, while others were free in the cytoplasm. Furthermore, the strains were seen to infiltrate the adjacent lamina propria of some villi. No infiltration with inflammatory cells were observed in the mucosa after the 2 h of incubation as demonstrated in Fig. 3.

FIG. 3.

Chicken intestinal villus with an intraepithelial apical (A) or basal (B) location of invading Salmonella serotype Typhimurium 4/74 (wild type). Salmonella-specific immunohistochemical PAP staining was used.

In mammals, salmonellae have been reported to invade through specialized M cells in the Peyer's patches (5), but invasion outside the lymphoid tissue can also take place (12, 13) and may be more common than previously anticipated. Peyer's patch structures are present in chickens (2, 7). They undergo development from two Peyer's patch structures in day-old chicks, increasing to up to five structures diffusely scattered in the intestine of 12-week-old chickens with a subsequent decline to a single Peyer's patch structure in the ileocecal junction at the age of 56 weeks (2). In the present study the histopathological findings demonstrated internalization in enterocytes of both the wild-type strain and the invH mutant (Fig. 3), supporting the finding of Turnbull and Richmond (12) that invasion can take place through normal enterocytes. The Peyer's patch-like structures in the jejunum are not visible by the naked eye (2), and it was therefore not possible to evaluate whether Peyer's patches in a loop could influence the invasion rate due to the presence of phagocytic M-cell-like cells. Whether the surprising observation of the highest invasion of Salmonella in the anterior part of jejunum could be caused by the presence of more lymphoid tissue in the anterior part of the intestine compared to the posterior part is at present unknown.

The invasion of the wild-type serotype Typhimurium and its isogenic invH mutant was compared in order to evaluate the ability of the model to demonstrate differences between strains. The mean uptakes (log₁₀ CFU) were 4.98 ± 0.16 from 10 loops dosed with serotype Typhimurium 4/74 and 4.1 ± 0.13 from 35 loops dosed with the invH mutant, giving a difference of 0.84 ± 0.29 equal to a sevenfold difference in invasiveness. The data were analyzed according to a model, including the random effect of test strain, loop position, and animal as follows: log₁₀ (count_slh) = μ + α_s + β_l + c_h + e_slh, where μ represents a general level, α_s represents effect of the strains, β_l represents the effect of the loop position, c_h is the random effect of animal (chicken), and e_slh is the residual effect, which covers effects that cannot be explained by the factors included in the model (9). From the statistical model, estimated differences in invasiveness between the two strains (α_i − α_j) were derived using best linear unbiased predictors. This gives a confidence interval for the estimated difference in addition to a correction for the effect of the loop position. The results from this analysis can be interpreted as prediction intervals for mean differences between two strains, whether they are tested in the same or different animals provided that the same reference strain is present in all animals. When this statistical analysis was performed, a sixfold difference in invasiveness between the wild type and the invH mutant was observed that corresponded to a difference in counts of 0.78 log₁₀ CFU with a 95% confidence interval between 0.37 and 1.19 (P < 0.001). The counts of loop 10 were included in the analysis in order not to underestimate the variance. The invH strain has previously been shown to have a decreased level of invasion compared to the wild-type strain in an intestinal loop assay in calves and in cell cultures (11, 13), and the result gives direct evidence for a role of invH, which is part of the type III secretion system encoded by Salmonella pathogenicity island I (3) in the invasion process of Salmonella in chickens.

The small variations in dose did not significantly influence the uptake of any of the strains, which may be explained by a threshold effect, i.e., that the loop dose exceeded a biological level of what can possibly be internalized in the loop epithelium during the 2 h of incubation. Also, the small variation in age from 10 to 12 weeks did not have an apparent effect on the bacterial counts.

In conclusion, the differences in invasion demonstrated between the wild type and the corresponding invH gene highlight the potential of this model to show even small differences in the invasive potential of Salmonella serotypes. Another perspective of the model is the ability to show the mean difference in invasion between two strains even if they are not tested in the same animal. The model may further be used to study whether or not the same cellular mechanisms in mammals are important during Salmonella invasion in the avian host.