Bread Feeding Is a Robust and More Physiological Enteropathogen Administration Method Compared to Oral Gavage

Oral administration is a preferred model for studying infection by bacterial enteropathogens such as Yersinia spp. In the mouse model, the most frequent method for oral infection consists of oral gavage with a feeding needle directly introduced in the animal stomach via the esophagus. In this study, we compared needle gavage to bread feeding as an alternative mode of bacterial administration. Using bioluminescence-expressing strains of Yersinia pseudotuberculosis and Yersinia enterocolitica, we detected very early upon needle gavage a bioluminescent signal in the neck area together with a signal in the abdominal region, highlighting the presence of two independent sites of bacterial colonization and multiplication.

tion on the overall infectious process is not known, we cannot exclude the possibility of bacterial dissemination in the bloodstream independently from the intestinal infection. In order to avoid this potential problem, alternative modes of pathogen administration have been described, such as bread feeding (7,8) or drinking water delivery (9,10).
Enteropathogenic Yersinia spp. are the third leading bacterial cause of human gastrointestinal infections in Europe, with Yersinia enterocolitica being the major agent responsible for this disease (11). Although isolated much less often than Y. enterocolitica, Yersinia pseudotuberculosis is responsible for acute gastroenteritis and mesenteric lymphadenitis in a wide variety of animals, including rodents, domestic animals, nonhuman primates, and humans (12). After oral ingestion, Y. pseudotuberculosis localizes to the ileum and proximal colon and can pass/cross the intestinal barrier, invading its host through gut-associated lymphoid tissues and Peyer's patches (13). In humans, dissemination to deeper tissues and the bloodstream is frequent in patients with underlying disease conditions (such as diabetes mellitus, liver cirrhosis, and hemochromatosis) and can lead to septicemia as a severe outcome of the infection (14,15).
While performing mouse oral administration of Y. pseudotuberculosis with a feeding needle, it has been regularly noticed that some animals exhibit spleen infections without displaying the presence of bacteria in Peyer's patches (PPs) or mesenteric lymph nodes (MLNs), suggesting the passage of bacteria in the bloodstream independently of the colonization of the gut-associated lymphoid tissues (16). Since it has been reported that needle gavage can cause lesions in the oropharyngeal region (3), we decided to reexamine this mode of administration and compare it to an alternative administration method (bread feeding) with fully virulent Y. pseudotuberculosis and Y. enterocolitica strains expressing bioluminescence in order to follow bacterial dissemination over time in the whole animal body. Our results clearly illustrate that needle feeding promotes bacterial colonization not only of the intestinal tract but also the neck region with a tropism for salivary glands and lymph nodes. On the contrary, bread feeding induces a much more robust intestinal tract infection that is never associated with neck region infections. Moreover, bread feeding allows bacteria to be protected from the acidic environment of the stomach. We conclude, therefore, that bread feeding is a better oral administration route to investigate the pathophysiology of bacterial enteropathogen infection.

RESULTS
Construction of constitutively bioluminescent Y. pseudotuberculosis and Y. enterocolitica strains. The IP32953 Y. pseudotuberculosis strain, isolated from the stools of a human patient, was chosen for this study due to its capacity to express two major virulence factors, the type 3 secretion system (T3SS) and the yersiniabactin (Ybt) iron uptake machinery. This strain has been previously shown to be fully virulent in laboratory mice (17,18). The WA Y. enterocolitica strain belongs to biotype 1B serotype O:8, and it was isolated from the blood of a human patient and was shown to be virulent in rodents (19).
To perform a comparative analysis of bacterial dissemination in vivo after oral infection, strains IP32953 and WA were genetically engineered to constitutively express bioluminescence. In order to correlate bioluminescence to bacterial numbers during animal infection, the luxCDABE operon (placed under the control of the constitutive rplN promoter) was stably introduced into the Yersinia chromosome using the mini-Tn7 transposon technology (20,21).
The resulting strains, IP32953-lux and WA-lux, harbor the construct mini-Tn7-Km-P rplN -luxCDABE between the two housekeeping genes glmS and pstS (see Fig. S1A in the supplemental material) (21). The stability of the bioluminescence signal over time was determined by performing 10 subcultures of IP32953-lux in LB without antibiotics over 17 days. At different time points, bacterial aliquots were streaked on LB plates, and individual colonies were checked for their capacity to emit bioluminescence. After 10 subcultures, 100% of the CFU expressed bioluminescence, confirming the stability of the bioluminescent phenotype in the bacterial population.
The IP32953-lux and WA-lux strains displayed in vitro growth similar to that of the parental strains (Fig. S2). The IP32953-lux strain exhibits virulence similar to that of the parental wild-type IP32953 strain (Fig. S3). Therefore, IP32953-lux was used to determine the characteristics and relevant differences between the two infection methods applied, oral gavage with a feeding needle versus bread feeding. The WA-lux strain was used as a second species to confirm some of the results obtained with Y. pseudotuberculosis.
Pathogenic Yersinia spp. colonize the neck and the abdominal region of infected animals upon needle feeding. We first set up the bread feeding protocol in OF1 mice based on a previously reported methodology (7). In our study, instead of using melted butter as a vehicle to deliver bacteria and to attract mice to the bread, mice were first trained to feed on bread only (see Materials and Methods) and then were infected with bread supplemented with a bacterial suspension in phosphatebuffered saline (PBS). A pilot experiment using this new method allowed us to validate that within less than 5 min, mice consumed bread supplemented with 8E7 CFU of IP32953-lux, subsequently exhibiting bioluminescence in the PPs and MLNs, therefore confirming Y. pseudotuberculosis colonization and dissemination in the intestinal tract as expected for an enteropathogen (Fig. S1B).
We then proceeded to compare needle versus bread feeding as oral administration routes. Twenty-four hours after oral administration of 3.5E8 CFU of IP32953-lux, mice were monitored for bioluminescence, and regions of interest (ROI) were identified for bioluminescence imaging (BLI) measurements. Upon needle gavage, up to 80% of the mice displayed a BLI signal in two distinct regions of the body, one in the expected abdominal region and a second in the neck, whereas upon bread feeding, none of the mice exhibited a neck signal, and bioluminescence was restricted to the abdomen (Fig.  1A). ROI measurements in the neck region of needle-infected mice indicate that the level of BLI is similar to that coming from the abdominal region, suggesting that bacteria colonize, disseminate, and multiply in the neck as efficiently as in the abdomen (Fig. 1B). This comparative analysis indicates that oral administration of Y. pseudotuberculosis using needle feeding results in infection of the neck region, a phenotype that is not observed upon bread feeding. A similar phenotype was observed using the Y. enterocolitica WA-lux strain, indicating that this phenomenon could be extended to other pathogens (Fig. S4).
Upon needle feeding, pathogenic Yersinia spp. multiply in the intersection of the esophagus and trachea and disseminate to the draining lymph nodes of the submandibular area. To identify which specific neck region is colonized by Y. pseudotuberculosis, mice were first orally gavaged with 4E8 CFU of IP32953-lux; then, at 24 and 48 h postinfection, mice exhibiting a BLI signal in the neck were euthanized and dissected in the cervical ventral region (see Materials and Methods). After skin removal, all of the mice exhibited a BLI signal coming from the cervical soft tissue composed of salivary glands, lymph nodes (LNs), and adipose tissues (Fig. 2). Among 10 dissected mice, nine exhibited bioluminescent signals in LNs from the salivary gland region ( Fig.  2A, C 1 , and D 1 ). Although there was no preferential right or left LN colonization, we often noticed that after removal of a first LN producing large amounts of photons, it was possible to identify secondary LNs producing less light, indicating various levels of bacterial colonization among LNs. Each time a BLI-positive LN was isolated, we verified its bacterial content by homogenization and CFU counting on agar plates (Fig. 2C 1 and D 1 ). In addition to LNs, we noticed in 70% of the mice a strong BLI signal in the esophagus and/or trachea ( Fig. 2B and C). Dissection of the esophagus and tracheal sections associated with BLI allowed us to localize bacterial colonization at the junction of these two structures ( Fig. 2C 2 ), suggesting a probable deposition of bacteria in the tissue consecutive to the introduction of the needle in the esophagus. Finally, half of the mice emitted a BLI signal in the oral cavity corresponding most of the time to the skin associated with the lip (Fig. 2D and D 2 ). A similar analysis performed on mice infected by needle gavage with the strain Y. enterocolitica WA-lux led to the same conclusions, where the BLI signal was detected at the laryngopharynx region, draining LNs and skin from the oral cavity (Fig. S4).
From these results, we speculate that the use of a feeding needle entails a risk of creating lesions in the skin of the mouth as well as in the tissue at the junction of the esophagus and trachea, from which pathogenic Yersinia spp. can disseminate to and multiply in the draining LNs located in the salivary gland region.
Although we do not know the impact of the neck colonization by Y. enterocolitica and Y. pseudotuberculosis in the overall infectious process, we cannot exclude the possibility of bacterial dissemination in the bloodstream occurring independently of the intestinal colonization and translocation at the gastrointestinal barrier.
Bread feeding protects Y. pseudotuberculosis from the acidic gastric environment and promotes efficient intestinal colonization. We next compared the kinetics of bacterial dissemination upon bread feeding versus needle feeding. Mice were orally infected with 3.5E8 CFU of IP32953-lux and imaged at 0.5, 6, 24, 48, and 72 h postinfection ( Fig. 3). At each time point, photon emission was measured from the abdominal region and neck region. As indicated in Fig. 3A and B, upon bread feeding, the BLI signal in the abdominal region is detected as early as 30 min postinfection; it then decreases to an almost undetectable signal at 6 hours postinfection (h p.i.) and in a second phase increases continuously. Upon needle feeding, the BLI signal is barely detectable before 24 h p.i.; after that time point, the BLI signal is detected in the neck ( Fig. 3A and D) and the abdominal region ( Fig. 3A and B), where it continuously increases over time. It is noteworthy that the neck region signal was never detected in bread-infected animals ( Fig. 3D) even at later time points. Statistical analysis indicates a significantly higher light emission signal in the abdominal region at early (0.5 h) and later (72 h) time points when using bread than when using a needle (Fig. 3B). This observation correlates with the smaller amount of Y. pseudotuberculosis found in the , P Ͼ 0.05; **, P Ͻ 0.0015) (a), and for nonparametric Wilcoxon matched-pairs signed-rank test (ns, P Ͼ 0.05; **, P Ͻ 0.003) (b). The median of the values is indicated by a horizontal bar. Bioluminescence signal is detected in the neck of 80% of the needle-infected animals, with an average radiance not statistically different than the signal from the abdominal region, whereas none of the mice infected with bread presented a signal in the neck. Although the median of the abdominal region is higher when mice are infected with bread, the overall signal is not statistically different due to the variability between mice. feces 6 h post-needle feeding, as shown by CFU counting (Fig. 3C), suggesting that a lower number of bacteria survive passage through the stomach.
When needle feeding is used, bacteria are delivered directly to the stomach, a compartment known to have an acidic pH (between 3 and 4) (22) that is aggressive to microorganisms. To evaluate whether the smaller amount of light observed in the abdominal region of needle-infected mice was due to bacterial killing by the acidity of the stomach, we infected animals using the needle or bread with either bacterium resuspended in PBS only or in PBS buffered with CaCO 3 (23). As shown in Fig. 4A and B, buffering the bacterial inoculating medium significantly increases the BLI signal in the abdominal region, as well as the amount of bacterial CFU recovered in the feces (Fig. 4C) of mice infected with the needle but not with the bread (Fig. S5). Our results therefore indicate that bacterial bread delivery protects Y. pseudotuberculosis from the acidic gastric environment and promotes efficient intestinal colonization without the requirement of an additional buffering compound. CFU of IP32953-lux) were euthanized using CO 2 , and a dissection procedure associated with BLI acquisition was performed step by step. Two representative mice are shown. After a first step of skin removal, the salivary gland region was dissected, and lymph nodes (LN1 to LN4) found in this region were collected. Then, the esophagus and trachea section were collected as indicated by the schematized red scissors. Regions of interest (ROI) were drawn, and the average bioluminescence (photons/s/cm 2 /sr) was calculated for each dissected organ. Organs exhibiting BLI signal were homogenized in PBS and enumerated for bacterial count (CFU). (C 1 ) Lymph nodes LN1, 2E5 ROI, 2E4 CFU; and LN2, 5.4E4 ROI, 4E3 CFU. (C 2 ) Esophagus, 2E6 ROI, 4E5 CFU; and trachea, 6E6 ROI, 6E5 CFU. (D 1 ) Lymph nodes LN3, 2.5E7 ROI, 4E6 CFU; and LN4, 1E6 ROI, 4E5 CFU. (D 2 ) Skin from the mouth, 9.3E5 ROI, CFU not done. Among 10 dissected mice, nine exhibited bioluminescent signals in lymph nodes from the salivary gland region (C 1 and D 1 ), 7 in the esophagus and/or trachea (C 2 ), and five close to the mouth/lip region (D 2 ).

Bread Feeding To Study Enteropathogens Infection and Immunity
Bread feeding is a more physiological oral administration method to study Y. pseudotuberculosis infection. The better survival of Y. pseudotuberculosis when associated with CaCO 3 prior to needle feeding led us to evaluate whether it would change the overall 50% lethal dose (LD 50 ) of strain IP32953-lux when administered with bread. Thus, mice were infected using three conditions (needle with or without supplementation of CaCO 3 and bread) with four different Y. pseudotuberculosis IP32953-lux concentrations (2.5E5, 2.5E6, 2.5E7, and 2.5E8 CFU). The doses were chosen where the lowest and the highest doses are expected to kill 0% and 100% of the mice, respectively. Animals were monitored for lethality up to 21 days postinfection to allow measurements of LD 50 . The use of the needle without supplementation of CaCO 3 gave the highest LD 50 (5.2E6 CFU). When bacteria are mixed with CaCO 3 , the LD 50 is lower (1.8E6 CFU), suggesting a protective effect of CaCO 3 when the needle-feeding administration is used. However, the lowest calculated LD 50 was obtained when mice were infected with bread (LD 50 , 9.5E5). Even though the statistical comparison (log rank test) of the mouse survival did not reveal significant differences among the three bacterial administration routes (Fig. S6), measurement of the abdominal BLI signal at the lowest infectious dose (2.5E5 bacteria, Fig. 5) shows a clear difference between needle PBS (no increase in the abdominal signal) versus bread and needle PBS plus CaCO 3 treatments (significant increase in the abdominal signal). Interestingly, this difference in the increase of the abdominal BLI signal between treatments is not observed any more at a higher infectious dose (2.5E7, Fig. 6), suggesting that the bread and the needle PBS plus CaCO 3 treatments allow a more sensitive monitoring of bacterial infection progression at low infectious doses (Fig. S6). It is noteworthy that even at low bacterial concentrations (2.5E5 CFU), needle feeding (with or without CaCO 3 ) leads to neck colonization, whereas none of the mice showed a neck BLI signal upon bread feeding (Fig. 5). Our results therefore suggest that infection using bread feeding is more physiological for Y. pseudotuberculosis delivery to the intestinal tract, resulting in an efficient colonization even at low inoculum concentrations.

DISCUSSION
When studying host pathogen interactions, the choice of a laboratory experimental model of infection is crucial. Besides the choice of the animal species and the pathogen to study, important parameters that have to be taken into account include the control of the dose of the infectious agent administered to each animal, the rapidity to handle the animal, and the stress induced by the animal manipulation. For decades, laboratories interested in studying animal oral infection have used needle feeding to orally deliver infectious agents. Although needle feeding allows control of the delivered dose, there are drawbacks, such as the requirement of a specific manipulator training, the use of anesthesia when animals are particularly agitated, or the stress induced by the handling of the animal. In the present work, we show that bread feeding allows a

Bread Feeding To Study Enteropathogens
Infection and Immunity nontraumatic pathogen administration, where animal handling is minimized and therefore is less stressful for both the manipulator and the animal. We found that the habituation step to feed on bread few days prior to the infection is crucial for effective bread feeding, with no need of additional melted butter as a vehicle as proposed by others (7,8). Bread feeding constitutes a very good administration method since it allows control of the dose administered and is a fast procedure, where all bread is generally consumed by mice in less than 5 min, as opposed to the drinking water method (within several hours) described by others (10). Importantly, in this study, we demonstrate that the use of needle gavage induces damage to the oral cavity of infected animals, leading to pathogen colonization in tissues and organs distant from the intestinal tract. Although it is not known whether such artificial colonization affects the intestinal infection process, it is reasonable to question its impact on the overall host immune response. For example, Barnes and collaborators have reported Y. pseudotuberculosis translocation to organs such as the liver and spleen shortly after oral inoculation with a feeding needle (16). Although the authors did not mention neck injury, it is possible that some lesions caused by the needle participated in this early translocation. In this context, it is important to mention that our results suggesting no significant differences in the LD 50 between mice infected via needle versus bread feeding may potentially mask important differences in organ colonization.
Our study also shows that bacteria directly delivered in PBS via needle feeding are susceptible to the acidic environment of the stomach. Even though this susceptibility can be rescued by buffering the bacterial suspension with CaCO 3 , this infection method OF1 mice were orally infected with 2.5E5 CFU of Y. pseudotuberculosis strain IP32953-lux and monitored over time for BLI signal after needle feeding of bacteria resuspended in PBS (black), in PBS supplemented with CaCO 3 (green), or after bread feeding of bacteria resuspended in PBS only (red). Quantification of the bioluminescent signal from the abdominal region for each individual mouse is shown below the mouse pictures under each condition at each time point (1, 2, 3 and 7 days p.i.). A linear mixed model was used to test for the differences in the BLI signal evolution (on the log 10 scale) between the three protocols of administration. A mean slope indicated as a blue dotted line of the BLI signal evolution was calculated with the linear mixed model for each protocol. The comparative analysis of the mean slopes indicates significant statistical differences in the BLI signal evolution between needle feeding with bacteria in PBS supplemented with CaCO 3 and needle feeding with bacteria in PBS only (P ϭ 0.012), and between bread and needle feeding with bacteria in PBS only (P ϭ 0.03). Red dotted lines indicate the average radiance of ROI measurements above which animals are in terminal illness. Bread feeding allows an efficient infection of the abdominal region by Y. pseudotuberculosis without neck colonization compared to the needle feeding. still produces artificial neck colonization. On the contrary, bread feeding provides bacterial protection against the stomach acidic environment and fully eliminates colonization of the neck region. Overall, our results encourage us to reinvestigate the infectious process by Y. pseudotuberculosis and Y. enterocolitica employing bread feeding, a very robust and more physiological administration method that can be potentially extended to other enteropathogen infection models.

MATERIALS AND METHODS
Culture conditions and bacterial strain construction. Bacteria were grown at 28°C in lysogeny broth (LB) or on LB agar (LBA) plates. For recombinant bacterial growth, control strains were also grown at 28°C in minimal medium containing 1ϫ M9 salts, 2 mM MgSO 4 , 0,1 mM CaCl 2 , 1% glucose, and 1% Casamino Acids. Growth of parental and recombinant lux strains in liquid culture was analyzed by diluting overnight cultures to fresh medium after being washed three times in PBS when minimum medium was used. Bacterial concentrations were evaluated by spectrometry at 600 nm and plating on LBA. Kanamycin (Km; 30 g/ml) or Irgasan (0.1 g/ml) was added to the media when necessary.
The fully virulent serotype 1b Y. pseudotuberculosis IP32953 strain, isolated from a human stool sample in France, and the biotype 1B serotype O:8 Y. enterocolitica WA strain, isolated from the blood of a human patient, were used as parental strains to generate their bioluminescence-expressing derivative strains, IP32953-lux and WA-lux, respectively. As previously described (21), the Photorhabdus luminescens luxCDABE operon controlled by the rplN constitutive Yersinia promoter was introduced into the Yersinia chromosome via the Tn7 technology (20). Plasmid pUCR6K-mini-Tn7-Km r -luxCDABE and the transposaseencoding plasmid pTNS2 were coelectroporated to IP32953 or conjugated to WA cells. Animal experiments. Female 7-week-old OF1 mice were purchased from Charles River France and allowed to acclimate for 1 week before infection. Prior to all infections (bread or needle), mice were fasted for 16 h and had continuous access to water. Only bread-fed mice were trained to feed on bread prior to the infection. All oral infections were performed on nonanesthetized animals. Serial dilution of bacterial suspensions was performed in phosphate-buffered saline (PBS without CaCl 2 and MgCl 2 ) from cultures grown for 48 h at 28°C in LB agar plates.
For oral gavage, mice were administered a 0.2-ml bacterial suspension using an animal feeding stainless steel bulbous-ended needle (0.9 mm by 38 mm, 20 G by 1.5", catalog no. 9921; Cadence Science). The bulbous-ended needle was inserted over the tongue into the esophagus and stomach, as previously described (23). When required, the 0.2-ml bacterial suspension used for oral gavage (feeding needle) was mixed with 0.3 ml of a 50-mg ml Ϫ1 suspension of CaCO 3 in PBS without CaCl 2 and MgCl 2 , and the 0.5-ml mixture was administered. Since CaCO 3 is not soluble at this concentration, the bacterial suspension was mixed with CaCO 3 at once before each feeding-needle injection.
For bread feeding, mice were first adapted to feed on bread prior to the infection. Thus, 3 days before infection, the food was replaced by small pieces of white bread (approximately 9 mm 2 ) to allow mice to feed on bread for a 2-h period. The same bread adaptation was repeated once 24 h before infection. Then, 16 h prior to the infection, the food was removed, and mice were fasted with access to water. A 20-l bacterial suspension (with or without supplementation of CaCO 3 ) was deposited on one piece of bread and placed in an empty and clean cage, where one mouse was introduced. Each mouse was visually monitored until complete bread feeding. Generally, bread feeding took from 30 s up to 10 min per mouse. After feeding, animals were housed in a cage with new litter and access to food and water ad libitum.
After infection, animals were monitored daily for 21 days, and every day, the litter was renewed in order to limit accumulation of feces in the cage and avoid cross-contamination between mice.
BLI imaging and dissection. In vivo imaging was performed with the In Vivo Imaging System (IVIS 100; Caliper Life Sciences). Animals were anesthetized using a constant flow of 2.5% isoflurane mixed with oxygen. Images were acquired with binning of 4 and an exposure time from 10 s to 2 min, according to the signal intensity. To quantify the luminescence signal, regions of interest (ROI) were drawn, and measurements of the ROI are given as average radiance (photons/s/cm 2 /steradian [sr]). Uninfected mice were used to set the light emission background.
Analysis of the cervical region was performed sequentially starting with the removal of the skin on euthanized animals and imaging, followed by removal of the most intense bioluminescent tissue and imaging again of the animal to identify a lower intensity signal. When signal was still detected after removal of the first identified bioluminescent tissue, the same sequence was repeated until no signal could be detected. Organs and tissues were aseptically removed, placed in glass beads containing tubes of 0.5 ml PBS, and subjected to homogenization to determine bacterial loads. Feces were collected from live mice and were homogenized in PBS using disposable homogenizers (Kimble Chase piston pellet; Fisher Scientific), and serial dilutions were performed to determine bacterial loads.
Statistical analysis and LD 50 calculation. Statistical analyses were performed with nonparametric Mann-Whitney t test and with the GraphPad Prism 5.0 software (San Diego, CA, USA). For comparison of BLI signals within the same animal, data were analyzed with a nonparametric Wilcoxon matched-pairs signed-rank test. Survival curves were compared two by two using a log rank (mantel-Cox) test. A linear mixed model was fitted using the nlme R package to test for the difference of the BLI signal evolution (on the log 10 scale) between the three protocols of administration (bread feeding or needle feeding with or without CaCO 3 ). The mouse effect was included as random so that a random intercept and slope were attributed for each animal. The calculated mean slopes were compared, and P values of Յ0.05 were considered significant.
To determine the 50% lethal dose (LD 50 ), mice (six to seven per dose) were infected with 10-fold serial dilutions of bacterial suspensions (from 2E5 to 2E8 CFU) and were monitored for 3 weeks. The LD 50 was calculated using the Reed and Muench method (24), where the survival (y 1 ; . . .; y i ; . . . ; y n ) and death (z 1 ; . . . ; z i ; . . . ; z n ) numbers for each dose (

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.