Yersinia enterocolitica Infection of Mice Reveals Clonal Invasion and Abscess Formation

Abstract

Yersinia enterocolitica is a common cause of food-borne gastrointestinal disease leading to self-limiting diarrhea and mesenteric lymphadenitis. Occasionally, focal abscess formation in the livers and spleens of certain predisposed patients (those with iron overload states such as hemochromatosis) is observed. In the mouse oral infection model, yersiniae produce a similar disease involving the replication of yersiniae in the small intestine, the invasion of Peyer's patches, and dissemination to the liver and spleen. In these tissues and organs, yersiniae are known to replicate predominately extracellularly and to form microcolonies. By infecting mice orally with a mixture of equal amounts of green- and red-fluorescing yersiniae (yersiniae expressing green or red fluorescent protein), we were able to show for the first time that yersiniae produce exclusively monoclonal microcolonies in Peyer's patches, the liver, and the spleen, indicating that a single bacterium is sufficient to induce microcolony and microabscess formation in vivo. Furthermore, we present evidence for the clonal invasion of Peyer's patches from the small intestine. The finding that only very few yersiniae are required to establish microcolonies in Peyer's patches is due to both Yersinia-specific and host-specific factors. We demonstrate that yersiniae growing in the small intestinal lumen show strongly reduced levels of invasin, the most important factor for the early invasion of Peyer's patches. Furthermore, we show that the host severely restricts sequential microcolony formation in previously infected Peyer's patches.

Yersinia enterocolitica is a common cause of food-borne gastrointestinal disease in the moderate and subtropical climates of the world. Y. enterocolitica infection may present as enteritis, terminal ileitis, or mesenteric lymphadenitis (pseudoappendicitis) with watery or sometimes bloody diarrhea. In patients with severe underlying diseases, such as hemolytic anemia and hemochromatosis, Yersinia may cause septicemia with focal abscess formation in the liver and spleen (3). In the mouse oral infection model, a similar disease results, with yersiniae replicating in the small intestine (SI), invading Peyer's patches (PPs) of the distal ileum, and disseminating to the liver and spleen. In these tissues and organs, yersiniae replicate predominately extracellularly and form microabscesses (32, 35). Extracellular replication is made possible by the type III secretion system-mediated injection of Yersinia outer proteins (Yops) that paralyze phagocytes of the innate immune system (reviewed in reference 15). Mechanisms that lead to Yersinia microabscess formation have not been described.

The invasion of PPs is made possible by the expression of several nonfimbrial adhesins, such as invasin (Inv) and possibly Yersinia adhesin A (YadA), that can potentially interact with β1 integrins and may mediate adherence to and the invasion of M cells (reviewed in references 15 and 17). Inv directly interacts with β1 integrins of host cells (18, 22), whereas YadA interacts with extracellular matrix proteins such as collagen and fibronectin as well as host cell integrins by extracellular matrix bridging (9, 16). M cells but not enterocytes express β1 integrins on their apical surfaces (5). Therefore, the invasion of PPs is believed to occur via M cells overlying the follicle-associated epithelium. In vivo, Inv is the predominant invasion factor of Yersinia and has been shown to be essential for the early invasion of PPs in the mouse oral infection model (23, 27). Inv of Y. pseudotuberculosis is composed of five globular domains (D1 to D5) that protrude 18 nm from the bacterial surface (11). Inv of Y. enterocolitica is missing the D2 self-association domain. The two C-terminal surface-exposed domains of Inv form an adhesion unit that is responsible for high-affinity interaction with β1 integrins.

The supposed preferential invasion of M cells by Y. pseudotuberculosis is supported by detailed microscopic evidence (6). M cells were frequently found to carry multiple adherent and invading yersiniae in the mouse ligated gut loop model, suggesting the translocation of many bacteria to the submucosal tissue (5). After translocation across the mucosal barrier by M cells, yersiniae disseminate from PPs to mesenteric lymph nodes (10, 13, 32). Further dissemination to the spleen and liver probably does not occur via PPs and lymph nodes. It was recently shown that organized intestinal lymphoid tissue is dispensable for the dissemination of Y. enterocolitica to internal organs (12), and Y. pseudotuberculosis colonization of the spleen and liver was shown to be derived from the gut lumen but not mesenteric lymph nodes (2). The precise mechanism of dissemination from the gut lumen to the liver and spleen is, however, obscure at present. Initially, microabscesses formed by Y. enterocolitica in livers and spleens of mice consist primarily of neutrophils (1, 4). During later stages of mouse infection, lesions are populated by mononuclear cells and exhibit a granulomatous character (1). In the present work, we have studied abscess formation in the mouse oral infection model by using yersiniae that express either red fluorescent protein (RFP) or green fluorescent protein (GFP). We were able to show that Y. enterocolitica infection of mice leads to monoclonal microcolony formation in PPs, spleens, and livers. Furthermore, we present evidence for the clonal invasion of PPs from the gut lumen and demonstrate that both Yersinia and the host contribute to this phenomenon.

MATERIALS AND METHODS

Bacterial strains and plasmids.

Y. enterocolitica WA-314 is a clinical isolate of serotype O:8 (14). WA-C is a plasmidless derivative of WA-314 (14). WA-C(pYV-kan) harbors a kanamycin resistance cassette in the noncoding region of the pYV plasmid between the YadA gene and the insertion sequence Yen1 and is as virulent as WA-314. WA-C-inv is an invasin mutant that has been described previously (30). Yersiniae were cultured aerobically in Luria-Bertani (LB) broth, on LB agar plates (Difco Laboratories), or on the Yersinia-selective agar CIN (Beckton Dickinson) at 27°C. Kanamycin and chloramphenicol at concentrations of 25 and 20 μg/ml, respectively, were used for plasmid selection. Constitutively RFP- and GFP-expressing bacteria were constructed by transforming WA-C(pYV) with the low-copy-number vector pACYC184 (New England Biolabs) harboring a translational fusion between the lac promoter and the gfp or rfp gene. The gfp and rfp DNA fragments were generated by PCR amplification using pGFPmut2 (8) and pDsRedexpress (BD Clontech) as templates. In each case, the lac promoter was cloned into pACYC184 by using the HindIII and BamHI restriction sites and the rfp or gfp gene was cloned by using the BamHI and SalI restriction sites. The pACYC184 plasmids harboring pGFPmut2 gfp and pDsRedexpress rfp under the lac promoter are referred to as pLACGFP and pLACRFP.

Oral infection of mice.

Six- to 8-week-old female BALB/c mice (Harlan Winkelmann) or oxidative-burst-deficient p47^phox−/− mice (19) were kept under specific-pathogen-free conditions (in a positive-pressure cabinet) and were provided with food and water ad libitum. Mice were infected orally with yersiniae from frozen stock suspensions. These suspensions were prepared by growing bacteria to stationary phase in LB medium at 27°C and freezing the bacteria in 15% glycerol at −80°C. After appropriate dilutions, bacteria were washed twice with phosphate-buffered saline and mice were fed 15 μl by using a microliter pipette. Mice were subjected to fasting for 16 h prior to the oral infection. The actual administered dose was determined by plating serial dilutions onto Mueller-Hinton agar for 36 h at 27°C. Mice were sacrificed by CO₂ asphyxiation, and SIs, PPs, spleens, and livers were aseptically removed. The levels of colonization of mouse organs and small intestinal lumina were determined as described previously (33). All mouse experiments were approved by government authorities (Regierung von Oberbayern). The statistical significance of data was evaluated with an unpaired two-tailed Student t test for quantitative data on microcolonies. A two-tailed Mann-Whitney test was used for colonization (CFU) data. P of ≤0.05 was considered significant.

Depletion of granulocytes.

To deplete mice of granulocytes, 0.25 mg of monoclonal antibody (MAb) RB6-8C5 (BD Pharmingen) was injected intraperitoneally 1 day before oral Yersinia infection as described previously (7). Control mice were injected with normal rat immunoglobulin G (Sigma). The depletion of granulocytes was verified by determining the total leukocyte count in tail vein blood and by determining differential leukocyte counts in smears of whole blood stained according to the method of Pappenheim (16a). To verify the depletion of granulocytes in PPs, RB6-8C5-treated and control mice were infected with 10⁹ yersiniae. Three days after oral infection, cryosections of PPs were immunostained with rat anti-mouse Ly6C/G antibody (Caltag Laboratories) and observed under a fluorescence microscope.

Cryosection preparation, immunohistochemical staining, and fluorescence microscopy.

All visible PPs and the liver and spleen from each mouse were embedded in Tissue-Tek (Sakura Finetek) and shock frozen in liquid nitrogen. Cryosections of 10 μm in thickness were prepared using a Leica cryomicrotome CM3050. These sections were mounted on SuperFrostPlus slides (Menzel) and covered with Fluoprep (Biomerieux) and a coverslip. Some cryosections were stained with 1 μg/ml DAPI (4′,6′-diamidino-2-phenylindole; Sigma-Aldrich) or immunostained with rat anti-mouse Ly6C/G antibody (Caltag Laboratories). Immunostaining was performed as described previously (25). For the detection of yersiniae by immunofluorescence, cryosections were fixed with 3.7% paraformaldehyde and incubated for 30 min in 2% bovine serum albumin to block nonspecific binding. Bacteria were stained with a primary monoclonal YadA antibody and a fluorescein isothiocyanate- or tetramethyl rhodamine isocyanate-labeled secondary antibody as previously described (20). Composite images of PPs, livers, and spleens were automatically assembled using a motorized Olympus BX61 fluorescence microscope with analysis software (Olympus Soft Imaging System). For the quantitative analysis of microcolonies in PPs, all visible PPs were harvested and completely sectioned. Every 50 μm, a 10-μm section was saved for analysis. A microcolony was defined as a clustered community of bacteria growing in tissue. Clusters of yersiniae fluorescing the same color, along with smaller satellite clusters of the same color, were assessed as a single microcolony.

RESULTS

Construction of red- and green-fluorescing yersiniae and stability of strains in vivo.

To study invasion by and the dissemination of Y. enterocolitica in the oral mouse model and to characterize microcolonies (polyclonal versus monoclonal), we labeled bacteria with plasmid pLACRFP or pLACGFP encoding RFP or GFP, respectively. To determine if these plasmids were sufficiently stable in vivo, mice were infected orally with 10⁹ CFU of fluorescing bacteria. Five days postinfection (p.i.), mice were sacrificed and tissue homogenates were prepared and plated. Of the bacteria recovered from intestinal lavage fluids, PPs, spleens, and livers, 99.9% grew as either green- or red-fluorescing bacteria, indicating that they had not lost the plasmids. To determine if the presence of pLACGFP or pLACRFP would lead to the attenuation of yersiniae, we compared bacterial burdens in the spleens and PPs of infected mice 5 days p.i. This experiment revealed no significant difference in bacterial burden between mice infected with the wild-type strains (mean log numbers of CFU ± standard deviations: spleen, 6.7 ± 0.6; PPs, 6.0 ± 0.6) and those infected with plasmid-harboring strains (mean log numbers of CFU ± standard deviations: spleen, 6.1 ± 0.5; PPs, 6.2 ± 0.7). To determine if yersiniae no longer expressing RFP or GFP were present in microcolonies, we also immunostained cryosections of Y. enterocolitica-infected mouse organs 5 days p.i. with anti-YadA rabbit antiserum followed by an anti-rabbit fluorescein isothiocyanate- or tetramethyl rhodamine isocyanate-conjugated antibody as previously described (20). This immunostaining revealed that all yersiniae and microcolonies in the tissue sections were detectable by RFP or GFP fluorescence (data not shown), indicating the reliable expression of the reporter genes and the maintenance of the reporter plasmids during the infection experiments.

Yersiniae form monoclonal microcolonies in mice.

To determine if Yersinia microabscesses in mouse organs originate from a single bacterium (monoclonal) or if microabscess formation requires many yersiniae, we infected female BALB/c mice (6 to 8 weeks old; Harlan Winkelmann) orally with a mixture of equal amounts of Y. enterocolitica expressing GFP and Y. enterocolitica expressing RFP. Mice received a dose of 10⁹ red- and green-fluorescing bacteria and were sacrificed 1, 2, and 5 days p.i. Subsequently, 10-μm cryosections of all PPs, livers, and spleens were prepared (each PP and the organs were completely sectioned), stained with DAPI, and examined under a fluorescence microscope. A microabscess is an area of localized suppurative inflammation produced by the seeding of pyogenic bacteria into tissue. Since we are considering fluorescing bacteria, it is probably more correct to talk about microcolonies than microabscesses, at least for the very early stage of infection. A microcolony would be defined as a clustered community of bacteria growing in tissue. As can be seen in Fig. 1, 2, and 3, microcolonies in PPs, livers, and spleens fluoresced either green or red, indicating clonal origin. Mixed red- and green-fluorescing microcolony populations were not observed in any organ or tissue studied. At 6 and 12 h p.i., we were not able to detect yersiniae or microcolonies in PPs by fluorescence microscopy. At 1 day p.i., microcolonies in PPs were tiny and located close to the dome region (Fig. 1A and B). Over the course of 5 days, microcolonies increased in size and spread over the entire PP, with different monoclonal microcolonies touching but not mixing (Fig. 1A to K). On day 5, microcolonies ulcerating into the gut lumen were seen (Fig. 1I and J). Microcolonies in the PP-free region of the SI were not observed. Interestingly, the numbers of microcolonies seen in each PP were very small (one to four), whereas hundreds of monoclonal microcolonies per 10-μm cross section of the spleen (Fig. 2) and liver (Fig. 3) were seen. These microcolonies were distributed homogeneously throughout the organs, without a preferential location. In addition, we observed approximately the same numbers of red and green microcolonies in each organ, indicating that the reporter proteins did not differentially affect the pathogenesis of yersiniae. Infecting mice via the intraperitoneal or intravenous route with green- and red-fluorescing Yersinia resulted in the same pattern of clonal microcolonies in spleens and livers as that observed after oral infection (data not shown).

FIG. 1.

Typical cryosections (10 μm) of PPs from BALB/c mice infected orally with a mixture of 10⁹ CFU of red- and green-fluorescing Y. enterocolitica. Panel A shows one red-fluorescing monoclonal microcolony (arrow) on day 1 p.i. Panel B shows a magnification of the microcolony seen in panel A. Panels C, E, and G show three consecutive cryosections of a PP on day 2 p.i. Two monoclonal microcolonies in this PP can be seen. Panels D, F, and H show magnifications of the microcolonies seen in panels C, E, and G, respectively. Panel I shows yersiniae supprating into (arrow) the gut lumen on day 5 p.i. Panel J is a magnification of the green-fluorescing microcolony seen in panel I. Panel K shows two large monoclonal microcolonies that are touching but not mixing (arrow) on day 5 p.i. Panel L shows a closeup view of a single green-fluorescing bacterium (arrow) in a typical green-fluorescing monoclonal microcolony.

FIG. 2.

Typical cryosection (10 μm) of the spleen from a BALB/c mouse infected orally with a mixture of 10⁹ CFU of red- and green-fluorescing Y. enterocolitica for 5 days. (A) Hundreds of red- and green-fluorescing monoclonal microcolonies throughout the spleen can be seen. The inset shows a cryosection of an uninfected spleen. A magnification of a typical microcolony is shown in panel B. Bottom panels C and D show single fluorescing bacteria from typical red- and green-fluorescing microcolonies, respectively.

FIG. 3.

Typical cryosection (10 μm) of the liver from a BALB/c mouse infected orally with a mixture of 10⁹ CFU of red- and green-fluorescing Y. enterocolitica for 5 days. (A) Hundreds of red- and green-fluorescing monoclonal microcolonies throughout the liver can be seen. The inset shows a cryosection of an uninfected liver. A magnification of typical microcolonies is shown in panel B. Panel C shows single red-fluorescing bacteria from the red-fluorescing microcolony.

Clonal invasion of PPs.

The observation of only one to four monoclonal microcolonies per PP suggested that only very few yersiniae invaded the gut epithelium and established monoclonal microcolonies in PPs. To analyze this suggestion in more detail, groups of five mice were orally infected with increasing bacterial doses ranging from 10⁶ to 10⁹ CFU. Mice were sacrificed on day 4 p.i., and cryosections of every macroscopically visible PP were analyzed using a fluorescence microscope. The total numbers of monoclonal red and green microcolonies in each PP were determined. Clusters of yersiniae fluorescing the same color, along with smaller satellite clusters of the same color, were assessed as a single microcolony. For quantification, the entire PP was sectioned and every 50 μm a 10-μm section was saved for microscopy. At an infection dose of 10⁶ CFU (below the 50% lethal dose), most PPs did not reveal any microcolonies at all and PPs with abscesses were observed only sporadically (Fig. 4A). As the infection dose was increased from 10⁶ to 10⁹ CFU, there was an increase in the percentage of PPs with microcolonies (Fig. 4F). Only at a very high dose (10⁹ CFU, above the 50% lethal dose) did all visible PPs actually reveal Yersinia microcolonies (Fig. 4D and F). The increases in the infection dose from 10⁷ to 10⁸ CFU and from 10⁸ to 10⁹ CFU were reflected by a significant increase in the number of yersiniae recovered from the small intestinal lumina on day 4 p.i. (P < 0.01) (Fig. 4G). Increasing the infection dose by 2 logs led to a corresponding 2-log increase in SI colonization (P < 0.01) within 24 h p.i. (Fig. 5). Later during the infection (between days 2 and 5), this difference decreased to about 1 log (P ≤ 0.01). These results suggest that a very high initial inoculum is necessary for yersiniae to reach and interact with the M cells overlying PPs in the mouse model and that the invasion of PPs by yersiniae in vivo is a rare event. Furthermore, it was evident that as the infection dose was increased logarithmically from 10⁶ to 10⁹ CFU, the number of monoclonal microcolonies per PP increased only linearly (Fig. 4E). The mean number of monoclonal microcolonies ± the standard deviation per infected PP increased only slightly, from 1.3 ± 0.5 (after infection with 10⁶ CFU) to 2.6 ± 1 (after infection with 10⁹ CFU) microcolonies per infected PP. Even at the high infection dose of 10⁹ CFU, most PPs revealed only between two and four microcolonies. Of course, many yersiniae may have invaded these PPs, with only a few surviving the initial encounter with the innate immune system. If this was the case, we would, however, expect a logarithmic increase in the number of microcolonies per PP with a logarithmically increasing infection dose.

FIG. 4.

Numbers of red (○) and green (•) monoclonal microcolonies in PPs of BALB/c mice orally infected with a 1:1 mixture of green- and red-fluorescing yersiniae. Mice were infected with 10⁶ (A), 10⁷ (B), 10⁸ (C), or 10⁹ (D) CFU and were sacrificed 4 days p.i. ∼ indicates that no microcolonies were observed in that PP. Data for PPs in sequential order from the stomach (1) to the cecum (10) are shown. The numbers of microcolonies per PP ± the standard deviations, the percentages of colonized PPs per mouse ± the standard deviations, and the numbers of yersiniae colonizing the SI ± the standard deviations after the oral infection of BALB/c mice are shown in panels E, F, and G, respectively. Significantly more microcolonies per PP were observed for mice infected with 10⁸ versus 10⁷ CFU (P < 0.01) and 10⁹ versus 10⁸ CFU (P < 0.01) but not for mice infected with 10⁷ versus 10⁶ CFU (P = 0.21). The increase in percentages of infected PPs per mouse is significant only when comparing infections with 10⁸ versus 10⁷ CFU (P < 0.01) and 10⁹ versus 10⁸ CFU (P < 0.05) but not 10⁷ versus 10⁶ CFU (P = 0.21). Differences in the numbers of colonizing yersiniae in the SI after infection with 10⁷ versus 10⁸ CFU (P < 0.01) and 10⁸ versus 10⁹ CFU (P < 0.01) are significant, but the difference in these numbers after infection with 10⁶ and 10⁷ CFU (P = 0.75) is not.

FIG. 5.

Colonization of the small intestinal lumen by Y. enterocolitica over the course of 5 days. Groups of five BALB/c mice were orally infected with either 10⁹ (A) or 10⁷ (B) yersiniae. The bacterial load in the SI was determined by plating serial dilutions of intestinal lavage specimens and is depicted as the log number of CFU ± the standard deviation. The level of colonization of the SI after infection with 10⁹ CFU was significantly higher than that after infection with 10⁷ CFU on all days tested (P ≤ 0.01).

Course of oral Yersinia infection in mice lacking granulocytes.

The results of the experiments described above strongly suggest that only very few yersiniae invaded a given PP. However, it is possible that many yersiniae initially invaded PPs but that only very few survived the initial host response. To address this question, mice were rendered severely granulopenic by treatment with MAb RB6-8C5 (BD Pharmingen) as described previously (7). Three granulopenic mice and three immunocompetent mice were infected orally with a mixture of 10⁹ red- and green-fluorescing yersiniae, and the numbers of monoclonal microcolonies in PPs were determined 2 days p.i. as described above. These experiments revealed similar situations for PPs of granulopenic mice and immunocompetent mice. Thirty-one infected PPs harvested from granulopenic mice showed an average of 3.5 microcolonies per PP, whereas 26 PPs from immunocompetent mice showed an average of 3.9 microcolonies per PP. To demonstrate the lack of granulocytes in PPs of mice treated with MAb RB6-8C5, cryosections of PPs from these mice were stained with rat anti-mouse Ly6C/G (data not shown). If many yersiniae were to invade PPs but only very few were to survive due to an influx of granulocytes, which are known to be important in early killing of yersiniae (7), we expect an increase in the number of microcolonies seen in PPs of granulopenic versus immunocompetent mice.

Yersinia infection in oxidative-burst-deficient p47^phox−/− mice.

To further support our hypothesis, three oxidative-burst-deficient p47^phox−/− mice (19) were orally infected with a mixture of 10⁹ CFU of red- and green-fluorescing yersiniae in equal amounts. Four days postinfection, mice were sacrificed and microcolonies in PPs were analyzed as described above. This experiment revealed a situation similar to that for wild-type mice. A total of 21 PPs were harvested from the three mice. Seventeen PPs showed abscesses and revealed on average 2.12 ± 0.93 monoclonal microcolonies per PP. Three C57BL/6 mice used as controls revealed 2.33 ± 0.76 abscesses per PP (P = 0.77). This result shows that the potential early killing of yersiniae by phagocyte oxidative burst cannot explain the phenomenon of clonal abscess formation in PPs.

Invasin levels in yersiniae recovered from the small intestinal lumen.

In vitro, yersiniae are known to express Inv at ambient temperatures but to down-regulate Inv expression at the host temperature of 37°C (26). One obvious bacterial mechanism leading to clonal invasion may therefore be a lack of Inv expression in the small intestinal lumen. If this were the case, the early (Inv-dependent) invasion of PPs by Yersinia would be limited to a short time frame after infection (during which invasin is still present on the bacterial surface). To determine if Inv is expressed in vivo in the SI, 10 mice were infected with 10⁹ yersiniae and 5 mice each were sacrificed on days 2 and 5 p.i. Small intestinal lavage specimens from five mice were pooled and filtered with 5-μm-pore-size Durapore filters (Millipore) to remove particles. Serial dilutions were plated onto CIN selective agar to determine the concentration of yersiniae in the intestinal lavage fluid. In order to detect Inv, Western blotting with 3 × 10⁶ bacteria from frozen stock suspensions was performed using a polyclonal rabbit anti-Inv antibody. As can be seen in Fig. 6, Inv is strongly expressed by stationary-phase yersiniae grown at 27°C in LB medium (those used to infect mice orally) [Fig. 6, lane labeled “WA-C(pYV-kan)”] but is not detectable by Western blotting in bacteria washed from the small intestinal lumina of mice on day 2 [Fig. 6, lane labeled “WA-C(pYV-kan) 2 days p.i.”] or day 5 [Fig. 6, lane labeled “WA-C(pYV-kan) 5 days p.i.”]. An equivalent amount of intestinal lavage fluid from uninfected mice and an isogenic inv-deficient Y. enterocolitica mutant (30) grown at 27°C were loaded as controls into lanes labeled “SI lavage not infected” and “WA-C(pYV)inv−,” respectively. To demonstrate the loading of equal amounts of bacteria, Western blotting with a polyclonal anti-Hsp60 antibody was also performed (21). Hsp60 was chosen since it is expressed by Y. enterocolitica in similar amounts at 27 and 37°C. Therefore, the down-regulation of Inv in the small intestinal lumen may be an important factor contributing to the clonal invasion of PPs.

FIG. 6.

Levels of Y. enterocolitica invasin are strongly reduced in the small intestinal lumen. Western blotting was performed with 3 × 10⁶ CFU of WA-C(pYV) either grown at 27°C in LB medium [WA-C(pYV-kan)] or washed from the SIs of mice 2 days [WA-C(pYV-kan) 2 days p.i.] or 5 days [WA-C(pYV) 5 days p.i.] after oral infection. As a control, an isogenic inv mutant grown at 27°C in LB medium was loaded into one lane [WA-C(pYV-kan)inv−], and a corresponding amount of intestinal lavage fluid from uninfected mice was loaded into another lane (SI lavage not infected). Blotting was performed with rabbit anti-Inv (α Inv) and anti-Hsp60 (α Hsp60) polyclonal antibodies.

Local host response inhibits the sequential infection of PPs.

The finding that only very few yersiniae establish microcolonies in PPs may be interpreted as a feature involving not only Yersinia mechanisms (such as the down-regulation of Inv in the SI) but also the host response to Yersinia infection. Although bacteria replicate in the gut lumen over the course of several days (Fig. 5), our results suggest that even at a very high oral infection dose, not many yersiniae go on to invade PPs. If yersiniae replicating in the gut lumen were to invade PPs continuously, we would expect many more microcolonies per PP after several days of infection. Therefore, we wondered whether yersiniae were able to invade and establish further abscesses in PPs in which yersiniae had already formed abscesses. For this purpose, we sequentially infected groups of 10 mice orally with 10⁹ green-fluorescing yersiniae followed by 10⁹ red-fluorescing yersiniae 2 days later. Mice were sacrificed on day 4, and all macroscopically visible PPs were analyzed for microcolony formation by using a fluorescence microscope as described above (Fig. 7 and Table 1). As a control, three mice were infected with red-fluorescing yersiniae on day 2 only and sacrificed on day 4. These experiments revealed that the initial infection with green-fluorescing yersiniae resulted in the establishment of microcolonies in 79 of 86 PPs taken from 10 mice. Subsequent infection with red-fluorescing yersiniae 2 days later resulted in microcolony formation in 11 of 86 PPs. A closer look revealed that yersiniae from the second inoculation (red) preferentially established abscesses in PPs that lacked abscesses after the initial infection with yersiniae (green). Of initially uninfected PPs from six mice, 100% (7 of 7) exhibited abscesses established by bacteria from the second inoculation, compared to only 5.1% of PPs (4 of 79 PPs from 10 mice) that already had abscesses from the first infection (P < 0.01). To demonstrate that this result was not due to reduced virulence or invasiveness of the red-fluorescing bacteria, we repeated this experiment with a low initial inoculum (10⁷ CFU) of green-fluorescing yersiniae (leaving many PPs without abscesses) followed by subsequent infection with a high inoculum (10⁹ CFU) of red-fluorescing yersiniae 2 days later. In this experiment, 23 out of 86 PPs from 10 mice showed abscesses from the initial infection with green-fluorescing bacteria. Subsequent infection with red-fluorescing yersiniae 2 days later resulted in the infection of 53 out of 86 PPs with red-fluorescing microcolonies. Of the PPs that lacked abscesses formed by green yersiniae, 81% (51 of 63) exhibited red-fluorescing microcolonies, but only 8.7% (2 of 23) of the PPs infected with green-fluorescing yersiniae developed red-fluorescing microcolonies (P < 0.01). The control experiment with three mice infected only on day 2 with red-fluorescing bacteria revealed only that all PPs had abscesses, with an average of 1.99 ± 0.30 red microcolonies/PP. To determine if the inhibition of the sequential infection of PPs was due to an influx of granulocytes, we repeated the sequential infections with 10⁹ green-fluorescing yersiniae followed by 10⁹ red-fluorescing yersiniae 2 days later using granulopenic mice (treated with MAb RB6-8C5 as described above). This experiment revealed a situation similar to that for immunocompetent mice. Red-fluorescing yersiniae from the second inoculation invaded and established abscesses in predominately those PPs that were not infected by the primary inoculum (data not shown). These results indicate that microcolony formation in PPs by yersiniae is inhibited presumably by the local host response (other than granulocytes) to Yersinia infection in only the PPs with abscesses.

FIG. 7.

Numbers of red and green monoclonal microcolonies in PPs of BALB/c mice infected sequentially with green (•)- and red (○)-fluorescing Y. enterocolitica. Mice were orally infected with 10⁹ CFU (B) or 10⁷ CFU (E) of green-fluorescing yersiniae followed by infection with 10⁹ red-fluorescing yersiniae 2 days later. Mice were sacrificed on day 4 p.i. Data for PPs in sequential order from the stomach (1) to the cecum (10) are shown. ∼ indicates that no microcolonies were observed in that PP. Panels A and D illustrate the infection procedure. Panels C and F show the numbers of red (white bars)- and green (black bars)-fluorescing yersiniae in the lumen of the SI on day 4 after sequential infection and control infection with just red-fluorescing bacteria (no preinfection with green-fluorescing yersiniae). The difference between the numbers of colonizing red bacteria after consecutive infection versus a single infection with just red yersiniae as shown in panel C is significant (P < 0.01), and that shown in panel F is marginally significant (P < 0.05).

TABLE 1.

Microcolonies observed in PPs of mice after primary infection with GFP-expressing yersiniae followed 2 days later by infection with RFP-expressing yersiniae

Initial dosea	No. of PPs with any microcolonies/total no. of PPs (%)	No. of PPs with only green microcolonies	No. of PPs with green and red microcolonies	No. of PPs with only red microcolonies	% of PPs with red microcolonies among PPs with green microcolonies (no. with red microcolonies/no. with green microcolonies)	% of PPs with red microcolonies among PPs without green microcolonies (no. with red microcolonies/no. without green microcolonies)
High (109 CFU)	86/86 (100)	75	4	7	5.1 (4/79)	100 (7/7)
Low (107 CFU)	74/86 (86)	21	2	51	8.7 (2/23)	81 (51/63)

^aData for the high-dose experiment correspond to Fig. 7A to C, and those for the low-dose experiment correspond to Fig. 7D to F.

Dynamics of abscess formation in PPs.

Since the host limits further abscess formation in PPs that already have Yersinia abscesses, the number of microcolonies per PP seen after oral infection should not increase over time. To test if this is the case, we infected 10 mice with a mixture of 10⁹ CFU of red- and green-fluorescing bacteria in equal proportions. Five mice each were sacrificed on days 2 and 5 p.i. Cryosections of each PP were analyzed as described above. This experiment revealed the development of abscesses in 84% ± 21% of 43 PPs on day 2, whereas abscesses had developed in 100% of 63 PPs by day 5 (P = 0.12). This result shows that most PPs already have abscesses at an early stage during infection. The number of microcolonies seen per PP remained constant (3.1 ± 0.8 on day 2 versus 2.8 ± 0.4 on day 5; P = 0.38) over this time period (Table 2). These data show that further invasion and microcolony formation by replicating yersiniae from the gut lumen are rare events and support the finding that the host restricts the invasion of previously infected PPs.

TABLE 2.

Comparison of levels of PP colonization by Yersinia 2 and 5 days after oral infection of mice

Day p.i.	No. of PPs harvested from 5 mice	No. (%) of PPs with microcolonies	No. of microcolonies/PP (mean ± SD)
2	43	36 (83.7)	3.1 ± 0.8
5	63	63 (100)	2.8 ± 0.4

DISCUSSION

Y. enterocolitica is a common cause of self-limiting gastroenteritis and mesenteric lymphadenitis. In patients with iron overload states (e.g., hemochromatosis), a systemic infection can result, leading to focal abscess formation in the liver and spleen (3). The mouse oral Yersinia infection model shows a similar course, with the dissemination of yersiniae to PPs of the SI, the lymph nodes, the liver, and the spleen (33). In these tissues and organs, yersiniae are known to replicate predominately extracellularly and to form microcolonies (32, 35). We wondered whether many bacteria are required for microcolony formation or if single yersiniae are sufficient to initiate abscess formation in PPs, the liver, and the spleen. By infecting mice orally with a mixture of equal amounts of red- and green-fluorescing yersiniae, we were able to draw several novel conclusions regarding abscess formation and the dissemination of Y. enterocolitica from the gut to PPs. (i) Microcolonies formed by yersiniae in PPs, the liver, and the spleen after oral infection are exclusively monoclonal, indicating that single yersiniae are able to initiate abscess formation. (ii) Only very few Yersinia cells establish microcolonies in PPs of the SI, due presumably to the clonal invasion of PPs. (iii) The observed reduction in invasin levels in yersiniae growing in the lumen of the SI may contribute to clonal invasion. (iv) The host severely limits the establishment of further Yersinia microcolonies in PPs in which yersiniae have already generated abscesses. (v) Most PPs develop abscesses early during infection, and further invasion and abscess formation by yersiniae growing in the lumen of the SI are rare events.

The main conclusion of this study is that Y. enterocolitica microcolonies in PPs, the liver, and the spleen are monoclonal as evidenced by the fact that individual microcolonies are exclusively single colored after oral infection with a mixture of red- and green-fluorescing bacteria. Two scenarios are conceivable when multiple yersiniae invade a PP or disseminate to the liver and spleen: either multiple bacteria need to associate prior to forming an abscess (Fig. 8A) or single bacteria may be sufficient to initiate abscess formation (Fig. 8B). Here, we demonstrate monoclonal microcolony formation according to the latter scenario (Fig. 8B). Only very few monoclonal microcolonies in any given PP were observed, whereas hundreds of monoclonal microcolonies in cross sections of liver and spleen tissue after 5 days of infection were seen. These results suggest that bacterial dissemination from the gut lumen to the liver and spleen may be much more efficient than the dissemination of bacteria from the gut lumen to PPs. Monoclonal microcolony formation in the liver and spleen may be due to single yersiniae disseminating from the gut lumen to these organs. Single yersiniae may be trapped in the capillary vessels of the liver and spleen, which may subsequently be plugged by the proliferating yersiniae, leading to monoclonal abscess formation. However, it is possible that only very few yersiniae reach the spleen and liver but that these bacteria replicate and disseminate efficiently within the respective organ, as has been suggested for Y. pseudotuberculosis (2). It was demonstrated previously that Y. pseudotuberculosis cells colonizing the spleen and liver are derived from a replicating pool of bacteria in the intestine rather than disseminating via PPs and lymph nodes (2). The number of clones present in the spleen and liver is very small and remains unchanged over time, suggesting the efficient dissemination of Y. pseudotuberculosis within the spleen and liver (2). For Salmonella enterica serovar Typhimurium, it was recently shown that foci of infection in the livers of mice result from the clonal expansion of individual bacteria. Unlike yersiniae, however, salmonellae are located intracellularly. Each focus of infection reaches a critical threshold beyond which bacteria redistribute to uninfected cells, forming new foci (31).

FIG. 8.

Schematic illustration of abscess formation and invasion of PPs. When more than one bacterium invades a PP, either microcolonies can be polyclonal (A) or several monoclonal abscesses can form (B). Here, we showed that monoclonal microcolonies develop according to the diagram in panel B. Furthermore, we showed that the host inhibits the sequential infection of PPs. This inhibition may be due to a local signal induced by Yersinia that shuts off antigen sampling and Yersinia uptake by M cells (C). Invading yersiniae may be eliminated by the activated PP (D). Alternatively, interepithelial dendritic cells may transport yersiniae to the subepithelium (E). Black symbols indicate green-fluorescing bacteria, and white symbols indicate red-fluorescing bacteria. +, no inhibition; −, inhibition.

The finding that only very few monoclonal microcolonies are seen in each PP even at a high oral infection dose suggests the clonal invasion of PPs by Y. enterocolitica. Of course, this observation may be the result of the low rate of survival among bacteria after the initial encounter with the host immune response following the invasion of a given PP by many yersiniae. This possibility has been suggested by the electron microscopic observation of multiple yersiniae invading M cells (5, 6). These experiments were, however, performed using the murine ligated gut loop model, which cannot be compared to in vivo oral infections. To support the finding that it is the limited invasion of PPs that is responsible for our observation, we performed infection experiments with granulopenic and oxidative-burst-deficient mice. Granulocytes have been shown to be critical in early host defense against Yersinia. It has been shown previously that the initial inactivation of yersiniae that implant in the liver and spleen during the first few hours of infection is primarily the feat of granulocytes (7). Besides accumulating in the liver and spleen early after infection, granulocytes have been demonstrated previously to infiltrate PPs within 24 h p.i. (4). Therefore, granulopenic mice were infected with a mixture of red- and green-fluorescing bacteria. If many yersiniae were to invade PPs and only a few were to survive the influx of granulocytes, we would expect an increase in the number of microcolonies per PP from granulopenic versus immunocompetent mice. The same should hold for mice impaired in the oxidative burst (p47^phox−/− mice). The infection of these mice with Y. enterocolitica resulted in essentially the same picture seen with immunocompetent mice. These findings support the clonal invasion hypothesis. If the observation of microcolony formation in a certain PP by just a few clones of Y. enterocolitica was solely the result of the invasion of the PP by many yersiniae, of which just a few survived the initial immune response, then we would expect a 10-fold-higher infection dose to lead to a 10-fold-higher number of monoclonal microcolonies per PP. This is, however, not the case, which lends support to the clonal invasion hypothesis.

Clonal invasion implies that signature-tagged mutagenesis is not a suitable tool to identify attenuated yersiniae in mouse PPs. In fact, a study of the dissemination of signature-tagged Y. pseudotuberculosis mutants in the mouse model previously noted that barriers that limit the number of bacteria that are able to reach the mesenteric lymph nodes and the spleen must exist. Such barriers were considered to be responsible for the failure of signature-tagged mutagenesis to identify attenuated mutants of Y. pseudotuberculosis in mesenteric lymph nodes and the spleen (24). The dissemination of signature-tagged mutants to PPs was, however, not studied.

The clonal invasion of PPs by Y. enterocolitica may reflect a Yersinia-specific characteristic, the host response to Yersinia infection, or a combination of both. Among Yersinia-specific factors, the simplest explanation for clonal invasion would be limited contact between Y. enterocolitica and M cells in the mouse model. Inv is the most important factor for the invasion of PPs (27) and is known to bind β integrins of M cells (5). From previous in vitro studies, it is known that Inv expression is high at ambient temperatures and down-regulated at the host temperature of 37°C (26). We therefore wondered whether yersiniae also down-regulate Inv in the gut lumen after oral infection. Here, we demonstrated by Western blotting that this is the case. The lack of Inv on the surface of Y. enterocolitica in the SI obviously restricts early (Inv-dependent) bacterial invasion to a short time period after oral uptake (when Inv is still present on the bacterial surface). Inv expression by Y. enterocolitica in only the PPs of mice was previously demonstrated, but the SI lumen was not studied at this time (29). Another explanation for clonal invasion may be that Yersinia actively prevents the invasion of PPs by injecting type III secretion system-dependent Yops into M cells, thereby paralyzing these cells and preventing Yersinia uptake. To gain some insight into this possibility, we performed coinfection experiments with RFP- and GFP-harboring YopH, YopO, YopP, YopE, YopM, YopT, and YopQ mutants (33). These experiments revealed numbers of monoclonal microcolonies per PP similar to those seen with wild-type yersiniae (our unpublished results). Besides several conceivable Yersinia-specific factors, the host response to infection may limit Yersinia invasion of PPs. To look into this possibility and to determine if Y. enterocolitica was able to invade and form microcolonies in previously infected PPs, we performed sequential infection experiments by orally inoculating mice with green-fluorescing yersiniae followed by red-fluorescing yersiniae 2 days later. These experiments revealed that yersiniae orally inoculated into mice 2 days after a primary Yersinia infection preferentially invaded those PPs that did not initially develop abscesses. The freshly inoculated yersiniae of the successive infection were invasion competent since they invaded and replicated in naïve PPs, but they showed a severely reduced ability to establish microcolonies in previously infected PPs. This finding indicates that the host severely limits the sequential infection of PPs and is obviously one important reason why only very few monoclonal microcolonies are seen in a certain PP after infection with a high bacterial dose (Fig. 8C and D). Granulocytes were not responsible for limiting the sequential infection of PPs. Presumably, PPs are permissive of invasion by multiple yersiniae only if they are invaded concomitantly. This hypothesis is supported by the results of the experiment demonstrating that at logarithmically increasing infection doses the number of microcolonies per PP increases only linearly. Possibly, a signal generated locally in a certain PP shuts off antigen sampling and Yersinia uptake by M cells of that PP only (Fig. 8C). Alternatively, it is possible that further yersiniae invade PPs but are rapidly eliminated by the “activated” PP (Fig. 8D). Very rarely were PPs in which abscesses had previously been established invaded and subjected to abscess formation by a subsequent Yersinia infection. Possibly only the invasin-/β integrin-mediated invasion process is inhibited by the host, with residual invasion taking place by alternate mechanisms such as YadA, Ail (28), or the transport of yersiniae to the subepithelium by interepithelial dendritic cells (Fig. 8E), which has been demonstrated previously for Salmonella (34). Finally, we showed that most PPs develop abscesses early during infection and that further invasion and microcolony formation by yersiniae replicating in the gut lumen are rare events. Furthermore, the number of microcolonies per PP remains constant over time, supporting the finding that the host limits the invasion of previously infected PPs. In summary, we demonstrate monoclonal microcolony formation by Y. enterocolitica in the mouse oral infection model and show that the clonal invasion of PPs is due probably to the host's severe limitation of the sequential infection of PPs and to reduced levels of invasin observed in Y. enterocolitica bacteria recovered from the SI.