Kinetics of Innate Immune Response to Yersinia pestis after Intradermal Infection in a Mouse Model

Abstract

A hallmark of Yersinia pestis infection is a delayed inflammatory response early in infection. In this study, we use an intradermal model of infection to study early innate immune cell recruitment. Mice were injected intradermally in the ear with wild-type (WT) or attenuated Y. pestis lacking the pYV virulence plasmid (pYV⁻). The inflammatory responses in ear and draining lymph node samples were evaluated by flow cytometry and immunohistochemistry. As measured by flow cytometry, total neutrophil and macrophage recruitment to the ear in WT-infected mice did not differ from phosphate-buffered saline (PBS) controls or mice infected with pYV⁻, except for a transient increase in macrophages at 6 h compared to the PBS control. Limited inflammation was apparent even in animals with high bacterial loads (10⁵ to 10⁶ CFU). In addition, activation of inflammatory cells was significantly reduced in WT-infected mice as measured by CD11b and major histocompatibility complex class II (MHC-II) expression. When mice infected with WT were injected 12 h later at the same intradermal site with purified LPS, Y. pestis did not prevent recruitment of neutrophils. However, significant reduction in neutrophil activation remained compared to that of PBS and pYV⁻ controls. Immunohistochemistry revealed qualitative differences in neutrophil recruitment to the skin and draining lymph node, with WT-infected mice producing a diffuse inflammatory response. In contrast, focal sites of neutrophil recruitment were sustained through 48 h postinfection in pYV⁻-infected mice. Thus, an important feature of Y. pestis infection is reduced activation and organization of inflammatory cells that is at least partially dependent on the pYV virulence plasmid.

INTRODUCTION

Yersinia pestis, the bacterial agent of plague, is an international public health concern from the perspective of both natural transmission and its potential use as a bioterrorist weapon (20, 41). Two other pathogenic Yersinia species, Yersinia pseudotuberculosis and Yersinia enterocolitica, cause a mild, self-limiting enteric disease. However, Y. pestis is unique among the yersiniae as a flea-borne pathogen of rodents (17). Yersinia pestis is highly invasive, causing severe septicemia which, if untreated, is usually fatal to its host. This high bacteremia is necessary to infect fleas feeding on the host and to propagate the transmission cycle (15, 32).

Transmission of Y. pestis by fleas typically results in bubonic plague, characterized by painful swelling of the lymph node proximal to the feeding site. A critical first step in determining the outcome of bubonic plague occurs in the dermis of the host, where Y. pestis is deposited by the flea vector. If the bacterium is contained and killed by the innate inflammatory response in the skin, the disease cannot progress. These events occur very early after transmission. Within about 12 to 24 h, Y. pestis has colonized the draining lymph node and from there can disseminate to other target organs in the host, such as the spleen and liver.

All three pathogenic Yersinia species carry a homologous virulence plasmid, pYV (often referred to as pCD1 in Y. pestis), which is necessary for disease pathogenesis in mammals. This plasmid encodes regulatory and structural proteins for a type III secretion system (T3SS), as well as a number of effector proteins which are injected into host cells upon contact or secreted into the extracellular space. These proteins modulate the host innate immune response in several ways, including inhibition of phagocytosis, reduction in proinflammatory cytokine production, and triggering apoptosis in host cells (13).

Evasion of an early inflammatory response is a key component of Y. pestis pathology, although initial immune events in the dermis have not been well elucidated. In order to study the inflammatory response in the dermis, Belkaid et al. (3) developed a model based on mouse ear explants (28). Mice were injected intradermally in the ear with purified lipopolysaccharide (LPS) or Listeria monocytogenes. Cells migrating out of the ear tissue into cell culture medium were collected and characterized with a combination of immunocytochemistry and flow cytometry. In the present study, we developed a dermal model to compare and contrast the inflammatory response in mice at early time points (≤48 h) after intradermal infection with the fully virulent 195/P strain of Y. pestis and an attenuated strain lacking the entire pYV virulence plasmid (11). Multiparameter flow cytometry was used to distinguish neutrophils and macrophages present in the ear and draining lymph node and to assess the expression of activation markers on these cells. We used immunohistochemistry to localize neutrophils and Y. pestis in thin sections of ear and lymph node tissue. This model clarifies early events of Y. pestis infection and the dynamics of the inflammatory response in the skin.

MATERIALS AND METHODS

Bacteria.

Two strains of Y. pestis were used: fully virulent wild-type (WT) Y. pestis strain 195/P (10) and an attenuated, isogenic pYV⁻ derivative. Strains were grown from frozen stocks in brain heart infusion broth overnight at 28°C, transferred into LB broth, and grown for 24 h at 28°C without aeration. These cultures were brought to 20% glycerol and stored in aliquots at −80°C. Titer of the frozen stock was determined by limiting dilution on blood agar plates in triplicate. The titer of aliquots was tested periodically, and there was no change in CFU/ml of the stock over the course of the study.

Mice.

Specific-pathogen-free, 6- to 12-week-old female BALB/c mice purchased from Jackson Laboratories (Bar Harbor, ME) were used for all experiments. All research involving animals was conducted in compliance with the Animal Care and Use Committee of NIAID/NIH.

Mouse infections.

Aliquots of Y. pestis were thawed and diluted in phosphate-buffered saline (PBS) to the indicated concentration. Mice were injected intradermally (i.d.) in the ear pinna with ∼500 CFU (WT) or ∼1,000 CFU (pYV⁻) bacteria in a total volume of 10 μl. Inoculum titers were confirmed by limiting dilution on blood agar plates. Inocula for virulent WT Y. pestis ranged from 310 to 730 CFU per dose (mean ± standard deviation [SD] = 515 ± 138 CFU); inocula for the attenuated pYV⁻ strain ranged from 701 to 1,350 CFU per dose (mean ± SD = 1,022 ± 256 CFU). In an initial experiment, groups of mice (n = 6 or 7) were euthanized at 3, 6, 12, 24, 48, and 72 h postinfection (p.i.). The injected ears, draining superficial parotid lymph nodes (using the nomenclature of Van den Broeck et al. [50]), spleens, and blood samples were taken to determine bacterial dissemination kinetics. Ears were collected into tubes with 70% EtOH (3). Superficial parotid lymph nodes were collected into tubes with 2 ml PBS without Ca or Mg. Ears and lymph nodes were homogenized by pressing through a 70-μm nylon cell strainer into 2 ml PBS. Spleens were collected into bead tubes (Lysing Matrix H; MP Biomedicals) with 1 ml PBS and processed with a bead homogenizer (BioSpec Products, Inc.). Bacterial load of homogenates and blood was determined by limiting dilution on blood agar plates. These mice were also monitored for clinical signs of infection. Mice were euthanized at the first sign of irreversible illness (lethargy, ruffled fur, hunched posture, reluctance to respond to external stimuli). Time to death was recorded in hours p.i. at the time of euthanasia.

For flow cytometry analyses, mice were infected i.d. in the ear as described above with either WT or pYV⁻. Control mice were injected with 10 μl PBS alone, and as a positive inflammatory control, mice were injected with 10 μl of PBS containing 1 μg ultrapure Escherichia coli serotype 0111:B4 LPS (InvivoGen). At 3, 6, 12, 24, and 48 h p.i., mice were euthanized. Ears were collected into tubes with 70% EtOH (3). Superficial parotid lymph nodes were collected into tubes with 2 ml PBS without Ca or Mg. Spleen and blood were also taken to measure bacterial load and processed as described above. For each treatment group at each time point, 5 to 15 mice were sampled.

In a separate experiment, 10 mice were injected i.d. in the ear with WT, pYV⁻, or PBS alone as described above. Twelve hours after the initial injection, mice were injected in the same ear with 1 μg ultrapure E. coli 0111:B4 LPS (InvivoGen) as a positive inflammatory stimulus. At 24 h after the initial injection, mice were euthanized; the injected ear and superficial parotid lymph node were collected and processed as described above.

Isolation of cells.

Ears were removed from EtOH and blotted dry. Ears were carefully peeled apart, separating the two skin layers, and floated dermal side down in a 6-well non-tissue-culture-treated plate. Wells contained 3 ml RPMI medium (Sigma) with 25 mM HEPES (pH 7.5), 1.5% NaHCO₃, 50 μg/ml DNase I (Worthington Biochemical Corporation), and 0.5 mg/ml Liberase PI (Roche Diagnostics). After 30 min of incubation at 37°C, ears were gently pressed against a 70-μm cell strainer into the RPMI medium to disperse cells. At this point, 50 μl of ear sample was removed to measure bacterial load by limiting dilution on blood agar plates. Wells were rinsed with 4 ml PBS without Ca or Mg and pooled with the ear sample. Each sample was then brought to a final volume of 10 ml with PBS.

Lymph nodes were gently pressed through a 70-μm cell strainer into 2 ml PBS, and a 50-μl aliquot was taken to measure bacterial load by limiting dilution on blood agar plates. The strainer was rinsed with an additional 8 ml PBS. Ear and lymph node samples were spun at 290 × g for 5 min at 10°C. Supernatant was carefully removed, and samples were resuspended in 2 ml 0.16 M ammonium chloride, pH 7.2, to lyse red blood cells. After 2 min at room temperature, 8 ml PBS was added to each sample and spun at 290 × g for 5 min at 10°C. Supernatant was carefully pipetted off, and samples were resuspended in 250 μl (ear) or 350 μl (lymph node) flow cytometry staining buffer (eBioscience).

Flow cytometry.

Fifty-microliter aliquots of cells from each sample were dispensed into 96-well round-bottom microtiter plates and stained with 1:200 dilutions of antibodies purchased from BD Pharmingen or eBiosciences. Antibodies that recognize Ly-6G (clone 1A8, fluorescein isothiocyanate [FITC] labeled), CD11b (clone M1/70, labeled with phycoerythrin-Cy7), F4/80 (clone BM8, allophycocyanin labeled), and major histocompatibility complex class II (MHC-II) (clone M5/114.15.2, phycoerythrin labeled) were used, along with rat IgG2a and IgG2b isotype controls. Cells were stained for 30 min at 4°C, spun at 650 × g for 1 min, and fixed with IC fixation buffer (eBioscience) for 1 h. Cells were spun at 650 × g for 1 min and resuspended in flow cytometry staining buffer. Cell phenotype data were acquired on a Partec CyFlow ML flow cytometer and analyzed with FloMax (Partec) and FloJo (Tree Star) software. Neutrophils were defined as Ly-6G⁺ F4/80⁻ cells, and CD11b was used as an activation marker for neutrophils (12, 14, 26, 45). Macrophages were defined as F4/80⁺ Ly6G⁻ cells. Expression of both CD11b and MHC-II was measured on these cells to determine activation status (24, 27, 35).

Histology and immunohistochemistry.

Mice were injected i.d. in the ear with WT Y. pestis 195/P, attenuated pYV⁻ Y. pestis, PBS, or LPS as described above. At 3, 6, 12, 24, and 48 h p.i., mice were euthanized; the injected ear and superficial parotid lymph node were collected and fixed in 10% neutral-buffered formalin. Fixed samples were embedded in paraffin, sectioned, and double-stained with a rabbit polyclonal antibody to Y. pestis (46) and a goat polyclonal antibody to myeloperoxidase (R & D Systems) to localize bacteria and neutrophils, respectively (11). Myeloperoxidase was visualized with diaminobenzidine (giving a brown color); the chromogen used for Y. pestis was Fast Red. Samples were counterstained with hematoxylin.

Histology slides were evaluated subjectively and assigned a numerical inflammation severity score from 1 to 4 according to the relative numbers of aggregated neutrophils. Very few aggregated cells (similar to PBS-injected controls at the same time point) indicated a score of 1 (minimal inflammation). “Mild” (2) and “moderate” (3) inflammation was indicated by increasingly larger infiltrates of stained cells (low to moderate numbers of cells), and a score of “severe” (4) was reserved for very large inflammatory cell aggregates (most of the tissue section infiltrated with stained inflammatory cells). In the case of animals infected with virulent Y. pestis, severe inflammation was also accompanied by visible staining of bacteria. At each time point, 4 or 5 individuals were evaluated based on 6 to 9 stained sections of tissue.

Statistical analyses.

Bacterial loads in WT-infected mice and pYV⁻-infected mice were compared using Student's t test. For flow cytometric data, groups at each time point were compared by Kruskal-Wallis nonparametric analysis of variance (ANOVA), followed by Dunn's multiple comparison test to detect differences between treatments or time points. The association between bacterial loads and neutrophil recruitment in the ear and draining lymph node was tested by correlation analysis. Analyses were done using GraphPad Prism software (version 5.01).

RESULTS

Survival and kinetics of infection of Y. pestis in vivo after i.d. infection in the ear.

At a dose of 895 CFU (determined by titration of the inoculum on blood agar plates), WT Y. pestis 195/P was lethal to 31 of 35 mice by 72 h p.i. (Fig. 1), whereas the attenuated pYV⁻ strain caused no morbidity or mortality in these experiments at a dose of 1,000 CFU and does not kill mice even at higher challenge doses (11, 29, 41).

Fig 1.

Survival curve of mice (1 infection trial, n = 35) following intradermal injection of 895 CFU virulent Y. pestis 195/P into the ear.

Numbers of WT Y. pestis 195/P CFU in the ear remained steady through 12 h and then increased rapidly through 48 h p.i. (Fig. 2A). Bacteria were detected as early as 6 h p.i. in the draining lymph node, although most mice did not have detectable lymph node colonization until 12 h p.i. (Fig. 2B). Bacteria were first detected in the spleen at 24 h p.i. in a few individuals, but by 48 h, 73% of mice had ≥10⁷ CFU in the spleen (Fig. 2C). Titers remained below the detectable level in the blood through 24 h, but between 24 and 48 h, septicemia rose to an average of ∼10⁶ CFU/ml blood (Fig. 2D).

Fig 2.

Bacterial loads in mice infected intradermally in the ear with ∼500 CFU WT Y. pestis 195/P or ∼1,000 CFU attenuated pYV⁻ Y. pestis. Bacterial load in the ear (A), draining superficial parotid lymph node (B), spleen (C), and blood (D) at different time points postinfection. Each circle or square symbol represents an individual. Samples sizes of WT-infected mice were 21 (3, 6, and 24 h), 11 (12 h), 22 (48 h), 6 (72 h); sample size of pYV⁻-infected mice was 5 at each time point (no 72-h time point taken). Data for WT-infected mice include both the samples taken from the initial bacterial load/survivorship trial and samples taken during the flow cytometry time course studies. Horizontal black lines indicate mean WT or pYV⁻ CFU per tissue or ml of blood. Dashed lines indicate the lower limit of detection of the assay. Means compared by Student's t test: *, 0.001 < P ≤ 0.01; **, P < 0.001.

In contrast to WT Y. pestis, attenuated pYV⁻ Y. pestis showed no net increase in the ear, and some of the pYV⁻-infected mice did not have detectable bacteria in the ear at most time points (Fig. 2A). Although bacteria could be detected in some individuals out to 48 h p.i., titers never exceeded the initial inoculum of 10³ CFU. Most mice had no detectable dissemination of bacteria to the draining lymph node, and titers did not exceed 2.38 log CFU (Fig. 2B). While pYV⁻ bacteria persisted in the lymph node out to 48 h p.i. in some individuals, in contrast to virulent Y. pestis, an increase in bacterial load over time was not seen. The pYV⁻ treatment group had significantly lower bacterial load in the ear and lymph node at 48 h p.i. than the wild-type strain. No mice had detectable dissemination of pYV⁻ bacteria beyond the lymph node to the spleen or blood (Fig. 2C and D), and no signs of illness were noted in any pYV⁻-infected mice.

Kinetics of neutrophil recruitment and activation in the ear.

The neutrophil response to intradermal infection was examined by flow cytometry of ear cell suspensions and by immunohistochemistry staining of intact ears. The median number of total neutrophils (Ly-6G⁺ cells) as a percentage of all cells in the ear suspension counted by flow cytometry did not differ among treatments at any time point, except at 48 h p.i., where WT-infected mice in the bubonic phase of disease (no detectable bacteria in the blood) had a significantly lower neutrophil response than the LPS treatment group (Fig. 3A). However, variability among individuals was observed, and immunohistochemical staining of ear sections revealed qualitative differences, especially between the virulent WT Y. pestis and attenuated pYV⁻ treatments (Fig. 4). At 24 h p.i., neutrophil recruitment in WT mice (Fig. 4D) typically appeared no more extensive than in the LPS (Fig. 4B) and pYV⁻ (Fig. 4C) mice. At 48 h p.i., neutrophils continued to accumulate in the LPS-injected ears (Fig. 3A and 4F), with Ly-6G⁺ cells significantly greater than at earlier time points, but with a large variability among individuals. In the pYV⁻ Y. pestis-infected group, focal areas of neutrophil recruitment intensified (Fig. 4E) without a significant increase in total Ly-6G⁺ cells in the entire ear (Fig. 3A). By 48 h p.i., the WT Y. pestis-infected group contained some individuals who still showed only low to moderate influx of neutrophils (≤10.7%) and others who showed signs of morbidity with neutrophils at 15 to 32% of cells. Because of the large animal-to-animal variation in the rate of disease progression, we separated the 48-h WT samples into individuals with bubonic plague (B; bacteria in the superficial parotid lymph node but not in blood) or with septicemic plague (S; bacteria present in blood). All mice with >7% Ly-6G⁺ cells were septicemic (Fig. 3A). This variation also was reflected in the immunohistochemical staining of ear tissue: individuals ranged from little (Fig. 4G) to moderate (Fig. 4H) recruitment of neutrophils, and these cells were much more diffuse in distribution, unlike the focal accumulations seen in the Y. pestis pYV⁻ ear samples. No bacteria were seen in any of the pYV⁻-infected ear sections. WT-infected individuals late in disease progression showed severe inflammation, with dense areas of Y. pestis growth surrounded by neutrophils (Fig. 4I). These data indicate that the influx of large numbers of neutrophils is delayed in mice infected with fully virulent Y. pestis 195/P until the late stages of disease, after the onset of septicemia.

Fig 3.

Time course of the neutrophil response in the ear of BALB/c mice. Total neutrophils, expressed as the percentage of Ly-6G⁺ cells out of all events counted (A); activated neutrophils, expressed as the percentage of Ly-6G⁺ CD11b⁺ cells out of all events counted (B); and % neutrophils activated, expressed as the percentage of all Ly-6G⁺ cells also expressing CD11b (C). Each symbol represents an individual; horizontal black bars indicate the median. Sample sizes were 5 mice at each time point for pYV⁻ and LPS, 5 to 10 for PBS, and 10 to 15 for WT. Median percentages were compared by Kruskal-Wallis nonparametric ANOVA followed by Dunn's multiple comparison test. *, 0.01 < P ≤ 0.05; †, 0.001 < P ≤ 0.01; ‡, P ≤ 0.001.

Fig 4.

Immunohistochemical staining of ear sections. Sections were stained with antimyeloperoxidase antibody (brown color) to label neutrophils and a polyclonal anti-Yersinia pestis antibody (red color). Panels show representative samples at 24 h after injection with PBS (severity score [SS] of 1) (A), E. coli LPS (SS of 2) (B), pYV⁻ Y. pestis (SS of 2) (C), or WT Y. pestis 195/P (SS of 2) (D) and at 48 h postinjection with pYV⁻ (SS of 3) (E), LPS (SS of 3) (F), or WT Y. pestis 195/P (G, H, and I). The bottom three panels show the variation in neutrophil response in mice infected with wild-type Y. pestis 195/P at 48 h, ranging from normal to mild (SS of 1) (G), moderate (SS of 2) (H), or severe (SS of 4) (I), with white arrows indicating examples of extensive bacterial growth. Scale bar = 50 μm.

In order to assess the activation state of recruited neutrophils, we used expression of the α-integrin CD11b as an activation marker (12, 14, 26, 45). The presence of activated neutrophils (Ly-6G⁺ CD11b⁺ cells) in the ear was significantly lower in WT Y. pestis-infected mice at 3 h and 6 h p.i. than in the pYV⁻ Y. pestis and LPS treatment groups but not significantly different from the PBS controls (Fig. 3B). At 24 h p.i., the percentages of activated neutrophils in the WT, pYV⁻, and LPS samples were significantly greater than that in the PBS control. However, the WT group did not exceed the pYV⁻ or LPS groups in activated neutrophils, even though WT bacteria, unlike pYV⁻ bacteria, were actively replicating and accumulating in the ear (Fig. 1A). By 48 h p.i., individuals in the WT group with no signs of disease showed very low numbers of CD11b⁺ neutrophils despite continued growth of bacteria in the ear, whereas most septic individuals showed a large influx of activated neutrophils. All mice with >3% Ly-6G⁺ CD11b⁺ cells were septicemic.

The reduced neutrophil activation in the ears of mice infected with WT Y. pestis can be seen in Fig. 3C (% neutrophils activated; the percentage of Ly-6G⁺ cells also expressing CD11b). At all time points through 24 h, infection with WT resulted in lower rates of neutrophil activation than the pYV⁻ and LPS treatments, which remained at a median CD11b⁺ rate of 60 to 80% throughout the experiment. These differences, although large, were not always statistically significant by Dunn's multiple comparison test. Activation rates in WT samples significantly decreased between 3 h and 24 h p.i. (P < 0.01) and were not significantly different from the PBS control. The increase seen at 48 h is skewed by septicemic individuals in the late stages of disease; bubonic individuals did not differ significantly from the PBS control. This is reflected in the representative frequency histograms of CD11b expression on Ly-6G⁺ cells presented in Fig. 5. There is little change in distribution in the PBS and LPS treatments at 3 h and 48 h p.i., although there is a large increase in the numbers of activated neutrophils in the LPS treatment at 48 h p.i. Expression of CD11b goes down in the virulent 195/P Y. pestis treatment between 3 h p.i. and 48 h in bubonic mice, with the addition of a large number of neutrophils expressing high levels of CD11b with the onset of septicemia.

Fig 5.

Representative smoothed frequency histograms of CD11b expression on neutrophils (Ly-6G⁺ cells) based on flow cytometry data from individual ear samples. Histograms at 3 h (light line) and 48 h (heavy line) p.i. from the PBS (A), LPS (B), and attenuated pYV⁻ Y. pestis (C) treatments. The panel from the virulent 195/P Y. pestis treatment (D) compares 3 h p.i. (light line) to 48 h p.i. (heavy line) for bubonic and 48 h p.i. for septicemic (shaded) mice.

The effect of WT infection on neutrophil activation in relation to bacterial load could be seen in both the ear (Fig. 6A) and the draining lymph node (Fig. 6B). The number of activated (CD11b⁺) neutrophils was variable in the ear, but even individuals with 10⁴ to 10⁶ bacteria showed only a low to moderate influx of activated neutrophils compared to that of the PBS control. The correlation between bacterial load and percentage of activated neutrophils was significant (Pearson r = 0.341, P = 0.031), although the correlation explains a small proportion of the variation in the data set (r² = 0.116). Three septicemic individuals at 48 h p.i. had a large influx of activated neutrophils in the ear (≥13.5%), and eliminating just these three outliers rendered the correlation insignificant (Pearson r = 0.309, P = 0.063). In the lymph node, bacterial loads of 4 to 7 log₁₀ per organ showed only moderate influx of activated neutrophils in most individuals. A statistically significant correlation between bacterial load and percentage of activated neutrophils was observed (Pearson r = 0.677, P = 0.0003). This correlation remained significant even without the three outliers that had 14.6 to 29.0% activated neutrophils in the lymph node (Pearson r = 0.623, P = 0.0026). Individuals at this stage of disease were distinct in immunohistochemical staining, showing neutrophil accumulation and most of the lymph node space filled with bacteria (Fig. 7H). Other individuals at 24 and 48 h p.i. with no clinical signs of infection showed little inflammation or a moderate, diffuse influx of neutrophils (Fig. 7F, G). Similar to what we observed in the ear, all mice with neutrophils at >7% of cells in the lymph node were septicemic (Fig. 3D).

Fig 6.

Correlation between bacterial load and activated neutrophil recruitment in BALB/c mice infected with virulent Y. pestis 195/P. The log₁₀ CFU is plotted against the percentage of Ly-6G⁺ CD11b⁺ cells out of all events counted in the ear (A) and draining superficial parotid lymph node (B) 12 to 48 h postinfection. Each point represents an individual. The shaded areas represent the range of percentage of activated neutrophils obtained from the PBS-injected controls.

Fig 7.

Immunohistochemical staining of superficial parotid lymph node sections. Sections were stained with antimyeloperoxidase antibody (brown color) to label neutrophils and a polyclonal anti-Yersinia pestis antibody (red color). Panels show the range in neutrophil response in animals infected with the attenuated pYV⁻ strain at 24 (A and B) and 48 h (C and D) postinjection (A and C given a severity score [SS] of 2; B and D, SS of 3) compared with the PBS treatment (SS of 1) (E) and WT Y. pestis 195/P (SS of 2) (F) at 24 h p.i. The bottom two panels are examples at 48 h p.i. with wild-type Y. pestis in mice before the appearance of clinical disease (SS of 1) (G) and in moribund individuals (SS of 4) (H). Scale bar = 50 μm.

Kinetics of macrophage recruitment and activation in the ear.

In the LPS treatment group, we observed a significant (P < 0.01) peak in the presence of macrophages (F4/80⁺ cells) in the ear at 6 h p.i. (Fig. 8A). This was also reported by Belkaid et al. (3) and may include maturation of resident monocytes in the ear after exposure to LPS. The PBS treatment group reached a peak in F4/80⁺ cells at 12 h p.i. (P < 0.01) and declined at later time points. Recruitment of macrophages in the WT group also saw an early increase at 6 h versus 3 h p.i. (P < 0.05) and remained steady for the rest of the experiment. However, the WT group did not exceed the PBS treatment at 12 h, 24 h, or 48 h p.i. or the pYV⁻ treatment at any time point. Similar to what we observed with activation of neutrophils, infection with WT was associated with reduced activation of macrophages as measured by expression of MHC-II (Fig. 8B), CD11b (Fig. 8C), or both of these markers (Fig. 8D). This effect is at least in part due to the presence of the pYV virulence plasmid, since the pYV⁻ treatment group had significantly higher rates of activation of neutrophils at 3, 12, and 24 h p.i. than the PBS control and significantly exceeded rates of macrophage activation compared to virulent Y. pestis 195/P at 24 h p.i. In contrast to observations of neutrophil recruitment, there was no large influx of macrophages in septic animals. Macrophage recruitments were similar between mice in the bubonic and septicemic phases of disease (Fig. 8).

Fig 8.

Time course of the macrophage response in the ear of BALB/c mice. Total macrophages, expressed as the percentage of F4/80⁺ cells out of all events counted (A), the percentage of F4/80⁺ also expressing MHC-II (B), the percentage of F4/80⁺ cells also expressing CD11b (C), and the percentage of F4/80⁺ cells also expressing both CD11b and MHC-II (D). Each symbol represents an individual; horizontal black bars indicate the median. Sample sizes were 5 at each time point for pYV⁻ Y. pestis, LPS, and PBS and 5 to 15 for WT Y. pestis 195/P. Median percentages were compared by Kruskal-Wallis nonparametric ANOVA followed by Dunn's multiple comparison test. *, 0.01 < P ≤ 0.05; †, 0.001 < P ≤ 0.01; ‡, P ≤ 0.001.

Ability of wild-type Y. pestis to suppress activation of neutrophils in the presence of an inflammatory stimulus.

The effects of Y. pestis on neutrophils could be the result of active suppression of neutrophil recruitment or a lack of induction of a neutrophil response by the bacteria. In order to address this question, we injected mice intradermally in the ear with WT, pYV⁻, or PBS. Twelve hours later, we injected the same ear with E. coli LPS to induce an inflammatory response. Twelve hours after that, we measured recruitment of neutrophils to the ear. Figure 9A shows that the WT is unable to prevent recruitment of neutrophils in response to LPS, indicating that Y. pestis cannot actively suppress neutrophil recruitment. However, as with injection with WT alone, we observed reduced rates of neutrophil activation in response to LPS stimulus in the presence of virulent Y. pestis (Fig. 9B and C), significantly lower than observed with the pYV⁻-LPS and PBS-LPS treatments.

Fig 9.

Effect of Y. pestis infection on inflammatory cell recruitment with exposure to E. coli LPS. Groups of 9 or 10 mice were injected with Y. pestis or PBS and reinjected 12 h later with LPS. Total neutrophils, expressed as the percentage of Ly-6G⁺ cells out of all events counted (A); activated neutrophils, expressed as the percentage of Ly-6G⁺CD11b⁺ cells out of all events counted (B); and % neutrophils activated, expressed as the percentage of Ly-6G⁺ cells also CD11b⁺ (C) were quantitated 12 h after LPS injection. Bars represent the median, with interquartile range indicated. Median percentages were compared by Kruskal-Wallis nonparametric ANOVA followed by Dunn's multiple comparison test. *, 0.01 < P ≤ 0.05; **, 0.001 < P ≤ 0.01; ***, P ≤ 0.001.

DISCUSSION

Transmission of Y. pestis by fleas most frequently results in intradermal deposition of bacteria (47). Many virulence factors necessary for Y. pestis infection and dissemination in the mammalian host are temperature regulated and are minimally expressed at the ambient temperature of the flea (36). Early events at the transmission site in the skin of the host, before the induction of the full complement of virulence factors, are likely to be critical to survival of Y. pestis, but this in vivo environment has never been examined. In this study, we adapted and applied a dermal infection model in a first attempt to characterize and compare the intradermal innate immune response to a fully virulent and an avirulent strain of Y. pestis.

After intradermal injection in the ear with WT Y. pestis, the mortality of mice and bacterial loads in different organs measured over time (Fig. 1) are comparable to published data using mice as a model (21, 41, 49). Viable pYV⁻ bacteria persisted, but did not increase in numbers, for at least 48 h in the ears of most mice, and dissemination to the lymph node occurred in some. In contrast, mean numbers of WT bacteria increased by ∼3 logs between 6 and 48 h. However, a large range in bacterial load was observed among individuals within the same time point. The kinetics of dissemination to lymph node, spleen, and blood also varied greatly, as did the time to the appearance of clinical signs of disease: some mice showed signs of morbidity at 48 to 72 h, while others appeared normal. This variation in disease progression affected interpretation of data on inflammatory cell recruitment, as will be discussed below.

The mouse ear represents a compartment of skin tissue which can be sampled after intradermal injection for the recruitment of inflammatory cells (3, 28). It is believed that immediately after infection of a mammalian host, Y. pestis is susceptible to phagocytosis by both neutrophils and macrophages and is killed when phagocytosed by neutrophils but can survive and replicate when phagocytosed by macrophages (43). Bacteria emerging from macrophages are then highly resistant to phagocytosis by both macrophages and neutrophils (9). Both in vivo and in vitro experiments have shown the ability of Y. pestis to interfere with innate immune functions of macrophages (23, 43). Surprisingly little has been published on the interaction of Y. pestis with neutrophils, although Spinner et al. demonstrated inhibition of reactive oxygen species production in human neutrophils infected with Y. pestis (48).

A hallmark of Y. pestis infection is inhibition of the innate immune response until the late stages of disease (6, 8, 11, 18, 42). In this study, neutrophil recruitment to the dermis in mice infected with either WT or pYV⁻ Y. pestis as measured by flow cytometry is not significantly greater than the response induced by injection of sterile PBS, even in animals with high bacterial loads (Fig. 3A). This might be attributable to the poorly immunostimulatory, tetra-acylated form of the lipid A component of LPS made by Y. pestis at 37°C (22, 25). The change from hexa-acylated to tetra-acylated lipid A upon temperature upshift from 21°C to 37°C is essential for pathogenesis, because Y. pestis mutants unable to switch to the tetra-acylated form stimulate a protective inflammatory response in the host and are unable to cause systemic disease even at high challenge doses (34). Robinson et al. (44) showed that Y. pestis producing tetra-acylated lipid A evades early dendritic cell activation. Y. pestis expressing the E. coli LpxL acyltransferase produced hexa-acylated lipid A at 37°C and stimulated dendritic cell activation and migration. Montminy et al. (34) demonstrated that tetra-acylated Y. pestis LPS failed to activate human peripheral blood mononuclear cells in vitro and also inhibited the ability of these cells to respond to E. coli LPS. Consistent with these findings, in our studies in the presence of an inflammatory agent (E. coli LPS), virulent WT Y. pestis was not able to prevent recruitment of neutrophils to the ear of mice, although significant reduction of activation of these cells was still observed (Fig. 8A). The observation that WT alone in the absence of E. coli LPS exposure after infection does not elicit such an influx of neutrophils until late in the disease progression suggests that Y. pestis fails to stimulate a strong innate immune response early in infection.

In addition, the reduced activation of neutrophils and macrophages in WT-infected ears observed in this study is also mediated through the T3SS encoded on the pYV virulence plasmid (4, 39, 51), since pYV⁻-infected mice had higher rates of activation of inflammatory cells. In the presence of E. coli LPS, similar to infection with pYV⁻ Y. pestis alone, the pYV⁻ strain (which retains the ability to switch to tetra-acylated LPS at 37°C) was unable to suppress activation of neutrophils. In addition to general cytotoxic effects of the T3SS on innate immune cells, V antigen has been correlated with a systemic increase in IL-10 production (7, 40) and suppression of gamma interferon (IFN-γ) and tumor necrosis factor alpha (TNF-α) (5, 31, 37, 38). Despite the fact that bacteria were multiplying to high numbers in the WT-infected animals, the median percent activation was <50%, except in mice that were septic and symptomatic (Fig. 3C, Fig. 5D). A similar picture was seen in the draining lymph node, where a large influx of activated neutrophils was seen only late in the disease progression, in animals with >4.5 log₁₀ CFU in this organ. These results are consistent with studies using the rat model of bubonic plague (11, 46).

A previous report showed that the V antigen of Y. pestis was able to inhibit neutrophil chemotaxis in mice to subcutaneously implanted sponges containing LPS (52), which seems at odds with our results. There are several differences between the two studies which could explain this. For example, Welkos et al. injected mice with V antigen specifically to test the effect of this virulence factor, while in our experiment mice had an active Y. pestis infection. Also, by measuring neutrophils in implanted sponges, they were able to look at recruitment to a final extravascular destination. In our flow cytometry data, we identified neutrophils from the entire ear, which would contain neutrophils at various stages of extravasation. Still, the histological findings did show a poorly organized neutrophil response in animals infected with WT Y. pestis, which expresses V antigen, and a more focal neutrophil response in mice infected with the pYV⁻ strain, which lacks V antigen.

It is important to note that flow cytometry evaluates all cells recovered from the entire ear but does not preserve information on focal accumulation of inflammatory cells within the tissue. Histology revealed that in the pYV⁻ group, local foci of neutrophils formed (Fig. 4C and E). This was also observed in the draining lymph node, where in seven out of nine samples taken at 24 or 48 h p.i., local, intense recruitment of neutrophils was found (Fig. 7B and D). In a study by Comer et al. (11), rats were infected i.d. with 10⁸ CFU pYV⁻ Y. pestis. This was done in order to achieve bacterial loads in the lymph node comparable to that seen after i.d. infection with 10³ CFU of fully virulent Y. pestis, and they observed the same focal recruitment of neutrophils sustained out to 60 h p.i. in pYV⁻-infected rats, consistent with this study. We injected a much lower dose of 10³ CFU pYV⁻ Y. pestis, and only a third of mice had detectable levels of pYV⁻ bacteria in the lymph node at 12 h p.i. or later. Thus, the pYV⁻ strain is unable to inhibit inflammation even at low bacterial loads.

Previous work has suggested that virulent Y. pestis alters neutrophil recruitment and function in vitro (48) and in vivo. Guinet et al. (18) observed a lack of organized neutrophil response in the lymph nodes of infected mice even after bacteria invaded the organ. Later in the infection, they observed a large recruitment of neutrophils which was unable to control the bacteria, suggesting a functional alteration or deficit in these cells. Sebbane et al. (46) characterized virulent Y. pestis disease progression in a rat model following intradermal infection. Similar to the present study, they observed dissemination to the draining lymph node as early as 6 h p.i. but saw little recruitment of neutrophils until after 36 h p.i., with low numbers of inflammatory cells compared to bacterial loads. At the end stages of disease, they found a large number of neutrophils, tissue destruction, and expansion of bacteria to fill the lymph node. Both of these studies noted a lack of the inflammatory response to form an abscess or a granuloma-type lesion. In the present study, in WT-infected mice, the appearance of large numbers of inflammatory cells in the ear or draining lymph node was associated with morbidity and disease progression rather than a static time point. Specifically, all mice showing an influx of neutrophils in the ear or lymph node of >7% have a bacteremia of >100 CFU/ml of blood, the lower limit of detection in this study. Similarly, in a transcriptomic study of Y. pestis infection in rats, Comer et al. (11) did not observe upregulation of proinflammatory genes in the lymph node in the bubonic phase of disease. In septicemic animals, however, elevated extracellular levels of multiple proinflammatory cytokines and chemokines were found. Principal component analysis of gene expression profiles from bubonic rats grouped with PBS-injected controls, whereas septicemic animals had a distinct expression pattern.

Macrophage recruitment and activation in the ear following infection with WT and pYV⁻ Y. pestis was analogous to what was seen for neutrophils. Again, despite bacterial replication during the 48 h, mice infected with WT had rates of expression of both CD11b and MHC-II equivalent to or even lower than that of other treatments at all time points (Fig. 8B to D). Inhibition of MHC-II expression (as well as other activation markers and cytokine production) by Y. enterocolitica on murine dendritic cells has been reported (2, 16). This was explained as a possible mechanism to interrupt T-cell activation, since Y. enterocolitica-specific CD4⁺ and CD8⁺ cells are necessary for protection (1). Y. pestis, with its typically acute disease progression, may not benefit from inhibiting adaptive immune responses in the host. However, suppression of CD11b expression could inhibit proper chemotaxis of both neutrophils and macrophages to sites of Y. pestis infection.

Zhou et al. (54) found that neutrophils from TLR4-deficient C57BL/6 mice were unable to increase expression of CD11b upon exposure to LPS. These cells were significantly impaired in their ability to adhere to and transmigrate cultured endothelial cells compared to wild-type neutrophils. In addition, suppression of CD11b on wild-type neutrophils via nuclear factor κB or c-Jun NH₂-terminal kinase inhibition, or addition of anti-CD11b antibody, significantly reduced cellular adhesion. In vivo, a recent study showed that treatment of mice with CD11b/CD18 antibody significantly reduced the percentage of adherent neutrophils that migrated toward a focus of tissue damage, as well as reducing neutrophil crawling velocity (33). Therefore, suppression of CD11b could significantly reduce the efficiency of recruitment of neutrophils and macrophages to foci of infection, disrupting an organized inflammatory response.

The delay of inflammation early in infection in animals infected with virulent Y. pestis seen in this study, followed by a sudden intense inflammatory response at later time points, is similar to the well-characterized biphasic response in primary pneumonic plague (8). The lung model also showed a lack of net replication of pYV⁻ bacteria, a failure to disseminate to the spleen, and no mortality in mice infected with pYV⁻ Y. pestis (29). Establishing an anti-inflammatory state in the lung was shown to be an active process rather than just a lack of virulent Y. pestis to provoke an inflammatory response (42). This suppressive environment may also be important in bubonic plague to establish active replication of bacteria in the skin and allow for dissemination to the draining lymph node in the absence of a protective inflammatory response. We observed reduced CD11b expression on neutrophils in our experiments (Fig. 3). Therefore, neutrophils are recruited to the ear but may be unable to migrate properly to their final destination. This effect was seen even in the presence of highly inflammatory E. coli LPS (Fig. 9C), suggesting that Y. pestis is able to actively suppress neutrophil activation locally.

In conclusion, in the ear of mice infected with WT Y. pestis, recruitment of activated neutrophils was variable, similar to PBS-injected controls, and not related to bacterial load; mice with no detectable bacteria had a range of percent activated neutrophils very similar to that of mice with 1.8 to >6 log₁₀ CFU in the ear (Fig. 6A). At later time points (24 to 48 h p.i.), some individuals showed a large influx of activated neutrophils in the ear, bacterial loads of >4.5 log₁₀ to >6 log₁₀, and signs of morbidity (Fig. 4I). There were a number of other individuals at these time points with similar bacterial loads, but very little increase in inflammation above that of PBS controls (Fig. 4D, G, H). Similarly, in the draining lymph node (Fig. 6B), with the exception of a few moribund individuals with ∼8 log₁₀ CFU showing a large influx of activated neutrophils, mice with 4.8 to 7 log₁₀ CFU showed very little inflammation. In mice with <4 log₁₀, the neutrophil response was no different from that of PBS-injected controls. At 48 h p.i., severe inflammation in the ear and lymph node was always associated with septicemia. Histology showed that the neutrophil response in WT-infected mice was weak and diffuse, in contrast to the intense, focal accumulation of neutrophils seen in the lymph node of pYV⁻-infected mice. Thus, WT Y. pestis shows a remarkable ability to locally inhibit inflammation even when replicating to high titers. In addition, neutrophils and macrophages recruited early in infection have lower rates of expression of activation markers and fail to form an organized and protective inflammatory response.

In nature, plague is transmitted by flea bite to vertebrate hosts. Flea bites and components of flea saliva cause both immediate and delayed (12 to 24 h) inflammatory reactions in vertebrates (19, 30, 53). It is unknown what effect infection in the context of a flea bite has for the disease progression of Y. pestis. In this study, in the presence of an inflammatory agent (E. coli LPS), Y. pestis did not inhibit recruitment of neutrophils, although there was a significant reduction of activation in recruited cells. Applying the model presented here, future studies will look at early events of Y. pestis infection in skin after transmission by fleas and the consequences of flea salivary antigens for transmission and infection kinetics.