Modification of the Pulmonary MyD88 Inflammatory Response Underlies the Role of the Yersinia pestis Pigmentation Locus in Primary Pneumonic Plague

Pneumonic plague, caused by Yersinia pestis, is a rapidly progressing bronchopneumonia involving focal bacterial growth, neutrophilic congestion, and alveolar necrosis. Within a short time after inhalation of Y. pestis, inflammatory cytokines are expressed via the Toll/interleukin-1 (IL-1) adaptor myeloid differentiation primary response 88 (MyD88), which facilitates the primary lung infection.

ABSTRACT

Pneumonic plague, caused by Yersinia pestis, is a rapidly progressing bronchopneumonia involving focal bacterial growth, neutrophilic congestion, and alveolar necrosis. Within a short time after inhalation of Y. pestis, inflammatory cytokines are expressed via the Toll/interleukin-1 (IL-1) adaptor myeloid differentiation primary response 88 (MyD88), which facilitates the primary lung infection. We previously showed that Y. pestis lacking the 102-kb chromosomal pigmentation locus (pgm) is unable to cause inflammatory damage in the lungs, whereas the wild-type (WT) strain induces the toxic MyD88 pulmonary inflammatory response. In this work, we investigated the involvement of the pgm in skewing the inflammatory response during pneumonic plague. We show that the early MyD88-dependent and -independent cytokine responses to pgm− Y. pestis infection of the lungs are similar yet distinct from those that occur during pgm+ infection. Furthermore, we found that MyD88 was necessary to prevent growth of the iron-starved pgm− Y. pestis despite the presence of iron chelators lactoferrin and transferrin. However, while this induced neutrophil recruitment, there was no hyperinflammatory response, and pulmonary disease was mild without MyD88. In contrast, growth in blood and tissues progressed rapidly in the absence of MyD88, due to an almost total loss of serum interferon gamma (IFN-γ). We further show that the expression of MyD88 by myeloid cells is important to control bacteremia but not the primary lung infection. The combined data indicate distinct roles for myeloid and nonmyeloid MyD88 and suggest that expression of the pgm is necessary to skew the inflammatory response in the lungs to cause pneumonic plague.

INTRODUCTION

Plague is a flea-borne disease caused by the Gram-negative bacterium Yersinia pestis. Following infection of mammals, including humans, bacterial invasion of tissues and vascular spread can cause death in as few as a few days without antibiotic intervention (1, 2). The bubonic form of plague occurs following transmission by fleas, and bacteria that have migrated from the peripheral bite site to the draining lymph node replicate and cause necrosis. The pneumonic form of plague occurs following inhalation of bacteria and rapidly progresses to fulminant bronchopneumonia. Each form of the disease typically involves the vascular spread of bacteria and the development of secondary septicemic plague. Modern day plague still carries a high mortality rate; antibiotic treatment, even in the United States where it can be rapidly implemented, is often insufficient to prevent death (3). While historically plague was notorious for causing death in 1 to 3 days after symptom onset, modern day plague is not always rapid, and in some cases, the patient has been hospitalized, receiving antibiotic treatment for more than a week before succumbing. A greater understanding of this disease is needed to develop improved methods of treatment that can augment antibiotic therapy.

Yersinia sp. is an invasive organism that induces a tissue-damaging hyperinflammatory response, and its expression of virulence genes destroys the innate immune response and reduces adaptive immunity (4). Several independent pathogenic mechanisms modify the innate immune response. For example, injection of macrophages and other cells with immune modulators and toxins by the bacterial type III secretion system (T3SS) is a major mechanism by which Y. pestis is thought to suppress inflammation in the lungs following pulmonary infection (5). The effector proteins of the T3SS regulate programmed cell death, regulate inflammatory signaling, and block phagocytosis (6). The early host inflammatory response is thought to be suppressed in a T3SS-dependent manner until bacterial growth induces a second wave of tissue-damaging inflammation (5, 7). We recently showed that pulmonary inflammation is mediated by the host signaling protein Toll/interleukin-1 (IL-1) receptor signaling adapter myeloid differentiation primary response (MyD88) (8). Following pulmonary infection of mice with wild-type Y. pestis, early low-level inflammation triggered by MyD88 ultimately facilitated bacterial colonization of the alveoli and a tissue-damaging inflammatory response. MyD88 also mediated the inflammatory phase of disease in a more positive way such that the robust MyD88 response reduced systemic bacterial growth.

The pigmentation locus (pgm) of Y. pestis is a 102-kb unstable element that is spontaneously lost when cultured in laboratory media due to the presence of inverted repeats on each end of the locus (9, 10). It has long been recognized that this mutation resulted in attenuation of virulence from the subcutaneous route of infection, a phenotype that could be partially rescued by providing supplemental iron to the host (11). Although this strain is highly attenuated, it has been associated with human infection and death (12). In mice, pgm− Y. pestis strains are virulent in intravenous infection models, with a 50% lethal dose (LD₅₀) that is similar to pgm+ strains (13). These observations led to the hypothesis that pgm− Y. pestis strains are not substantially attenuated for growth in blood. Within the pgm is a high pathogenicity island that encodes the biosynthesis and transport of yersiniabactin (Ybt), a siderophore with strong affinity for ferric iron such that it chelates Fe³⁺ from host lactoferrin (LF) and transferrin (TF) (14, 15). In the lungs, LF and TF are highly expressed, thereby creating an environment where would-be pathogens are starved for iron, but the secretion of Ybt allows Y. pestis to chelate Fe³⁺ from LF and TF such that growth is accelerated. Although loss of Ybt strongly attenuates virulence, only very high levels of iron supplementation can rescue the growth defect in the lungs (16, 17). Furthermore, Ybt biosynthesis mutants are significantly more virulent than loss of the entire pgm in the mouse pneumonic plague model (18). These observations suggest additional pathogenesis mechanisms are provided by the expression of the pgm. In this work, we sought to explore this hypothesis and gain an understanding of the attenuation of the pgm mutant in the lungs by studying the MyD88 response against pgm− Y. pestis infection in a murine model of pulmonary infection.

RESULTS

Expression of Y. pestis pigmentation locus fine tunes MyD88-dependent cytokine expression from infected macrophages in vitro.

Previous work showed that purified Ybt induced proinflammatory cytokine expression in cultured respiratory epithelial cells (19). To measure this effect and determine the role of MyD88 during Y. pestis infection, we isolated thioglycolate-elicited peritoneal macrophages from wild-type (WT) and Myd88^−/− mice and infected them with Y. pestis KIM6+pCD1^Ap (wild-type, T3SS+, pgm+ strain, referred to herein as pgm+) or KIM6-pCD1^Ap (T3SS+, pgm−, referred to herein as pgm−) and measured cytokines in the culture supernatant. Production of IL-6, tumor necrosis factor alpha (TNF-α), and IL-1β was dependent on MyD88 regardless of the presence or absence of the pgm (Fig. 1A to C). Furthermore, there was a small but significant impact of the pgm on the amplitude of inflammatory cytokine production in WT macrophages that was absent in the Myd88^−/− macrophages. Expression of TNF-α and IL-1β was approximately 20% higher in the macrophages infected with the pgm+ strain than that in the pgm− strain. These data suggest that the pgm has a small impact on the MyD88 response in macrophages, where it may contribute to amplifying inflammation.

FIG 1.

MyD88 is required for cytokine expression by Y. pestis-infected macrophages. Peritoneal macrophages were harvested from wild-type C57BL/6 (filled bars) or Myd88^−/− (open bars) mice and then infected with Y. pestis KIM6+pCD1^Ap (pgm+) or KIM6-pCD1^Ap (pgm−) at a multiplicity of infection (MOI) of 20. Control macrophages were uninfected (UI). Supernatants were collected at 8 hours postinfection (hpi) and quantified for IL-6 (A), TNF-α (B), and IL-1β (C). Data shown were pooled from three independent trials of three replicates per condition (n = 9). Bars indicate standard error; data were evaluated by two-way ANOVA with Sidak’s multiple-comparison test; *, P < 0.05; ***, P < 0.001; ****, P < 0.0001; NS, not significant.

Early cytokine responses to pgm− Y. pestis are MyD88 dependent and independent.

We therefore looked at the early stage of infection to determine if the loss of the pgm affected the pulmonary MyD88 response to Y. pestis in vivo. Wild-type and Myd88^−/− mice were challenged by intranasal infection, and bacterial load and cytokine production in the lungs were measured at 18 hours postinfection (hpi). The median bacterial titer for wild-type and Myd88^−/− mice was 7 × 10⁴ CFU, which suggests little bacterial growth had occurred in either strain, and in many mice in both groups, bacteria were undergoing clearance (Fig. 2A). This finding indicates that MyD88 is not playing a critical role in early pulmonary defense, which appears opposite to the response to the pgm+ strain, where MyD88 contributed to bacterial clearance in the early stage (18 hpi) (20). In addition, inflammatory cytokines, including IL-1β, IL-6, IFN-γ, and keratinocyte-derived chemokine (KC), were present in higher concentrations in the lungs of WT mice at 18 hpi than in Myd88^−/− mice (Fig. 2B to E). Therefore, MyD88 contributed to the production of inflammatory cytokines, whereas it was not critical for limiting bacterial growth.

FIG 2.

Early MyD88-dependent cytokine response in the lungs against nonpigmented Y. pestis. Wild-type C57BL/6 and Myd88^−/− mice were challenged by intranasal infection with the indicated volume (10 or 30 μl) of 1 × 10⁶ CFU (CFU) Y. pestis KIMD27 and analyzed for bacterial loads and cytokine responses. Mice were euthanized after 18 h; lungs and spleen were collected and homogenized for enumeration of bacterial titer (A, F). Lung homogenates were additionally analyzed for IL-1β (B, G), IL-6 (C, H), IFN-γ (D, I), and KC (E, J). Bacterial titer is shown as CFU per whole lung or spleen; dashed line indicates limit of detection; number of mice with undetectable bacteria is indicated below the line; bars indicate median. Cytokines are shown as mean + SEM. Data were collected in 2 to 3 independent trials, with 19 (10 μl) or 10 to 11 (30 μl) per group; data were analyzed by Mann-Whitney; *, P < 0.05; **, P < 0.01; ****, P < 0.0001; NS, not significant.

The differences of this response compared with what we had seen for the pgm+ infection, however, could be caused by a dosing effect, since more than a 100× higher dose of pgm− Y. pestis is required for lethal infection in this model. The intranasal infectious dose of pgm− Y. pestis results in deposition of 10⁴ to 10⁵ bacteria in the lungs, whereas the dose for pgm+ Y. pestis was 10³ CFU in the previous study (20). To address this potential variable, we lowered the volume of the inoculating dose. When the intranasal dosing volume of Y. pestis is reduced (10 μl compared to 30 μl), less than 1% of the inoculum reaches the lungs; yet, the dose is still lethal due to systemic spread and secondary septicemic plague (21). As expected, only 10² to 10³ CFU was recovered at 18 hpi in WT or Myd88^−/− mice (Fig. 2F). These data are consistent with the high volume/high deposition model, suggesting MyD88 may not play a major role in limiting growth of the pgm mutant in the lungs. Consistent with the reduced bacterial load, lower levels of inflammatory cytokines were present at 18 hpi than that we observed in the high-volume challenge model (Fig. 2G to J). Low levels of pulmonary IL-6 appeared to be at least partially dependent on MyD88, whereas IL-1β and IFN-γ appeared lower in Myd88^−/− mice but were not detectably different between groups. In contrast, KC was not MyD88 dependent (Fig. 2J). Overall, these results contrast what we found during the high-volume infection and suggest that the host response to Y. pestis could be altogether different in the presence and absence of the pgm in vivo.

The pigmentation locus is required to suppress early MyD88-independent pulmonary host defense.

Our previous work included analysis of a different biovar (CO92, Orientalis biovar) compared to that used here (KIMD27, Mediaevalis biovar). To confirm that the observed differences were not caused by biovar-specific responses, we conducted a side-by-side comparison of pgm+ and pgm− infection from the Mediaevalis biovar. At a challenge dose of approximately 2× LD₅₀, we recovered an increased titer of the pgm+ strain in the lungs and bronchial alveolar lavage fluid (BALF) of the Myd88^−/− mice at 18 hpi compared with that of WT mice, indicating MyD88 contributes to an initial protective response (Fig. 3A). In contrast, the pgm− strain was present at similar levels as the pgm+ strain in the lungs and BALF of WT mice. In Myd88^−/− mice, however, there was no increase in titer of pgm− Y. pestis in lungs or BALF, indicating MyD88-independent host defense can control growth of this strain in the early phase.

FIG 3.

Role for the pigmentation locus in modulation of the MyD88 inflammatory response. Groups of 5 WT (filled) and Myd88^−/− (open) mice were challenged by intranasal infection with 1,000 CFU Y. pestis KIM6+pCD1^Ap (pgm+) or KIM6-pCD1^Ap (pgm−). At 18 hpi, mice were euthanized, alveolar lavage fluid (BALF) collected, and lungs homogenized in 1 ml sterile PBS (LH). (A) Bacterial titer; (B) IL-1β; (C) TGF-β; (D) IL-10; (E) TNF-α; (F) IL-6; (G) IFN-γ. Data shown are expressed per ml of BALF or LH and were collected in two independent trials; dotted line indicates lower limit of detection; the number of mice in the group that had undetectable bacteria are indicated below the lower limit of detection. Data were evaluated by two-way ANOVA; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

We compared cytokine and chemokine levels in these mice, looking for those that were dependent on MyD88 and expressed more prominently in the lungs of pgm+-infected mice. Consistent with previous data, KC was MyD88 dependent in the lungs, but not the BALF, of pgm+-infected mice (Fig. 3B). In contrast, no difference in KC was observed in the lungs of WT and Myd88^−/− mice infected with pgm− Y. pestis at 18 hpi. Likewise, TNF-α also appeared to be reduced in the lungs of Myd88^−/− mice infected with pgm+ Y. pestis compared with that of pgm− Y. pestis (Fig. 3C). These data indicate that pgm+ induces a distinct MyD88 response that includes KC and TNF-α whereas the pgm− strain does not. The inflammatory response in the lungs may therefore be critical to the observed differences in bacterial growth and disease in Myd88^−/− mice infected with pgm+ or pgm− Y. pestis strains. Strikingly, a similar pattern, although a different result, was present for TGF-β. In the lungs of the pgm+-infected Myd88^−/− mice, elevated TGF-β was present (Fig. 3D). In contrast, this finding was not true in the lungs of Myd88^−/− mice that were infected with the pgm− strain, suggesting the pgm may cause reduced anti-inflammatory cytokine production in the lungs. However, this effect was limited to TGF-β. The other major anti-inflammatory cytokine IL-10 was present in very small amounts and appeared to be MyD88 independent (Fig. 3E). The overall amounts of IL-1β were also relatively low for both pgm+ and pgm− WT lungs, but this amount appeared to be dependent on MyD88 in both cases (Fig. 3F). In addition, IL-6 and IFN-γ were present in small amounts in the infected mice in all groups (Fig. 3G and H). This analysis strongly suggests that the pgm allows Y. pestis to modify the inflammatory environment in the lungs through MyD88.

In the airway, however, the results were different. When we measured the mean concentration of IL-1β, TGF-β, and IL-10 in BALF recovered from WT mice infected with the pgm− mutant, we found increased cytokine production compared with that of the group infected with the pgm+ strain which may be indicative of a controlled inflammatory response (Fig. 3D to F). As with the lung homogenate, IFN-γ and IL-6 were present in very low concentrations. These data suggest that the inflammatory response in the BALF has a different pattern than that in the lung tissue and is overall similar between infected pgm+ and pgm− mice with little dependence on MyD88. This result is consistent with the similar bacterial growth in the airway at this time point and may indicate tissue-specific MyD88 responses involving the pgm.

Contributions from myeloid and nonmyeloid MyD88 control growth of pgm− Y. pestis.

Even though there was no difference in the early control over pgm− infection of WT and Myd88^−/− mice, the mutant mice were highly susceptible to lethality and succumbed to disease much more rapidly, with 100% mortality on days 3 and 4 postinfection (dpi) compared with 80% mortality in WT mice on days 5 to 9 postinfection (Fig. 4A). In contrast, Y. pestis KIM6−, which lacks the type III secretion system, were avirulent in WT and Myd88^−/− mice; even at a very high challenge dose (>10⁷ CFU), all the mice in both groups survived (data not shown). This finding indicates, not unexpectedly, that expression of the T3SS is essential for overcoming MyD88-independent immunity.

FIG 4.

MyD88 restricts growth of iron-starved, nonpigmented Y. pestis. (A, B) Groups of 5 to 7 WT or Myd88^−/− mice (A) and WT or LysM^Cre-Myd88^Flox (LysM) mice (B) were challenged by intranasal infection with 1 × 10⁶ CFU Y. pestis KIMD27 (pgm−) and monitored for survival over 14 days. Data shown were collected in 2 independent trials; the number per group is indicated in parentheses. (C to H) Groups of 4 to 6 WT, LysM^Cre-Myd88^Flox, and Myd88^−/− mice were challenged by intranasal infection with Y. pestis KIMD27. On day 3 postinfection, mice were euthanized and lungs, liver, spleen, and blood were collected. (C to I) Lung homogenates were quantified for bacterial load (C), lactoferrin (D), and transferrin (E); and cytokines KC (F), IFN-γ (G), IL-6 (H), and IL-1β (I) (plus 6 other cytokines shown in Fig. S1). Data shown are expressed per organ (1 ml of homogenate) and were collected in 2 independent trials; n = 9 WT, 8 Myd88^−/−, and 7 LysM mice (3 Myd88^−/− mice and 1 LysM mouse succumbed to infection prior to harvest); dotted line indicates limit of detection (number undetectable indicated below the line); bars indicate median. Data were analyzed by Mantel-Cox log rank test (A, B) or Kruskal-Wallis test, followed by Dunn’s multiple-comparison test (C to I); *, P < 0.05; ***, P < 0.001; NS, not significant.

To determine whether myeloid expression of MyD88 is necessary for protection, we asked if mice lacking MyD88 in myeloid cells were more susceptible to infection. We tested LysM^Cre-Myd88^Flox mice, which lack MyD88 in myeloid (lysozyme-expressing) cells, for susceptibility to intranasal infection with Y. pestis KIMD27 (22). The LysM^Cre-Myd88^Flox mice were significantly more susceptible to infection than wild-type mice, indicating that the loss of MyD88 in myeloid cells significantly impaired host defense against pgm− Y. pestis (Fig. 4B). However, the kinetics (5 to 8 days) did not resemble the full Myd88 knockout (3 to 4 days), indicating important contributions from nonmyeloid cells in the MyD88 response.

We compared bacterial growth in the lungs of WT, LysM^Cre-Myd88^Flox, and Myd88^−/− mice on day 3 postinfection. Strikingly, 100% of the Myd88^−/− mice harbored greater than 10⁵ CFU in the lungs, which is nearly 10,000-fold greater than the median titer recovered from WT mice or that recovered from Myd88^−/− mice at 18 hpi, indicating substantial growth of the iron-starved strain (Fig. 4C). In contrast, control over growth of pgm− Y. pestis in the lungs was retained in the LysM^Cre-Myd88^Flox mice, indicating nonmyeloid cells are necessary for MyD88-dependent bacterial clearance in the lungs. Previous work has shown that TLR signaling induces the expression of the iron chelator lactoferrin by bronchial epithelial cells, thereby contributing to nutritional immunity against infection by microbes that do not express siderophores (23). Since nonmyeloid MyD88 was critical for controlling growth of the iron-starved pgm− strain, we asked if lactoferrin was reduced in the Myd88^−/− mice. All 3 groups of mice harbored high levels of lactoferrin in the lungs (Fig. 4D). Similarly, transferrin was high in the lungs of Myd88^−/− mice (Fig. 4E). These data indicate that nutritional immunity in the lungs of Myd88^−/− mice, although functional, was unable to prevent growth of the iron-starved pgm− strain.

The increase in bacterial load in the lungs of Myd88^−/− mice was associated with a significant increase in KC compared with that of WT or LysM^Cre-Myd88^Flox mice (Fig. 4F). While neutrophil chemokine MIP2 was also significantly higher in the lungs of Myd88^−/− mice, IL-6, IFN-γ, and IL-1β were not detectably different between any group of mice (Fig. 4G to I, see Fig. S1 in the supplemental material). Since the bacterial population in the lungs of Myd88^−/− mice was much higher than that of the other groups, these data suggest that MyD88^−/− mice do not ramp up cytokine expression in the lungs as the infection progresses. Yet, since neutrophil chemokines KC and MIP2 were MyD88 independent, the combined analysis suggests that Myd88^−/− mice may have recruited neutrophils to the infected lungs that failed to control bacterial growth.

Similar to the lung, bacterial titers in the liver, spleen, and blood on day 3 postinfection were substantially higher in Myd88^−/− mice than in WT or LysM^Cre-Myd88^Flox mice (Fig. 5A). To understand tissue-specific roles for MyD88, we measured cytokines in the liver, spleen and serum and found different cytokine responses. For example, IFN-γ was strictly dependent on MyD88 in the serum and not in the liver and spleen (Fig. 5B). However, similar levels of IFN-γ were recovered in LysM^Cre-Myd88^Flox mice and in the WT. In addition, liver and serum IL-10 levels were higher in Myd88^−/− mice, suggesting a MyD88-independent anti-inflammatory response occurred, although IL-10 was not elevated in the lungs or spleen (Fig. 5C, Fig. S1). This was not true of LysM^Cre-Myd88^Flox mice, which might indicate that the anti-inflammatory response occurs in the resident macrophages in the liver (which do not express lysozyme), leaving WT levels of MyD88 expression in LysM^Cre-Myd88^Flox mice (24). Levels of KC, IL-6, IL-1α, RANTES, and TNF-α were similar in all three mouse strains in the serum, liver, and spleen at 72 hpi, whereas IL-1β appeared to be MyD88 independent in the serum (Fig. 5D to F, see Fig. S2 in the supplemental material). In the spleen of the Myd88^−/− mice, but not the other tissues or blood, IFN-β was elevated compared with that in WT or LysM^Cre-Myd88^Flox mice (Fig. S2). Collectively, these data show tissue specificity in MyD88-independent cytokine expression of KC, MIP2 (lungs), IL-10 (liver and blood), and IL-1β (spleen). As none of the cytokines were changed in the LysM^Cre-Myd88^Flox mice compared with the WT, it appears that myeloid and nonmyeloid cells make important contributions to the MyD88-inflammatory response during the disease phase. Notably, however, the Myd88^−/− mice did not appear to have a systemic hyperinflammatory response at 72 hpi even though 100% of these mice would be expected to succumb to infection within 24 h. This finding suggests that the overwhelming bacterial growth leads a very rapid progression of disease to lethality in Myd88^−/− mice. Similarly, it appears that MyD88 is necessary for the development of sepsis.

FIG 5.

Requirement for MyD88 for IFN-γ expression during disease phase. Additional samples from mice challenged in Fig. 3, groups of 4 to 6 WT, LysM^Cre-Myd88^flox, and Myd88^−/− mice challenged by intranasal infection with 1 × 10⁶ CFU Y. pestis KIM D27, were analyzed for bacterial loads, cytokines, and chemokines. On day 3 postinfection, the liver, spleen, and serum were quantified for CFU (A), IL-6 (B), IFN-γ (C), KC (D), IL-10 (E), and IL-1β (F) (plus 5 other cytokines shown in Fig. S2). Data shown are expressed as the amount per tissue or 1 ml serum and were collected in 2 independent trials; n = 9 WT, 8 Myd88^−/−, and 7 LysM (1 LysM and 3 Myd88^−/− mice succumbed to infection prior to 72 hpi); bars indicate standard error. Data were analyzed by Kruskal-Wallis test, followed by Dunn’s multiple-comparison test; *, P < 0.05; **, P < 0.01; NS, not significant.

Tissue-specific roles for MyD88 in neutrophil recruitment and pathology.

To determine if Myd88^−/− mice were able to recruit neutrophils to the lungs and develop primary lung disease, we assessed histopathology and immunohistochemistry of the lungs, liver, and spleen. At 72 hpi, the lungs of Myd88^−/− mice harbored multiple inflammatory foci, which appeared with granular morphology and stained positive for Gr1, that were essentially absent in WT mice (Fig. 6A). Although the number of inflammatory foci per low power field in the lungs of Myd88^−/− mice was significantly increased, there was minimal associated alveolar necrosis and the intra-alveolar congestion appeared mild and acellular. Therefore, without MyD88, only mild immunopathology was present even though there was a high degree of bacterial growth.

FIG 6.

Tissue-specific dependence on MyD88 for recruitment of neutrophils during infection by pgm− Y. pestis. Groups of 4 to 5 wild-type C57BL/6 (left) and Myd88^−/− (right) mice were challenged by intranasal infection with 1 × 10⁶ CFU Y. pestis KIMD27 (pgm−). On day 3 postinfection, mice were euthanized and tissues processed for histochemistry, staining with hematoxylin and eosin (H&E, top) or immunohistochemistry with anti-Gr1 (IHC, bottom). Shown are representative lesions commonly observed in the lungs (A), liver (B), and spleen (C). Scale bar indicates 50 μm (top) or 200 μm (bottom). (D, E) Mean severity scores for anti-Gr-1 IHC (D) and H&E (E). Data shown were collected in 2 independent trials, n = 8 to 9; data were analyzed by Student’s t test; **, P < 0.01; ****, P < 0.0001; NS, not significant.

In contrast, in the livers of WT mice, there were multiple, sometimes large, neutrophilic foci; they were rare in the livers of Myd88^−/− mice, consistent with the presence of elevated anti-inflammatory cytokine IL-10 (Fig. 6B). In the spleen, there were no detectable differences in Gr1+ stained cells or in lesion severity between the two groups of mice (Fig. 6C). Quantification of these lesions showed a significant increase in neutrophils in the lungs of Myd88^−/− mice, whereas neutrophil recruitment to the liver was significantly reduced (Fig. 6D and E). These data are consistent with the cytokine data, indicating that MyD88-independent inflammatory and neutrophilic responses to infection by pgm− Y. pestis are tissue specific.

To verify that the cytokine responses we observed were not simply an indication of disease progression, we evaluated the tissues of WT and LysM^Cre-Myd88^Flox mice at a later time point. On day 5 postinfection, there were no significant differences in the bacterial titer in the lungs, liver, or spleen between groups of mice (see Fig. S3A in the supplemental material). There was a decrease (rather than the increase seen during the disease phase of the full knockout) in serum IL-10 in the LysM^Cre-Myd88^Fl mice compared with the WT, while other cytokines, including TNF-α and IFN-γ, were not detectably different (Fig. S3B to D). We also examined tissues from infected LysM^Cre-Myd88^Fl mice for pathological lesion severity in the lungs and liver. Unlike the full knockout on day 3, both groups of mice harbored neutrophilic inflammatory lesions in the liver with many bacteria, whereas in the lungs, there was mild pathology, few neutrophils, and no bacteria found (Fig. S3E). These data are consistent with the interpretation of day 3, suggesting tissue-specific roles of the MyD88 response may be initiated by nonmyeloid cells. Furthermore, it appears the greatest impact of MyD88 may be to defend against the blood infection, perhaps due to the production of IFN-γ.

IFN-γ protects mice from secondary septicemic plague.

The role of IFN-γ during Y. pestis infection has been previously addressed but remains unclear. Early work showed that treatment of mice with IFN-γ protected them from lethality following intravenous (i.v.) infection with Y. pestis KIMD27, suggesting that the production of IFN-γ was normally delayed during the infection (25). A later study showed that anti-IFN-γ accelerated time to death in mice challenged by intranasal (i.n.) infection with KIMD27, suggesting there could be a small role for the cytokine in host defense (26). From these studies, we anticipated that MyD88-dependent serum IFN-γ may not contribute substantially to the phenotype of the Myd88^−/− mice. To verify this idea, we characterized Ifng^−/− mice with Y. pestis KIMD27. Strikingly, there was a significant increase in the lethality of Ifng^−/− mice compared with the WT (Fig. 7A). Histopathology was conducted on lungs, liver, and spleen of moribund mice. As with wild-type mice, moribund Ifng^−/− mice developed severe pathology in the liver with numerous bacteria and neutrophilic inflammatory foci (Fig. 7B and C). No detectable difference in the severity of inflammatory and necrotic lesions were found in the liver, lungs, or spleen between moribund WT and Ifng^−/− mice (Fig. 7D). This result is consistent with a more rapid progression of plague due to overwhelming systemic bacterial growth.

FIG 7.

IFN-γ is required for host defense against intranasal challenge by Y. pestis pgm−. Groups of 5 to 7 WT and Ifng^−/− mice were challenged by intranasal infection with 1 × 10⁶ CFU Y. pestis KIMD27 (pgm−). (A) Mice were monitored for survival over 14 days. Data shown were collected in 2 independent trials; number per group is indicated in parentheses; data were analyzed by Mantel-Cox log rank test; ****, P < 0.0001. (B to D) Tissues were assessed by histopathology postmortem; (B, C) representative lesions from the liver of WT (B) and Ifng^−/− (C) mice; arrowheads show necrotic lesions with bacteria; scale bar indicates 100 μm. (D) Lesion severity was scored 0 to 3 in each tissue, and mean scores with standard error are shown (n = 6 per group). Data were analyzed by t test; NS, not significant.

In the lungs, a small, but not significant, increase in the median titer was found at 72 hpi, suggesting a minor role at most for IFN-γ in controlling pgm− bacterial growth the lungs (Fig. 8A). Similar results were found in the spleen. In contrast, the median titer in the liver and blood was significantly higher in the Ifng^−/− mice, with up to 5-orders of magnitude higher bacterial titer compared with that of WT mice at this time point. This indicates a major role for IFN-γ in controlling bacterial growth in the blood and in the liver. Opposing the bacterial loads, however, IL-1β was significantly decreased in the lungs of Ifng^−/− mice and suggests IFN-γ may contribute to the production of IL-1β in the lungs (Fig. 8B). Other cytokines, including IL-6, also appeared reduced in Ifng^−/− mice, although the difference was not significant (Fig. 8C and D). No major impact on cytokine expression was observed in the serum of Ifng^−/− mice despite increases in bacterial titer, suggesting a possible role for IFN-γ in the hyperinflammatory response (Fig. 8E to G). Histopathology of the liver at 72 hpi showed increased lesion severity of the Ifng^−/− mice, consistent with more rapid progression of infection (Fig. 8H). Together, these data suggest that the loss of control over growth in the blood and liver led to an increased rate of progression to lethality in the Ifng^−/− mice.

FIG 8.

Histopathology of Ifng^−/− mice suggests more rapid progression of secondary septicemic plague. Groups of 5 to 6 wild type (C57BL/6) and Ifng^−/− mice were challenged by intranasal infection with 1 × 10⁶ CFU Y. pestis KIMD27 (pgm−), and the infection was assessed at 72 hpi. Lungs, liver, spleen, and blood were collected for (A) quantification of bacterial load (A) as well as IL-1β (B, D) and (C, E) IL-6 (C, E) in the lung (B, C) or serum (D, E). Data shown are expressed as CFU or cytokine/organ or ml of blood or serum and were collected in 2 independent trials (1 Ifng^−/− mouse succumbed to infection prior to harvest); n = 11 WT, 10 Ifng^−/− mice; bars indicate median (A to C) or mean with standard error (D, E). Data were analyzed by Mann-Whitney test; *, P < 0.05; ***, P < 0.001; ****, P < 0.0001; NS, not significant. (F) Histopathology analysis of lesion severity in each tissue harvested at 72 hpi (n = 3 per group); mean scores with standard error are shown; data were analyzed by t test; NS, not significant.

DISCUSSION

The prevailing model for pulmonary Y. pestis infection is that the T3SS suppresses inflammatory cytokines in the early stage of infection, setting up a biphasic inflammatory response (7). Nonpathogenic bacteria could survive in this environment and ultimately grow, suggesting an immune-privileged niche is created by the T3SS (5). When Y. pestis carry the T3SS but lack the pgm, however, there is no immune-privileged niche in the lungs, and we found that MyD88-dependent inflammation is able to control bacterial growth. The accepted dogma is that the lack of growth of the pgm mutant in the lungs is due to a loss of the iron chelator yersiniabactin, which was thought to facilitate the evasion of nutritional immunity. In this work, we show that, in fact, the pgm− strain needs to be cleared from the lungs and that without the Toll/IL-1 receptor signaling adapter MyD88, the strain is able to grow to a high titer even in the iron-depleted lung tissue. However, Myd88^−/− mice did not develop a hyperinflammatory response, and the overall apparent damage due to immunopathology in the infected lungs and liver was less severe.

When we compared the pulmonary host response to pgm+ and pgm− Y. pestis, we identified differences as early as 18 hpi. Although it is true that the pgm− strain is expected to be iron starved for growth in the lungs, there was no significant difference between early growth of pgm+ and pgm− in WT mice, suggesting that there is either little growth of the pgm+ strain in the early phase and/or that there is little clearance of the pgm− strain. In Myd88^−/− mice, however, the early growth of pgm+ was enhanced, while that of pgm− was not, suggesting immune evasion mechanisms may be encoded within the pgm. Further support for this conclusion comes from a comparison of the corresponding cytokines. While MyD88 was largely dispensable for early inflammatory cytokine production following pgm− infection, the pgm+ Y. pestis strain induced MyD88-dependent activation of KC and TNF-α production. These observations suggest a proinflammatory effect is caused by pgm-expressing Y. pestis that may ultimately redirect host defense toward immunopathology.

The only virulence factor in the pgm that is known to be critical to pneumonic plague is yersiniabactin; although, there is evidence for other virulence factors encoded in the locus (18). Yersiniabactin is necessary for Y. pestis to acquire iron from lactoferrin and transferrin. In the lungs of Myd88^−/− mice, we found abundant lactoferrin and transferrin; yet, the pgm mutant was able to grow. This finding suggests alternative iron sources or the existence of a second iron chelator that can partially compensate for Ybt. In the genome of Y. pestis, there is a second genetic site, the ysu locus, that may encode this redundancy. The ysu cluster codes for the biosynthesis of the hydroxymate siderophore yersiniachelin (Ych), which has been purified from the surface of pgm− Y. pestis, suggesting it is expressed and could play a role in chelating iron (27, 28). To date, however, there is no known role for Ych in virulence. Even if Ych is upregulated, it would not explain the cytokine differences. One other pgm-carried gene ripA has been suggested to be important to the virulence of Y. pestis since the ripA mutant appears defective for intracellular growth (29). It is intriguing to speculate that the host response to intracellular bacteria may influence the inflammatory pathology; however, ripA has not yet been demonstrated to have a role in the development of plague.

Beyond the issue of iron chelation and inflammation in the lungs, we found that the pgm mutant is restricted for growth in the liver and blood by MyD88-dependent IFN-γ. This observation contradicts a previous report suggesting no role for IFN-γ against intravenous infection by KIMD27 where the liver and blood become heavily colonized (25). It may be that the IFN-γ produced from infected organs is the primary mechanism whereby serum IFN-γ is derived or that the lung infection by the pgm− strain may prime the host. One effect observed in the Ifng^−/− mice was a significant decrease in lung IL-1β, suggesting IFN-γ signaling may influence inflammation. In the mucosal tissue of the gut, epithelial IFN-γ receptor signaling was shown to sensitize cells to undergo Salmonella sp.-induced programmed cell lysis, resulting in the release of inflammatory cytokines, such as IL-1β (30). It is reasonable to hypothesize that the respiratory mucosa responds similarly. Furthermore, previous work suggests that IL-1β may play a role in the inflammatory phase of infection by pgm+ Y. pestis; although, in those experiments, high challenge doses were used which likely amplified the inflammatory response. Perhaps it would be informative in the future to reexamine IL-1β signaling at lower doses of Y. pestis (pgm+ and pgm−) or perhaps a low volume i.n. dose to determine whether IL-1β is an important modulator of the MyD88 response that contributes to or protects from pneumonic and septicemic plague.

MATERIALS AND METHODS

Bacterial strains.

Yersinia pestis KIMD27 (T3SS+, pgm−) was isolated as a spontaneous mutation that resulted in the loss of pigmentation on Congo red agar (31). Y. pestis KIM6+pCD1^Ap (T3SS+, pgm+) is a wild-type strain carrying a recombinant pCD1 plasmid that confers ampicillin resistance; Y. pestis KIM6-pCD1^Ap (T3SS+, pgm−) is nonpigmented mutant carrying ampicillin resistance on pCD1 (18, 32). Bacteria were routinely grown fresh from frozen stock by streaking for isolation onto heart infusion agar (HIA) plates. For intranasal challenge studies, a single colony was used to inoculate calcium-supplemented heart infusion broth (Ca-HIB) and grown for 18 to 20 hours at 37°C and 125 rpm. Bacteria were diluted in sterile phosphate-buffered saline (PBS) for challenge in mice. Following overnight growth at 26°C, cultures were diluted 1:10 in HIB, grown for 1 hour at 26°C, and then shifted to 37°C for 1 hour prior to infection of macrophages.

Vertebrate animals.

All animal procedures were performed in compliance with the Office of Laboratory Animal Welfare and the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the University of Missouri Animal Care and Use Committee.

C57BL/6J mice were the inbred strain background for all mice. Mice with conditional deficiencies of MyD88 in myeloid cells (LysM^Cre-Myd88^flox) were previously generated by breeding mice with floxed Myd88 alleles to mice expressing Cre recombinase under the LysM promoter (33, 34). Wild-type, Ifng^−/−, Myd88^−/−, and conditional knockout parental strains were originally from Jackson Laboratories (Bar Harbor, ME, USA). Mice were bred and maintained in-house at the University of Missouri. Male and female wild-type and mutant mice, ranging from 15 to 30 g, were used for challenge experiments. For challenge, mice were lightly anesthetized with inhaled isoflurane prior to intranasal instillation of Y. pestis KIMD27. All infected mice were monitored by daily assignment of health scores, which involved assessments of their appearance and activity. Animals that survived to the end of the 14-day observation period or were identified as moribund (defined by pronounced neurologic signs, inactivity and severe weakness) were euthanized by CO₂ asphyxiation, followed by bilateral pneumothorax or cervical dislocation, which are methods approved by the American Veterinary Medical Association Guidelines on Euthanasia.

Infection studies.

Doses used in the challenge studies are indicated in the figure legends, and they were delivered in 10- to 30-μl instillation volumes, as previously described (21). At 18 hours, 3 days, or 5 days postinfection (as indicated in the text and figures), mice were euthanized and blood removed. In some experiments, lungs were lavaged in situ with 3 ml sterile PBS to collect bronchial alveolar lavage fluid (BALF). Lungs as well as liver and/or spleen were collected as indicated. Tissues were homogenized in sterile PBS and then, along with blood, diluted and plated on HIA in duplicate for enumeration of bacteria. Serum was collected following centrifugation and stored with tissue homogenates at –80°C until analysis. Serum and tissue homogenates were analyzed with a multiplex cytokine assay (Millipore-Sigma, Burlington, MA, USA) or by enzyme-linked immunosorbent assay (ELISA) for IFN-β (PBL Assay Science, Piscataway, NJ, USA), TGF-β (R&D Systems, Minneapolis, MN, USA), lactoferrin (G Biosciences, St. Louis, MO, USA), and transferrin (GenWay, San Diego CA, USA).

Histopathology.

Lungs were perfused and fixed in 10% formalin along with liver and spleen. Organs were further processed for paraffin embedment, blocked in wax, and cut into 5-μm sections. Tissue sections were stained with hematoxylin and eosin (H&E) or by immunohistochemistry, and coverslips were permanently affixed to stained slides. Researchers were blind to sample identities for the analysis of lesion severity. Severity scoring was based on increasing size of necrosis, inflammatory foci, edema, hemorrhage, or thrombosis; and the frequency and lesion size as well as percentage of the tissue affected. Lesions were assigned severity scores ranging from 0 to 3, and bacteria were noted when visible by H&E.

Cellular infection assay.

Peritoneal macrophages were elicited by intraperitoneal (i.p.) injection with 1 ml BBL Fluid Thioglycolate Medium (Becton, Dickinson, Sparks, MD, USA) (35). Macrophages were harvested by peritoneal lavage 4 days after thioglycolate treatment and cultured in Dulbecco modified Eagle medium (DMEM) with 10% fetal bovine serum (FBS) at 37°C and 5% CO₂ for in vitro studies. Macrophages were seeded at 5 × 10⁵ cells per well in 24-well plates 1 hour prior to infection. Yersinia sp. was added at a 20:1 multiplicity of infection (MOI) and centrifuged at 41 × g for 5 minutes at room temperature to initiate infection (36). At 8 hours postinfection, cell supernatants were collected and stored at −80°C until assayed for cytokine production by a multiplex cytokine assay (Millipore-Sigma, Burlington, MA, USA). Triplicate wells were infected for all trials.

Statistical evaluation.

Data from the controls in each trial were first tested by analysis of variance (ANOVA) to verify the similarity between trials. In all cases, the control samples indicated similarity between trials, allowing data from independent trials to be pooled for analysis. For macrophage infections, each trial was conducted in triplicate using 3 independent bacterial cultures for each infection condition and combined for analysis of statistical significance. Statistical significance was evaluated using Prism 7 (GraphPad Software, La Jolla, CA). Specific statistical evaluations are indicated in the figure legends; significance was generally concluded when the P value was <0.05.