Histopathology in a murine model of anthrax

Abstract

Systemic anthrax infection is usually fatal even with optimal medical care. Further insights into anthrax pathogenesis are therefore urgently needed to develop more effective therapies. Animal models that reproduce human disease will facilitate this research. Here, we describe the detailed histopathology of systemic anthrax infection in A/J mice infected with Bacillus anthracis Sterne, a strain with reduced virulence for humans. Subcutaneous infection leads to systemic disease with multiple pathologies including oedema, haemorrhage, secondary pneumonia and lymphocytolysis. These pathologies bear marked similarity to primary pathologies observed during human disease. Therefore, this simple, small animal model will allow researchers to study the major pathologies observed in humans, while permitting experimentation in more widely available Biosafety Level 2 facilities.

Bacillus anthracis is a gram-positive, spore-forming bacterium that causes anthrax in animals and humans (Dixon et al. 1999). Anthrax takes on three clinically distinct forms determined by the route of exposure: cutaneous, gastrointestinal and inhalation anthrax. Without treatment, the most deadly form, inhalation anthrax, almost invariably proceeds to fatal systemic infection. Even with modern antibiotics and supportive care, half of the people, who developed inhalation anthrax after exposure to B. anthracis-laden letters in 2001, died (Jernigan et al. 2002). Furthermore, many of those who recovered were left with long-term disability (Reissman et al. 2004).

Because of its ease of production and dispersal, B. anthracis remains a potent threat as a biological weapon (Jernigan et al. 2001). To develop more effective therapies for such a lethal disease, a better understanding of host–pathogen interaction is urgently needed. A simple, reproducible small animal model will greatly facilitate this research effort.

Here, we investigate the potential of a murine model of systemic anthrax to simulate human disease. We take advantage of previous observations by Welkos et al. (1986) regarding differential susceptibility of murine strains to anthrax. Importantly, they report that the readily available A/J and DBA/2J murine strains are susceptible to a capsule-deficient anthrax strain known as Sterne. Based on the absence of the capsule-encoding virulence plasmid, pXO2, Sterne is attenuated in its ability to infect humans, cattle and most murine strains. Therefore, it can be handled safely using standard Biosafety Level 2 precautions (Richmond & McKinney 1999), greatly simplifying experimentation.

Later studies by Welkos & Friedlander (1988) determined that the known C5-complement deficiency in A/J and DBA/2J murine strains accounts for their susceptibility to capsule-deficient B. anthracis. On the basis of the role of C5 in formation of the bacteriocidal membrane-attack complex, their studies imply that the B. anthracis capsule normally counteracts C5-dependent destruction of bacteria in murine hosts. As a result, they have identified directly compensatory mutations in pathogen and host that lead to disease approximating the course and lethality of wild-type B. anthracis infection. Here, we provide the first detailed description of the histopathology of terminal B. anthracis Sterne infection in C5-deficient mice and show that this host–pathogen combination reproduces primary pathologies observed during human anthrax infection.

Methods

Strains

Eight-week-old female A/J mice were from Harlan Sprague Dawley (Indianapolis, IN, USA) and DBA/2J and Balb/c mice were from Charles River Laboratories (Wilmington, MA, USA). B. anthracis Sterne strain 7702 was used in all experiments (Pezard et al. 1991).

Murine infection

Bacteria were replated from frozen stocks onto trypticase soy agar with 5% defibrinated sheep blood (TSA II; Becton Dickinson, Franklin Lakes, NJ, USA) and grown at 37 °C with 5% CO₂ for 18 h prior to each experiment. To prepare the infectious inoculum, we resuspended colonies of bacteria in sterile phosphate-buffered saline (PBS) (Cellgro, Herndon, VA, USA). Bacterial concentration was estimated by optical density based on previously determined correlations with colony-forming units (CFU) (data not shown). The bacterial suspension was then diluted in sterile PBS to obtain the desired infectious inoculum. After anaesthesia with 2,2,2-tribromoethanol (Sigma-Aldrich, St. Louis, MO, USA), 50 µl of bacterial suspension was injected subcutaneously into the shaven right flank of each mouse. Groups of five A/J mice were infected with 10², 10⁴ and 10⁶ CFU, respectively. A small number of DBA/2J mice was also similarly infected.

All procedures and care were carried out in accordance with the Institutional Animal Care and Use Committee at Beth Israel Deaconess Medical Center in an AAALAC-approved facility.

Tissue processing

Tissues were collected either shortly after death from infection or from terminally moribund animals after euthanasia with phenobarbital. Harvested tissue was fixed in PBS containing 3.7% formaldehyde (Fisher Scientific, Pittsburgh, PA, USA), dehydrated, embedded in paraffin, sectioned and stained with haematoxylin and eosin or orcein (Poly Scientific R&D Corp, Bayshore, NY, USA).

Image acquisition

Photomicrographic images were acquired using an inverted Nikon Eclipse TE100 microscope equipped with a colour CCD camera and IPLab imaging software (Scanalytics, Rockville, MD, USA). All images shown in figures are from A/J mice except for Figure 3(a,c), which are from DBA/2J mice.

Figure 3.

Lung and skin. (a) Section of lung tissue affected by secondary pneumonia. The alveolar septum and visceral pleura are expanded with neutrophils, macrophages and haemorrhage. Adjacent apoptotic thymus is visible in the lower right. (b) At higher power, chains of bacilli are visible in interstitial lung tissue. (c) An orcein stain marks the alveolar basement membrane (brown linear structures), demonstrates the expansion of the interstitial tissue and makes apparent the breaching and destruction of some alveoli by the inflammatory infiltrate. Bacteria are evident as blue linear structures. (d) Flank skin around site of inoculation. The epidermis is intact (e.g. no eschar); however, there is extensive underlying dermal oedema, indicated by the wide separation of collagen fibres. Scattered neutrophils and haemorrhage are present. (e) The hypodermis at the site of inoculation shows sheets of bacilli superficial and deep to the striated muscle layer of murine skin. Note the minimal associated inflammatory response. Edema and vascular congestion are also evident. (f) Abdominal skin distant to site of inoculation showing extensive haemorrhage in the hypodermis and subcutaneous tissue. H&E, original magnification: (a) ×200, (b) ×600, (d) ×400, (e) ×400 and (f) ×40. Orcein, original magnification: (c) ×200.

Results

Clinically, the infected A/J mice all became moribund or died as a result of subcutaneous infection with doses of vegetative Sterne bacteria ranging from 10² to 10⁶ organisms. While mice infected with 10² CFU died on day 3 or 4 post-infection, mice infected with 10⁴ or 10⁶ CFU died on day 2 or 3. Terminal signs of illness consisted of a staggering gait, decreased mobility and tachypnea, ending with deep agonal breathing. Therefore, lethal infection was consistently established with this infectious route and host-pathogen combination, even with low doses of vegetative bacilli. As observed previously (Welkos & Friedlander 1988), infection with 10⁶ CFU in a non-complement deficient mouse strain (i.e. Balb/c), neither led to death nor obvious signs of illness, highlighting the importance of the specific host–pathogen combination in this model.

Terminal histopathological findings were similar for all A/J mice infected with different doses of bacteria and therefore are described collectively below. Results from a small number of likewise infected DBA/2J mice, an alternative C5-deficient mouse strain, and A/J mice infected with spores rather than vegetative bacilli were also similar.

Lymphoid tissue

Thoracic (pulmonary and mediastinal) lymph node pathologies ranged from mild involvement to architectural effacement and severe haemorrhagic lymphadenitis. In more severely affected nodes, subcapsular sinuses were filled with clusters of rod-shaped bacilli and macrophages (Figure 1a); many of the macrophages contained bacterial fragments and haemosiderin. Macrophages, haemorrhage, bacilli and occasional neutrophils were also scattered throughout the parenchyma. Characterized by condensed and fragmented basophilic nuclei, morphologically apoptotic lymphocytes were present in mild to markedly increased numbers (Figure 1a) in contrast to their rarity in uninfected controls.

Figure 1.

Thoracic lymph nodes, thymus and spleen. (a) In a representative thoracic lymph node, bacilli and macrophages (arrows) fill a subcapsular sinus. Bacterial fragments are present within many of the macrophages. Haemorrhage is visible in both the sinus and the cortex. Condensed and fragmented nuclear debri, consistent with apoptotic lymphocytes (arrowhead), are visible throughout the lymph node. (b) Left panel, thymus from infected mouse: almost all thymic lymphocytes are apoptotic. There is significant lymphocyte depletion allowing epithelial cells to appear more prominently. Right panel, thymus from uninfected control: note closely apposed lymphocytes. (c) Left panel, spleen from infected mouse: lymphocyte depletion in periarteriolar lymphoid sheath leads to a starry sky pattern. Note also the absence of a sheath marginal zone, prominent in uninfected controls (right panel, arrow) and depletion of lymphocytes in the red pulp (left panel, infected mouse). (d) Junction of splenic red pulp and white pulp in an infected animal. Note lymphocyte depletion, absence of monocytoid appearing lymphocytes associated with the splenic marginal zone and multiple apoptotic figures (arrowhead) in the white pulp. Faintly staining bacilli are present in the red pulp. H&E, original magnification: (a) ×400, (b) ×400, (c) ×100 and (d) ×600.

Strikingly, thymic lymphocytes were almost uniformly destroyed in all infected mice and bore the above-mentioned morphological hallmarks of apoptotic cells (Figures 1b and 3a). In comparison to uninfected animals, there was also significant lymphocyte depletion with loss of the geographical molding of lymphocytes on one another observed in controls. In contrast to thoracic nodes, thymus and spleen (see below), other lymphoid tissue examined (Peyer's patches, mesenteric lymph nodes and bronchial-associated lymphoid tissue) showed neither haemorrhage nor inflammation and only rare apoptotic cells.

Spleen

Terminal infection led to dramatic microscopic changes in the architecture of the spleen (Figure 1c,d). In the white pulp, the periarteriolar lymphoid sheath showed a loss of lymphocytes, visible at low power as a starry sky pattern (Figure 1c). In contrast to controls, the marginal lymphocyte zone was also no longer apparent. In the red pulp, there was also significant depletion of lymphocytes with relative sparing of macrophage numbers (Figure 1c). In addition, occasional neutrophils and clusters of bacilli were observed (Figure 1d). Here, bacilli were often surrounded by macrophages laden with bacterial fragments. Many of the splenic macrophages contained haemosiderin suggesting haemorrhage. Lastly, in both red and white pulp, many morphologically apoptotic lymphocytes were also evident (Figure 1d).

Liver, kidney and gastrointestinal tract

There was marked congestion within the hepatic portal vessels, sinuses and central veins (Figure 2a). Small clusters of bacilli were focally present within sinuses, Kupffer cells, liver parenchyma and portal veins (Figure 2b). There was also mild periportal, sinusoidal and centrilobular neutrophil infiltration.

Figure 2.

Liver and kidney. (a) Hepatic portal triad contains a congested portal vein with fibrinous deposits (arrow), suggesting vascular damage. Adjacent hepatic parenchyma shows vascular congestion and scattered neutrophil infiltration (arrow heads). (b) Section of liver showing bacilli within a sinusoid (arrow). (c) Renal glomerulus filled with bacilli and a minimal associated inflammatory response. (d) A section of renal cortex shows a cord of bacilli within the interstitium and an absence of associated inflammation. H&E, original magnification: (a–d) ×400.

Within the kidneys, numerous bacilli and areas of haemorrhage were present within glomeruli and interstitial tissue (Figure 2c,d). A mild accompanying inflammatory response consisted primarily of scattered neutrophils. In addition, a layer of bacilli often covered the renal capsule surfaces, suggesting spread of bacteria along tissue planes in the retroperitoneum.

The gastrointestinal tract was remarkable for the absence of pathologies in the bowel wall, although bacteria were occasionally observed on serosal surfaces and within Peyer's patches.

Lungs

Grossly, lungs from infected animals had a reddish hue in contrast to the light, pink-tan lungs from uninfected controls. Multiple dark geographical areas were visible, suggesting haemorrhage in the underlying lung. Microscopically, most of the lung was involved by secondary pneumonia. In these areas, both interstitial tissue and alveoli were expanded with congested capillaries, oedema, haemorrhage and infiltrates of neutrophils and macrophages (Figure 3a,b). With an orcein stain (a stain for collagen and elastin), the outlines of alveoli become apparent (Figure 3c). The stain highlights the expansion of the interstitial tissue and infiltration of bacilli and inflammatory cells through compromised alveolar walls.

Skin

Grossly, there was neither swelling nor eschar formation at the cutaneous infection site, in contrast to the dramatic gross pathology in human cutaneous anthrax. However, microscopically, oedema was often evident in the wide splaying of dermal collagen fibres (Figure 3d). Bacteria were also observed individually and forming large sheets within the hypodermis surrounding the subcutaneous striated muscle layer of murine skin (Figure 3e). Despite the large numbers of bacteria, the inflammatory response was disproportionately mild, consisting of a few scattered neutrophils and macrophages (Figure 3e). In addition, massive haemorrhage was occasionally observed in skin distant from the site of bacterial inoculation (Figure 3f).

Discussion

Here, we describe detailed gross and histological findings in a murine model of systemic anthrax infection. As detailed below, we found that the model recapitulates the major pathologies observed during human disease.

A signature pathology seen in human inhalational anthrax is thoracic (pulmonary and mediastinal) haemorrhagic lymphadenitis, accounting for the widened mediastinum observable on chest X-ray (Albrink 1961; Abramova et al. 1993; Dixon et al. 1999; Guarner et al. 2003). Histologically, lymph node pathology in humans consists of morphologically apoptotic lymphocytes, oedema, haemorrhage and infiltration by bacteria, neutrophils and macrophages (Albrink 1961; Dutz & Kohout 1971; Suffin et al. 1978; Dixon et al. 1999; Grinberg et al. 2001; Jernigan et al. 2001; Guarner et al. 2003). All of these histological findings were present in the thoracic lymph nodes of infected mice. In addition, we observed massive apoptosis of thymic lymphocytes, a phenomenon not previously described in human anthrax, perhaps because the thymus is often not examined in autopsy studies. On the basis of our results, it is plausible that B. anthracis also destroys human thymic lymphocytes and, in this manner, may contribute to human mediastinal disease and to defects in cellular immunity related to thymic function.

The lymphocytic apoptosis observed in humans (Grinberg et al. 2001) and mice may result from the activity of anthrax lethal toxin. Lethal toxin is a metallo-protease that specifically cleaves and inactivates mitogen-activated protein kinase kinases (Duesbery et al. 1998). It thereby blocks mitogen-activated protein kinase signal transduction pathways known to play a role in cell survival. Through its inhibition of the p38 and extra cellular regulated kinase (ERK) map kinase pathways, lethal toxin induces apoptosis of cultured macrophages (Park et al. 2002; Popov et al. 2002a, b) and endothelial cells (Kirby 2004), respectively. It may similarly contribute to lymphocytolysis in vivo. Interestingly, a recent study noted that injection of purified oedema toxin led to both thymic and splenic lymphocyte depletion qualitatively similar to observations made here (Firoved et al. 2005). Therefore, edema-toxin (ET) might also account for Sterne-induced lymphocytolysis.

In contrast to expectations from in vitro models (Park et al. 2002; Popov et al. 2002b), we did not observe apoptosis of macrophages in lymph nodes, spleen or the liver (e.g. Kupffer cells). In fact, macrophages were observed engulfing bacteria and lying in close proximity to apoptotic lymphocytes (Figure 1a,d). Therefore, perhaps only certain populations of macrophages, not observed here, undergo apoptosis, or conditions were not appropriate for observing this phenomenon (Park et al. 2002). Further work is necessary to define more fully target cell populations and toxin effects during infection.

Following cutaneous inoculation in mice, anthrax infection resulted in thymic ablation and modest to severe effects in thoracic lymph nodes and spleen (i.e. apoptosis of lymphocytes), while sparing other lymphoid tissue. These results point to potential physiological differences in susceptibility among different lymphoid populations, an area of interest for future investigation. These results also suggest a potential alternative model for the development of mediastinal disease in human inhalational anthrax. The generally accepted model is that direct lymphatic tracking of inhaled spores leads to severe pathology during early primary infection (Dixon et al. 1999). Because we observe similar pathology after cutaneous infection, we suggest that severe mediastinal disease may also result from infection of mediastinal lymphoid tissue after systemic dissemination and reflect a unique biological susceptibility of lymphoid tissue in the mediastinal region.

Interestingly, as in humans, despite large numbers of bacteria in the liver, kidneys and skin, the inflammatory response was disproportionately mild (Albrink 1961; Dutz & Kohout 1971; Suffin et al. 1978; Dixon et al. 1999; Grinberg et al. 2001; Jernigan et al. 2001; Guarner et al. 2003). For instance, both in murine skin and cutaneous human anthrax, there is a lack of pus formation (e.g. copious neutrophil influx) as would be expected in response to a typical bacterial infection. These observations suggest that B. anthracis suppresses the cellular immune response and underscores the potential relevance of a number of in vitro findings (Moayeri & Leppla 2004). First, anthrax oedema toxin, an adenylate cyclase, inhibits neutrophil phagocytosis and superoxide generation (O'Brien et al. 1985). Second, anthrax lethal and oedema toxins prevent dendritic cell cytokine secretion and expression of stimulatory surface proteins (Agrawal et al. 2003; Tournier et al. 2005), potentially thereby suppressing adaptive immunity as well. Finally, lethal and oedema toxins inhibit lymphocyte activation and cytokine secretion (Fang et al. 2005; Paccani et al. 2005). Therefore, as for lymphocyte destruction, immune suppression may be anthrax toxin mediated, a hypothesis that deserves further scrutiny in animal models such as the one described here.

Although pneumonia was described in about half of the inhalation cases from Sverdlovsk (Abramova et al. 1993; Grinberg et al. 2001), it occurred consistently in our murine studies. Microscopically, mice show substantial interstitial and alveolar pathology similar to the severest pathology found in humans (Grinberg et al. 2001). The comparatively milder pathology observed in some recent human cases may reflect widespread use of antibiotics (Guarner et al. 2003). Significant pleural effusions, often described in human disease (Dutz & Kohout 1971; Suffin et al. 1978; Abramova et al. 1993; Guarner et al. 2003), were also not found, perhaps related to difficulty in observing small fluid collections in the murine thorax or to innate or therapy-induced differences in host physiology.

In human inhalation anthrax, B. anthracis spores are not thought to infect the lungs directly. Rather, after being taken up into pulmonary lymphatics, they secondarily infect the lungs after spread through the blood (Dixon et al. 1999). Pathological support for this model includes the observation that bacilli appear at higher density in septal capillaries and interstitium and lower density in alveoli (Grinberg et al. 2001), suggesting directional spread from the bloodstream to air spaces. Similar pathologies were observed here (Figure 3a–c). Therefore, the pathophysiology of pulmonary involvement in humans and our murine model, where a cutaneous route of infection precludes direct infection of the lungs, are likely to be similar, evolving from a shared common pathway of infection from the blood.

The pathology of the spleen, kidneys and liver was also similar to that found in humans (Albrink 1961; Dutz & Kohout 1971; Suffin et al. 1978; Dixon et al. 1999; Grinberg et al. 2001; Jernigan et al. 2001; Guarner et al. 2003). The observed lymphocyte apoptosis, bacterial and neutrophil infiltration and haemosiderin deposition are consistent with previous descriptions of haemorrhagic splenitis in human anthrax.

Furthermore, as noted in humans, pathologies associated with septic shock, such as centrilobular necrosis of the liver, acute tubular necrosis of the kidney and diffuse alveolar damage of the lung, did not occur (Grinberg et al. 2001). A potential reason for this is that a precipitous death leaves inadequate time for these pathologies to develop (Grinberg et al. 2001). The sudden preterminal clinical deterioration observed in our experiments and in humans (Albrink et al. 1960) is consistent with this scenario.

A number of potentially instructive differences were also apparent. Because we infected mice subcutaneously, we were able to examine the consequences of localized skin infection. Neither an eschar nor a gross swelling, characteristic of human cutaneous anthrax, was observed (Dutz & Kohout 1971). Perhaps absence of gross pathology in the murine host relates to the relatively rapid course of illness or to anatomic differences in the skin. Also, in contrast to humans, A/J mice rapidly and consistently developed systemic disease after cutaneous infection, indicating an inability to localize cutaneous infection. A final difference was the lack of secondary gastrointestinal disease. These observations contrast, for example, with the Sverdlovsk case series, in which many patients demonstrated secondary gastrointestinal involvement (Grinberg et al. 2001). The pathophysiological bases for these differences are of great interest, as they may suggest ways to modify host physiology therapeutically to limit spread of disease.

This model also demonstrates similarities to and differences from pathologies described in other murine models using different host-pathogen combinations. Similar to our findings, intratracheal or intranasal infection of non-complement deficient, BALB/c mice with the virulent B. anthracis, Ames strain led to apoptosis in lymphoid organs, septic splenitis and renal glomerular infiltration by bacteria without an associated inflammatory response (Lyons et al. 2004). These results provide further support for the validity of using the Sterne strain as substitute for virulent anthrax in a susceptible murine host. Interestingly, in contrast to our experiments and a subset of human cases (Abramova et al. 1993; Grinberg et al. 2001), lungs were not involved with secondary pneumonia. However, intratracheal infection of sublethally irradiated B6D2F1/J mice (a heterozygote for C5 complement and presumably complement sufficient) with Sterne did lead to severe pneumonia and also kidney glomerular infiltration by bacteria (Brook et al. 2001). These results suggest that radiation may overcome the non-permissiveness of a complement-sufficient murine host and leads to lung pathologies similar to those in A/J mice.

In summary, we found that subcutaneous infection was a consistent and simple way to establish systemic anthrax in a mouse model. Because, unlike in humans, the skin in mice does not appear to be a significant barrier to systemic disease, we did not need more complex or invasive infection procedures used in other models such as aerosolization chambers or intratracheal injection (Albrink 1961; Dalldorf et al. 1971; Fritz et al. 1995; Elliott et al. 2002; Vasconcelos et al. 2003). Similarly, we did not need to use previously described intraperitoneal infection routes (Karginov et al. 2004; Popov et al. 2004), thereby avoiding direct and potentially confounding interactions of bacteria with intraperitoneally dosed anaesthetics and therapeutic compounds under study in our laboratory.

Importantly, we were also able to recapitulate human pathologies using Sterne, an anthrax strain that has not been reported to cause disease in humans, and is safe to use under standard Biosafety Level 2 conditions. One important caveat, however, is that we will not be able to detect potential contributing effects of capsule (lacking in Sterne), outside of its proposed role in protecting against C5 complement and effects of additional loci located on the capsule-encoding virulence plasmid. However, our data indicate that infection of C5-deficient mice with B. anthracis Sterne is otherwise a compelling model of human infection. The detailed description of murine pathology provided here should serve as a basis for further experimentation with this model.