Bacillus anthracis lethal toxin, but not edema toxin, increases pulmonary artery pressure and permeability in isolated perfused rat lungs

Abstract

Although lethal toxin (LT) and edema toxin (ET) contribute to lethality during Bacillus anthracis infection, whether they increase vascular permeability and the extravascular fluid accumulation characterizing this infection is unclear. We employed an isolated perfused Sprague-Dawley rat lung model to investigate LT and ET effects on pulmonary vascular permeability. Lungs (n ≥ 6 per experimental group) were isolated, ventilated, suspended from a force transducer, and perfused. Lung weight and pulmonary artery (P_pa) and left atrial pressures were measured over 4 h, after which pulmonary capillary filtration coefficients (Kf.c) and lung wet-to-dry weight ratios (W/D) were determined. When compared with controls, LT increased P_pa over 4 h and Kf.c and W/D at 4 h (P < 0.0001). ET decreased P_pa in a significant trend (P = 0.09) but did not significantly alter Kf.c or W/D (P ≥ 0.29). Edema toxin actually blocked LT increases in P_pa but not LT increases in Kf.c and W/D. When P_pa was maintained at control levels, LT still increased Kf.c and W/D (P ≤ 0.004). Increasing the dose of each toxin five times significantly increased and a toxin-directed monoclonal antibody decreased the effects of each toxin (P ≤ 0.05). Two rho-kinase inhibitors (GSK269962 and Y27632) decreased LT increases in P_pa (P ≤ 0.02) but actually increased Kf.c and W/D in LT and control lungs (P ≤ 0.05). A vascular endothelial growth factor receptor inhibitor (ZM323881) had no significant effect (P ≥ 0.63) with LT. Thus, LT but not ET can increase pulmonary vascular permeability independent of increased P_pa and could contribute to pulmonary fluid accumulation during anthrax infection. However, pulmonary vascular dilation with ET could disrupt protective hypoxic vasoconstriction.

NEW & NOTEWORTHY The most important findings from the present study are that Bacillus anthracis lethal toxin increases pulmonary artery pressure and pulmonary permeability independently in the isolated rat lung, whereas edema toxin decreases the former and does not increase permeability. Each effect could be a basis for organ dysfunction in patients with this lethal infection. These findings further support the need for adjunctive therapies that limit the effects of both toxins during infection.

INTRODUCTION

Although invasive bacterial infection and sepsis are associated with reductions in vascular endothelial integrity and extravasation of proteinaceous fluid, these changes are pronounced with Bacillus anthracis infection. Pathology from patients dying with inhalational anthrax in Sverdlovsk, Russia, in 1979 was notable for extravasation of fluid, protein, and blood cells (1, 20). Recurrent pleural effusions have been prominent in patients with inhalational anthrax and respiratory failure in the United States and Europe (7, 24). Cutaneous and soft tissue anthrax infection produces marked tissue edema (15, 55). Finally, gastrointestinal anthrax is associated with intestinal and retroperitoneal edema (49, 55). Since loss of vascular integrity during B. anthracis infection could contribute to organ dysfunction directly as well as to the resistant shock patients demonstrate, understanding its basis may improve management.

B. anthracis produces two toxins, edema and lethal toxin (ET and LT), consisting of protective antigen (PA), the component necessary for host uptake of each of the toxin’s toxic moieties, edema and lethal factor (EF and LF) (22, 55). Selective inhibition of either toxin is protective in bacteria-challenged animal models, and the administration of each toxin alone in animals produces hypotension, organ injury, and lethality (2, 11, 12, 23, 33, 37, 38, 53). Despite their pathogenic importance, whether ET or LT contribute to increased vascular permeability during infection is unclear. Whereas ET produces localized tissue edema when injected subcutaneously in animals, EF has potent adenylate-cyclase activity that increases intracellular cAMP levels (32), an action potentially increasing endothelial barrier function (47). By contrast, in vitro studies now suggest that LT, but not ET, increases the permeability of vascular endothelial and lung epithelial cell monolayers (6, 14, 29, 31, 52, 57).

To investigate the effects of ET and LT on vascular permeability at the organ level, we employed an isolated perfused rat lung model. Sprague-Dawley rats, which are sensitive to lethal effects of LT, were used as lung donors for most experiments. Studies examined the effects of each toxin alone or together and in low or high doses; LT under perfusion conditions of either constant flow or pressure; a PA-directed monoclonal (PA-mAb) when combined with either toxin alone; two Rho-kinase inhibitors (GSK269962 and Y27632); and a VEGF receptor inhibitor (VEGFR-I; ZM323881) when combined with LT. The choice of agents for these later investigations was based in part on a recent review of in vitro and zebra fish embryo studies that implicated Rho-kinase and VEGF pathways in permeability effects of LT (6, 44, 63). In a final study, we also examined LT in lungs prepared from other rat strains sensitive or insensitive to lethal effects of LT.

MATERIALS AND METHODS

Animal care.

The protocol for this study (CCM 16-02) was reviewed and approved by the Animal Care and Use Committee at the Clinical Center, National Institutes of Health.

Sprague-Dawley rat isolated perfused lung model studies.

Table 1 summarizes challenge and treatment doses and numbers of lungs from Sprague-Dawley (studies 1 to 9), Wistar, Brown Norway, and Lewis (study 10) rats (all males weighing 300–350 g) tested in 10 studies. Figure 1 shows a general timeline for studies. Study 1 examined the effects of: PA alone (control, 2 μg/ml, n = 10); LT alone (1 μg/ml LF + 2 μg/ml PA, n = 17); ET alone (1 μg/ml EF + 2 μg/ml PA, n = 11); or the same doses of LT and ET together (LT + ET, n = 19). Additional LT (n = 7) and LT + ET (n = 8) lungs employed to measure effluent cAMP levels were included in study 1 analysis. Toxin concentrations were based on the total recirculating volume in the perfusion system and were comparable to ones reported in live B. anthracis-challenged animals and human patients (36, 40, 46, 61). After lungs were isolated and equilibrated (see Lung isolation and perfusion below), PA alone or toxin was added to the perfusion system, and lungs were perfused at a constant flow rate for 4 h while lung weight (WT), pulmonary artery pressure (P_pa), and left atrial pressure (P_la) were monitored. At 30-min intervals, perfusion and effluent samples were removed for analysis. After 4 h, P_la was increased 7 cmH₂O, and change in lung weight was assessed over 15 min and a lung capillary filtration coefficient (Kf.c) calculated (see Measurements and calculations below) (16, 59, 60). Before and following the increase in P_la, pulmonary capillary pressure (P_pc) was calculated. Lungs were then removed for wet-to-dry weight ratio determination (W/D, see below). Measures in subsequent studies were similar to study 1.

Table 1.

Summary of the studies

Study	Perfusion	Challenge, μg/ml	Treatment/Strain	Number of Lungs
1. Effect of LT and ET, alone or together	Constant flow	PA, 2		10
		LT, 1		17
		ET, 1		11
		LT + ET, 1		19
2. High-dose LT	Constant flow	PA, 10		6
		LT, 5		6
3. Perfusion pressure in LT Challenge	Constant pressure	PA, 2		9
	Constant pressure	LT, 1		14
	Constant flow	LT, 1		8
4. PA-mAb in LT	Constant flow	PA, 2		10
		LT, 1	NS-mAb	10
		LT, 1	PA-mAb ×10	8
5. GSK269962 in LT	Constant flow		DMSO	3
			GSK269962, 3 μM	3
		LT, 1	DMSO	6
		LT, 1	GSK269962, 3 μM	6
6. Y-27632 in LT	Constant flow		DMSO	3
			Y-27632, 5 μM	3
		LT, 1	DMSO	8
		LT, 1	Y-27632, 5 μM	8
7. ZM 323881 in LT	Constant flow	LT, 1	DMSO	7
		LT, 1	Y-27632, 5 μM	9
8. High-dose ET	Constant flow	PA, 10		12
		ET, 5		12
9. PA-mAb in ET	Constant flow	PA, 10		12
		ET, 5	NS-mAb	8
		ET, 5	PA-mAb ×10	8
10. Effect of rat strain in LT	Constant flow	PA, 2	Wistar	7
		LT, 1	Wistar	8
		PA, 2	Brown Norway	8
		LT, 1	Brown Norway	7
		PA, 2	Lewis	7
		LT, 1	Lewis	7

ET, edema toxin; LT, lethal toxin; NS-mAb, nonspecific monoclonal antibody; PA, protective antigen; PA-mAb, monoclonal antibody directed to PA.

Fig. 1.

Timeline for the procedures and measures performed and the challenges and interventions employed in the 11 studies described. See Table 1 for further details regarding each study, including the number of lungs examined in each study group. ET, edema toxin; FITC-albumin, albumin, fluorescein isothiocyanate conjugate; Kf.c, pulmonary capillary filtration coefficient; LT, lethal toxin; NA, not applicable; NS-mAb, nonspecific monoclonal antibody; PA, protective antigen; PA-mAb, monoclonal antibody directed to PA; P_pa, pulmonary artery pressure; P_pc, pulmonary capillary pressure; TV, tidal volume; W/D, lung wet-to-dry weight ratio.

The next six studies investigated LT alone. Study 2 compared a dose of LT (5 μg/ml) (n = 6) five times greater than in study 1 to PA only (n = 6). Study 3 examined whether LT would increase Kf.c if P_pa was kept constant throughout. After PA (n = 9) or LT (n = 14) was added to the perfusion circuit in doses similar to study 1, flow rate was adjusted every 1–3 min starting at 90 min to maintain P_pa at baseline levels (see Lung isolation and perfusion below). Pressure in PA lungs remained constant throughout, and flow rate was not altered. For comparison, additional lungs (n = 8) were perfused with LT at a constant flow rate. Study 4 examined whether prevention of LT uptake by host cells with a PA-mAb (dose of 10× the molar PA dose included in LT) would block lung effects of LT (see Toxin and treatment preparation below). Lungs were perfused under constant flow with PA only (n = 10), LT and nonspecific mAb (NS-mAb; n = 10), or LT and PA-mAb (n = 8). Study 5 compared the effect of GSK2969629 (a rho-kinase inhibitor, n = 6) versus diluent (control, n = 6), and study 6 compared the effect of Y27632 (another rho-kinase inhibitor, n = 8) versus diluent (n = 8), both in LT-challenged lungs. Each agent was also compared with diluent in lungs not challenged with LT (3 diluent vs. 3 GSK269962 lungs and 3 diluent vs. 3 Y-27632 lungs). Study 7 compared the effect of ZM323881 (VEGFR-I, n = 9) versus diluent control (n = 7) in LT-challenged lungs.

Two studies evaluated ET further. Study 8 compared an ET dose (5 μg/ml, n = 12) five times greater than in study 1 to PA only (10 μg/ml, n = 12) under constant flow. Study 9 examined the effects of a PA-mAb (dose of 10× the molar PA dose included in ET) with ET. Lungs were perfused with PA only (n = 12), ET and NS-mAb (n = 9), or ET and PA-mAb (n = 9). The ET and PA doses were similar to study 8.

Wistar, Brown Norway, and Lewis rat isolated lung studies.

A tenth study (study 10) examined whether LT had similar effects on lungs from Wistar and Brown Norway rats sensitive to lethal effects of LT and from Lewis rats that are resistant (42). Lungs were prepared and challenged and measures made as in study 1.

Lung isolation and perfusion.

Animals were anesthetized with 60 mg/kg ketamine and 5 mg/kg xylazine. The trachea was cannulated with a 14-gauge blunt needle, and the animal was mechanically ventilated (Model 683 Animal Ventilator, Harvard Apparatus, Holliston, MA) with a 1.2-ml tidal volume, 2 cmH₂O positive end-expiratory pressure, 60 breaths/min respiratory rate, and gas mixture of 95% O₂ and 5% CO₂. Two minutes after a midline laparotomy and administration of 0.2 ml of 10,000 μ/ml heparin via the inferior vena cava, animals were exsanguinated after inferior vena cava and descending aorta transection. The lung and heart were exposed with a midline sternotomy, the pulmonary artery and left atrium (LA) were cannulated, and the lung was perfused free of blood with 20 ml of 4°C perfusate (Pulmonary Arterial and Left Atrial Cannulas, Harvard Apparatus). The PA and LA cannulas were immediately connected to the perfusion system, and the lung and heart block (termed lung-block) were dissected free and suspended from a force transducer (MLT0201, Colorado Springs, CO) in a heated water jacket (37°C) and humidified chamber while ventilation continued. Perfusion was immediately started with a closed perfusion circuit and variable speed peristaltic pump (MasterFlex L/S, Vernon Hills, IL). Perfusate was circulated from a closed reservoir to the PA and then returned from the LA to the reservoir. The perfusate consisted of a modified Krebs-Henseleit buffer containing 118.5 mM NaCl, 25 mM NaHCO₃, 4.7 mM KCl, 1.5 mM MgSO₄, 1.5 mM KH₂PO₄, 1.92 mM CaCl₂, and 5.74 mM glucose with BSA (4%), dextran (1.67%), and (13 μM) 21 amino acids. P_pa (i.e., the perfusion pressure) and P_la were continuously measured via pressure transducers. Perfusion rate was set at 3 ml·min⁻¹·100 g body wt⁻¹, tidal volume at 8 ml/kg, positive end-expiratory pressure at 2 cmH₂O, and P_la at 3–4 cmH₂O. Lungs were then equilibrated for 15 min. The perfusate pH was maintained at 7.35 to 7.45 with adjustment of gas flow rate. Any lung with an increase in either P_pa ≥1 cmH₂O or lung-block weight ≥0.05 g during this equilibration period was excluded from study.

Measurements and calculations.

Lung WT, P_pa, and P_la measures were recorded every 15 min. In studies with constant flow, the perfusion flow rate was not changed during the experiment. In the constant pressure study (study 3), the flow rate was reduced 1 revolution/min (0.167 ml/min) every 1 to 3 min when P_pa increased higher than the pressure at 90 min, the earliest time P_pa was observed to increase with LT under constant flow. Perfusate and effluent samples (0.5 ml) were drawn at 30-min intervals to assess pH, gas tensions, electrolytes, and lactate. In some lungs challenged with LT, ET, or LT + ET (n = 8, 11, and 8 lungs, respectively) and lungs challenged with PA or ET with either PA-mAb or NS-mAb (n = 12, 9, and 9 lungs, respectively), effluent was collected at 60, 120, 180, and 240 min for cAMP measurements.

After 240 min of perfusion, pulmonary capillary permeability was measured as the capillary filtration coefficient (Kf.c, ml/min/cmH₂O/100 g). After 10 s of baseline lung WT measures, P_la was increased 7 cmH₂O by raising the LA reservoir and maintained at that level for 15 min. During this period, lung WT was recorded every minute. As previously described, lung weight gain during the initial 5 min was attributed to filling of the pulmonary vasculature, whereas weight gain during the final 10 min was attributed to extravasation of perfusate (16, 59, 60). P_pc was measured before and after P_la was increased by occluding both perfusion inlet and outlet tubing for 4 s and recording the pressure after P_pa and P_la had become equal. Kf.c was calculated using the formula: Kf.c = [(dWT/dt)/ΔP_pc/WT_e × 100)], in which dWT/dt is the slope of the lung weight (WT) change per min from 5 to 15 min; ΔP_pc is the change in P_pc related to the increase in P_la; and WT_e is the estimated baseline lung weight (0.00472 × animal’s body weight) (60). Following Kf.c measures, lungs were removed and W/D subsequently calculated.

An FITC-albumin assay was used to further assess pulmonary vascular permeability in six lungs challenged with LT and six challenged with PA using toxin doses and perfusion methods similar to study 1. At 120 min after starting perfusion, FITC-albumin (Sigma-Aldrich, St. Louis, MO) was added to the perfusate at a final concentration of 0.16 mg/ml. At 240 min, rather than calculating Kf.c, lung uptake of FITC-albumin was measured. After a sample of perfusate was collected, lungs were perfused with 20 ml 1 × PBS (pH = 7.4) to remove labeled albumin from the pulmonary circulation. The lung-block was removed from the perfusion circuit, the right lower lung lobe was collected for W/D determination, and the right middle lobe was weighed and homogenized in 0.5 ml PBS. The left lobe was lavaged through the tracheal cannula three times with 1.5 ml PBS, and the return volume was measured. The lung homogenate and lavage were then centrifuged at 8,000 revolutions/min for 10 min and the supernatants collected. The FITC-albumin concentrations in perfusate and the supernatants of the lung homogenate and lavage were then measured by fluorescence (excitation 485 nm and emission 520 nm wavelengths) with a plate reader (Biotek, Winooski, VT). The lung tissue and lavage FITC-albumin concentrations were then normalized based on the concentration of perfusate FITC-albumin concentration and expressed as either (mg/g lung tissue)/(mg/ml perfusate) or (mg/ml lung lavage)/(mg/ml perfusate), respectively.

Perfusate and effluent Po₂, Pco₂, sodium, potassium, chloride, glucose, and lactate levels were measured using the Stat Profile Critical Care Express System (NOVA Biomedical, Waltham, MA). cAMP was measured using a chemiluminescent direct immunoassay (Arbor Assays, Ann Arbor, MI).

Toxin and treatments.

Toxin components (PA, LF, and EF) were recombinant proteins prepared from Escherichia coli as previously described (11). LT and ET consisted of LF and EF, respectively, with PA in ratios of 1:2 on the basis of weight. The lipopolysaccharide (LPS) content of the PA, LF, and EF was 0.001, 0.002, and 0.006 ng/μg, respectively, as measured with the Limulus Amoebocyte Lysate Chromogenic endotoxin assay (Clongen, Gaithersburg, MD). The PA-mAb was Raxibacumab (Human Genome Sciences), and the control was an inactive, NS-mAb (38). Rho-kinase inhibitors GSK269962 and Y27632 and VEGFR-I ZM323881 (Selleck Chemicals, Houston, TX) were endotoxin free and diluted in dimethyl sulfoxide (DMSO) (final concentrations of DMSO in the perfusion circuit were 0.03%, 0.0125%, and 0.02%, respectively). When compared with perfusion over 4 h with buffer alone, perfusion with buffer with PA or with DMSO alone in the high and low concentrations employed did not have significantly different effects on any of the lung parameters measured (P ≥ 0.10 for all).

Statistics.

Data were analyzed using PROC MIXED in SAS version 9.4 software (SAS Institute, Cary, NC). Two-way repeated measure ANOVAs accounting either for challenge (toxin vs. PA) or treatment type (treatment vs. diluent) and time after challenge or treatment were conducted to assess the effects of challenge or treatment over time, respectively, on pulmonary arterial pressure, perfusion rate, and effluent cAMP level. Two-way ANOVAs were used to determine the dose effect (low vs. high) of LT or ET (toxin vs. PA). One-way ANOVAs were then used to assess the effects of challenge or treatment on other parameters. Serial changes from baseline and the final change from baseline for P_pa and perfusion rate were used for analysis and presentation. The Kf.c was analyzed after log-transformation because of its non-normal distribution. Throughout, data are presented as means ± SE. A P value ≤0.05 was considered significant. Lungs were randomized to a particular challenge and/or treatment before preparation, and lung numbers vary in comparison groups in part related to exclusion of lungs during the equilibration based on criteria outlined above.

RESULTS

Sprague-Dawley rat isolated lung model studies.

When compared with control lungs (group study sizes shown in figures) perfused with PA only, LT (LF and PA) progressively increased serial changes from baseline in P_pa (cmH₂O) starting midway through the 240-min observation period (P < 0.0001 for the time interaction) (Fig. 2). By 240 min, LT increased the final mean change (±SE) in P_pa (5.16 ± 0.78 vs. 1.15 ± 1.02 cmH₂O) and the mean (±SE) P_pc before permeability measure (7.31 ± 0.21 vs. 5.99 ± 0.11 cmH₂O), Kf.c [4.48 ± 0.78 vs. 2.48 ± 0.07 log(ml/min/cmH₂O/100 g)], and W/D (8.24 ± 0.53 vs. 5.53 ± 0.08) (P ≤ 0.004). ET produced small reductions in P_pa over time approaching significance (P = 0.09 for the time interaction) but did not significantly alter the final mean change in P_pa (−0.23 ± 0.49 cmH₂O, P = 0.22) or the P_pc (6.13 ± 0.18 cmH₂O), Kf.c (2.36 ± 0.08 [log(ml/min/cmH₂O/100 g)]) or W/D (5.58 ± 0.10) (all P ≥ 0.29). When combined, LT and ET (LT + ET) did not significantly alter P_pa over time (P = 0.44 for the time interaction) or the final P_pa change (−0.08 ± 0.29 cmH₂O) or the P_pc (6.11 ± 0.10 cmH₂O) (P ≥ 0.15) but did increase Kf.c [4.31 ± 0.32 log(ml/min/cmH₂O/100 g)] and W/D (7.78 ± 0.35) (P ≤ 0.0004). Thus, ET counteracted pulmonary vascular constrictor effects of LT but not its permeability effects. Additional studies were then conducted of LT (Studies 2–7) and ET (Studies 8 and 9) with Sprague-Dawley lungs.

Fig. 2.

Serial mean (±SE) changes in pulmonary artery pressure (P_pa) from baseline to 240 min for lungs challenged with lethal toxin (LT), edema toxin (ET), or LT + ET vs. those challenged with protective antigen alone [control, protective antigen (PA)] are compared (A). P values in A are for the overall effect of either ET or LT + ET challenges compared with PA (designated challenge) and for the change in the effect of each of these three challenges compared with PA over time (i.e., the time interaction and designated challenge × time). P value for the overall effect of LT is not applicable (NA), since the time interaction was significant. For the same groups, the mean (±SE) overall changes in P_pa from baseline to 240 min (B), initial pulmonary capillary pressure (P_pc) measures at 240 min (C), pulmonary capillary filtration coefficients (Kf.c) (D), and subsequently measured wet-to-dry weight lung ratios (W/D) (E) are compared. Chall, challenge.

In study 2, compared with PA, a dose of LT (5 μg/ml) five times greater than in study 1 also produced progressive increases in P_pa over time (P < 0.0001 for the time interaction) and at 240 min increased the final change in P_pa (13.10 ± 2.38 vs. 1.09 ± 1.29 cmH₂O) and the P_pc (9.79 ± 0.67 vs. 6.52 ± 0.20 cmH₂O), Kf.c [4.33 ± 0.15 vs. 2.55 ± 0.12 log(ml/min/cmH₂O/100 g)]) and W/D (7.68 ± 0.32 vs. 5.61 ± 0.10) (all P ≤ 0.001) (Fig. 3). When compared with low-dose LT, high-dose LT significantly increased P_pa and P_pc (P ≤ 0.003), but not Kf.c and W/D (P ≥ 0.56).

Fig. 3.

Serial mean (±SE) changes in pulmonary artery pressure (P_pa) from baseline to 240 min for lungs challenged with a low dose of lethal toxin (LT, 1 μg/ml) vs. a low dose of protective antigen (PA; 2 μg/ml) and for those challenged with a high dose of LT (5 μg/ml) vs. a high dose of PA (control; PA, 10 μg/ml) are compared (A). P values in A are for the change in the effects over time of either 1 μg/ml LT compared with 2 μg/ml PA or for 5 μg/ml LT compared with 10 μg/ml PA (i.e., the time interactions and designated challenge × time). For the same groups, the mean (±SE) overall changes in P_pa from baseline to 240 min (B), initial pulmonary capillary pressure (P_pc) measures at 240 min (C), pulmonary capillary filtration coefficients (Kf.c) (D), and subsequently measured wet-to-dry weight lung ratios (W/D) (E) are compared. B–E also compare the effects of the low and high LT doses (P values designated with #). The data shown for the low LT and PA doses is the same as the data shown in Fig. 2. The concentrations of LT shown here represent the amounts of lethal factor employed in the toxins, which is 50% of the concentration of PA included in the toxin. Chall, challenge.

To further exclude the possibility that increases in P_pa with LT contributed to the permeability effects of this toxin, study 3 investigated LT while P_pa was maintained at constant baseline levels (see materials and methods). When compared with PA controls also perfused under constant pressure (but that didn’t require changes in flow to maintain constant pressure), lungs perfused under constant pressure and challenged with LT required progressive reductions in flow (P < 0.0001) to maintain pressures not different from baseline or controls (Fig. 4). At 4 h, compared with the overall change from baseline in P_pa in control lungs (−1.14 ± 0.38 cmH₂O), this change was not significantly different in LT lungs perfused under constant pressure (−0.66 ± 0.15 cmH₂O, P = 0.18) but was increased in lungs perfused under constant flow (3.94 ± 1.49 cmH₂O, P = 0.003), as seen previously. By contrast, compared with Kf.c and W/D in control lungs [2.86 ± 0.12 log(ml/min/cmH₂O/100 g) and 5.95 ± 0.12], with LT both parameters were increased in lungs perfused either under constant pressure [3.95 ± 0.09 log(ml/min/cmH₂O/100 g) and 7.37 ± 0.24, P ≤ 0.004] or constant flow [4.67 ± 0.45 log(ml/min/cmH₂O/100 g) and 7.78 ± 0.45, P ≤ 0.001]. Thus, increases in Kf.c and W/D with LT did not appear related to increasing P_pa with LT.

Fig. 4.

Serial mean (±SE) changes in pulmonary artery pressure (P_pa) from baseline to 240 min for lungs challenged with lethal toxin (LT) and perfused either under constant pressure (LT + CP) or constant flow (LT + CF) vs. lungs challenged with protective antigen alone under constant flow [control, protective antigen (PA)] are compared (A). Although PA lungs were perfused with constant flow, pressure remained constant in these lungs. Therefore, PA lungs are denoted as constant flow and pressure (PA + CF/CP). The same comparisons for changes in perfusion rates from baseline to 240 min are shown (B). P values in A are for the overall effect of LT + CP compared with PA (designated challenge) and for the change in the effect of LT + CP or LT + CF compared with PA over time (i.e., the time interaction and designated challenge × time). P value for the overall effect of LT + CF is not applicable (NA), since the time interaction was significant. P value in B is for the change in the effect of LT + CP vs. PA over time, and there is no comparison of PA and LT + CF because flow was constant for each. For the same groups, the mean (±SE) overall changes in P_pa from baseline to 240 min (C), pulmonary capillary filtration coefficients (Kf.c) measured at 240 min (D), and subsequently measured wet-to-dry weight lung ratios (W/D) (E) are compared. Chall, challenge.

In study 4 (Fig. 5), compared with PA, perfusion with LT and treatment with an NS-mAb produced progressive increases in P_pa (P < 0.0001 for the time interaction) and increases in the final change in P_pa (4.73 ± 1.43 vs. 1.15 ± 1.02 cmH₂O, P = 0.02), P_pc (7.88 ± 0.35 vs. 6.40 ± 0.48 cmH₂O, P = 0.0002), Kf.c [4.04 ± 0.18 vs. 2.48 ± 0.07 log(ml/min/cmH₂O/100 g), P < 0.0001)] and W/D (7.05 ± 0.42 vs. 5.53 ± 0.08, P = 0.001). In contrast, perfusion with LT when combined with PA-mAb to inhibit host cell uptake of toxin produced small but progressive decreases in P_pa over time (P = 0.02 for the time interaction) and did not significantly alter the final change in P_pa (−0.75 ± 0.29 cmH₂O), P_pc (6.81 ± 0.45 cmH₂O), Kf.c [2.81 ± 0.16 log(ml/min/cmH₂O/100 g)] or W/D (5.61 ± 0.23) (P ≥ 0.09). When compared with LT with NS-mAb, LT with PA-mAb progressively lowered P_pa (P < 0.0001 for the time interaction) and decreased the final change in P_pa and in P_pc, Kf.c, and W/D (P ≤ 0.03).

Fig. 5.

Serial mean (±SE) changes in pulmonary artery pressure (P_pa) from baseline to 240 min for lungs challenged with lethal toxin (LT) and treated with either nonspecific monoclonal antibody (LT + NS-mAb) or protective antigen-directed mAb (PA-mAb) vs. those challenged with PA without treatment (control, PA) are compared (A). P values in A are for the change in the effects of LT + NS-mAb or LT + PA-mAb compared with PA over time (i.e., the challenge time interaction and designated challenge × time) and for the change in the effect of LT + PA-mAb compared with LT + NS-mAb over time (i.e., the treatment time interaction and designated Rx × time). For the same groups, the mean (±SE) overall changes in P_pa from baseline to 240 min (B), initial pulmonary capillary pressure (P_pc) measures at 240 min (C), pulmonary capillary filtration coefficients (Kf.c) (D), and subsequently measured wet-to-dry weight lung ratios (W/D) (E) are compared. Chall, challenge; Rx, treatment.

Studies 5 and 6 examined whether either of two rho-kinase inhibitors, GSK269962 or Y27632, would alter effects of LT. In LT-challenged lungs, compared with diluent, both agents reduced increases in P_pa (P ≤ 0.03 for the time interaction for both) and the final change in P_pa at 240 min significantly or in a trend approaching significance (P = 0.004 and 0.07) (Fig. 6), but neither altered P_pc significantly (P ≥ 0.21). Surprisingly, however, both agents actually increased Kf.c and W/D with LT in patterns that were or that approached significance (P ≤ 0.07). These agents were also compared with diluent in normal lungs (6 Y-27632 vs. 6 diluent lungs and 2 GSK269962 vs. 2 diluent lungs). The effects of the two agents did not differ significantly for any parameter (P ≥ 0.15), and these were combined for analysis. When compared with diluent, the rho-kinase inhibitors did not alter P_pa or P_pc significantly (P ≥ 0.44) but did increase Kf.c [4.30 ± 0.66 vs. 2.95 ± 0.17 log(ml/min/cmH₂O/100 g)] and W/D (8.58 ± 1.16 vs. 6.52 ± 0.42) in trends approaching significance (P = 0.08 and 0.09, respectively; data not shown). In study 7, in LT challenged lungs compared with diluent, the VEGFR-I ZM323881 did not produce significant changes in any measured parameter (P ≥ 0.77, data not shown).

Fig. 6.

Serial mean (±SE) changes in pulmonary artery pressure (P_pa) from baseline to 240 min for lungs challenged with lethal toxin (LT) and treated with either the Rho-kinase inhibitors GSK269962 (A) or Y27632 (B) vs. lungs challenged with LT and treated with diluent alone (controls) are compared. P values in A and B are for the changes in the effects of each treatment vs. the diluent control over time. For the same groups, the mean (±SE) overall changes in P_pa from baseline to 240 min (C), initial pulmonary capillary pressure (P_pc) measures at 240 min (D), pulmonary capillary filtration coefficients (Kf.c) (E), and subsequently measured wet-to-dry weight lung ratios (W/D) (F) are compared.

In study 8 (Fig. 7), compared with PA, perfusion with an ET dose (5 μg/ml) five times greater than in study 1 produced progressive decreases in P_pa over time (P = 0.0004 for time interaction) and decreased the final change in P_pa (−1.12 ± 0.36 vs. −0.39 ± 0.24 cmH₂O, P = 0.07) and in P_pc (5.32 ± 0.11 vs. −5.86 ± 0.12 cmH₂O, P = 0.003). The higher ET dose also produced a decrease in Kf.c that approached significance [2.40 ± 0.27 vs. 2.79 ± 0.13 log(ml/min/cmH₂O/100 g), P = 0.14] but did not alter W/D (5.86 ± 0.22 vs. 5.90 ± 0.14) (P ≥ 0.86). The effects of the high- and low-dose ET did not differ significantly for any of the parameters (P ≥ 0.44).

Fig. 7.

Serial mean (±SE) changes in pulmonary artery pressure (P_pa) from baseline to 240 min for lungs challenged with a low dose of edema toxin (ET, 1 μg/ml) vs. a low dose of protective antigen (PA, 2 μg/ml) or a high dose of ET (5 μg/ml) vs. a high dose of PA (10 μg/ml) are compared (A). P values in A are for the overall effect of either 1 μg/ml ET vs. 2 μg/ml PA (designated challenge) and for the change in the effects over time of either 1 μg/ml ET compared with 2 μg/ml PA over time or for 5 μg/ml ET compared with 10 μg/ml PA (i.e., the time interactions and designated challenge × time). For the same groups, the mean (±SE) overall changes in P_pa from baseline to 240 min (B), initial pulmonary capillary pressure (P_pc) measures at 240 min (C), pulmonary capillary filtration coefficients (Kf.c) (D), and subsequently measured wet-to-dry weight lung ratios (W/D) (E) are compared. B–E also compare the effects of the low and high ET doses (P values designated with #). The data shown for the low ET and PA doses is the same as the data shown in Fig. 2. The concentrations of ET shown here represent the amounts of edema factor employed in the toxin, which is 50% of the concentration of PA included in the toxin. Chall, challenge.

In study 9 (Fig. 8), compared with PA, perfusion with ET combined with NS-mAb treatment progressively decreased P_pa (P < 0.0001 for the time interaction) and decreased the final change in P_pa in a significant trend (−0.97 ± 015 vs. −0.39 ± 0.24 cmH₂O, P = 0.10) and Kf.c [2.40 ± 0.06 vs. −2.79 ± 0.13 log(ml/min/cmH₂O/100 g), P = 0.03] but did not alter P_pc (5.90 ± 0.19 vs. 5.86 ± 0.13 cmH₂O, P = 0.84) or W/D (5.95 ± 0.14 vs. 5.90 ± 0.0.14, P = 0.82). When compared with PA control, perfusion with ET combined with PA-mAb did not alter P_pa over time (P = 0.94 for the time interaction) or the final change in P_pa (−0.45 ± 0.23 cmH₂O), P_pc (5.86 ± 0.09 cmH₂O), W/D (5.73 ± 0.18), or Kf.c [2.27 ± 0.08 log(ml/min/cmH₂O/100 g) (P ≥ 0.09)]. When compared with ET with NS-mAb, ET with PA-mAb produced progressively higher P_pa (P < 0.04 for the time interaction) but did not significantly alter the final change in P_pa (P = 0.09) or in P_pc, Kf.c, or W/D (P ≥ 0.35).

Fig. 8.

Serial mean (±SE) changes in pulmonary artery pressure (P_pa) from baseline to 240 min for lungs challenged with edema toxin (ET) and treated with either nonspecific monoclonal antibody (ET + NS-mAb) or protective antigen-directed mAb (ET + PA-mAb) vs. those challenged with protective antigen without treatment (control, PA) are compared (A). P values in A are for the overall effect of ET + NS-mAb vs. PA and for the change in the effect of ET + NS-mAb or ET + PA-mAb compared with PA over time (i.e., the challenge time interaction and designated challenge × time) and for the change in the effect of ET + PA-mAb compared with ET + NS-mAb over time (i.e., the treatment time interaction and designated Rx × time). For the same groups, the mean (±SE) overall changes in P_pa from baseline to 240 min (B), initial pulmonary capillary pressure (P_pc) measures at 240 min (C), pulmonary capillary filtration coefficients (Kf.c) (D), and subsequently measured wet-to-dry weight lung ratios (W/D) (E) are compared. Chall, challenge; Rx, treatment.

FITC-albumin determinations in lung tissue and lavage following LT or PA challenge.

When compared with PA-challenged lungs (n = 6), LT challenge (n = 6) significantly increased the lung tissue FITC-albumin concentration normalized to the perfusate concentration [0.62 ± 0.13 vs. 0.25 ± 0.03 (mg/g)/(mg/ml), P = 0.03] (Fig. 9). The lung lavage FITC-albumin was not detectible in either PA- or LT-challenged lungs. Similar to other studies, LT in these experiments also increased W/D (P ≤ 0.01) (data not shown).

Fig. 9.

Serial mean (±SE) lung tissue albumin-fluorescein isothiocyanate conjugate (FITC-albumin) concentration normalized by perfusate FITC-albumin concentration at 240 min in lungs challenged with lethal toxin (LT, 1 μg/ml) or protective antigen (PA, 2 μg/ml) (n = 6 for all) is compared.

cAMP level determinations in perfusate following ET, LT, or PA challenge in lungs from studies 1, 8, and 9.

In effluent samples collected at 60, 120, 180, and 240 min from studies 1 and 8, compared with lungs perfused with PA alone, LT alone did not increase cAMP levels (P = 0.85), whereas ET alone and ET combined with LT increased cAMP levels, although these increases varied over time (P < 0.0001 for the time interaction) (Fig. 10). In study 9, compared with PA alone, ET with NS-mAb also produced significant increases in cAMP that varied over time (P = 0.0002 for the time interaction), but ET with PA-mAb did not. cAMP levels were also significantly increased with ET with NS-mAb compared with ET with PA-mAb (P = 0.005 for the time interaction).

Fig. 10.

Serial mean (±SE) cAMP levels at 60, 120, 180, and 240 min from lungs challenged with low-dose lethal toxin (LT, 1 μg/ml), high-dose edema toxin (ET, 5 μg/ml), and LT + ET (both 1 μg/ml) vs. lungs challenged with protective antigen (PA, 5 μg/ml, control) (A) and lungs challenged with 5 μg/ml ET and treated with either nonspecific monoclonal antibody (NS-mAb) or PA-directed mAb (PA-mAb) vs. 10 μg/ml PA (B) are compared. P values are for the overall effect of toxin with or without treatment vs. PA (designated challenge) and for the change in the effects of toxin with or without treatment compared with PA over time (i.e., the time interaction and designated challenge × time). Chall, challenge.

Wistar, Brown Norway, and Lewis rat isolated lung model studies.

In study 10, in Wistar and Brown Norway rats sensitive to lethal effects of LT, compared with control lungs, LT produced progressive increases in serial mean (±SE) changes from baseline in P_pa (P < 0.0001) and increased the final change in P_pa at 240 min and in mean P_pc, Kf.c, and W/D (P ≤ 0.06) (Fig. 11). In Lewis rats resistant to lethal effects of LT, the toxin did not produce significant changes in any of the parameters measured (P ≥ 0.35).

Fig. 11.

Serial mean (±SE) changes in pulmonary artery pressure (P_pa) from baseline to 240 min for lungs isolated from Wistar (A), Brown Norway (B), or Lewis (C) rats and challenged with lethal toxin (LT) vs. those challenged with protective antigen alone (control, PA) are compared. P values in A and B are for the change in the effects of LT compared with PA over time (i.e., the time interaction and designated challenge × time), whereas P values in C are for the overall effect of LT and for the time interaction. Effect of LT vs. PA on the mean (±SE) overall changes in P_pa from baseline to 240 min, initial pulmonary capillary pressure (P_pc) measures at 240 min, pulmonary capillary filtration coefficients (Kf.c), and subsequently measured wet-to-dry weight lung ratios in lungs from Wistar (D), Brown Norway (E), and Lewis (F) rats. Chall, challenge.

DISCUSSION

B. anthracis LT and ET had very different effects on pulmonary vascular pressures and permeability in isolated perfused Sprague-Dawley rat lungs. LT produced pulmonary vascular constriction evidenced by increases in both pulmonary artery and capillary pressures (P_pa and P_pc), and it increased the measured lung Kf.c and W/D ratios. Increases in Kf.c and W/D with LT were not related to effects of the toxin on increasing vascular pressure. Increased lung permeability with LT was also noted when assessed with a FITC-albumin assay. Although the effects of ET were less pronounced, in contrast to LT, ET reduced P_pa and P_pc and produced small but nonsignificant reductions in permeability. ET strongly blocked pulmonary artery vasoconstrictor effects of LT but did not alter its permeability ones. Inhibition of host cell uptake of LF or EF with PA-mAb negated each the effects of each toxin.

Findings here add to others suggesting that LT can increase vascular permeability. In prior studies, LT increased permeability across monolayers constructed from human pulmonary, brain, and umbilical vein microvascular endothelial cells and rat pulmonary microvascular endothelial cells using either electrical resistance or labeled protein or dextran measurements (52, 63). These changes with LT in rat pulmonary microvascular cells occurred in as little as 1 h, a time consistent with the permeability increases noted at 4 h with LT in the present study (35). Injection of LT in vivo increased extravasation of fluorescent microspheres in zebra fish embryos and fluorescein-labeled dextran in mouse lungs (5, 6). Different from effects of LT on permeability, this is the first report we are aware of showing that this toxin can cause direct pulmonary artery constriction, effects that were dose related and dependent on PA-mediated host cell uptake of LF.

On the one hand, the effects of LT on increasing both pulmonary vascular permeability and pressures are consistent with an effect of LF on activating intracellular actin-myosin elements in either endothelial or smooth muscle cells (52). LT has been shown to stimulate endothelial actin-myosin elements and, in so doing, to disrupt adherens junctional complex proteins, and inhibition of ERK1/2, a known action of LF, has been associated with this disruption (26, 52). No study has examined effects of LT on pulmonary artery smooth muscle cells, but ERK-1/2 activation has been associated with vascular smooth muscle relaxation related to some stimuli (51, 56), and ERK-1/2 inhibitory effects of LT would counter this relaxation. Also consistent with the present findings, in another study we conducted in aortic rings, LT increased the maximal contractile force rings developed with phenylephrine (34).

On the other hand, some mechanisms leading to increased permeability and vascular pressures in this lung model appear to differ. Two different rho-kinase inhibitors negated the effects of LT on P_pa but did not alter P_pa in lungs challenged with diluent alone. Involvement of rho-kinase in pulmonary vasoconstrictor effects of LT are very consistent with the role rho-kinase plays in regulating pulmonary vascular tone and producing pulmonary artery smooth muscle contraction related to stimuli such as hypoxia and endothelin, actions which Y27362 has been shown to inhibit (3, 27, 41). However, neither rho-kinase inhibitor decreased the permeability effect of the toxin, and both increased permeability in LT-challenged as well as in normal lungs. These findings differ from in vitro ones with LT. In one of the first studies to suggest that LT increased endothelial permeability, two different rho-kinase inhibitors (H1152 and Y27632) both decreased LT-stimulated myosin-light chain phosphorylation and permeability in human pulmonary microvascular cell monolayers (62). In another study, Y27632 disrupted LT-stimulated actin cable formation in human umbilical vein endothelial cells (45). These prior findings combined with the present ones suggest that rho-kinase has a different role in the changes in permeability measured in endothelial cell monolayer systems and this ex vivo lung system. The findings also suggest that LF triggers non-rho-kinase-associated mechanisms in this ex vivo whole organ system that contribute to the permeability effects of the toxin (4). Although activation of VEGFR has also been associated with permeability effects of LT in the zebrafish, a VEGFR-I administered at reportedly inhibitory concentrations did not alter effects of LT here (5, 6). Mechanisms underlying permeability of LT are likely not inflammatory in nature as well. Administration of LT in vivo had minimal effects on inflammatory mediator release in this Sprague-Dawley rat model (13) and actually inhibited this release in endotoxin-challenged animals (10). Other mechanisms and pathways have been associated with effects of LT on permeability in in vitro and in other models including activation of apoptosis (25), inhibition of p38-MK-2-stimulated heat shock protein-27 (35), inhibition of DE-cadherin transport to adheren junctions (21), inhibition of angiopoietin-1 signaling through Tie-2 (18), and histamine activation (19, 52). Resource limitations prevented our investigating these additional pathways in the present studies.

In the present study, LT also produced increases in P_pa and permeability in Wistar and Brown Norway rats, two other strains sensitive to the lethal effects of the toxin, but not in Lewis rats that are insensitive to these effects (42). These findings raise the possibility that pulmonary effects of LT here relate to some rat strain-specific genes such as isoforms of the Nlrp1 gene. Polymorphisms in this gene in the rat have been linked to an association between specific rat strains and both macrophage sensitivity to lysis by LT and to lethal effects of the toxin. However, since nonrat species (e.g., mice, rabbits, and canines) are also sensitive to the lethal effects of LT and LT increased lung permeability in a mouse study and across human endothelial cell monolayers, the present findings in the isolated rat lung appear relevant to the pathogenic effects of LT in other nonrat species (30, 54).

Endotoxin (LPS), a bacterial toxin commonly associated with the development of sepsis-associated lung injury and shock, has also been investigated in isolated perfused rat lung models. In studies we reviewed, for comparison with our findings with LT, the effects of LPS on lung function in these models appear to have varied based on the LPS dose and methodology employed. Two to three hours of perfusion with LPS concentrations ≤50 μg/ml in perfusate with buffer alone produced no effect on either P_pa or lung permeability (17, 58). Concentrations of LPS of 200–300 μg/ml either did not alter P_pa but increased permeability after 6 h of perfusion with a perfusate including buffer and blood (9, 50) or produced an immediate transient increase in the former without a change in the latter with whole blood perfusion (8). Finally, a concentration of LPS of 400 μg/ml increased permeability but not P_pa after 20 min in lungs perfused with buffer alone (48). The concentrations of LPS used in these models are comparable to ones producing lethality in vivo in animals (5–20 mg/kg) (10) but are notably much higher than the concentration of LT producing pulmonary effects or lethality in the lung model or in vivo (50 μg/kg) (11–13). Also, although pulmonary and lethal effects of LPS in both ex vivo and in vivo lung models have frequently been associated with inflammatory mediator release (e.g., TNF, IL-1, lipid metabolites), lethal LT doses in the rat may actually suppress this release (10, 11, 13).

Reductions in pulmonary artery and capillary pressures with ET both by itself and in combination with LT in the present study are most likely related to the effect of this toxin on increasing intracellular cAMP levels. We have shown in an isolated aortic ring model and in in vivo studies that ET causes potent systemic arterial vasodilation in association with increases in both tissue and circulating cAMP levels (11, 34, 53, 54). Adefovir, a nucleoside that selectively inhibits cAMP production by EF, blocks these vasodilatory effects of ET both in the aortic ring model and in vivo (34, 53). Since increases in cAMP stimulate pulmonary arterial relaxation, ET likely has similar effects on pulmonary and systemic vascular tone (43). Although the vasodilatory effects of ET were less pronounced in the high-compliance pulmonary vasculature, once tension was increased with LT in isolated lungs, the effects of ET were much more apparent. Uptake of EF was necessary for these vasodilatory effects of ET as PA-mAb reversed reductions in P_pa. Consistent with their effects on P_pa, LT increased, and at the higher dose, ET decreased P_pc. These opposing effects likely reflect the differing mechanisms, potentially mediating the effects of these toxins on P_pa. Effects of ET on permeability in this lung model are less clear. Different from what its name would imply, ET did not increase Kf.c or W/D in any study. In fact, in each of three studies, compared with a PA control, ET was associated with decreases in Kf.c that were not significant in study 1 and study 8 with low- and high-dose ET but were significant in study 9 with a high dose. However, PA-mAb did not alter small decreases of ET in Kf.c, and ET had no measurable effect on increases in Kf.c with LT. Overall though, the present findings with ET are more consistent with the strengthening effects that increased cellular cAMP can have on endothelial barrier function (47).

The present findings with LT and ET in this isolated lung model provide insights into how these toxins potentially alter lung function during B. anthracis infection. Increased permeability with LT exposure at either the alveolar endothelial or epithelial levels or both could produce direct alveolar protein and fluid accumulation, ventilation and perfusion mismatching, and hypoxia and reduced lung compliance. Histopathology studies have demonstrated alveolar protein and fluid accumulation in some but not all models (13, 28, 39, 64). Permeability increases in the pleural visceral or parietal tissues, could also contribute to the recurrent pleural effusions that characterize inhalational anthrax disease. LT challenge has been shown to produce pleural effusions in animal models (13, 39). When clinically large, these effusions produce atelectasis and indirectly impair oxygenation and compliance. Finally, permeability changes with LT within the mediastinal tissues could contribute to the marked mediastinal edema observed in patients with inhalational disease and which in turn could alter mediastinal lymphatic flow and aggravate pleural fluid accumulation. However, although histopathology study in humans dying with inhalational anthrax demonstrate protein-rich fluid accumulation in all of these lung tissues, to what extent these changes relate to LT versus other bacterial components or some combination is unknown (20). Notably, the necrotizing-hemorrhagic mediastinitis and disrupted lymphatic drainage seen in patients is thought to make an important contribution to the recurrent pleural effusions (1, 24). Different from its permeability effects, pulmonary vasoconstrictor effects of LT in this isolated lung model, although highly consistent and significant across experiments, were relatively small and not likely sufficient alone to produce consequential increases in pulmonary vascular resistance and secondary abnormalities in cardiac performance. These constrictor effects could add to the cardiovascular effects of other bacterial components such as cell wall constituents. Finally, vasodilatory effects of ET in this model, although potentially countering vasoconstrictor effects of LT, could contribute to dysregulated pulmonary blood flow. Importantly, ET-associated vasodilation in the pulmonary vasculature could impair the protective effects of hypoxic vasoconstriction in patients with anthrax-associated pulmonary infection, edema, or pleural effusions.

This study has limitations. Changes were observed in the model over a relatively brief 4-h period, and how long they would persist either after a single exposure to toxin or with repeated challenges is unknown. Also, it is unknown whether the changes noted in this ex vivo lung and toxin system will also occur in lungs exposed to toxin in vivo. Finally, B. anthracis produces other components that contribute to its pathogenesis, and whether these components would synergize with or antagonize the pulmonary vascular effects of LT and ET is unknown.

The present findings support the possibility that LT contributes to the increased vascular permeability and extravasation of fluid and protein that characterizes B. anthracis infection. Although other bacterial components such as B. anthracis cell wall with its robust inflammatory properties likely also participate in this process, our findings do not support a role for ET in this increased vascular permeability. However, ET may have pulmonary vasodilator effects, which could counter protective adaptive mechanisms like hypoxic pulmonary vasoconstriction. These findings with LT and ET together add to others providing a basis for administration of agents that target both toxins during the development of shock and organ injury with B. anthracis infection.

GRANTS

This research was supported by the Intramural Research Programs of the National Institutes of Health, Clinical Center, Critical Care Medicine Department, and the National Institute of Allergy and Infectious Diseases.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

X.C., M.M., S.H.L., and P.Q.E. conceived and designed research; X.C., W.X., P.N., A.W.-S., R.W., B.P., Y.L., Y.F., and P.Q.E. performed experiments; X.C., M.M., S.H.L., and P.Q.E. analyzed data; X.C., M.M., S.H.L., and P.Q.E. interpreted results of experiments; X.C. and P.Q.E. prepared figures; X.C. and P.Q.E. drafted manuscript; X.C., W.X., P.N., A.W.-S., R.W., B.P., Y.L., M.M., S.H.L., Y.F., and P.Q.E. edited and revised manuscript; X.C., W.X., P.N., A.W.-S., R.W., B.P., Y.L., M.M., S.H.L., Y.F., and P.Q.E. approved final version of manuscript.