The Role of Cyclooxygenase Products in Lung Injury Induced by Tumor Necrosis Factor in Sheep1-4

ARTHUR P. WHEELER, WILLIAM D. HARDIE, and GORDON R. BERNARD

Introduction

Sepsis is the most common precipitating cause of acute lung injury (1). Endotoxin, or lipopolysaccharide (LPS), is a major cause of acute lung injury, and it has been postulated that tumor necrosis factor-alpha (TNFa), a peptide elaborated by macrophages and monoeytes,is one mediator of endotoxernia (2, 3). Endotoxin and TNFa infusions in sheep are characterized by pulmonary hypertension, altered lung mechanics, disordered gas exchange, and the exudation of protein-rich lymph, changes similar to those that occur acutely in human sepsisinduced acute lung injury (4, 5). Data supporting the role of TNFa in LPS-induced injury come from four areas. First, animals resistant to LPS are deficient in the ability to produce TNFa (6). Second, TNFa elaboration is coincident with the onset of physiologic changes after LPS (7). Third, recombinant TNFa can reproduce the effects of endotoxin on the lung, albeit more rapidly than LPS (3, 8). Finally, anti-TNFa antibodies improve survival after the infusion of LPS or live bacteria in some models of sepsis and can attenuate the production ofinterleukins 1 and 6 (9-11). In various models, TNFa like endotoxin causesthe releaseofeyclooxygenase and lipoxygenase products, plateletactivating factor, interleukin-l (IL-l) and interleukin-2 (12, 13).Some of these same inflammatory mediators in turn exert negative feedback on the inflammatory response. For example, prostaglandin E2 (PGE 2 ) attenuates leukotriene B4 (LTB4 ) , TNFa, and IL-l synthesis (14-17). Although some arachidonic acid metabolites may haveinflammatory actions (e.g., leukotriene B4 ) , or direct detrimental physiologic effects (e.g., thromboxane A 2 , prostaeyclin) (15), PGE2 and prostaeyclin (P0I 2) may also favorably modulate the adverse effects of cytokines (14, 16).Hence, inhibition of PGE 2 and P0I 2 production could worsen physiologic derangements by augmenting LTB4 , TNFa, 632

SUMMARY Tumor necrosis factor-alpha (TNFa) has been proposed as a mediator of endotoxinInduced lung Inlury. When given to sheep, TNFa mimics endotoxin (LPS) causing hypoxemia, pulmonary hypertension, leukopenia, redUced dynamic compliance (Cdyn), Increased resistance to airflow (RL), exudation of lung lymph, and enhanced airway reactiVity. TNFa also Induces rapid releaee of thromboxane A, (1XA,), prostaglandin E, (PGE,), and prostacyclln (PGI,). We hypothesized thst the Inflammatory effects of TNFa are due at least In part to cyclooxygenase prodUcts, and therefore cyclooxygenase Inhibition would have similar effects on TNFa-Induced lung Injury as has previously been demonstrated for LP8-lnduced lung damage. Using awake sheep with ehrenIc lung lymph fistulas, _ measured Cdyn, RL, and FRC using a whole-body plethysmograph. Pulmonary artery (Ppa), left atrial (PLA), and systemic arterial (Pas) preasures were recorded continuously. Arterial blood gasas (for celculatlng AaPO,),leukocyte counts, and lymph semplas (for prostanold lavela)werecollected avery 30 min. EllMln animals underwent paired random-order experiments receiving Ibuprofen (14 mgllcg) 1 h before human recombinant TNFa (10 119/1cg), or an Identical dose of TNFa alone. Within 15 min of Initiating TNFa, Ppa doubled and remained elevated for 4 h. Ibuprofen prevented the early rise In Ppa aftar TNFa. In the group receiVing TNFa alone, Increases In Ppa _re accompanied by a 60% dacllne In leukocyte count and a 50% Increase In AaPO, within 30 min. Ibuprofen prevented Increases In AaPO" but It had no effect on leukopanla or late Increases In lymph flow. By Itself, TNFa caused a 35% reduction In Cdyn and a fourfold Increase In RL within 15 min. Ibuprofen prevented Increases In RL and delayed reductions In Cdyn. Increases In 1XA, and PGE, metabolites after TNFa _re prevented by Ibuprofen, but Increases In PGI, metabolites _re Incompletely blocked. Weconclude thst the early Increases In Ppa, RL, Cdyn, and AaPO,after TNFa are at least partially mediated by cyclooxygenase prodUCts and can be diminished by Ibuprofen. The effects of Ibuprofen on TNFa-lnduced physiologic changes are similar to those of other cyclooxygenase Inhibitors on endotoxin-Induced lung Injury In sheep that have been previously reported. AM REV RESPIR DIS 1992; 145:632-639

and IL-l production and by removing po- sis (26-31), the importance of eyclooxtent vasodilators (17). Conversely, re- ygenase products and the net physiologduced production of vasoactive and ic effect of eyclooxygenase inhibition in bronchoactive substances could amelio- TNFa-induced lung injury is unknown. rate the lung injury caused by TNFa. Therefore, we investigated the effects of There is also interesting but conflict- ibuprofen on TNFa-induced lung injury ing data regarding the relationship of in chronically instrumented, awake sheep the drug ibuprofen to LPS-induced cytokine generation. In humans, data indicate that ibuprofen augments TNFa pro- (Received in original form April 22, 1991 and in duction after endotoxin administration revised form October 2, 1991) (18), whereas in anesthetized pigs, ibuFrom the Center for Lung Research, Division profen appears to reduce TNFa genera- of Pulmonary and Critical Care Medicine, Departtion in response to endotoxin (19).There ment of Medicine, Vanderbilt University,Nashville, are also numerous reports documenting Tennessee. 1 Supported by TrainingGrants HL-07I23, SCOR the effectiveness of E series prostaglandins in limiting inflammation and cyto- Grant for Pulmonary Research HL-I9IS3, and Grant HL-27274 from the National Institutes of kine generation in vitro and in animal Health and by the Upjohn and Cetus Corporations. models (20, 21); however, such com3 Presented in part at the Second International pounds have not proved effective in hu- Conference on Cytokines, Hilton Head, South Carolina, December 1989. man clinical trials (22-25). • Correspondence and requests for reprints Despite reports of enhanced survival should be addressed to Arthur P. Wheeler, M.D., and improved physiology afforded by ey- B-1308 Medical Center North, Vanderbilt Univerclooxygenaseinhibitors in models of sep- sity, Nashville, TN 37232. 1

633

CYCLOOXYGENASE INHIBITION IN TNFa WNG INJURY

to define the role ofcyclooxygenaseproducts and to contrast this with previous reports of cyclooxygenase inhibition in LPS-induced lung injury.

ville, MD). Ibuprofen was supplied as a sterile solution of 50 mg/ml in 0.9070 nonpyrogenic saline and was infused intravenously over 10min via the jugular catheter in a dose of 14 mg/kg.

tinued, the change in plethysmograph volume versus change in Pao was graphed. Anatomic and circuit dead space were subtracted for FRC calculations and adjusted for barometric pressure.

Methods

TNFa Preparation Human recombinant TNFa was generously supplied by Cetus Corporation (Emeryville, CA). On the day of each experiment 312 ug (7.5 x 106 units) ofTNFo. were reconstituted in 1.2ml of pyrogen-free water by gentle mixing. TNFa exceeded 99% purity as determined by SDS-PAGE and contained less than 0.02 ng of endotoxin/mg protein as determined by limulus lysate assay. A dose of 10 ug/kg was infused intravenously over 30 min. Any reconstituted TNFa not immediately used was stored at - 70 0 C.

Hemodynamic Measurements Left atrial (PLA), pulmonary arterial (Ppa), and systemicarterial pressures(Psa) weremeasured using Hewlett-Packard 1280Cpressure transducers (Hewlett-Packard, Palo Alto, CA) through a Validyne CD-19 carrier demodulator (Validyne Engineering, Northridge, CA) and recorded onto a Gould 2800 multichannel recorder (Gould Instruments, Cleveland, OH). Mean pressures were determined electrically; however, the dynamic signal was inspected frequently to assure an undamped waveform.

Measurement of Lung Mechanics Awake sheep were studied while standing in a Plexiglas'" 285-L pressure-compensated, integrated-flow whole-body plethysmograph as previously described (3-5, 33-35). The sheep were loosely restrained by a sling placed under them to discourage lying down during the experiment. The tracheostomy tube was connected to an external valve by noncollapsible tubing, allowing occlusion of the airway for determination of FRC. A constant 10 L/min bias flow of humidified room air was used to reduce the effective dead space of the circuit tubing. Tidal volume (VT) was measured by pressure-compensating the integrated pressure signal from the plethysmograph. Flow (V) was derived by electrically differentiating the volume signal. Pressure at the airway opening (Pao) was measured using a multiple side-hole catheter placed 0.5 em beyond the end of the tracheostomy tube. Pleural pressure (Ppl) was measured using the pleural envelopes. Transpulmonary pressure (Ptp) was calculated as the difference between Ppl and Pao, Pressure signals from the pleural envelopes,airway,and vascular catheters were tuned to eliminate phasic distortion. To provide a constant-volume history, one minute before each set of measurements the lungs wereinflated to 40 em H 20 airway opening pressure using the bias gas flow. Simultaneous VT/V and VT/Ptp tracings were recorded and photographed on a Tektronix 5A18N dual-beam storage oscilloscope (Tektronix; Beaverton, OR) to allow calculation of dynamic lung compliance (Cdyn) and resistance to airflow across the lung (RL). Points of zero flow were electronically marked on the oscilloscope tracings. Cdyn was calculated as VTdivided by Ptp at points of zero flow, expressed as L/cm H 20 at BTPS. RL was calculated using the method of von Neergaard and Wirz (36) by dividing Ptp by V at midtidal volume, expressed as em H 2 0 / L / s at BTPS. FRC was measured using the Boyleslaw method of Dubois and coworkers (37). To calculate FRC, the airway was obstructed at endexpiration and, while inspiratory efforts con-

Eicosanoid Measurements Lymph samples werecollectedover EDfA and meclofenamate, centrifuged to remove cells, and stored at -70 0 C until analyzed. Eicosanoid analysis was performed as described by Pradelles and coworkers (38) and by Westcott and colleagues (39). All monoclonal antiIgG antibodies, enzymatic tracers, and specific eicosanoid antisera were obtained from AlA reagents (Aurora, CO). A 96-well microtiter plate (Nunc, Roskilde, Denmark) was coated with an appropriate mouse monoclonal antirabbit IgG. To each sample well was added enzyme tracer (50 ~I) composed of PGE2 , PGF" or lXB2 covalently linked to purified acetylcholinesterase. Fifty microliters of known standard or sample wereintroduced into each well. Specific rabbit eicosanoid antiserum (50 ~I) was added to each well to initiate competitive binding. After 16 to 24 h of incubation, plates were washed four times with phosphate buffer (10-2 M; 0.05% Tween 20; pH, 7.4) using an automated washer (Flow Laboratories, McLean, VA).Wells were then filled with 200 ~I of Ellman's reagent (2 ug/ml acetylthiocholine iodide and 2.15 ug/ml of 5-5"-dithiobis'-2-nitrobenzoic acid in 10-2 M phosphate buffer). Enzymatic production of a yellow product was monitored at 414 nm by an automatic plate reader (Titertek Multiskan MC; Flow Laboratories), Each sample was assayed in duplicate. Maximal tracer binding in the absence of eicosanoid (Bs) was determined by replacing sample with an equal volume of assay buffer (0.1 M potassium phosphate with 0.01% NaN 3 , 0.4 NaCI, 1mM tetrasodium EDfA, and 0.1% BSAat pH 7.4). Nonspecific binding was measured by replacing the specific antisera with assay buffer. A standard curve was constructed by graphing B/Bo% (ratio of bound fraction absorbance in the presence of eicosanoid to absorbance in the absence of eicosanoid) against picograms of eicosanoid/well. Construction of the standard curve and quantitation of eicosanoid in the lymph samples was accomplished with a nonlinear curve-fitting program (40). Cross-reactivities of the lXB2 antisera were as follows: 1XB" 17%; dinor lXB2 , 11%; 11-

Sheep Preparation Sheep weighing 28 to 40 kg were chronically instrumented as previously described (3-5) to study the impact of ibuprofen on TNFo.induced changes in airway mechanics, gas exchange, lymph flow and composition, circulating leukocyte profile, and hemodynamics. After induction of anesthesia intravenously with short-acting barbiturate (Thiamalayl; Parke Davis, Morris Plains, NJ) orotracheal intubation was performed, and anesthesia was maintained using halothane and oxygen. A right lateral thoracotomy was performed to cannulate the efferent duct of the caudal mediastinal lymph node with a heparinized silastic catheter (32). Through a low right thoracotomy incision, the tail of the lymph node was ligated, and the hemidiaphragms were cauterized to prevent contamination of lung lymph with that from an abdominal source. Two silastic envelopes for measurement of pleural pressure wereinserted through the same thoracotomy incision. The silastic catheter entering each envelop was exteriorized and secured. Silastic catheters were introduced in a retrograde manner into the superior vena cava and aorta from the jugular vein and carotid artery. A left lateral thoracotomy was performed for the placement of catheters into the main pulmonary artery and left atrium. Before closure of the thoracotomy incisions, the lungs were inflated to 40 em H 20 pressure, and the incision was closed while 20 cm H 20 airway pressure was maintained to prevent atelectasisand residual pneumothorax. A permanent tracheostomy was created 4 em below the larynx into which a no. 10 cuffed tracheostomy tube (Shiley, Irvine, CA) was inserted before each study and removed immediately after each experiment. After surgery the sheep routinely were awake within 30 min and were able to stand and eat within 1 h. Vascular catheters were flushed twice daily with a mixture of normal saline, heparin (100 U/ml), and gentamicin (1 mg/ml). The tracheostomy stoma was cleaned daily with a 1070 solution of hydrogen peroxide and covered with a loose-fitting collar to prevent contamination. Each animal was allowed 5 to 7 days to fully recover from surgery before any experiment. Animal protocols were approved by the institutional animal care committee in accordance with National Institutes of Health guidelines. Ibuprofen Preparation and Dosing An intravenous preparation of ibuprofen was a generous gift of the Upjohn Corporation (Kalamazoo, MI). The drug was endotoxinfree « 0.1 EU/ml) when tested using a chromogenic modification of the limulus lysate assay (Whittaker; MA Bioproducts Walkers-

634

WHEELER, HARDIE, AND BERNARD

dehydro-Txlsj, 0.7070; PGD., 0.4%; PGF.ll , 0.2%; PGE. < 0.1%; 6-keto-PGF," < 0.1%; 6,15-diketo-PGF Ill < 0.1%.

recorded every 15 min, collected, and stored at 30-min intervals.

Other Measurements

All results are expressed as mean ± SEM. The effects of ibuprofen pretreatment on TNFa-induced changes were compared with baseline values and with each other using oneand two-way analysis of variance and paired t tests when differences were demonstrated. Comparison of ibuprofen-treated and -untreated groups with respect to prostanoid levels was performed using a single t test of area under the curve. In each case, the null hypothesis was rejected if p < 0.05.

Arterial blood gases were obtained anaerobically from the aortic catheter using less than 0.1 ml of heparan sulfate as an anticoagulant. Temperature-corrected gas tensions and pH were measured within 1 h of collection using a Corning 158pH/blood gas analyzer (Corning Medical and Scientific, Medfield, MA). Alveolar to arterial Po. difference (AaPo.) was calculated using the alveolar gas equation, assuming a respiratory quotient of 0.8. Blood leukocyte counts were made using a Coulter counter (Coulter Electronics, Hialeah, FL). Differential leukocyte counts were done manually using Wright-stained smears. Lymph flow was recorded at 15-min intervals, and pooled samples were collected at 30min intervals. Lymph was collected on EDTA and meclofenamate to prevent clotting and to inhibit prostaglandin metabolism, respectively. Lymph and plasma protein concentrations were determined using an automated Biuret method (Technicon Autoanalyzer; Technicon Instruments, Tarrytown, NY).

Sheep Baseline Criteria Prior to initiation of any experiment, sheep met all of the following criteria: pulmonary artery pressure ~ 20 cm H.O, Pao, ~ 80 mm Hg, peripheral leukocyte count ~ 5,OOO/mm', and ~ 15,OOO/mm'. Additionally, there could be no overt signs of sepsis or evidence of local infection.

Experimental Protocol Eleven sheep underwent paired, random-order experiments in which each sheep received either a 15-min infusion of ibuprofen (14 mg/kg) followed 1 h later by TNFa (10ug/kg) or a 30-min infusion of TNFa alone. Sheep were monitored for 4 h after TNFa administration. The paired experiments were separated by 3 to 5 days. Eight of the 11 sheep had functioning lung lymph fistulas for both arms of the experiment. On the day of each experiment, sheep were placed in the whole-body plethysmograph, and 60 to 90 min of baseline hemodynamic and airway data were collected until a l-h stable baseline was achieved. During the studies, vascular pressures were continuously measured and recorded with pressure transducers zero-referenced to the level of the left atrium. At least four measurements of Cdyn, RL, FRC, and lymph flow were made at 15-min intervals during the baseline period. Arterial blood gases and leukocyte counts were determined at 30-min intervals during the baseline. The baseline value for each variable used for statistical analysis was computed as the mean of all of the baseline measurements. After infusion of ibuprofen or TNFa, vascular pressures and lung mechanics measurements were made at 15-min intervals. Arterial blood gas and leukocyte counts were performed every 30 min. Lung lymph flow was

Statistical Analysis

r: p < 0.05 n. BASELINE ..,

PA

Results

Control Studies One sheep received TNFa boiled for 20 min before infusion. Such treatment inactivates TNFa, but it does not destroy endotoxin. After infusion, there were no changes in any measured variable over 4 h of monitoring (data not shown). This sheep subsequently received unboiled TNFa and demonstrated the typical response described below. The infusion of 14 mg/kg of ibuprofen alone in four sheep did not significantly alter any measured variable during 1 h of monitoring (table 1). General Observations Within 15min of initiating the infusion, animals receivingTNFa alone developed rigors and mild agitation. There were no changes seen in the ibuprofen-treated animals. Hemodynamics Ibuprofen alone had no significant effect on Ppa or PLA (table 1). In the group receiving TNFa only, Ppa peaked 15min after beginning the TNFa infusion at ap-

PRESSURE 30 (em

- -s!---s!

HzO)

* 15

oJ

'--p

!jill ~

,

i

-2

-1

,

"

•

0 1 2 TIME lin hoursl

proximately twice baseline and remained elevated for 4 h (figure 1). Ibuprofen prevented the early peak in Ppa (p < 0.05, ibuprofen-treated versus TNFa only groups); however, within 1 h of beginning the TNFa, Ppa was similar in both groups (figure 1). During the TNFa infusion, PLA transiently decreased in the group receiving TNFa only, but it promptly returned to baseline. PLA values were not different from baseline values in either group and did not differ from each other at any time point (table 2).

Lung Lymph Ibuprofen infusion alone had no effect on lung lymph flow (QL) (table 1). There were no significant differences in QL or lymph protein clearance (QLP) calculated as (lymph flow x [lymph protein]/[plasma protein]) at baseline between the groups (figure 2). QLwas lower in the ibuprofen-treated group 30 and 45 min after initiating the TNFa infusion; however, 1 h after TNFa, both

TABLE 1 4)*

Time after Ibuprofen Infusion

Ppa, cm H20 PLA, cm H20 OL, ml/15 min Cdyn L1cm H20 RL, cm H20/L/s WBC, x lQ3/mm' AaP0 2, mm Hg PGE2 , pglml TxB 2 , pglml 6-keto-PGF,., pglml

Baseline

17.7 1.75 1.04 0.100 1.6 8.1 28.2 66 660 169

± 0.7 ± 2.2 ± 0.2

± 0.01 ± 0.3

± 0.8 ± 2

± 23 ± 314 ± 49

4

Fig. 1. Pulmonary artery pressure response to tumor necrosis factor. Characteristic pulmonary hypertension occurred in the TNFa control animals (closed circles) (n = 11); however, the early rise in Ppa was blunted by pretreatment with ibuprofen (open circles) (n • 11). Asterisks indicate p < 0.05, ibuprofen-treated versus TNFa control animals.

EFFECT OF IBUPROFEN ADMINISTRATION ALONE (n

Variable

*

-e 0.05 .1. BASELlNE....J

30 Min

17.8 ± 0.9 0.0 ± 1.8 1.07 ± 0.2 0.095 ± 0.01 2.3 ± 0.6 7.6 ± 0.7 27.0 ± 3 34 ± 10 305 ± 81 380 ± 220

60 Min 18.5 -1.0 1.2 0.090 2.4 7.8 30.0 45 588 197

± 0.1

± 1.8 ± 0.4 ± 0.01 0.5

± ± ± ± ± ±

0.9 5 19 384 110

Definition of abbreviations: Ppa = pulmonary artery pressure; PLA = left atrial pressure; QL = lung lymph flow; Cdyn = dynamic compliance; RL = pulmonary resistance; wee • white blood cells; AaPo, = alveolar-arterial Po, difference; PGE, • prostaglandin E,; TxB, = thromboxane B,. • Values are mean ± SEM. t p < 0.05 time point compared with group bBSeline.

635

CVCLOOXYGENASE INHIBITION IN TNFa WNG INJURY

TABLE 2

There was no significant difference in baseline FRC between the two groups and no change in FRC from baseline in either group (data not shown).

LEFT ATRIAL PRESSURE (n = 11)* Time after TNFa Infusion Group TNFa Ibuprofen

Baseline

30 Min

60 Min

120 Min

180 Min

240 Min

-0.1 ± 1.0 - 1.0 ± 1.8

-3.0 ± 2.4 - 3.2 ± 2.1

-3.7 ± 2.4 - 3.3 ± 1.8

-2.2 ± 1.1 - 3.8 ± 1.0

-2.7 ± 1.4 - 2.8 ± 1.1

-1.0 ± 1.2 - 1.5 ± 1.4

Definitionof Bbbrev/ations: TNFa = tumor necrosis lactor-alpha. • Measured in em H20. Values are mean :i: SEM. t p < 0.05 time point comparedwith group baseline.

groups had equivalent increases in QLto 2.5 to 3.5 times baseline. QLand QLP remained approximately twofold to threefold above baseline for 4 h in both groups (figure 2).

Lung Mechanics The effects of TNFa infusion on lung mechanics with and without ibuprofen are shown in figure 3. Ibuprofen alone caused an approximately 10070 decline in mean Cdyn, a change that failed to reach statistical significance. Cdyn declined rapidly after beginning the TNFa, ultimately falling to approximately 60% of baseline in both control and treated groups. Cdyn declined less rapidly in the ibuprofen-treated group, but it reached a similar nadir. Cdyn remained below

baseline for 4 h for both groups (figure 3). No statistical difference could be demonstrated between groups with respect to Cdyn after completion of the TNFa infusion. Resistance to RL increased fourfold within 30 min in the TNFa group and remained approximately twofold above baseline for 4 h (figure 3). In the ibuprofen-treated group there was a slow and nonstatistically significant increase in RLseen 2 to 4 h after the TNFa infusion; however, the early profound rise in RL was prevented (p < 0.05, TNFa-only group versus ibuprofen-treated group). RL in the ibuprofen-treated group was significantly lower than RL in the group receiving TNFa only during the TNFa infusion (figure 3).

p < 0.05 VS. TNF BASELINE

r-

5

---,

4

LYMPH

3

FLOW (m1/15 minutes) 2

0

Fig. 2. Lymph flow and lymph protein clearance after TNFa challenge. Early increases in lymph flow were reduced by ibuprofen (open circles) treatment; however. the latersustainedelevationin flow rates seen in TNFa control animals (closed circles) (n = 8) were not attenuated by ibuprofen (n = 8). Lymph protein clearance did not differ between groups at any point.

[@ll -2

ITNFI

I

I

-1

0

1

3

2

4

TIME lin hours)

p

.-

< 0.05 vs. TNF BASELINE

-.

3.5 3 LYMPH PROTEIN 2.5 CLEARANCE 2

r

ITT

TT1 1 1'-0 T,.0--0 r o TIl /1 6 0--0--0

1.5

T l: /1

L - p < 0.05 vs. ---.J

0--0--0--0

-2

I

i

-1

0

IBU+TNF BASELINE

1

2

TIME.(in hoursl

3

4

Gas Exchange There wereno differences betweengroups with regard to AaP02 at baseline. Ibuprofen alone had no significant effect on AaP02 (table 1). Gas exchange data are graphically depicted in figure 4. The mean AaP02 at baseline increased by approximately 50% within 30 min in the TNFa control animals and remained above baseline throughout the monitoring period. Increases in AaPo2 in the group receiving TNFa only averaged 10 to 12 mm Hg AaPo2 did not increase in the ibuprofen-treated group (p < 0.05, TNFa-only group versus ibuprofen treated group) (figure 4). WBC Counts and Differential There wereno differences betweengroups in leukocyte count at baseline and no effect of ibuprofen alone on WBC counts (table 1).As shown in figure 5, all animals exhibited leukopenia and neutropenia after TNFa infusion. Total leukocyte counts reached a nadir of 35 to 40% of baseline within 30 min of beginning the TNFa infusion in both groups (p < 0.05, nadir WBC versus baseline) (figure 5). Reductions in neutrophil counts paralleled the fall in total leukocyte count. There were no significant differences among the groups with regard to number of circulating neutrophils at baseline, or at any time point after infusion of the TNFa with or without ibuprofen pretreatment (table 3). Prostaglandin Results The pattern of eicosanoid release into lung lymph is shown in figure 6. lXB2 , the stable metabolite of thromboxane A 2 , did not differ between groups at baseline. lXB2 rapidly increased in the group receivingTNFa only, rising approximately fivefold above baseline to peak 30 min after initiating the TNFa infusion (p < 0.05 versus baseline). lXB 2 slowly returned toward baseline levels in the group receiving TNFa only, but it remained 2-fold above baseline at 4 h. There was no increase in lXB2 in the ibuprofen-treated group (p = NS versus baseline). Differences in 1XB2 levels as determined by area under the curve analysis were significant between the group receivingTNFa only (5,835 ± 590, mean ± SD), versus 2,205 ± 171 in the ibuprofen-treated group (p < 0.05).

636

WHEELER, HARDIE, AND BERNARD

120

p < 0.05 vs. BASELINE

r

T

T

BOTH GROUPS

~-~- -

100

...,

T

CO'fN

80 IX BASELINE)

60

-2

-1

0

1

2

4

3

TIME (in hours)

p < 0.05 vs, BASELINE

..

600

,.-,

Fig. 3. Dynamic compliance (Cdyn) and resistance to airflow across the lung (RL) after TNFa infusion. The rapid decline in Cdyn induced by TNF alone (closed circles) (n = 9) was not changed by ibuprofen treatment (open circles) (n = 9); however, increases in RL after TNFa were attenuated by ibuprofen. Asterisks indicate p < 0.05, ibuprofen-treated versus TNFa control animals.

400

~

T T __0"

IX BASELINE)

200

.,01

T

2

3

........

T .1.

100

!i6ill mill

o

I i i

-2

-1

0

1

4

TIME (in hours)

There were no differences in lymph PGE z levels between the groups at baseline. PGE z rapidly increased in the group receiving TNFa only, reaching maximal p < 0.05 vS.TNF ,--

BASELINE

---,

18

12

CHANGE IN A.PO. FROM BASELINE (mm Hg)

00

-6

illffi

I i i

-2

-1

0

1

2

4

TIME (ln hoursl

Fig. 4. AaPo, after TN Fa . There was no significant difference at baseline in AaPo, between the TN Fa control (closed circles) (n = 11) and the ibuprofen-treated groups (open circles) (n = 11). Ibuprofen prevented the increase in AaPo. seen in TNFa control animals. Asterisks indicate p < 0.05, ibuprofen-treated versus TNFa control animals.

12

levels 1 h after beginning the TNFa infusion (p < 0.05 versus baseline). PGE z levels remained above baseline for 4 h in the group receivingTNFa only; however, there was no significant increase in PGE z in the ibuprofen-treated animals (p = NS). Area under views curve analysis demonstrated differences in PGE z metabolite levels at p < 0.05 (2,427 ± 259 in the group receiving TNFa alone versus 414 ± 28 in the ibuprofen-treated group). The stable prostacyclin metabolite 6-keto-PGF ta was not different between groups at baseline. 6-keto-PGF ta levels increased in both groups when compared with baseline values (p < 0.05versus baseline for both groups). Differences in 6-keto-PGF t u levels between groups by area under the curve analysis were also significant at p < 0.05 (3,068 ± 330 in the group receiving TNFa only versus 1,607 ± 130inibuprofen-treatedgroup).

T T 1GR6u;sO~i. ~~~~LlNEl

Discussion

There is a growing body of evidence to

WBe

suggest that TNFa and other cytokines mediate endotoxemia; however, the critical distal physiologic mediators generated by LPS and TNFa remain uncertain (3, 8). In this study we have characterized the effect of cyclooxygenase inhibition on the physiologic effects induced by one cytokine, TNFa. By inhibiting one of several potential pathways of the action of TNFa, i.e., the cyclooxygenase pathway, we hoped to gain insight into the mechanism of TNFa-induced injury. We believe this to be the first largeanimal study to simultaneously measure the effects of TNFa on lung mechanics, hemodynamics, gas exchange, lung lymph flow and composition, and peripheral white blood cell count as well as the effect of a cyclooxygenase inhibitor on those changes . We have previously shown that infusion of 10 ug/kg of human recombinant TNFa into awake sheep causes the rapid onset of systemic and pulmonary arterial hypertension, hypoxemia, prompt profound and sustained reductions in Cdyn, and shorter-lived increases in RL (3). TNFa also causes leukopenia, granulocytopenia, and the production of proteinrich lung lymph. In addition, TNFa causes increases in prostaglandin E z , thromboxane A z , and prostacyclin metabolites in lung lymph (8) while producing only a modest degree of perivascular pulmonary edema (41). Such changes are qualitatively similar to those seen after low doses of LPS (1, 4, 5). In animal models of endotoxemia, cyclooxygenase inhibitors prevent increases in Ppa and pulmonary vascular resistance in the first 60 to 90 min after endotoxin; however, they do not affect later gradual increases of Ppa (I, 5). Reduction in LPSinduced pulmonary hypertension seen after treatment with cyclooygenase inhibitors or specific thromboxane antagonists correlates well with reductions in the levels of the potent vasoconstrictor Txa, (5, 33, 42, 43). We have now demonstrated similar findings using human recombinant TNFa and ibuprofen. We speculate that the early pulmonary hypertension seen after TNFa may be mediated

T

6 (cells x 103/mm3)

TABLE 3

3

i

-2

lffiUJ -1

Pw!

0

PERIPHERAL BLOOD GRANULOCYTES (n i

,

1

2

TIME (in hours)

Fig. 5. Circulating white blood cell count after TNFa. There was no attenuation of the characteristic leukopenia or neutropenia afforded by pretreatment with ibuprofen (open circles) (n = 11) when compared with the TNFa control group (closed circles) (n = 11).

11)*

Time after TNFa Infusion Group TNFa Ibuprofen

Baseline

30 Min

60 Min

120 Min

180 Min

240 Min

5.5 ± 1.1 4.9 + 0.7

0.8 ± 0.2 0.6 + 0.2

1.0 ± 0.4 1.0 ± 0.3

1.7 ± 0.4 1.3 ± 0.3

3.1 ± 0.7 1.7 ± 0.5

3.8 ± 0.7 3.0 ± 0.8

* Neutrophils x 10'/mm'. Values are mean ± SEM.

t p < 0.05 lime

point compared with group baseline.

637

CYCLOQXlGENASE INHIBITION IN TNFa WNG INJURY

0.05 TNF r--- p S

0 1 2 TIME (in hours)

p

3

4

< 0.05 BOTH GROUPS

...-VS BASEllNE---. 1250 1000

6-keto

750

PGF,. lpg/mil

500

250

o p

i

i

-2

-1

< 0.05 BETWEEN GROlJ>S

i

0 1 2 TIME (in hours)

3

4

by 1XA2 ; however, the later pulmonary In sheep, lung lymph flow is a useful hypertension is likely due to another combined index of vascular permeabilimediator not measured in this study. ty and surface area. In the control group, Ibuprofen has effects on TNFa-induced 3- to 4-fold increases in QLwere typical pulmonary hypertension similar to those and are similar to those seen after low seen when cyclooxygenase inhibitors are dose LPS in other studies (4, 5, 33, 34, used in experimental endotoxemia (5). 42, 43). Ibuprofen had no significant efMany of the detrimental effects of 1XA2 fect on peak levels of lung lymph flow; may be balanced by prostacyclin, a po- however, it delayed increases in QL. We tent vasodilator and platelet antiaggre- believe that the early reductions in QL gant. In this study, the PGI 2 metabolite were due to reductions in the initial pulwe measured peaked later and was more monary hypertension. Equivalent insustained in the duration than 1XA2 af- creases in lymph flow seen in both groups ter TNFa administration. Our initial con- between 1 and 4 h after TNFa suggest cerns that the late pulmonary hyperten- no effect of cyclooxygenase inhibition on sion seen after TNFa might actually be the vascular permeability surface area increased by inhibition of PGI 2 produc- product. We believe it is most likely that tion were not realized. Although we were noncyclooxygenase products are responunable to directly evaluate the degree of sible for the late increases in lung lymph pulmonary edema formation as a factor flow. in the pulmonary hypertension, recent Altered gas exchange occurred rapidstudies indicate that perivascular edema ly (within 30 min) after beginning the is minimal or modest after the doses of TNFa infusion. Coincident with inTNFa used in these studies (41). creases in AaPo2 were a rise in RL and

a decline in Cdyn. The changes in lung mechanics and gas exchange correspond temporally to increases in QL and increases in PGE 2 , 1XA2 , and 6-ketoPGF la levels. It has been hypothesized that 1XA2 and PGI 2 may be responsible for gas exchange abnormalities by altering ventilation-perfusion matching (44). TxA2 is a potent vasoconstrictor and bronchoconstrictor and the effects of PGI2 include increased oxygenconsumption and disrupted hypoxic vasoconstriction (44, 45). Our data suggest that the roleof prostacyclinin late-phase increases in AaPo2 is minimal since ibuprofen did not completely prevent elevations in prostacyclin but did prevent elevation in AaPo2 • Previous studies in endotoxemic sheep have demonstrated effects of cyclooxygenase inhibitors on RLand AaP02 similar to those we observed (33, 34). Unlike prior studies of endotoxemia, ibuprofen delayed but did not ultimately prevent decreases in Cdyn. It is not clear whether reductions in Cdyn occurring after TNFa infusion are the result of a different mechanism than those seen after endotoxin or if the fall in Cdyn is related to lessthan complete suppression of PGI2 generation. Cdyn reflects not only changes in static lung compliance but also changes in airway resistgnce-« predominately small airways resistance. Wedo not believe,however,that changes in airway resistance alone account for the decline in Cdyn since the ibuprofentreated group had complete inhibition of RLincreases, but had decreases in Cdyn comparable to the untreated group. If bronchoconstriction were the predominate pathophysiologic mechanism causing reductions in Cdyn, increases in RL and FRC might be expected, but they werenot seen. If airway constriction was responsible for reductions in compliance, changes in Cdyn would be expected to correlate temporarily with increases in Rt. In contradistinction, the reductions in Cdyn seen in our study werelong-lived, whereas the increases in Rr, wereof short duration. Numerous studies in rats, dogs, and sheep have shown that cyclooxygenase inhibitors improve physiologic parameters and survival after endotoxin (28-34) or TNFa challenge (12, 13); however, the role of individual cyclooxygenase products in specific physiologic alterations is less clear. It is clear, however, that not all of the physiologicchanges of endotoxemia seen in vivo are due to cyclooxygenase products (33, 34). Similarly, it is

638

also apparent that not all of the effects of TNFa are mediated by cyclooxygenase products. For example, in vivo, cyclooxygenase inhibitors have consistently failed to prevent the leukopenia and neutropenia seen after administration of endotoxin or TNFa, and, in vitro, the cytolytic effect of TNFa on tumor cells or fibroblasts is not prevented by treatment with cyclooxygenase inhibitors (12, 13, 46). Recent evidence suggests that high doses of ibuprofen can attenuate the neutrophilic alveolitis induced by endotoxin (47) or combinations of IL-l and TNFa (48). Although we were unable to histologically examine the effect of ibuprofen on leukocyte accumulation in the lung or lung water because of the paired design of this study, our data suggest that at least the occurrence of leukopenia and granulocytopenia seen after TNFa are not dependent upon the cyclooxygenase products wemeasured. Our data further suggest that the mechanisms accounting for leukopenia and neutropenia differ from those altering Ppa, AaPo2 , and RL, since changes in circulating leukocyte counts were independent of these other physiologic alterations. Although increases in the chemoattractant and platelet aggregant 1XA2 were prevented by ibuprofen, we failed to show an impact on TNFa-induced neutropenia. This suggests fo us that some other chemoattractant, possibly activated complement, LTB4 , IL-8, or platelet-activating factor, is responsible for the neutropenia seen after TNFa. These results are similar to previous studies in which cyclooxygenase inhibitors, even when given over a wide dose range, have failed to prevent neutropenia after endotoxin (28, 31, 33) or TNFa (49). Recent reports that granulocyte depletion attenuates but does not prevent the lung injury of endotoxin or TNFa also supports the role of nonneutrophil-dependent mechanisms in TNFa-induced lung injury (35, 50, 51). TNFa also stimulates cells to produce and release interleukins and interferons. The secondary release of these eytokines may further amplify the inflammation caused by TNFa through both cyclooxygenase dependent and independent mechanisms such as protease, leukotriene, platelet-activating factor, or oxygen radical release. Ibuprofen is not known to be a protease inhibitor, oxygen radical scavenger, or leukotriene or platelet-activating factor antagonist in the doses employed in this study. Generation of these other toxic mediators and synergy between active compounds may

WHEELER, HARDIE, AND BERNARD

in part explain why cyclooxygenase inhibitors may not entirely prevent the injury of TNFa (52). In addition to the generation of secondary cytokines and lipid mediators, there is now evidence that TNFa stimulates the generation of oxidants such as hydroxyl radical (53). Previous work in the sheep endotoxemia model has suggested that oxidants play a role in endotoxin-induced lung injury, and antioxidants can decrease physiologic derangements and prostaglandin generation (54); however, any link between TNFa-induced generation of oxidants and endotoxemia remains speculative. In summary, we have demonstrated that infusion of sublethal doses of TNFa into awake sheep causes abrupt lung mechanics, hemodynamic, and gas exchange abnormalities that may be substantially modified by the use of the cyclooxygenase inhibitor ibuprofen. In this model, initial severe pulmonary hypertension, leukopenia, and hypoxemia are accompanied by profound decreases in dynamic compliance and increases in airway resistance. Changes in Ppa, RL, and AaPo2 wereprevented with ibuprofen. We conclude that many of the acute physiologic changes of the TNFa are caused by the prostaglandins measured in this study. Interestingly, despite reduced formation of 1XA2 , PGE 2 , and to a lesser extent prostacyclin, several of the physiologic changes induced by TNFa were not prevented. We speculate that leukopenia, increasesin capillary permeability (as reflected in lymph protein clearance), and changes in lung compliance are not predominately mediated by the prostaglandin products we measured. These results are similar in many ways to those seen with previously published studies of meclofenamate-treated endotoxemic sheep. Acknowledgment This work would not have been possible without the guidance and assistance of Drs. James Sheller and James Snapper. The writers are also pleased to acknowledge the assistance of Frank Bostic, Nancy Wickersham, and Gayle King. References 1. Brigham KL, Meyrick B. Endotoxin and lung injury. Am Rev Respir Dis 1986; 133:913-27. 2. Beutler B, Cerami A. Cachectin (tumor necrosis factor): a macrophage hormone governing cellular metabolism and inflammatory response. Endoer Rev 1988; 9:57-66. 3. Wheeler AP, Jesmok G, Brigham KL. Tumor necrosis factor's effects on lung mechanics, gas exchange and airway reactivity in sheep. J Appl Phys-

iol 1990; 68:2542-9. 4. Wheeler AP, Hardie WD, Bernard G. Studies of an antiendotoxin antibody in preventing the physiologic changes of endotoxemia in awake sheep. Am Rev Respir Dis 1990; 142:775-81. 5. Ogletree M, Begley C, King G, Brigham KL. Influence of steroidal and nonsteroidal antiinflammatory agents on accumulation of arachidonic acid metabolites in plasma and lung lymph after endotoxemia in awake sheep: measurements of prostacyclin and thromboxane metabolites and 12 HETE. Am Rev Respir Dis 1986; 133:55-61. 6. Beutler B, Krochin N, Milsark I, et al. Control of cachectin (tumor necrosis factor) synthesis: mechanisms of endotoxin resistance. Science 1986; 232:977-80. 7. Michie HR, Manogue KR. Detection of circulating tumor necrosis factor after endotoxin administration. N Engl J Med 1988; 318:1481-6. 8. Johnson J, Meyrick B, Jesmok G, Brigham KL. Human recombinant tumor necrosis factor-a mimics endotoxemia in awake sheep. J Appl Physiol 1989; 66:1448-54. 9. TraceyKJ, Fong Y,Hesse DG, et al. Anti-TNFa monoclonal antibodies prevent septic shock during lethal bacteremia. Nature 1987; 330:662. 10. Beutler B, Milsark IW, Cerami A. Passive immunization against cachectin/tumor necrosis factor protects mice from lethal effects of endotoxin. Science 1985; 229:869-71. 11. Fong Y, Tracey KJ, Moldawer LL, et at. Antibodies to cachectin/tumor necrosis factor reduce interleukin 113 and interleukin 6 appearance during lethal bacteremia. J Exp Med 1989;170:1627-33. 12. Evans DA, Jacobs DO, Revhaug A, Wilmore DW.The effects of tumor necrosis factor and their selective inhibition by ibuprofen. Ann Surg 1989; 209:312-21. 13. Kettlehut IC, Fiers W, Goldberg AL. The toxic effects of tumor necrosis factor in vivo and their prevention by cyclooxygenase inhibitors. Proc Nat! Acad Sci USA 1987; 84:4273-7. 14. Kunkel SL, Spengler M, May MA, Spengler R, Larrick J, Remick D. Prostaglandin E. regulates macrophage-derived tumor necrosis factor gene expression. J Bioi Chern 1988; 263:5380-4. 15. Brigham, KL, Serafin W, zadoff A, Blair I, Meyrick B, Oates JA. Prostaglandin E. attenuation of sheep lung responses to endotoxin. J Appl Physiol 1988; 64:2568-74. 16. Monick M, Glazier J, Hunninghake GW. Human alveolar macrophages suppress interleukin-I (IL-I) activity via the secretion of prostaglandin E a- Am Rev Respir Dis 1987; 135:72-7. 17. Ham EA, Soderman DD, Zanetti ME, Dougherty HW, McCauley E, Kuehl FA. Inhibition by prostaglandins of leukotriene B. release from activated neutrophils. Proc Nat! Acad Sci USA 1983; 80:4349-53. 18. Spinas GA, Bloescb D, Keller U, Zimmerli W, Cammisuli S. Pretreatment with ibuprofen augments circulating tumor necrosis factor-a, interleukin-6, and elastase during acute endotoxemia. J Infect Dis 1991; 163:89-95. 19. Leeper-Woodford SK, Carey D, Fisher BJ, Sugerman HJ, Fowler AA. Tumor necrosis factor activity of pulmonary alveolar macrophages is inhibited by ibuprofen (abstract). Am Rev Respir Dis 1991; 143:393. 20. Dervin G, Calvin JE. Role of prostaglandin E, in reducing pulmonary vascular resistance in an experimental model ofacute lung injury. Crit Care Med 1990; 18:1129-33. 21. Sibbald WJ, Campbell D, Raper RR, Rutledge FS, Cheung H. The effects of prostaglandin E, on lung injury complicating hyperdynamic sepsis in sheep. Am Rev Respir Dis 1989; 139:674-81.

639

CYCLOOXYGENASE INHIBITION IN TNFa WNG INJURY

22. Cox JW, Andreadis NA, Bone RC, et al. Pulmonary extraction and pharmacokinetics of prostaglandin E, during continuous intravenous infusion in patients with adult respiratory distress syndrome. Am Rev Respir Dis 1988; 137:5-12. 23. Melot C, Lejeune P, Leeman M, Moraine J, Naeije R. Prostaglandin E, in the adult respiratory distress syndrome. Am Rev Respir Dis 1989; 139:106-10. 24. Bone RC, Siotman 0, Maunder R, et al. Randomized double-blind multicenter study of prostaglandin E, in patients with the adult respiratory distress syndrome. Chest 1989; 96:114-9. 25. Russell JA, Ronco JJ, Dodek PM. Physiologic effects and side effects of prostaglandin E, in the adult respiratory distress syndrome. Chest 1990; 97:684-92. 26. Jenkins JK, Carey PD, Byrne K, Sugerman HJ, Fowler AA. Sepsis-induced lung injury and the effects of ibuprofen pretreatment. Am Rev Respir Dis 1991; 143:155-61. 27. Reines HD, Haluska PV, Cook JA, et al. Plasma thromboxane concentrations are raised in patients dying with septic shock. Lancet 1982; 1:174. 28. Jacobs ER, Soulsby ME, Bone RC, et al. Ibuprofen in canine endotoxic shock. J Clin Invest 1982; 70:536-41. 29. Jacobs ER, Bone RC, Balk RA, et al. Ibuprofen improves survival in sheep endotoxin shock. J Crit Care 1986; 3:142-7. 30. Almquist PM, Kuezig M, Schwartz SI. neatment of canine endotoxin shock with ibuprofen, a cyclooxygenase inhibitor. Circ Shock 1984; 13:227-32. 31. Balk RA, Tryka AF, Bone RC, et al. The effect of ibuprofen on endotoxin-induced injury in sheep. J Crit Care 1986; 1:230-5. 32. Staub NR, Bland R, Brigham KL, Demling R, Erdmann J, WolvertonW.Preparation ofchronic lung lymph fistulas in sheep. J Surg Res 1975; 19:315-20. 33. Begley C, Ogletree M, Meyrick B, Brigham K. Modification of pulmonary responses to endotoxemia in awake sheep by steroidal and nonsteroidal anti-inflammatory agents. Am Rev Respir

Dis 1984; 130:1140-4. 34. Snapper J, Hutchison A, Ogletree M, Brigham K. Effects of cyclooxygenase inhibitors on the alterations in lung mechanics caused by endotoxemia in the unanesthetized sheep. J Clin Invest 1983; 72:63-76. 35. Hinson J, Hutchison A, Ogletree M, Brigham K, Snapper J. Effect of granulocyte depletion on altered lung mechanics after endotoxemia in sheep. J Appl Physiol 1983; 55:92-9. 36. Von Neergaard K, Wirz K. Die messung der Stromungswiderstande in den Atemwegen des Menshen, insbesmdere ber Asthma und Emphysem. Z Klin Med 1927; 105:52-2. 37. DuBois AB, Botelho SY,Bedell ON, Marshall R, Comroe JH. A rapid plethysmographic method measuring thoracic gas volume: a comparison with nitrogen washout method for measuring functional residual capacity in normal subjects. J Clin Invest 1956; 35:322-6. 38. Pradelles P, Grassi J, Maclouf J. Enzyme immunoassays of eicosanoids using acetylcholinesterase as label: an alternative to radioimmunoassay. Anal Chern 1985; 57:1170-3. 39. Westcott JY, Chang S, Balazy M, et al. Analysisof 6-keto PGF,., 5-HETE, and LTC. in ratlung: comparison of GC/MS, RIA, and EIA. Prostaglandins 1986; 32:857-73. 40. Rodbard E, Bridson W, Rayford P. Rapid calculation of radioimmunoassay results. J Lab Clin Med 1969; 74:770-81. 41. Johnson J, Brigham KL, Jesmok G, Meyrick B. Morphologic changes in lungs of anesthetized sheep after intravenous infusion of recombinant . tumor necrosis factor-a. Am Rev Respir Dis 1991; 144:179-86. 42. Henry C, Ogletree M, Brigham K, Hammond J. Attenuation of the pulmonary vascular response to endotoxin by a thromboxane synthesis inhibitor (UK38485) in unanesthetized sheep. Circ Res 1991; 50:77-81. 43. Winn R, Harlan J, Nadir B, Harker J, Hildebrandt J. Thromboxane A z mediates lung vasoconstriction but not permeability after endotoxin. J Clin Invest 1983; 72:911-8.

44. Carmona RH, Tsao TC, 'Irunkey DD. The role ofprostacyclinand thromboxanein sepsisand septic shock. Arch Surg 1984; 119:189-92. 45. Haluska PV, Dollery CT, MacDermot J. Thromboxane and prostacyclin in disease: a review. Q J Med 1983; 52:461-70. 46. Elias JA. Thmor necrosis factor interacts with interleukin-I and interferons to inhibit fibroblast proliferation via fibroblast prostaglandin-dependent and independent mechanisms. Am Rev Respir Dis 1988; 138:652-8. 47. Rinaldo JE, Pennock B. Effects of ibuprofen on endotoxin-induced alveolitis: biphasic dose response and dissociation between inflammation and hypoxemia. Am J Med Sci 1986; 291:29-38. 48. Jenkins JK, Carey PD, Byrne K, Sugerman HJ, Fowler AA. Sepsis-induced lung injury and the effects of ibuprofen pretreatment. Am Rev Respir Dis 1991; 143:155-61. 49. Ulich TR, del Castillo J, Ni R, Bikhazi N, Calvin L. Mechanisms of tumor necrosis factor alphainduced lymphopenia, neutropenia, and biphasic neutrophilia. J Leukocyte BioI 1989; 45:155-67. 50. Heflin AC, Brigham KL. Prevention by granulocyte depletion of increased vascular permeability of sheep following endotoxemia. J Clin Invest 1981; 68:1253-60. 51. Stephens KE, Ishizaka A, Wu Z, Larrick JW, Raffin T. Granulocyte depletion prevents tumor necrosis factor-mediated acute lung injury in guinea pigs. Am Rev Respir Dis 1988; 138:1300-7. 52. Okusawa S, Gelfand JA, Ikejima T, Connolly RJ, Dinarello CA. Interleukin 1 induces a shocklike state in rabbits. J Clin Invest 1988;81:1162-72. 53. Yamauchi N, Watanabe N, Kuriyama H, et al. Suppressive effects of intracellular glutathione on hydroxyl radical production induced by tumor necrosis factor. Int J Cancer 1990; 46:884-8. 54. Bernard GR, Lucht WD, Niedermeyer ME, et al. Effect of N-acetylcysteine on the pulmonary response to endotoxin in the awake sheep and upon in vitrogranulocyte function. J Clin Invest 1984; 73:1772-84.