INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

One of the first observations in patients with ARDS was an increase in the minimum surface tension of surfactant recovered from lung specimens at autopsy (1). The best characterized function of surfactant is the ability to reduce surface tension at the alveolar air-liquid interface, stabilizing alveoli and terminal airways at low lung volumes. Surfactant is a complex mixture of phospholipids and proteins, and the importance of the surfactant-associated proteins (SP-A, SP-B, SP-C, and SP-D) extends beyond their contributions to surfactant function (4).

Surfactant protein A (SP-A) is the most abundant surfactant protein and is important in the formation of tubular myelin and antibacterial host defense (5). SP-A concentrations in bronchoalveolar lavage fluid (BAL) are abnormally low in pneumonia and chronic pulmonary diseases, and low SP-A concentrations predict survival in patients with idiopathic pulmonary fibrosis (8, 9). Surfactant protein-B (SP-B) enhances spreading and stabilizes surfactant phospholipids at the air-liquid interface (10). Hereditary SP-B deficiency causes acute respiratory failure and death in newborns and mice (11). Pulmonary SP-B content is reduced by 50% in heterozygous SP-B deficient mice, and is associated with increased sensitivity to hyperoxia (12). Surfactant protein-D (SP-D) is a member of the C-type lectin superfamily, along with mannose-binding protein (MBP) and SP-A. SP-D is secreted by Type II cells and nonciliated bronchiolar epithelium, but unlike SP-A and SP-B, SP-D does not localize to lamellar bodies (13). Like SP-A, SP-D may have a role in pulmonary host defense (6). SP-D is detectable in the serum of patients with idiopathic pulmonary fibrosis (IPF), pulmonary alveolar proteinosis (PAP), and interstitial pneumonia with collagen disease, and the serum SP-D concentrations reflect disease activity (14, 15). SP-D regulates surfactant phospholipids in vivo, as surfactant phospholipid pool size is markedly increased in SP-D deficient mice (16).

Abnormalities of surfactant proteins and phospholipids have been found in the BAL of patients with ARDS. Gregory and coworkers (17) reported that the concentrations of SP-A and SP-B were reduced in the BAL samples of patients with ARDS and patients who were at risk for ARDS after trauma (17). Pison and coworkers (18) found that the SP-A concentrations were low in BAL fluid from patients with ARDS following trauma. In patients with less severe lung injury, the BAL SP-A concentration increased during recovery, whereas it remained low in patients with more severe lung injury. Günther and coworkers (19) found that SP-A concentrations were low, whereas SP-B concentrations were normal in patients with ARDS. Patients with bacterial pneumonia had similar abnormalities, whereas the BAL SP-A and SP-B concentrations were normal in patients with cardiogenic pulmonary edema. The serum concentrations of SP-A and SP-B increase in patients with ARDS, and were proposed as serum markers of injury to the epithelial and endothelial barriers in the lungs (20, 21). SP-D has not been studied in ARDS.

The major goal of this study was to investigate the changes in surfactant protein concentrations that occur before and after the onset of ARDS in order to determine whether SP-A, SP-B, and SP-D change in a coordinated manner. A second goal was to determine whether changes in surfactant proteins in BAL and/or serum would predict the development of ARDS in patients at risk, or the severity of lung injury or clinical outcome in patients after the onset of ARDS. We measured surfactant protein concentrations in BAL and serum of patients at risk for ARDS and then serially throughout the course of the disease, and compared the results with clinical, physiologic, and outcome variables.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

The study population consisted of patients at Harborview Medical Center between January 1994 and April 1997 who either were at-risk for ARDS after sepsis or trauma, or who had clinically defined ARDS. Informed consent was obtained from each patient or the legal next of kin. The study was approved by the University of Washington Human Subjects Review Committee.

Patients at-risk for ARDS. All intubated patients in the intensive care units at Harborview Medical Center (Seattle, WA) were screened prospectively using predetermined risk criteria for trauma or sepsis (22). For trauma, risk criteria included either (1) two or more of the following signs of major trauma: multiple fractures, unstable pelvic fracture, pulmonary contusion, massive transfusion (> 15 units in 24 h) or (2) a single trauma risk and Injury Severity Score (ISS) > 20 (23). For sepsis, patients met the risk criteria in Category A and Category B. A. Two or more of the following: (1) positive blood culture or definite or suspected source of infection; (2) WBC > 14,000/mm³ or < 4,000/mm³; (3) T > 39° C or < 36° C. B. Two or more of the following signs of a severe systemic response: (1) systemic vascular resistance < 800 dyne cm/m²; (2) unexplained hypotension (systolic BP < 90 mm Hg for > 1 h); (3) metabolic acidosis with anion gap > 20 mEq/L; (4) vasopressor use to maintain systolic blood pressure > 90 mm Hg; (5) platelet count  80,000/mm³. Patients at risk for ARDS did not meet either radiographic or oxygenation criteria for ARDS.

Patients with documented ARDS. All intubated patients in the intensive care unit who were between 18 and 72 yr of age were screened prospectively for the onset of ARDS using the following criteria: (1) critical hypoxemia, with Pa_O2/FI_O2< 150 mm Hg, or < 200 mm Hg on  5 cm H₂O positive end-expiratory pressure (PEEP); (2) diffuse parenchymal infiltrates involving at least 50% of three quadrants on the chest radiograph; (3) pulmonary artery wedge pressure (when available) < 18 mm Hg or no clinical evidence of congestive heart failure; (4) no other obvious explanation for these findings. All of the patients with ARDS also met the American European Consensus Conference definition of ARDS (24). Patients were followed until death or hospital discharge. Survival was defined as discharge from the hospital.

Bronchoalveolar lavage procedure. Patients underwent BAL within 24 h of being identified as at-risk for ARDS. In patients with ARDS, BAL was performed within 24 h of the onset of ARDS, and then again on Days 3, 7, and 14 after the onset of ARDS unless the patient was extubated, died, or was too unstable to tolerate the BAL procedure.

All patients were intubated and ventilated at the time of BAL. They were preoxygenated with FI_O2 = 1.0 for 15 min, and continued receiving FI_O2 = 1.0 during the procedure. BAL was performed using five separate 30-ml aliquots of 0.89% sterile NaCl instilled into the right middle lobe or lingula. The BAL recovery averaged 75 ml (50% return), with a range of 23 to 110 ml. Patients were excluded from the BAL study for safety concerns if they met any of the following criteria (25): (1) Pa_O2 < 80 mm Hg with FI_O2 = 1.0; (2) evidence of acute myocardial ischemia; (3) systolic blood pressure < 90 mm Hg; (4) cardiac dysrhythmias (supraventricular tachycardia > 140 beats/min or complex ventricular ectopy); (5) uncontrolled intracranial hypertension (intracranial pressure > 20 mm Hg); (6) endotracheal tube internal diameter < 7.0 mm; (7) burns or inhalation injury. Clinical exclusion criteria included HIV infection, asthma or COPD requiring daily medication, sarcoidosis, or interstitial lung disease. Age was not an exclusion criterion.

Healthy volunteers. For comparison, we studied BAL fluids from 35 normal subjects who underwent peroral fiberoptic bronchoscopy and BAL using five separate 30-ml aliquots of 0.89% NaCl instilled into the right middle lobe or the lingula. These subjects had no respiratory symptoms and did not smoke cigarettes.

BAL Fluid Analysis

The BAL fluid was transported immediately to the laboratory for analysis. The fluid was poured through gauze moistened with 0.89% NaCl to remove mucus, and the total volume recovered was recorded. Total cell counts were performed in a hemacytometer using aliquots of the fresh BAL fluid. Differential cell counts were performed on cytospin preparations stained with Diff-Quik (American Scientific Products, McGaw Park, IL). The lavage fluid was spun at 200 × g for 30 min to pellet the cells, and the supernatant fluid was stored at 70° C. Total protein was measured on an aliquot of the supernatant using the bicinchoninic acid method (Pierce Chemical Co., Rockford, IL).

Measurement of SP-A

SP-A was measured using a sandwich enzyme-linked immunosorbant assay (ELISA) technique using two different monoclonal antibodies to human SP-A (designated 9.5 and 22.7). SP-A isolated from the lavage fluid of patients with alveolar proteinosis was used as the standard. To prepare the standard, the lavage fluid was spun at 100,0000 × g for 1 h, then the pellet was extracted with butanol. The butanol-insoluble proteins were resuspended in 30 mM N-octylglucoside, 150 mM NaCl, 5 mM TRIS (pH, 7.4). The SP-A, which remains insoluble, was then solubilized in 5 mM TRIS (pH, 7.40).

To measure SP-A, immunoassay plates (96-well EIA; Costar, Cambridge, MA) were coated overnight at 4° C with capture antibody (mAb 9.5) diluted in 0.1 M NaHCO₃ (pH, 8.3) to a final concentration of 5 µg/ml. The plates then were washed twice with phosphate-buffered saline (PBS)/0.2% bovine serum albumin (BSA)/0.05% Tween-20, and blocked using the same solution for 1 h at room temperature. The detergent Triton-X (1.0% final concentration) was added to each BAL or serum sample and standard, and the samples and standards were then sonicated for 5 s on ice. Preliminary studies showed that this method of sample preparation gave the maximal signal using BAL fluid. The addition of Triton-X to serum did not increase the serum signal. Samples and standards were diluted in PBS/0.2% BSA/ 0.05% Tween-20/1% Triton X and added to the wells. The plates were incubated at 37° C for 2 h, washed three times, and the biotinylated detecting antibody (mAb 22.7) (1:250 dilution) was added to each well. After a 2-h incubation at 37° C, the wells were washed three times, incubated with HRP/strepavidin (Zymed Laboratories, San Francisco, CA) for 30 min, and developed with TMB/H₂O₂ (Kirkegaard & Perry Laboratories, Gaithersburg, MD). The optical density in each well was read at 490 nm (OD₄₉₀) using a microtiter plate spectrophotometer (Dynatech Laboratories, Chantilly, VA). The standard curve is linear from 1 to 100 ng/ml SP-A. All assays were performed in triplicate dilutions.

The integrity of SP-A in BAL and serum was analyzed by Western blotting. Proteins in cell-free BAL from one healthy volunteer and one patient with ARDS were separated by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose. SP-A was detected with a murine monoclonal antibody to human SP-A (9.5). Mannose binding protein was detected with polyclonal rabbit serum IgG raised against purified rat mannose binding protein. Purified rat mannose binding protein was used as a control. The results confirmed that the SP-A in BAL fluid was intact and ran with the expected molecular weight (30 to 32 kD), and that there was no cross-reactivity with mannose binding protein or other proteins in BAL. Immunoprecipitation of serum samples from one patient with ARDS using a rabbit polyclonal IgG raised against human SP-A, followed by Western blotting showed a 32-kD band consistent with SP-A. Similar treatment of normal serum did not show this band. Immunoprecipitation and Western blotting of normal serum "spiked" with recombinant human SP-A showed only the SP-A band.

Measurement of SP-B

SP-B was measured by ELISA. The SP-B standard was purified from bovine lung lavage by the method of Kogishi and coworkers (26) and characterized by SDS-PAGE, silver staining, and Western blotting. The standard was quantitated by amino acid composition of a measured volume of undiluted standard. Microtiter wells were conditioned overnight at 4° C with 0.1 M NaHCO₃ (pH, 8.3), then washed twice with wash buffer (0.01 M TRIS, pH = 7.4; 0.05% Tween 20) and conditioned briefly with a buffer containing 0.15 M NaCl; 0.01 M TRIS (pH = 7.4); 5 mg/ml BSA (assay buffer). The BAL samples and SP-B standards were diluted in PBS containing 0.5% Nonidet P-40, added to the wells, and incubated for 1 to 2 h at 37° C. The wells were washed three times and then incubated with polyclonal rabbit anti-SP-B IgG in assay buffer for 1 to 2 h. After washing three times the wells were incubated with peroxidase-conjugated goat antirabbit IgG (Calbiochem, La Jolla, CA) in buffer containing 0.05 M NaPO₄ (pH, 6.3); 0.15 M NaCl; 0.05% Tween 20 for 1 h. Color was developed using orthophenylene diamine substrate system. All samples were measured in quadruplicate. To test the effect of human lavage fluid on this assay, standard curves were generated using the SP-B standard diluted in assay buffer only, or in human lavage fluid. The two standard curves were indistinguishable. The SP-B ELISA was linear over a range of 5 to 100 ng/ml.

Measurement of SP-D

Human SP-D was measured by ELISA as previously described using recombinant human SP-D (rhSP-D) as the standard (27). The recombinant human SP-D was produced using a baculovirus expression system as previously reported (14). The ELISA used two monoclonal antibodies against human SP-D, designated 6B2 and 7C6, which were raised against human SP-D purified from BAL fluids of patients with alveolar proteinosis (28). The rhSP-D standard or samples were diluted 1:10 to 1:200, added to wells coated with antibody 7C6, and incubated overnight at 4° C. After washing the HRP-conjugated 6B2 antibody was added and the plates were incubated at room temperature for 2 h. The plates were developed with TMB/H₂O₂ and the O.D.₄₅₀ was measured in each well. All assays were performed in duplicate.

The monoclonal antibodies used in this assay detect a single band of 43 kD by Western analysis of BAL fluid from a patient with pulmonary alveolar proteinosis (28). When serum from a patient with alveolar proteinosis was passed over a mannose-Sepharose column and the adherent proteins were analyzed by Western blotting, the assay antibodies detected a single band of 43 kD, consistent with SP-D.

Statistical Analysis

The data are reported as medians and ranges. Wilcoxon's Matched Pairs Sign Rank Test was used for all comparisons because the data were not normally distributed. Statistical significance was defined as p  0.05. Analysis of receiver operating characteristic (ROC) curves was used to determine the optimal sensitivity and specificity values for individual surfactant proteins in predicting either the onset of ARDS in patients at risk or survival after the onset of ARDS (29). Discriminant analysis was used to test the association of groups of surfactant proteins in BAL and serum with the onset of ARDS and survival (30).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Population

All patients with trauma or sepsis risk factors for ARDS and all patients with ARDS were identified prospectively between January 1994 and March 1997. During this period, 22 patients at-risk for ARDS and 41 patients with ARDS were enrolled in the study. Of the patients at risk for ARDS, 10 had a sepsis risk and 12 had a trauma risk. In patients with ARDS, the primary risk factors were sepsis (n = 12), major trauma (n = 16), or "other" (n = 13), including gastric aspiration, drug overdose, or massive transfusions. The study population was 68% male, and the mean age was 42.8 yr (range, 18 to 75 yr). The initial Pa_O2/FI_O2 ratio was 217.6 ± 21.4 (mean ± SE) in the "at- risk" group and 152.6 ± 8.1 in the ARDS group (p < 0.01). The mean APACHE II scores at the time of ICU admission were similar in the "at-risk" and ARDS groups (21.2 ± 1.3 for "at-risk" versus 21.2 ± 1.1 for ARDS, mean ± SE). The mean ISS score for patients with ARDS after trauma was 24.3. The hospital mortality was 9.1% in patients at risk and 22.5% in patients with ARDS. The average recovery of BAL fluid was 49.3 ± 2.9% (mean ± SE) in "at-risk" patients and 51.9 ± 2.2% on Day 1 of ARDS. The BAL recovery did not vary significantly during the course of ARDS.

Surfactant Protein Analysis

BAL. In patients at risk for ARDS, the SP-A concentration in BAL was significantly lower than normal (median = 880 ng/ml versus 4,710 ng/ml, respectively, p < 0.02) (Figure 1A). In patients with ARDS, the BAL SP-A concentration was low at the onset of ARDS (Day 1 of ARDS, median = 993 ng/ml) and remained low for as long as 14 d in patients with persistent ARDS. The BAL SP-B concentration followed this same pattern (Figure 1B). As with SP-A, the SP-B concentration was low in patients at risk for ARDS and remained low throughout the course of ARDS. In contrast to SP-A and SP-B, the median BAL SP-D concentration did not differ from normal either in patients at risk for ARDS or in patients with established ARDS (Figure 1C).

View larger version (20K): [in this window] [in a new window]   Figure 1.   Surfactant protein concentrations in BAL from normal volunteers, patients at-risk for ARDS ("Risk") and patients with established ARDS studied at sequential times. The horizontal axis shows the patient group. Numbers in parentheses show the number of subjects. The box plots show the 10th, 25th, 75th, and 90th percentiles and the median (40). *p  0.0001 versus normal subjects. NL = normal subjects.

Serum. The median serum SP-A concentration was normal in patients at risk for ARDS, although the range was broader than in normal volunteers (Figure 2). On Day 1 of ARDS, the median serum SP-A concentration was significantly elevated (ARDS median = 86 ng/ml versus normal median = 32 ng/ml, p < 0.02). The serum SP-A concentration was maximal on Day 3 of ARDS, with a median value that was approximately 20 times higher than normal (634 ng/ml, p < 0.006), and remained significantly elevated at Day 7 (223 ng/ml) and Day 14 (351 ng/ml) in patients with persistent ARDS. The serum SP-D concentrations showed similar trends as for SP-A. The serum SP-D concentrations were normal in patients at risk for ARDS, were significantly elevated on Day 1 of ARDS and were highest on Days 3 and 7 of ARDS. Unlike the SP-A concentrations, which remained elevated in nearly all patients with sustained ARDS, the serum SP-D concentrations returned to the normal range by Day 14 in many of the patients. The serum concentrations of SP-A and SP-D were similar in patients with sepsis and trauma as primary risk factors for ARDS.

View larger version (24K): [in this window] [in a new window]   Figure 2.   Surfactant protein concentrations in serum of patients at-risk for ARDS, or with established ARDS studied at sequential times. The format is the same as for Figure 1. *p  0.0001 versus normal subjects at all times, except for SP-D on Day 1 of ARDS, where p = 0.0028. NL = normal subjects.

Surfactant proteins and onset of ARDS. The BAL surfactant protein concentrations in the at-risk patients who developed ARDS and those who did not are shown in Figure 3. ARDS developed in eight of 22 patients at risk for ARDS (36.4%). The BAL concentration of SP-A was significantly lower in patients who developed ARDS. No patient with a BAL SP-A above 1,200 ng/ml developed ARDS. This value for SP-A had a 100% sensitivity and a 57% specificity for predicting the onset of ARDS. The median BAL SP-B and SP-D concentrations also were lower in patients who later developed ARDS, but these trends did not reach statistical significance. The serum SP-A and SP-D concentrations in patients at risk did not identify the patients who later developed ARDS.

View larger version (13K): [in this window] [in a new window]   Figure 3.   Surfactant protein concentrations in BAL of patients at-risk for ARDS who developed ARDS (closed circles) or who did not develop ARDS (open circles). Each point represents data from a single patient. The bars are the median values. n = number of patients. *p = 0.0347 for ARDS versus no ARDS

Surfactant proteins and survival in ARDS. The BAL SP-A, SP-B and SP-D concentrations on Day 1 of ARDS by outcome are shown in Figure 4. The BAL SP-A and SP-B concentrations were low in all patients with ARDS, but the values did not differ in patients who lived or died. In contrast, the BAL SP-D concentration on Day 1 of ARDS was significantly lower in patients who died than in patients who lived (medians = 406 versus 940 ng/ml for died versus lived, p = 0.025). This was also true on Day 3 of ARDS (medians = 482 versus 995 ng/ml for died versus lived, p = 0.038). Consistent with this, the BAL SP-D was directly related to the Pa_O2/FI_O2 ratio on Days 1 and 3 of ARDS (Day 1: Spearman's r = 0.34, p = 0.05; Day 3: r = 0.51, p < 0.001).

View larger version (16K): [in this window] [in a new window]   Figure 4.   Surfactant proteins in BAL by survival in patients on Day 1 of ARDS. Shown are box plots, as in Figure 1. n = number of patients. *p = 0.025 for lived versus died.

Analysis of ROC curves showed that BAL SP-D was sensitive and specific for predicting death after the onset of ARDS, whereas the SP-A and the SP-B concentrations were not predictive (Figure 5). The optimal value for BAL SP-D was 440 ng/ml, which had a sensitivity of 85.7%, specificity of 74.1%, and a likelihood ratio of 3.31, i.e., patients who died were 3.31 times more likely to have a low BAL SP-D concentration than were patients who lived. Of 13 patients with BAL SP-D < 440 ng/ml, 6 died (46.1%); whereas of 21 patients with BAL SP-D > 440 ng/ml, only 1 died (4.8%) (p = 0.007). The same trends were also significant on Day 3 after the onset of ARDS.

View larger version (19K): [in this window] [in a new window]   Figure 5.   Receiver operating characteristic curves relating BAL SP-A, SP-B, and SP-D on Day 1 of ARDS and survival. The vertical axis is the number of true positives (sensitivity). The horizontal axis is the number of false positives (100-specificity). AUC = area under the curve, which represents the fraction of patients with an outcome (death) that would have a positive test (low surfactant protein in BAL); the maximum value = 1.0.

The serum concentrations of SP-A and SP-D tended to be higher on Day 1 of ARDS in patients who died, but the ranges were broad, and the differences did not reach statistical significance (Figure 6). On Day 3 of ARDS, the serum SP-D was significantly related to the lung injury score (Spearman's r = 0.34, p = 0.03), but neither serum SP-A nor SP-D were significantly related to the Pa_O2/FI_O2 ratio. Serum SP-A and SP-D were neither sensitive nor specific predictors of survival at any time after the onset of ARDS.

View larger version (16K): [in this window] [in a new window]   Figure 6.   Serum SP-A and SP-D concentrations on Day 1 of ARDS in patients who lived or died. Shown are box plots, as in Figure 1. n = number of patients.

Discriminant analysis was used to determine whether combinations of BAL and serum surfactant proteins were significantly related to outcome after the onset of ARDS. The BAL SP-D was the only surfactant protein measurement that was strongly related to outcome, and this was true on Days 1 and 3 of ARDS. The best statistical models using BAL SP-A and SP-B with or without serum SP-A and SP-D only weakened the significant univariate relationship between BAL SP-D and survival.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The major goal of this study was to investigate the changes in surfactant proteins that occur in BAL and serum of patients before and after the onset of clinically defined ARDS. We found that SP-A and SP-B concentrations are reduced in the lungs of patients at risk for ARDS, even before the onset of clinically defined lung injury. Once ARDS was clinically apparent, SP-A and SP-B concentrations were low in BAL, and they remained low for as long as 14 d in patients with sustained ARDS. In contrast, the median concentration of SP-D in BAL was normal in most patients at risk, and normal in most patients with established ARDS. These findings confirm that major surfactant protein abnormalities occur in the lungs before and after the onset of ARDS, and show for the first time that the concentrations of SP-A, SP-B, and SP-D change differently in patients with acute lung injury. The low concentrations of SP-A and SP-B in BAL fluid do not result simply from dilution of alveolar fluids by plasma entering the alveolar spaces or from direct toxicity to alveolar Type II pneumocytes because the SP-D concentrations did not fall in parallel with SP-A and SP-B. Furthermore, the concentrations of immunoreactive SP-A and SP-D increased in serum after the SP-A concentration in BAL had declined and while the BAL SP-D concentration was in the normal range. Therefore, leakage of surfactant proteins into the bloodstream also is not a sufficient explanation for the decreased concentrations of surfactant proteins in BAL.

The surfactant-associated proteins are produced by Type II pneumocytes and are regulated differently from the surfactant phospholipids. SP-A is the most abundant surfactant protein and has a number of important roles, including enhancing surface-active properties of surfactant phospholipids, promoting the formation of tubular myelin, and facilitating recognition of bacteria by alveolar macrophages (4). SP-A is a member of the collectin family and is structurally homologous to C1q. Like C1q, SP-A enhances phagocytosis of bacteria by macrophages, and SP-A-deficient mice have enhanced susceptibility to bacterial infections (6, 7). Thus, the relative deficiency of SP-A that occurs before and after the onset of ARDS could disrupt surfactant function and impair antibacterial host defenses in the lung parenchyma.

SP-B is a smaller and extremely hydrophobic protein whose main function is to enhance the stability of surfactant phospholipids at the air-liquid interface (11). SP-B deficiency in mice and humans results in severe respiratory distress and death shortly after birth. SP-B knockout mice have no recognizable tubular myelin and virtually no surfactant activity in the lungs (31). A phospholipid mixture containing SP-B-like peptides improved gas exchange in infants with neonatal respiratory distress (32). The low concentrations of SP-B that occur before and after the onset of ARDS could result in profound changes in alveolar stability, / matching, and gas exchange.

We found that SP-A and SP-B concentrations are low in the BAL of patients at risk for ARDS, even before the onset of clinically defined lung injury, in agreement with the findings of Gregory and coworkers (17). The patients at risk for ARDS had increases in BAL total protein and PMN, but not to the levels that were present in patients with established ARDS (22). Thus, the inflammatory response that occurs in the lungs of patients at risk for ARDS is associated with major abnormalities in SP-A and SP-B concentrations even before ARDS is clinically apparent. In contrast, the median concentration of SP-D in patients at risk for ARDS was similar to that of normal volunteers. Once ARDS was clinically apparent, the SP-A and SP-B concentrations remained low, whereas the median SP-D concentrations remained in the normal range throughout the course of ARDS. Because the SP-A, SP-B, and SP-D measurements were made on the same fluids, the results suggest that the decline in SP-A and SP-B concentrations reflects a regulatory abnormality in SP-A and SP-B metabolism, rather than a global disruption of Type II pneumocyte function.

This interpretation is consistent with experimental data showing that TNF- and bacterial products have differential effects on the regulation of SP-A, SP-B, and SP-D. TNF- reduces SP-A and SP-B expression in a human epithelial cell line and in murine lungs (33, 34). In human fetal lung tissue, TNF-, bacterial lipopolysaccharide (LPS), and protein kinase C activation all reduce SP-A expression, but these treatments have no effect on SP-D (35). TNF- and LPS have been found in lung fluids of patients with ARDS, although less is known about patients at risk (36).

SP-D is a collagenous glycoprotein with a symmetric cruciform structure similar to that of bovine conglutinin (4). The C terminal carbohydrate recognition domain (CRD) is highly conserved and is similar for SP-D, SP-A, C1q, and bovine conglutinin. The CRD unit is primarily responsible for the host defense properties of SP-D and SP-A. As compared with SP-A, SP-D binds phospholipids weakly, is soluble in alveolar fluid, and can be removed from surfactant by centrifugation. Previous evidence has suggested that SP-D measurements in serum or lavage fluid might be a useful marker of outcome, as serum SP-D concentrations are increased in patients with idiopathic pulmonary fibrosis (IPF) and correlate with survival (14). Although the median concentration of SP-D in BAL was normal throughout ARDS, on Day 1 the concentration of SP-D was very low in the subgroup of patients who later died, and the BAL SP-D was directly related to the PI_O2/FI_O2 ratio on Days 1 and 3 of ARDS. These findings suggest that SP-D falls when lung injury is severe, consistent with destruction of Type II pneumocytes in the lungs.

The assays that we used detected SP-A and SP-D immunoreactivity in serum of patients at risk for ARDS, and during the course of established ARDS. Although the serum proteins detected by these antibodies have not been identified rigorously, the SP-D antibody that we used recognizes a protein in the serum of patients with alveolar proteinosis that binds to a mannose column and has the appropriate molecular weight for SP-D (43 kD). We detected an appropriate-sized band in ARDS serum using a polyclonal anti-SP-A antibody (32 kD). Kuroki and coworkers (15) used an antihuman SP-A monoclonal antibody to detect a protein in alveolar proteinosis serum that binds to a mannose-sepharose column and has an apparent molecular weight of 30 to 32 kD. It has been shown experimentally that human SP-A can leak from the lungs into the systemic circulation in ventilated newborn rabbits, consistent with the concept that surfactant proteins leak from the air spaces into the vascular compartment in the setting of acute lung injury (39). The precise molecular forms of SP-A and SP-D immunoreactivity detected in serum by the ELISAs used are unknown at present and need to be identified. Likewise, the mechanisms by which SP-A and SP-D reactive peptides enter the plasma compartment need to be determined.

We found that the median serum concentrations of immunoreactive SP-A and SP-D, as determined by the ELISA assays used, were not increased in patients at risk for ARDS. When ARDS occurred, the serum concentrations of immunoreactive SP-A and SP-D increased, and were highest from Days 3 to 7 of ARDS. The serum SP-A concentrations remained high for 14 d, whereas the SP-D concentrations tended to decline by 14 d in most patients. Doyle and coworkers (20, 21) found that SP-A and SP-B were increased in the serum of patients with ARDS, and they proposed that these surfactant proteins reflected lung epithelial injury. The plasma SP-B and the SP-B/SP-A ratio were inversely related to blood oxygenation and static respiratory system compliance (21). It appears that a major epithelial injury is needed before surfactant protein concentrations increase substantially in plasma, as we found that the serum concentrations of SP-A and SP-D were actually highest on Days 3 and 7 after the onset of ARDS.

An important question is whether surfactant protein measurements predict either the onset or the outcome of ARDS. We found that BAL SP-A was significantly lower in patients at risk who later developed ARDS, and that SP-B and SP-D showed similar trends. Although there was overlap between those who did and those who did not progress to ARDS, no patient with a BAL SP-A concentration above 1.2 µg/ml developed ARDS. Thus, the BAL SP-A concentration may be useful in identifying patients who are at low risk for developing ARDS. In addition, we found that patients with ARDS who later died had the lowest concentrations of SP-D in BAL on Days 1 and 3 of ARDS. All of these patients also had low concentrations of SP-A and SP-B in BAL fluid. The BAL SP-D concentration had an 85.7% sensitivity and a 74% specificity in predicting death, and patients who died were 3.31 times more likely to have a low SP-D on Day 1 of ARDS than those who lived. These findings suggest that a low BAL SP-D concentration identifies patients with the most severe epithelial injury in the lungs, and those with the worst overall lung injury. Immunoreactive SP-A and SP-D in serum were neither sensitive nor specific in distinguishing patients who lived from those who died. These results need to be confirmed in a larger prospective series of patients studied before and after the onset of ARDS.

Our attempt to study serial changes in surfactant proteins during the course of ARDS has some limitations. The patient population with ARDS included patients who survived for as long as 14 d. However, patients who improved rapidly and were extubated, and those who died dropped out of the serial study, so that the later times do not reflect either the sickest patients or those with the mildest disease. Nevertheless, the population is representative of patients with ARDS who remain critically ill for at least 14 d. An important strength of the study is that each of the patients was studied at each time, so that the data represent true serial measurements in an affected population. In addition, each of the at-risk patients who developed ARDS is included in the ARDS population.

In summary, the data show that SP-A, SP-B and SP-D are regulated differently in the lungs of patients with ARDS. The SP-A and SP-B concentrations declined significantly in the lungs of patients at risk for ARDS, and remained low throughout the course of established ARDS. In contrast, the BAL SP-D concentration was normal in most patients at risk for ARDS, and in most patients with established ARDS. SP-A and SP-D immunoreactivity appeared in the serum after the onset of ARDS and remained detectable in most patients throughout the course of the disease. The BAL SP-A concentration was significantly lower in at-risk patients who develop ARDS, and the BAL SP-D concentration was significantly lower in the patients with ARDS who died. In contrast, the serum concentrations of immunoreactive SP-A and SP-D did not predict either onset of ARDS or survival in this series of patients. The different responses of SP-A, SP-B, and SP-D before and after onset of ARDS are consistent with the hypothesis that SP-A and SP-B decline as a consequence of the inflammatory response in the alveolar milieu, whereas SP-D declines when a major injury occurs to Type II pneumocytes in the alveolar epithelium. More information is needed about the factors that regulate surfactant proteins in the air spaces of patients with lung injury. Strategies to increase the concentrations of surfactant proteins in the lungs could be useful in reducing the onset or the outcome of ARDS in some patients.