Pulmonary Edema Fluid from Patients with Early Lung Injury Stimulates Fibroblast Proliferation through IL-1β-Induced IL-6 Expression ¹

Current estimates suggest that the annual incidence of acute lung injury is 20–75 per 100,000 persons (1). A significant fraction of patients with acute inflammatory lung injury develop a severe fibroproliferative process that is physiologically and clinically evident within 10–14 days after clinical presentation. This fibroproliferative process is characterized by the formation of alveolar granulation tissue composed of mesenchymal cellular proliferation and products, new blood vessels, and the deposition of a collagen (types I and III) and a fibrinous, fibronectin-rich matrix (2, 3, 4, 5, 6). The development of the fibroproliferative response to lung injury identifies patients who are more likely to die with acute lung injury. Histologically, fibrosis is detectable as early as 5 days after the onset of clinically evident injury (7, 8). However, type III pro-collagen peptide levels in alveolar edema fluid are elevated within hours after intubation and predict the development of a fibroproliferative response, prolonged mechanical ventilation, and poor survival (9, 10, 11). Taken together, these observations indicate that lung fibroblast activation begins early after acute lung injury is clinically manifest.

Prior studies have identified several cytokines and chemokines in pulmonary edema fluid (or bronchoalveolar lavage) from patients at risk for or with established acute lung injury, including TNF-α, IL-1, IL-8, growth-related oncogene-β (gro-β) ³/macrophage-inflammatory protein-2-α (MIP-2-α), epithelial neutrophil-activating protein-78, and platelet-derived growth factor (PDGF) (12, 13). Antigenic levels of these factors have a modest predictive value for the development of clinical acute lung injury and its outcome (1, 12, 14). Pulmonary edema fluid from patients with early acute lung injury demonstrates clear increases in both TNF-α and IL-1β Ag levels. However, ascribing functional roles for these cytokines based on nonfunctional antigenic data is problematic. For example, increases in soluble receptors and soluble receptor antagonists for TNF-α and IL-1β in acute lung injury can alter the availability of the cytokines to bind to their receptors and/or can block intracellular signaling, respectively (15, 16). Furthermore, animal models amply demonstrate the biological importance of posttranslation modifications and/or activation of existing latent factors, such as TGF-β (17, 18). Although mesenchymal cells from patients dying of acute lung injury manifest an enhanced proliferative phenotype (19), the bioactive factors present in the alveolar compartment that initiate the fibroproliferative response in early clinical acute lung injury remain largely unknown.

Although IL-1β is generally not considered a potent profibrotic factor in lung injury, transient overexpression of IL-1β using an adenoviral vector in rat lungs has recently been shown to induce tissue injury, subsequent local production of PDGF and TGF-β, and resultant pulmonary fibrosis (20). In human acute lung injury, PDGF-related molecules (B chain isoforms) have been identified as one group of mitogens for fibroblasts in the alveolar compartment (21). However, prior work was largely accomplished using samples (bronchoalveolar lavage) that are 100-fold diluted relative to alveolar lining fluid in vivo and that were collected several days after the onset of clinical acute lung injury. Therefore, this study was undertaken to identify the soluble, mitogenically active factors present in early acute lung injury pulmonary edema fluid. The pulmonary edema fluid samples used in this study were obtained within the first few hours of clinical acute lung injury and were compared with pulmonary edema fluid from a well-defined control patient population with hydrostatic pulmonary edema. The primary objectives were to test the hypothesis that IL-1β is a potent mitogen and a significant modulator of fibroblast phenotype in early human acute lung injury and to begin to define downstream mechanisms for its effect.

All patient protocols were approved by the Committee for Human Research of the University of California at San Francisco and by the Institutional Review Board at the University of Alabama. Patients were eligible for inclusion in this study if they had acute pulmonary edema from either acute lung injury or hydrostatic causes and required endotracheal intubation for positive pressure ventilation. Pulmonary edema fluid was obtained from patients (1987–1998) within 4 h of intubation through gentle luminal suction applied to a 14 French catheter passed into the distal airways, as described previously (22, 23). Plasma samples were drawn simultaneously. Samples were centrifuged (3000 × g, 10 min, 4°C) and the supernatants were stored at −80°C until they were analyzed. Patients were segregated into two groups based on the edema fluid:plasma total protein ratio, as determined by the Biuret method, and by clinical criteria (22, 23, 24). The simplified acute physiology score II (SAPS II) was calculated as described (25). The acute lung injury group was defined as having an edema fluid:plasma total protein ratio of >0.65 and by clinical criteria of bilateral infiltrates on chest radiograph, PaO₂/FIO₂ of <300, and a pulmonary capillary wedge pressure of <18 mm Hg, if measured, with clinical risk factor(s) for the development of clinical acute lung injury (22, 23). The hydrostatic edema patient group was defined as having an edema fluid:plasma total protein ratio of <0.65, with a wedge pressure of >18 mm Hg or a two-dimensional echocardiogram demonstrating a reduced left ventricular ejection fraction and a clinical history consistent with cardiac dysfunction (22, 23). Due to limited patient-derived material, equal aliquots of each patient’s sample within the same group were pooled.

Normal diploid adult human lung fibroblasts (CCL-210) were obtained from the American Type Culture Collection (Gaithersburg, MD) and propagated in Eagle’s MEM with l-glutamine and penicillin/streptomycin (Mediatech/Cellgro; Birmingham, AL) with 10% FBS (Sigma-Aldrich, St. Louis, MO). For proliferation measurements, low-density fibroblasts were made quiescent (0.4% serum-containing medium (SCM) for 48 h), followed by addition of either pulmonary edema fluid diluted in serum-free medium (SFM) or the indicated cytokines/Abs diluted in SFM (with 1% BSA and 1× insulin-transferrin-selenium) for the indicated times. Conditioned medium was generated by incubation of pulmonary edema fluid or cytokines on subconfluent, quiescent fibroblasts for 24 h. The range of dilutions of pulmonary edema fluid was chosen based on initial experiments that defined the full range of the fibroblast proliferative response (from 0.4% to 10% SCM) from cells exposed to edema fluid for a total of 6 days. Serum titration growth curves demonstrate that 0.4% serum conditions resulted in <4% increase in cell number/day over 6 days and <5% trypan blue-excluding cells. Cell number was determined by direct counting of trypsinized, resuspended cells or by eyepiece reticle counting using a regression formula (total cell number = 0.0666 (cell number/mm²) − 0.5347; r² = 0.97) as validated in our laboratory. Where indicated, DNA synthetic rates were measured by 5-bromo-2′-deoxyuridine incorporation by ELISA according to the manufacturer’s protocol (Roche Applied Science, Indianapolis, IN). For growth inhibitory assays, either cell monolayers or test solutions (edema fluids/conditioned medium) were preincubated with the indicated reagents. The 50% neutralizing dose (ND₅₀) for TNF binding protein-1, recombinant human extracellular domain of TNFR-1 (TBP-1), is 15 ng/ml to neutralize the bioactivity of TNF-α (1 ng/ml) (26). The ND₅₀ of the IL-1R antagonist (IL-1ra) is 1 ng/ml. The ND₅₀ for IL-6 Ab is 15 ng/ml, and for the gp130 Ab, it is 20 ng/ml to neutralize the bioactivity of 50 pg/ml of IL-1β and 2.5 ng/ml of recombinant human IL-6, respectively (R&D Systems, Minneapolis, MN).

Total RNA was isolated from fibroblasts using the method of Chomczynski and Sacchi (27), followed by a second phenol/chloroform/isoamyl extraction and ethanol precipitation (5). The resultant RNA was treated with RNase-free DNase (Promega, Madison, WI) to remove trace genomic DNA contaminants. All RNA samples were verified as undegraded and free of DNA through quantitation of the 28S and 18S RNA bands generated by denaturing gel electrophoresis (1.2% agarose/2.2% formaldehyde, 1× MOPS, 80 V, 2 h) after ethidium bromide staining of 5 μg of RNA/lane. RT-PCR was performed on 1 μg of RNA/reaction using Superscript II (Life Technologies, Grand Island, NY) at 55°C for 30 min using gene-specific primer pair mixes provided by the manufacturer of the cDNA membrane array (Clontech, Palo Alto, CA) and 3000 Ci/mmol ³²P-tagged d-ATP (Amersham Pharmacia Biotech, Piscataway, NJ). Equal cpm/sample were hybridized to identical cDNA membrane arrays, and relative levels of individual genes were determined in duplicate by PhosphorImager analysis (Molecular Dynamics, Sunnyvale, CA) using ImageQuant software (Molecular Dynamics). The values for the average band densities of the nine housekeeping genes from the two samples were within 10%, indicating comparable ³²P-RT-PCR efficiency and hybridization. In support of a targeted approach that focuses on biological pathways and maximizes results with limited patient samples (1 μg of total RNA/hybridization), we used the 588 human gene array. This array contains duplicate cDNA spots from a broad spectrum of human genes that function in different biological pathways.

Total RNA was extracted by the acid guanidium isothiocyanate-phenol-chloroform method, and RNA (10 μg/lane) was electrophoresed under denaturing conditions and transferred to nylon membranes (Magnagraph; Osmonics, Minnetonka, MN) as described (5, 27). The membranes were then hybridized with 10⁶ cpm/ml of deoxyadenosine 5′-[α-³²P]triphosphate (3000 Ci/mmol; Amersham Pharmacia Biotech) and Klenow fragment-labeled DNA (>1 × 10⁸ cpm/μg DNA) and then were washed (1 h at 65°C with 5% SDS, 0.67× SSC) as published (5). The cDNA for human ICAM-1 was used as previously described (28). The cDNAs for human GAPDH and for rat 18S RNA were used as described (5). The cDNAs for human IL-6, α-catenin, MIP-2-α, PDGF-A chain, and G-CSF were generated by oligo(dT)-primed RT-PCR (Superscript II RT−; Life Technologies) of total fibroblast RNA followed by PCR using unique gene-specific primers (Clontech). The resultant products were size and sequence validated and subjected to topoisomerase I-isomerase dependent-ligation into pCR-TOPO (Invitrogen, San Diego, CA) using minor modifications of the manufacturer’s instructions. The intensity of the signal for specific mRNAs was quantified directly on a PhosphorImager using ImageQuant software. Equal loading and transfer of the RNA was verified by visualization of the ethidium bromide-stained RNA in the membrane after transfer and by normalizing for the 18S RNA signal intensity.

To quantify mRNA levels of G-CSF, PDGF-A chain, MIP-2-α (gro-β), IL-6, and GAPDH, quantitative RT-PCR was performed using a dsDNA-specific fluorochrome (SYBR Green-I) on the LightCycler Instrument (Roche Molecular Biochemicals, Mannheim, Germany) as described (29, 30). Briefly, the reverse transcriptase reaction was performed as described above. The gene-specific PCR primer sets were obtained commercially (Clontech). Reaction conditions were optimized for each primer set, and each reaction was individually validated for specific, single product (Tm/size homogeneity) using a melt curve and gel electrophoresis. To determine relative RNA levels, the second derivative maximum of the fluorescence/cycle number plot was calculated. The linear relationship (r² > 0.99) between the starting DNA concentration and the log of the concentration over a large concentration range (six to eight logs in duplicate) was verified for each primer set, and similar efficiencies of PCR among the fibroblast RNA samples and the individual cDNA-containing plasmids were verified. Recovery of spiked cDNA into the fibroblast RNA samples was routinely >95% with an intersample coefficient of variation of <5%. The difference in individual starting mRNA amounts in each sample was calculated by comparison with a known amount of cDNA standards and with serially diluted unknown samples run in parallel. These validation steps support robust and accurate quantification of RNA levels among samples.

IL-1β and IL-6 Ag levels were measured by double-Ab ELISA using reagents and protocols according to the manufacturer’s instructions with minor modifications (R&D Systems). Levels of IL-1β or IL-6 were determined by comparison with the absorbance (450–550 nm; V max; Molecular Devices, Menlo Park, CA) of known amounts of recombinant human IL-1β and IL-6 standards. The IL-1β ELISA is specific for mature, human IL-1β with negligible interference by 3000-fold excess of IL-1 soluble receptors and no interference or cross-reactivity by 3000-fold excess of IL-1ra and IL-1α. The IL-6 ELISA is specific for human IL-6 with negligible interference by 3000-fold excess of IL-6 soluble receptors (free or gp130 bound).

The continuous variables were compared by an unpaired Student’s t test or by a Mann-Whitney test if the variables were not normally distributed. The effect of acute lung injury and hydrostatic edema fluid dilutions and the effect of inhibitory molecules on fibroblast proliferation and IL-6 production were compared by an ANOVA followed by the Student-Newman-Keuls or the Dunnett’s multiple comparison procedure. All proportional values were compared by a Fisher’s Exact test (31). Statistical significance was accepted as p < 0.05.

Pulmonary edema fluid was obtained from 19 patients with acute lung injury and from 17 patients with hydrostatic pulmonary edema within 4 h of endotracheal intubation and initiation of mechanical ventilation. Patients were classified as having acute lung injury based on both clinical and physiological criteria from the North American-European Consensus Conference and by an edema fluid:plasma total protein ratio of >0.65. The acute lung injury group had a higher edema fluid:plasma total protein ratio (1.0 ± 0.2 vs 0.5 ± 0.1; p < 0.001), and an underlying infection was the most common cause of acute lung injury (Table I). Intrapulmonary infection was the most common infectious etiology, accounting for 60% of the total infectious causes of acute lung injury. The groups were otherwise matched for age, gender, smoking status, and SAPS II (Table I).

To test whether early acute lung injury edema fluid is mitogenic, normal human lung fibroblasts were incubated with pooled edema fluid (diluted in SFM) that was obtained from patients with either acute lung injury (ALI) or hydrostatic edema (HYDRO). The range of pulmonary edema fluid dilutions was chosen based on preliminary experiments that defined the range of serum-dependent fibroblast proliferative responses. Both the hydrostatic edema and acute lung injury edema fluids were mitogenic (cell number increment at 4 days), with the maximal acute lung injury fluid effect approximating that seen with complete 10% SCM. The edema fluid derived from the ALI patients exhibited a greater overall mitogenic effect and a different pattern of concentration dependency compared with edema fluid from the HYDRO group (Fig. 1; p < 0.00001 by ANOVA/Student-Newman-Keuls).

Because both TNF-α and IL-1β Ag levels have been reported as increased in ALI and may function as modulators of fibroblast proliferation, we tested the effects of the specific inhibitors of TNF-α/TNF-β (TBP-1) and IL-1 (IL-1ra or neutralizing IL-1β Ab) on pulmonary edema fluid-induced fibroblast proliferation. Inhibition of pulmonary edema fluid IL-1β activity with IL-1ra or neutralizing IL-1β Abs blocked acute lung injury edema fluid-induced fibroblast proliferation by 40 ± 7% (p < 0.05 by ANOVA/Dunnett’s test) in a concentration-dependent and specific manner (Fig. 2). Although TNF-α may interact with IL-1β in its influence on fibroblast proliferation (32, 33), there was no effect of inhibiting TNF-α in pulmonary edema fluid with TBP-1, with or without concomitant IL-1β signal blockade (Fig. 2). The observation that combined PDGF A and B chain neutralizing Abs also partially inhibited ALI-induced proliferation extends prior work (21) demonstrating a mitogenic role for PDGF B-derived molecules in adult respiratory distress syndrome (ARDS) bronchoalveolar lavage fluid (Fig. 2 A). Taken together, these data substantiate IL-1β’s important role in ALI edema fluid-induced proliferation and demonstrate the importance of other modulating factors on the mitogenic effects of IL-1β.

Because the ALI edema fluid demonstrated greater IL-1β-dependent mitogenic bioactivity, we compared IL-β Ag levels in ALI edema fluid with HYDRO fluid. IL-1β mean Ag levels in pulmonary edema fluid were eightfold higher in ALI patients compared with HYDRO patients (mean ± SD; ALI, 587 ± 690 pg/ml; HYDRO, 71 ± 77 pg/ml; p = 0.009) (Fig. 3).

To determine the potential contribution of autocrine-acting fibroblast products, we compared the mitogenic effect of conditioned medium from fibroblasts stimulated by ALI edema fluid to that of ALI edema fluid alone. There was a dose-dependent, threefold higher (p = 0.00003 by ANOVA) lung fibroblast cell number at 4 days when incubated with ALI edema fluid-stimulated conditioned medium, compared with equivalent dilutions of ALI edema fluid alone (nonconditioned medium) (Fig. 4). The autocrine, mitogenic activity of ALI fluid in fibroblasts was independent of serum supplementation (data not shown). As a second control, fibroblasts incubated with SFM, with or without 1% BSA, did not condition their medium with mitogenic factors. Thus, in response to ALI edema fluid, fibroblasts secrete soluble factor(s) that stimulate proliferation in an autocrine/paracrine manner.

To determine the nature of the differential biological response and to identify putative autocrine-acting mitogenic factors, the gene expression profiles of fibroblasts incubated with ALI edema fluid or HYDRO fluid were compared using cDNA membrane arrays. Total RNA was isolated from fibroblasts that were incubated under the identical conditions that demonstrated a quantitative difference in their proliferative response (as in Fig. 1) and was subjected to ³²P-RT-PCR using gene-specific primer mixes. A focused array (588 genes in duplicate/array) was chosen because it targets biologically active, known genes, can be used with limited starting material, and does not require an up-front, random-primed PCR step before hybridization. Several quality control issues were addressed. The efficiencies of the ³²P-RT-PCR and hybridizations were comparable, in that the average band density of the nine housekeeping genes from the two samples was within 10%, and the median ratio of the expression levels (ALI-exposed fibroblasts/HYDRO-exposed fibroblasts) of the analyzable genes was close to 1 (0.98) (Fig. 5 B). Genes with band densities of less than fourfold above background levels were excluded from the analysis (315 of 588 genes).

After logarithmic transformation of the absolute band density, six genes from the ALI edema fluid-exposed fibroblasts were identified that had an induction level greater than 2 SDs above the mean (>97.5 percentile, more than fourfold). They were PDGF-A chain, IL-6, the chemokine MIP-2-α (gro-β), G-CSF, and the adhesion-related genes ICAM-1 and α-catenin (Fig. 5 A). Interestingly, these are all inflammatory/growth-modulating gene products that may play pathogenic roles in acute lung injury/repair and have all been found to be IL-1β responsive in other systems and/or cell types (34). Furthermore, a number of additional genes that are both IL-1β responsive in other systems and are putative modulators of the inflammation/fibroproliferative response were moderately induced (twofold to fourfold) by ALI edema fluid. These include epithelial neutrophil-activating protein-78, IL-5, IL-8, IL-10, placental growth factor (vascular endothelial growth factor-related protein), integrin α5 subunit, and urokinase receptor, as well as several growth/inflammation-related transcription factors.

To confirm the array findings, precisely quantify the levels of gene induction, and determine the relative significance of IL-1β’s effect on highly induced genes, Northern blot analysis and/or quantitative RT-PCR was performed on RNA isolated from fibroblasts exposed to edema fluids in the absence or presence of IL-1ra (200 ng/ml; n = 4/condition). Quantitative RT-PCR and Northern blot analysis confirmed the array findings of ALI edema fluid induction of all six highly induced genes, and furthermore, the level of induction was quantitatively similar to that of the array (Fig. 6 and Table II). The specific IL-1ra blocked the ALI edema fluid gene response of IL-6 (by >90%) and the induction of all of the other highly induced genes by 35–70% (Fig. 6 and Table II). Taken together with the noted induction of a number of other IL-1β-responsive genes, these observations demonstrate the magnitude of the significance of IL-1β-dependent cell signaling on fibroblast phenotype in response to ALI edema fluid.

To determine whether the IL-1β mRNA-inducing activity of ALI edema fluid is biologically significant at the protein level, we tested whether fibroblasts secrete IL-6 protein in response to pulmonary edema fluid in an IL-1β-dependent manner. Fibroblasts secreted 3.5-fold more IL-6 into the medium in response to ALI edema fluid compared with their response to HYDRO fluid (25,200 ± 7,250 pg/ml vs 7,831 ± 2,566 pg/ml; p = 0.004 by ANOVA; Fig. 7,A). Interestingly, the IL-6 protein response to ALI edema fluid was 50-fold that of 10% SCM (Fig. 7,A). The in vivo relevance of this result is supported by the observation that edema fluid obtained from patients with ALI had an 11-fold higher IL-6 level compared with those with HYDRO (104 ± 10 ng/ml vs 9 ± 2 ng/ml; p < 0.00001). Concordant with the IL-6 gene expression dependency on IL-1β, IL-1ra (200 ng/ml) but not TBP-1 abrogated the IL-6 protein induction above baseline by ALI edema fluid in a dose-dependent and specific manner by >90% (Fig. 7 B). These data demonstrate that the IL-1β in ALI edema fluid is bioactive and is responsible for the majority of the marked induction of IL-6 gene expression and protein production by fibroblasts.

The observation that both fibroblast proliferation and IL-6 expression are dependent, in part, on the IL-1β bioactivity in ALI edema fluid provides the impetus to examine whether IL-6 is the IL-1β-dependent, autocrine-acting fibroblast mitogen. To this end, the mitogenic activity of exogenously added cytokines and of conditioned medium from IL-1β-stimulated fibroblasts was determined. Exogenously added IL-1β exhibited a concentration-dependent (1 pg/ml − 10 ng/ml; EC₅₀ = 50 pg/ml), modest mitogenic activity that approached 30% of that seen with 10% SCM (data not shown), whereas exogenously added IL-6 (10 pg/ml–100 ng/ml) alone had no mitogenic effect (Fig. 8,A). However, the addition of small amounts of IL-1β (10 pg/ml) along with IL-6 revealed an interactive, concentration-dependent mitogenic effect of IL-6 (Fig. 8,A). Conditioned medium from IL-1β-exposed cells (10 pg/ml for 24 h) exhibited a threefold greater mitogenic effect than did the same concentration of IL-1β in unconditioned medium, indicating that IL-1β induces fibroblasts to secrete autocrine/paracrine-acting mitogenic factors into the medium (Fig. 8,B). The mitogenic-enhancing effect of the IL-1β-conditioned medium was specifically blocked by neutralizing Abs to IL-6 protein or to the IL-6 cell signaling transducer gp130 (Fig. 8 B). Taken togther, these data demonstrate the physiological significance of IL-1β-stimulated, autocrine/paracrine-acting IL-6 on fibroblast proliferation.

The major findings in this study can be summarized as follows. Pulmonary edema fluid from patients with early ALI had a greater mitogenic effect on human lung fibroblasts than did edema fluid from control patients with HYDRO. Furthermore, human lung fibroblasts incubated with ALI edema fluid produced soluble mediators that augment this mitogenic effect in an autocrine/paracrine manner. This effect was demonstrated by the finding of a greater mitogenic effect of medium conditioned by exposure of fibroblasts to ALI edema fluid compared with the same edema fluid alone. There was a different gene expression response profile in human lung fibroblasts incubated with pulmonary edema fluid derived from ALI patients compared with edema fluid derived from HYDRO patients. Highly induced genes, confirmed by Northern blotting and/or RT-PCR, encode for both inflammation and growth-modulating molecules that have the potential to alter fibroproliferative events in lung injury, including cell adhesion, migration, cytokine/chemokine production, and proliferation. We found that IL-1β Ag levels and IL-1β-dependent mitogenic bioactivity were increased in pulmonary edema fluid obtained within hours of intubation, at a time when there is biochemical evidence for tissue remodeling (9, 10, 11). Not only was IL-1β Ag eightfold higher in ALI edema fluid compared with HYDRO fluid, but IL-1ra abrogated both the increase in expression of the highly induced genes and the mitogenic effect of ALI edema fluid in a concentration-dependent and specific manner. Most notably, IL-6 induction (20-fold) by ALI edema fluid was almost entirely dependent on IL-1β.

The biological significance of IL-1β-dependent IL-6 induction and autocrine action on fibroblast proliferation was experimentally demonstrable. First, when combined with submaximal concentrations of exogenous IL-1β, exogenous IL-6 produced a vigorous mitogenic response. Second, the mitogenic effect of conditioned medium from IL-1β-exposed cells was blocked by inhibitory Abs to IL-6 and its signal transducer, gp130. These lines of evidence identify IL-1β as an important, bioactive molecule in the alveolar compartment in early clinical ALI and demonstrate mechanisms whereby IL-1β-dependent activation of pulmonary fibroblasts can modulate subsequent inflammatory and fibroproliferative events in an autocrine/paracrine manner.

Studies in several different experimental animal models of acute inflammation have demonstrated a direct in vivo role for IL-1-dependent signaling (35, 36, 37). Experimental models of chronic pulmonary inflammation also demonstrate the pathophysiologic importance of IL-1β. For example, IL-1ra protects mice from the fibroproliferative effects of intratracheal bleomycin and silica instillation (38). Conversely, transient overexpression of IL-1β in rat lungs is followed by induction of the profibroproliferative cytokines PDGF and TGF-β and a fibroproliferative lung disorder (20). These data causally link increases in alveolar IL-1β activity with the subsequent development of pulmonary fibrosis, perhaps through autocrine/paracrine pathways.

We found that fibroblast IL-6 expression was strongly induced by ALI edema fluid and, concordantly, that IL-6 levels were 11-fold higher in ALI edema fluid compared with HYDRO fluid. A number of molecules present in the acutely injured alveolus have been reported to induce IL-6 in fibroblasts, including IL-1, IL-17, TNF, PDGF, PGE, platelet activating factor, and LPS (39, 40, 41). In this light, our finding of an almost complete dependency of IL-6 mRNA and protein induction on the IL-1β in pulmonary edema fluid derived from patients with ALI is remarkable. This dependence has been similarly observed in vivo. Overexpression of IL-1β in rat lungs induces a concomitant increase in alveolar IL-6 levels (20). This observation extends the concept of IL-1β dependency of IL-6 to the in vivo level, in the context of lung inflammation. The fractional contribution of pulmonary fibroblasts to the overall increase in pulmonary IL-6 during IL-1β-mediated experimental ALI remains conjectural, because endothelial cells, smooth muscle cells, and leukocytes can produce IL-6 in response to IL-1β (39). Furthermore, additional mediators may act to induce IL-6 in these cell types during clinical ALI. The findings clearly demonstrate one pathway whereby IL-1β activity in actual lung injury fluid will induce fibroblast-derived IL-6 and other gene products that modulate the inflammatory/fibroproliferative process.

The net action of IL-6 during lung injury/repair is not easily inferred based on the sum of its individual cell type-specific actions. Direct in vivo studies show that IL-6-deficient mice exhibit an enhanced neutrophil recruitment and an enhanced induction of proinflammatory cytokines (i.e., TNF-α, MIP-2, and GM-CSF) in response to exogenous LPS, suggesting that IL-6 acts to limit acute pulmonary inflammation (42). In contrast, overexpression of IL-6 along with IL-6R in rat lungs using an adenoviral vector or in transgenic mice induces an interstitial lymphocytic alveolitis, concordant with its known stimulatory effects on lymphocytes (43). These apparently contradictory pro- and anti-inflammatory in vivo findings highlight the pleiotropic nature of IL-6 and the importance of its interactions with other cytokines, and the particular cell type that expresses the IL-6R, in determining the ultimate response. The specific role of IL-6 in response to a pulmonary fibroproliferative challenge in vivo has yet to be directly tested in a loss of function experiment.

IL-6 has a positive, negative, or neutral effect on cell proliferation, depending on the cell type, its state of differentiation, and cross-talk with other cytokines (39, 44, 45). In the case of human fibroblasts, exogenous IL-6/IL-6R is mitogenic in rheumatoid arthritis patient-derived synovial fibroblasts, but suppresses proliferation in normal dermal fibroblasts (46, 47). Our data using purified IL-1β and IL-6 alone and in combination demonstrate that IL-1β sensitizes human lung fibroblasts to the mitogenic effects of IL-6. The mechanism for this sensitization may involve induction of IL-6 protein, changes in IL-6R by induction and/or proteolysis-induced shedding, induction of gp130, and/or cross-talk among intracellular signaling pathways (48, 49). There is precedence for more than one mechanism for the mitogenic effects of IL-1β/IL-6 interactions. IL-1β-dependent proliferation of human megakaryocytic leukemic cells is a consequence of secreted, extracellular IL-6, whereas PDGF-dependent proliferation of lung fibroblasts has been found to be mediated by IL-6 in a manner that was independent of extracelluar IL-6/IL-6R (50, 51). Our data are most consistent with the former pathway, in that IL-6 mRNA and secreted protein were induced by ALI edema fluid, and neutralizing Abs to IL-6 blocked the mitogenic effect of cell-conditioned medium from cells exposed to IL-1β.

The partial inhibitory responses of proliferation and gene expression to IL-1ra or IL-1β Abs in this study indicate that IL-1β is not the only fibroblast mitogen or phenotypic modulator in early ALI edema fluid. Prior work has implicated TNF-α and PDGF isoforms as soluble fibroproliferative mediators in early ALI. In this study, inhibition of TNF-α and TNF-β bioactivity with TBP-1 did not alter the fibroblast IL-6 protein induction or the mitogenic effect of early ALI edema fluid in a direct or synergistic (with IL-1β (32, 52)) manner. Similarly, inhibition of TNF-α signaling did not prevent the inducing effect of ARDS bronchoalveolar lavage on the surface expression of ICAM-1 on alveolar epithelial cells, nor did it have an effect on in vitro epithelial cell wound closure induced by ALI edema fluid on other reports (14, 26, 53, 54). Strong evidence points away from TNF-α as a major modulator of fibroblast phenotype in early ALI, despite its abundance. In contrast, prior work provides convincing evidence for an alveolar mitogenic predominance and a fibroblast mitogenic role for PDGF B chain-related molecules in early (within 3 days) ALI bronchoalveolar lavage fluid (21, 55). IL-1β can also up-regulate PDGF A chain and PDGFR-α in fibroblasts, thereby potentially sensitizing fibroblasts to the mitogenic effects of PDGF (56, 57). These results are complementary to our findings, in that we found that combined PDGF A/B chain neutralizing Abs did partially abrogate the mitogenic effect of ALI edema fluid. Our finding of a reduced bioactivity of IL-1β in the edema fluid compared with that in purified IL-1β is concordant with the observation in bronchoalveolar lavage from early ARDS patients of residual IL-1β bioactivity, despite a molar excess of the soluble antagonist IL-1ra (16).

We have shown that fibroblasts are induced to produce several soluble mediators and cell-cell contact receptors in response to ALI edema fluid that can modulate the migration and function of inflammatory cells (i.e., G-CSF, MIP-2-α, ICAM-1, etc.) (58). Fibroblasts support leukocyte adherance in an integrin-dependent manner in vitro, and electron microscopic studies demonstrate intimate (within 15 nm) connections between interstitial fibroblasts and migrating neutrophils during acute experimental lung injury in animals (59, 60). Thus, both contact-dependent and contact-independent cross-talk between fibroblasts and leukocytes may modulate the inflammatory response to lung injury. The fibroblast may function as an autocrine/paracrine modulator of early lung injury-related inflammation through its response to IL-1β and other soluble mediators present in the injured alveolus.

The frequency of ALI due to all infectious etiologies of 90% in our study was somewhat higher than that observed in recent trials (60%), whereas the frequency of underlying infectious pneumonia in our study (60%) is concordant with recent large scale ARDSnet trials (61). The well characterized common occurrence of underlying infection in patients with ALI raises several issues. First, we cannot rule out the possibility that some of our findings are a consequence of infection-induced, perhaps organism class-specific, alveolar cytokine profiles. However, levels of the key cytokine, IL-1β, in bronchoalveolar lavage from patients with ALI due to infection overlap with those from acute lung injury due to trauma or other causes (16). Second, the possibility that infection-related molecules mediate fibroblast responses in ALI should be considered. For example, IL-17 has recently been shown to be increased in experimental pulmonary infection and to possess a gene-induction pattern that overlaps with IL-1β in fibroblasts (40, 41). Lastly, although LPS is measurable in bronchoalveolar lavage from patients with ALI, the concentrations are 100- to 1000-fold lower (after accounting for dilution) than that reported to induce IL-6/G-CSF in fibroblasts (62, 63).

This study has some limitations. The volume of patient-derived pulmonary edema fluid was limited, which necessitated pooling of the samples. Thus, we are not able to make clinical-molecular correlations regarding individual patients. Furthermore, we chose a single time point to correspond with the growth factor-dependent phase of the cell cycle and to correspond to the time of maximal gene expression changes in serum-stimulated fibroblasts (64). Despite these considerations, the assay of bioactivity in pulmonary edema fluid is a strength, in that it supports the detection of trace molecules that may be below their bioactive concentrations in bronchoalveolar lavage (65). For example, the levels of IL-1β and IL-6 found in pulmonary edema fluid were ∼100-fold that seen in bronchoalveolar lavage from similar patients (16). The essential identity of the IL-6 mRNA findings using quantitative RT-PCR, Northern blotting, and the microarray, along with the other validation steps, provides a high degree of confidence in the molecular analysis. Because the IL-ra only partially inhibited proliferation and gene expression, it is clear that there are additional, perhaps interactive, mitogens present in the pulmonary edema fluid that will be identified in future work.

In summary, this study provides several independent lines of evidence demonstrating that IL-1β in the alveolar compartment amplifies the inflammatory and fibroproliferative process through regulation of fibroblast mitogenesis and fibroblast gene expression. Among the plethora of bioactive molecules within the acutely inflamed human alveolus, we have found for the first time that IL-1β is one important regulator of the fibroblast phenotype in patients with ALI. These novel observations demonstrate the capacity of IL-1β-driven human lung fibroblasts to both regulate the early inflammatory response and initiate the fibroproliferative response to clinical ALI.

We thank the National Institutes of Health/Cystic Fibrosis Foundation Gene Therapy Center for access to equipment and Terri Nelson for her secretarial assistance.