Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Restrictive lung disease develops in 30-60% of patients with systemic sclerosis (scleroderma) within the first three to five years of disease (1). In ~ 15% of patients, this fibrotic process progresses to severe restrictive lung disease (1), which is a major cause of death in scleroderma.

The mechanisms that lead to progressive lung fibrosis in scleroderma remain obscure. Lung inflammation strikes a subset of patients, and its presence, as assessed by neutrophilia or eosinophilia on bronchoalveolar lavage (BAL) cell differential, indicates patients at greater risk for progressive lung fibrosis and death (2, 3). A variety of cell types are increased in BAL fluids from scleroderma patients with lung inflammation. These cell types include alveolar macrophages, CD8⁺ T cells, mast cells, basophils, eosinophils, and neutrophils (2). Attempts have been made to link changes in expression of inflammatory mediators with lung fibrosis in scleroderma. Thrombin (7), fibronectin (8), transforming growth factor- (TGF-) (9), endothelin-1 (10), and type 2 cytokine mRNAs (11) all have been reported to be increased in BAL fluids or cells from scleroderma patients. The multiplicity of the potential mediators suggests that pathologic mechanisms of scleroderma lung disease are complex, and that approaches more robust than looking at single factors will be required to develop a better understanding of this complexity.

DNA array technology has been used to address changes in gene expression in a murine model of lung fibrosis secondary to bleomycin (12). In that work, analyses of whole lungs identified distinct clusters of genes that were induced by bleomycin. Genes associated with both the inflammatory and fibrotic responses included complement components, serine proteases, chemokines and chemokine receptors, genes restricted to leukocytes, and other genes associated with inflammation. Among the genes associated with the fibrotic response were genes involved in the formation of extracellular matrix, genes involved in the regulation of cellular responses to the extracellular matrix, genes induced by DNA damage, and genes induced by TGF-.

The purpose of this work was to gain insight into potential mechanisms of lung fibrosis in scleroderma by identifying abnormal patterns of gene expression in BAL cells from scleroderma patients at higher risk for progressive lung fibrosis, that is, patients with lung inflammation. The hypothesis of the study is that patterns of gene expression in BAL cells from these patients will reflect the importance of cellular immunity with chronic macrophage activation in the genesis of lung fibrosis in scleroderma.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Patients and Control Subjects

None of the patients and control subjects were current smokers. All patients were outpatients at the Johns Hopkins and University of Maryland Scleroderma Center, all met American College of Rheumatology criteria for classification of scleroderma (13), and none were receiving cyclophosphamide at the time of BAL. One patient with lung inflammation had two BAL procedures done 20 mo apart, and data from both BAL cell samples are included. Informed written consent was obtained from all patients and control subjects.

BAL

The BAL procedures and cell differential counts were done as previously described (2). All BAL procedures were done in an identical manner by the same physician in the same facility and processed by the same technician. Briefly, three subsegments of the right lung were each lavaged with five 20-ml aliquots of sterile saline, and the recovered lavage fluids were pooled for each individual. Cell differential counts were judged to indicate inflammation if neutrophils were 3.0% or greater or if eosinophils were 2.2% or greater of total cells (2). Thus, volunteers could be divided into patients with lung inflammation, patients without lung inflammation, and control subjects.

BAL Enzyme-Linked Immunosorbent Assay

For enzyme-linked immunosorbent assays (ELISAs), BAL fluids were separated from BAL cells by centrifugation and concentrated 10-fold using Macrosep 10K filters (Pall Filtron, Northborough, MA). ELISA kits for monocyte chemoattractant protein-1 (MCP-1), macrophage inflammatory protein (MIP)-1, interleukin-1 receptor antagonist (IL-1ra), and TGF-1 were purchased from R&D Systems (Minneapolis, MN), and protein quantification was performed according to the manufacturer's recommendations. For ELISA measurement of pulmonary and activated-response chemokine (PARC) protein, capture antibody and detection biotinylated anti-PARC antibody were purchased from R&D Systems.

DNA Arrays

Data on expression of 4,132 genes were collected from GF211 arrays (Research Genetics, Huntsville, AL), and data on expression of 375 genes were collected from Panorama Cytokine Gene Arrays (Sigma-Genosys, Woodland, TX). Procedures for RNA purification, cDNA preparation and labeling with ³³P, hybridization, and membrane washing were done according to manufacturers' instructions that come with the arrays. For the GF211 arrays, cDNA was prepared with oligo dT primers from Research Genetics. For the cytokine arrays, components of the cDNA preparation and labeling reactions were obtained from Sigma-Genosys in a kit that contained human cytokine labeling primers and all components for the reverse transcriptase reaction. The RNA was isolated from about three million cells per BAL sample. The RNA from BAL samples from 16 patients (17 samples, with 2 samples from one patient) and 7 control subjects was tested with the GF211 arrays, and RNA from 14 patients and 6 control subjects was tested with the cytokine array, with too little remaining in the other samples to test with the cytokine array.

Analysis of DNA Array Data

Hybridized membranes were exposed to phosphorimaging screens for 1-3 d. Images were scanned with a Storm Imaging System (Molecular Dynamics, Sunnyvale, CA). Pathways 2 software (Research Genetics) was used to assign gene names to spot densities and calculate background values on GF211 arrays. ImageQuant (Molecular Dynamics) template from Sigma-Genosys was used to identify genes on cytokine arrays. For every gene spot on the cytokine template, a set of an additional four background spots around each gene spot was created. The average of these four spots was used as the local background for the corresponding gene spot on the arrays. Spot densities were exported into Excel (Microsoft, Redmond, WA). Background values were subtracted, and data normalized to the combined density of all spots, with a mean spot intensity set at 100 for each individual array. Genes were selected if they were expressed on at least one array, with 2,071 genes on the GF211 arrays and 235 genes on the cytokine arrays remaining after such selection. Data were normalized again to the combined spot density of the selected genes, with a mean spot intensity of 100 for each array. Expression at or below the level of detection was assigned an arbitrary small value of 10 for calculation purposes for fold differences, Mann-Whitney, and clustering analyses, in a fashion similar to that reported by others (12, 14).

Further selection of the genes was done using significance analysis of microarrays (SAM) software (15), Mann-Whitney analyses, and fold differences. The SAM software uses repeated permutations of the data to determine whether the differences in the expression of a gene between groups are significant. The results obtained from DNA arrays using this technique correlate well with Northern blotting analyses (15). Stringency parameters were set such that falsely significant genes were ~ 10% of total selected genes. All genes identified by SAM analyses were subjected to Mann-Whitney analyses using Statistica (StatSoft, Tulsa, OK) software, and genes with significantly different expression (P  0.05) between groups were considered further. Then, fold differences in the level of expression of these selected genes were determined by dividing the mean intensity in one group by that of another. Genes were selected for final consideration if the mean level of expression was at least 2-fold different in the two groups being compared.

For hierarchical clustering, data were exported into Gene Cluster software (freeware from Eisen and colleagues [14]), and subjected to mean centering. Hierarchical clustering was done using uncentered correlation and average linkage clustering. Clusters were viewed using TreeView (freeware from Eisen and colleagues [14]).

Website

A website at http://www.bwresearch.com/AJRCMB supports the data presented in this paper. The GF211 and cytokine datasets themselves, information on the GF211 and cytokine arrays, all figures, and an expanded table that lists all genes that were stimulated or repressed in scleroderma patients with lung inflammation can be found at this site. The table includes GenBank links through the accession numbers.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Patients and Control Subjects

For DNA array experiments, 16 patients with scleroderma and seven healthy individuals were studied. Nine patients had lung inflammation, and seven did not. Patients with and without lung inflammation had similar age, sex, race, disease duration, disease type, pulmonary hypertension, and prednisone therapy; P > 0.05 for all comparisons (Table 1). Patients without lung inflammation had median forced vital capacity (FVC) and diffusing capacity for carbon monoxide (DLco) within normal limits at the time of BAL, with values  80% predicted. However, their DLco (% predicted) values were less than those of control subjects (P = 0.03, Mann-Whitney). Patients with lung inflammation had lower FVC and DLco than control subjects (P < 0.05, Mann- Whitney) for both absolute and % predicted values. As expected, BAL samples from patients with lung inflammation had higher cells/ml, % neutrophils, and % eosinophils than those from patients without lung inflammation and control subjects. Nonetheless, in all BAL samples, alveolar macrophages comprised the majority of BAL cells, ~ 90% of cells.

Stimulated and Repressed Genes in Scleroderma Patients without Lung Inflammation

Previous work has suggested that scleroderma patients without lung inflammation are at lower risk for progressive lung fibrosis than patients with lung inflammation (2, 3). For this reason, the gene expression profile in BAL cells from scleroderma patients without lung inflammation was compared with that in BAL cells from control subjects to determine what differences, if any, there were between the two groups. Analyses revealed that there were no genes on the GF211 arrays and only three genes on the cytokine arrays that had different expression in BAL cells in patients without lung inflammation and healthy control subjects (see Figures 1A and 1B, scatterplots for GF211 and cytokine arrays, respectively). Those genes were CD64, the high-affinity Fc receptor that mediates phagocytosis by macrophages, tumor necrosis factor (TNF) RI, an apoptosis signaling molecule, and B7.2, a T cell-co-stimulatory molecule expressed by antigen-presenting cells. All were reduced in patients without lung inflammation compared with control subjects, with fold reductions of 3.3 (P = 0.009), 7.7 (P = 0.015), and 4.5 (P = 0.025), respectively.

View larger version (44K): [in this window] [in a new window]   Figure 1.   Scatter plot analyses of gene expression in scleroderma patients and control subjects. Data are presented as logarithm of mean level of expression of genes in (A) patients with no lung inflammation versus healthy control subjects, GF211 arrays, (B) patients with no lung inflammation versus healthy control subjects, Panorama cytokine arrays, (C) patients with lung inflammation versus the combined control group of patients with no lung inflammation and healthy control subjects, GF211 arrays, and (D) patients with lung inflammation versus the combined control group of patients with no lung inflammation and healthy control subjects, Panorama cytokine arrays. The solid diagonal line in each panel indicates equal gene expression. The dashed lines indicate two-fold difference in expression. Expression of genes shown in color were either increased (red) or decreased (green) in the patient group presented on the vertical axis. The criteria for selection of genes with increased or decreased expression included identification by SAM, fold difference  2, and P value  0.05 by Mann-Whitney analyses. Numbering of the genes in panels C and D corresponds to genes listed in Figure 2 and Table 2.

View larger version (53K): [in this window] [in a new window]   Figure 2.   Clustergrams of stimulated and repressed genes in scleroderma patients with lung inflammation. Arrays from subjects are presented in columns and divided into groups of control subjects (C1- C7), patients with no lung inflammation (NI1- NI7), and patients with lung inflammation (I1- I10). Arrays I8 and I9 are from the same patient, but the BAL samples were obtained 18 mo apart. Genes were clustered by hierarchical clustering and are presented in rows, with data from GF211 arrays presented in genes 1- 76 and data from Panorama cytokine arrays presented in genes 77- 90. To indicate gene expression relative to the mean, red indicates increased expression and green indicates decreased expression. Numbering of the genes corresponds to Table 2.

Stimulated and Repressed Genes in BAL Cells from Scleroderma Patients with Lung Inflammation

Next, genes were identified that were stimulated or repressed in patients with lung inflammation, compared with a combined control group of patients without lung inflammation and healthy control subjects. The combined control group was used because gene expression profiles were similar in BAL cells from these two groups, with the exception of the few genes mentioned above. Seventy-six genes were identified from GF211 array data, and 16 genes were identified from cytokine array data. Figures 1C (GF211 array data) and 1D (cytokine array data) show scatter plots of the mean level of expression of these genes in patients with lung inflammation, compared with the combined control group without lung inflammation. Cluster analyses (Figure 2) show clear separation of abnormally expressed genes into those with increased or decreased expression in patients with lung inflammation. Table 2 provides a list of selected genes corresponding to those shown in Figure 2, with GenBank accession numbers, fold differences, and significance of differences by Mann-Whitney analyses.

Differences in gene expression profiles suggested that certain chemokines and chemokine receptors might be involved in the increase in cellularity in BAL fluids found in patients with lung inflammation. There was increased expression of genes for myeloid progenitor inhibitory factor-1, CXCR4, epithelial cell-derived neutrophil chemoattractant-78, and PARC (see Figure 2 and Table 2, genes 80-83). Because of the known restriction of PARC expression largely to the lung and its specificity for T cells (16) and the lack of data implicating dysregulation of PARC expression in lung disease in other human diseases, increased expression of PARC was confirmed by ELISA. As shown in Figure 3A, PARC protein levels were higher in BAL fluids from patients with lung inflammation, compared with the combined group of patients without lung inflammation and healthy control subjects, with mean ± standard deviation (SD) values of 5.8 ± 3.0 pg/ml versus 3.6 ± 1.6 pg/ml, P = 0.001, two-tailed Student's t test. Other chemokines that are produced by activated macrophages and have been associated with fibrotic lung disease, including MCP-1, MIP-1, and eotaxin, had increased gene expression in patients with lung inflammation, with fold increases in expression of 3.9, 3.5, and 2.2, respectively. Although these differences did not reach the stringent criteria used for final gene selection, because of the known stimulation of collagen production by MCP-1 (17), ELISA was done to determine whether protein levels of MCP-1 were also increased in patients with lung inflammation. As shown in Figure 3B, there was an increase in MCP-1 levels in patients with lung inflammation, compared with the combined control group, with mean ± SD values of 9.2 ± 7.0 pg/ml versus 5.0 ± 4.1, P = 0.04, two-tailed Student's t test. Consistent with chemoattraction and activation of monocytes by chemokines was the increased expression of the gene for acid phosphatase 5 (Figure 2 and Table 2, gene 35), which is associated with monocyte transformation into macrophages (18).

View larger version (17K): [in this window] [in a new window]   Figure 3.   ELISA analyses of cytokine levels for PARC (A), MCP-1 (B), IL-1ra (C), and TFG-1 (D) in BAL fluids of healthy control subjects, patients with no lung inflammation, and patients with lung inflammation. BAL fluid samples included in these analyses were selected randomly on the basis of availability and did not correspond precisely to those included in DNA array studies reported in this article. The number of samples tested in control, no inflammation, and inflammation groups in each panel were (A) 13, 22, 32, (B) 9, 17, 16, (C) 9, 10, 21, and (D) 11, 11, 11. In all cases, differences are significant (P  0.05) in patients with lung inflammation compared with combined control group of patients with no lung inflammation and healthy control subjects, two-tailed Student's t test.

Among the abnormally expressed genes were those that encode proteins involved in intracellular signaling. Changes suggested stimulation of nuclear factor (NF)-B, G-protein, Notch, and calmodulin-dependent pathways, with increased expression of the gene for NF-B-inducing kinase and decreased expression of three inhibitory genes, regulator of G-protein signaling 12, lunatic fringe, and neurogranin (Figure 2 and Table 2, genes 6, 51, 52, and 55). Other changes suggested reduced signaling through membrane inositol phospholipid metabolism, mitogen-activated protein kinase pathways, and stress-activated pathways. There was increased expression of the gene for diacyclglycerol kinase  and decreased expression of inositol 1,4,5-trisphosphate 3-kinase and cytohesin-1 (sec7) genes, as well as decreased expression of genes for extracellular signal-related kinase 1 and rac3, which activates c-jun (see Figure 2 and Table 2, genes 16, 70, 66, 60, and 61). Reduction in expression of genes for the purinergic receptor P2X4 and the voltage-dependent dihydropyridine-sensitive L-type Ca2+ channel  unit (Figure 2 and Table 2, genes 56 and 58) suggested reduced signaling through nonselective ion channels. Genes involved in gene transcription and translation and cell cycle were also dysregulated (see Figure 2 and Table 2, genes 1, 3, 5, 7, 11, 21, 24, 30, 36, 54, 62, and 74).

Other changes in gene expression suggested activation of phagocytes. There was increased expression of genes associated with phagocytic cell functions of opsonization (increased C1q), vesicle transport (amphiphysin, lethal giant larvae homolog 2, chromosome 3p25 membrane protein and Rab2), respiratory burst (annexin VI), and migration (actin-related protein Arp2) (see Figure 2 and Table 2, genes 2, 10, 8, 9, 13, 14, and 23, respectively). Of interest was a reduction in expression of genes involved in the processing of antigens presented in the context of major histocompatibility complex class I molecules (tapasin and low-density lipoprotein-related protein), but an increase in expression of antigen-presenting cell surface molecules involved in delivering T cell costimulation (intercellular adhesion molecule 1 [CD154] and adhesion molecule- activated leukocyte cell adhesion molecule [CD166]) (see Figure 2 and Table 2, genes 53, 69, 77, and 79).

Significant increases in expression of genes for proinflammatory cytokines often associated with macrophage activation were not seen. For example, the gene for TNF had a 2.5-fold reduction, not stimulation, in expression. Unexpectedly, changes were seen in expression of certain genes that might dampen inflammatory responses. Expression of genes for several protective proteins, apolipoprotein J (clusterin), and the antioxidant metallothionein were increased, and expression of the proinflammatory receptor for advanced glycosylation end-product-specific receptor was decreased (see Figure 2 and Table 2, genes 17, 19, and 49). Follistatin gene expression was reduced (Figure 2 and Table 2, gene 74). This secreted molecule binds and serves as an antagonist to activin A (19), a member of the TGF- family that has anti-inflammatory properties and has been implicated in pulmonary fibrosis and lung remodeling (20). Of note was an increase in expression of the gene IL-1ra (Figure 2 and Table 2, gene 84). An increase in IL-1ra has been reported in idiopathic interstitial fibrosis (21). ELISA was done to confirm that IL-1ra protein was increased in scleroderma patients with lung inflammation, compared with patients without lung inflammation and healthy control subjects. An increase was seen, with mean ± SD values of 17.4 ± 12.6 pg/ml versus 11.0 ± 6.9 pg/ml, P = 0.05, two-tailed Student's t test (see Figure 3C).

The group of selected genes was reviewed for those that might directly stimulate extracellular matrix production in patients with lung inflammation, but none were identified. It was noted, however, that expression of TGF-1 mRNA was increased 1.5-fold in patients with lung inflammation, although expression of this gene was not identified as significantly increased using SAM, Mann-Whitney analyses, or fold differences. Because of the important role of TGF-1 in animal models of lung fibrosis, total TGF-1 protein was measured in BAL fluids from patients and control subjects. Levels of TGF-1 protein were increased in some patients with lung inflammation, compared with the combined control group, with mean ± SD values of 5.6 ± 6.2 pg/ml versus 1.6 ± 0.6 pg/ml, P = 0.05, two-tailed Student's t test (see Figure 3D).

Two individual patients had arrays whose patterns of gene expression more closely resembled those of the other group. Array NI7 (see Figure 2) showed a pattern of gene expression more similar to that seen in patients with lung inflammation, although that patient had a normal BAL cell differential, with 1.0% neutrophils and 0.4% eosinophils. After a 21-mo follow-up, the patient was found on repeat BAL to have now developed lung inflammation, with 2.2% neutrophils and 5% eosinophils. During that interval, the patient had significant decline in lung function, with 15% decline in FVC and 37% decline in DLco. Conversely, array I10 (see Figure 2) from a patient with lung inflammation showed a pattern of gene expression more similar to that of patients without lung inflammation. This patient had stable lung function tests over a 28-mo follow-up, with 7% decline in FVC and 4% increase in DLco, without cyclophosphamide therapy.

One patient had two BAL samples, obtained 20 mo apart, before (Figure 2, array I9) and after (Figure 2, array I8) 18 mo of cyclophosphamide therapy, with the second BAL done 2 mo after cyclophosphamide was stopped. Lung inflammation persisted on the second BAL sample. Hierarchical cluster analyses were done by arrays, rather than genes, and showed that the two arrays from this patient were more closely related to each other than to other arrays (not shown). The correlation coefficient of levels of expression of abnormally expressed genes on the two arrays was 0.86.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Lung inflammation, as assessed by BAL neutrophilia or eosinophilia, is associated with increased risk of progressive restrictive lung disease (2, 3). This suggests that the inflammation itself is causally associated with progressive lung fibrosis, because it has been implicated in bleomycin-induced lung fibrosis (22), or that lung inflammation is triggered by a process that also causes lung fibrosis. The hypothesis of the study was that cell-mediated immunity with chronic macrophage activation occurs in the lungs of scleroderma patients with BAL neutrophilia or eosinophilia (lung inflammation). This was addressed by analyses of gene expression profiles in BAL cells from patients and healthy subjects. We speculate further that activation of cell-mediated immunity and macrophage activation are directly involved in the genesis of progressive lung fibrosis in the scleroderma.

The use of unfractionated BAL cells allowed assessment of overall changes in gene expression and emphasized dysregulation of gene expression in alveolar macrophages, which were ~ 90% of cells in BAL samples from all groups. It is unlikely that the observed changes in gene expression simply reflected a difference in cell populations. The only differences in cell differentials between patients with lung inflammation and a combined control group of patients without lung inflammation and healthy control subjects were a higher percentage of neutrophils and eosinophils, which together made up a small portion of the total cells, ~ 5%. The data were normalized to total spot intensities on the arrays, which would minimize differences caused by minor differences in cell differentials. Many of the differences in gene expression reflect cellular activation. In addition, 2-fold or greater increases or decreases in expression of those genes that are largely expressed by alveolar macrophages could not be explained by differences in cell populations in BAL samples.

Dysregulation of gene expression in BAL cells from patients without lung inflammation appeared very limited, when compared with healthy individuals. This is consistent with the clinical course in this group of patients, with most of these patients having stable lung function over time (2). The few changes in gene expression that were seen might contribute to the development of underlying autoimmunity in scleroderma patients and be unrelated to the development of lung fibrosis. Inhibition of apoptosis can lead to failure to remove autoimmune cells, causing prolonged autoimmune stimulation (23). In scleroderma patients, reduced expression of TNF receptor (TNFR)1 suggests possible suppression of the TNF-TNFR pathway of apoptosis. Lack of B7 costimulation has been reported to promote autoimmunity in animals, perhaps through defective regulation of autoreactive T cells (24), and is found in humans at risk for the development of type I diabetes, along with reduction in B7.2 expression (25). Because B7 molecules are largely expressed by antigen-presenting cells, and alveolar macrophages are the predominant cell type in BAL cell samples, it is likely that the alveolar macrophages were the cellular source of reduced B7.2 gene expression.

In contrast, dysregulation of gene expression was more blatant in patients with lung inflammation, when compared with patients without lung inflammation and healthy control subjects. Increases in expression of chemokine and chemokine receptor genes were prominent. Of particular interest was the increase in expression of PARC and CXCR4 genes. Monocyte-to-macrophage differentiation is a requisite for PARC expression (26), and PARC is induced in macrophages by Th2-associated cytokines interleukin (IL)-4, IL-13, and IL-10, and inhibited by interferon- (26). Its expression is largely restricted to the lungs, and it specifically recruits T cells (16). There is a single receptor on lymphocytes for PARC (16), which is CCR3 (27). PARC can inhibit effects of other CC chemokines that bind CCR3 (eotaxin, eotaxin-2, MCP-3, and RANTES) (27). Thus, PARC may play a particularly important role in scleroderma lung disease, as a chemoattractant responsible for T cell recruitment and as a regulator of leukocyte movement into the lungs (27). The chemokine receptor CXCR4 is expressed on higher levels on type 0 and type 2 T cells, and its expression is stimulated by IL-4 and inhibited by interferon- (28). An increase in its expression is consistent with our previous report that types 0 and 2 T cells are increased in BAL samples from scleroderma patients, especially those with lung inflammation (11).

The changes in chemokine gene expression in patients with lung inflammation (as defined by neutrophilia or eosinophilia) are consistent with activation of alveolar macrophages. Many of the changes are similar to those seen when the macrophage cell line Thp1 is activated by infection with an intracellular pathogen, Mycobacterium tuberculosis (29). That model and our results have in common increased expression of PARC, myeloid progenitor inhibitory factor-1, MCP-1, MIP-1, MIP-1, osteopontin, GRO-, and GRO-. Although the level of expression of some of these chemokine genes did not met the criteria for final selection in this report, review of the dataset showed that expression tended to be increased for each, with 2.2- to 4.4-fold increases in expression.

Several CC chemokines may play a bifunctional role in the development of lung fibrosis, both in recruitment of effector cells and in stimulation of collagen production. MCP-1 protein is increased in BAL fluids and in the lungs of humans with idiopathic pulmonary fibrosis, where alveolar macrophages are a source (30, 31). In scleroderma, elevated serum levels of MCP-1 are associated with pulmonary fibrosis (32). In a rodent model of fibrotic versus nonfibrotic pulmonary granulomas, procollagen production in fibroblasts from diseases with Th2-type T cells is dependent on MCP-1 production (33). MCP-1 stimulates collagen production in rat lung fibroblasts, a process that may involve paracrine stimulation of TGF- production in fibroblasts by MCP-1 (33). Preliminary data from our lab (not shown) suggests that PARC may share with MCP-1 the ability to stimulate collagen production by lung fibroblasts.

The increase in expression of the gene for NF-B- inducing kinase, which is required for phosphorylation, release, and degradation of IB, suggests that activity of NF-B signaling pathways may be increased in scleroderma patients with lung inflammation. NF-B signaling pathways are involved in a variety of proinflammatory processes, including chemokine production, and blocking NF-B signaling pathways can reduce inflammation and lung fibrosis in bleomycin-induced pulmonary disease (34). However, increased expression of the gene for the proinflammatory cytokine TNF-, whose production is stimulated by NF-B signaling pathways, was not seen. Instead, anti-inflammatory cytokines IL-1ra and TGF- were increased. These findings suggest that the mechanisms of lung disease in scleroderma differ from typical delayed-type hypersensitivity responses in which macrophage activation with excessive TNF- production is driven by type 1 interferon--producing T cells. It suggests instead that type 2 cytokine-producing T cells may be inhibiting macrophage production of TNF- and stimulating production of anti-inflammatory cytokines. Increased IL-1ra and TGF-1 expression occur in both idiopathic and bleomycin-induced pulmonary fibrosis (32), and some (9), but not all (35) other investigators have also found increased TGF-1 levels in BAL samples from scleroderma patients. These results suggest that increased production of the immunosuppressive cytokines IL-1ra and TGF-1 is common to fibrotic lung processes of different etiologies, including scleroderma.

Our results provide the first panoramic view of changes in gene expression in the lungs of scleroderma patients. Refinement of and additions to this first picture will come with analyses of fractionated BAL cell populations, which will be particularly useful as they add data from less-frequently occurring cell populations, which may be masked in studies of unfractionated cells. For example, similar analyses of isolated T cells from the lungs of scleroderma patients might provide critical information about changes in T cell functions that may drive downstream processes such as macrophage activation and fibrosis.

This study was not designed to address the clinical usefulness of gene expression profiling in scleroderma lung disease. However, clear-cut differences in gene expression were identified on cluster analyses when the group of patients with lung inflammation, as identified by BAL cell neutrophilia or eosinophilia, were compared with the group of patients without lung inflammation (Figure 2). Exceptions to this were noted in two individuals, in whom the pattern of gene expression more closely resembled that of the other group of patients and was matched by subsequent changes in lung function that were more typical of the other group. Additional studies in more patients are needed to determine whether there is indeed an association between the pattern of gene expression in BAL cells and subsequent clinical course. If this association holds true, then it may become realistic to combine information on expression of a selected set of genes in BAL cells with other clinical information to improve stratification of scleroderma patients into risk groups for progressive lung inflammation, and it may prove valuable to follow the pattern of gene expression in BAL cells to monitor response to therapy.