We used an integrative genomics approach to define the characteristics of macrophage activation in human smokers and in two transgenic mouse models of emphysema. We found that cigarette smoking had a remarkably potent and consistent effect on macrophage activation in human subjects, a finding that was not observed in subjects with asthma. Among the most highly induced genes in human smokers was MMP-12, a proteinase believed to be important in the pathogenesis of emphysema on the basis of mouse models. However, many of the gene expression changes seen in smokers would not have been predicted from analyses of the two mouse models of emphysema. For example, we found that osteopontin was highly induced in macrophages from smokers but not in the mouse models, and that increased expression levels correlated with lung function impairment in smokers. These analyses indicate that habitual cigarette smoking has a distinctive and reproducible effect on macrophage activation and identify specific genes that may contribute to the pathogenesis of COPD.
We found that the effects of cigarette smoking on alveolar macrophage gene expression were reproducible and distinctive despite a very conservative approach to data analysis. We performed hierarchic clustering of samples according to expression of the most variable genes to determine the predominant sources of variation in gene expression in an unbiased manner (without using information about smoking status). We found that all 15 smokers completely segregated from all 15 nonsmoking control subjects (), but that the 15 subjects with asthma could not be distinguished from control subjects (). This indicates that smoking status accounted for much of the observed variation in gene expression and that smoking induces a dramatic and highly consistent change in macrophage gene expression patterns. To identify differentially expressed genes, we applied the conservative Bonferroni approach to adjust for the very large number of probe sets on the microarrays. Even using this conservative approach, we identified 110 genes as differentially expressed in smokers versus nonsmoking control subjects. By contrast, only 10 genes were differentially expressed in macrophages from subjects with asthma, and none of these genes overlapped with findings in smokers. Quantitative PCR confirmed differential expression of genes in smokers identified by microarray. There are likely to be additional genes induced by smoking that were not identified in microarray analyses due to our conservative approach to data analysis and to the limitations of microarrays for detection of low abundance transcripts. In support of this idea, we used real-time PCR to analyze expression of proteinases not identified as differentially expressed in array studies and found evidence that some of these (MMP-2, MMP-9, and cathepsin K) were increased in smokers (data not shown). Although microarray-based studies have certain intrinsic limitations in sensitivity, we were still able to find clear and convincing evidence for extensive smoking-induced changes in macrophage gene expression using this approach.
We compared macrophage gene expression changes in human smokers with those seen in two transgenic mouse models of emphysema to identify common features that might be involved in emphysema pathogenesis. Previous work suggests that development of emphysema in these two mouse models is driven by very different mechanisms. IL-13–overexpressing mice develop emphysema very rapidly and have granulocytic inflammation and increased TGF-β levels (44
), whereas Itgb6−/−
mice develop emphysema more slowly, do not have granulocytic inflammation, and are deficient in functional TGF-β (22
). Given these differences, we were surprised to find that the pattern of macrophage activation was strikingly similar in these two models ( and ). In contrast, the degree of similarity between human smokers and the mouse models was much more restricted ( and ). There are inevitable limitations inherent in comparing expression microarray data between species. Different microarrays must be used for human and mouse samples, and it is not always possible to unambiguously identify mouse orthologs of human genes. Furthermore, our between-species comparison may be affected by technical differences in the way that mouse and human macrophages were collected and purified and in the methods used to amplify mouse and human macrophage mRNAs. Nonetheless, our finding that there were almost as many discordant changes as concordant changes shows that there are major differences between the human and mouse macrophage activation states. On the one hand, this finding focuses attention on a relatively small number of genes that could have a role in emphysema pathogenesis in both human smokers and in mouse models of emphysema. On the other hand, the finding helps to identify unique features of smoking-induced macrophage activation that may not be modeled by these transgenic systems.
One of the genes induced both in human smokers and in mouse models was the proteinase MMP-12. Using arrays with more than 50,000 probe sets, MMP-12 was identified as the third most highly induced gene in alveolar macrophages from smokers, and we confirmed previous reports that MMP-12 is highly induced in each of the mouse models (21
). MMP-12 is required for the development of emphysema in Itgb6−/−
mice and in cigarette smoke–exposed mice (20
). However, whether MMP-12 contributes to human emphysema has been uncertain, due in part to conflicting evidence about whether this proteinase is induced in human smokers (10
). Using quantitative PCR, which has a very large, dynamic range, we measured an even greater degree of induction of MMP-12 transcripts in alveolar macrophages from smokers than estimated by the arrays (41-fold; ). However, we, like some others (10
), were unable to reliably detect MMP-12 protein or activity by casein zymography, Western blotting, or fluorokine multi-analyte profiling assays with antibody-coated microparticles performed on BALF concentrated using Microcon filters (data not shown). In a recent report (23
), MMP-12 protein was detected in highly concentrated lyophilized BALF from subjects with COPD. Thus, we are uncertain whether the large and consistent smoking-induced increases in MMP-12 transcript levels that we observed lead to increased MMP-12 protein expression or function. In addition, our data suggest that increased MMP-12 expression may not in itself be sufficient to cause emphysema because expression was increased in all smokers independent of the degree of impairment in lung function (). Other genes induced in both human smokers and in mouse models include platelet-activating factor acetylhydrolase (PLA2G7, the second most highly induced gene in smokers), monocyte chemoattractant protein 1 (CCL2), and chemokine receptor 5 (CCR5; ). Mouse models will likely continue to be useful tools for helping to investigate the possible contributions of these genes to emphysema.
Many of the gene expression changes induced by smoking were not apparent in either of the transgenic mouse models. A notable example is the enzyme CYP1B1, the most highly induced gene in smokers. This cytochrome P450–family protein is known to be induced by cigarette smoke (37
) and produce carcinogenic DNA adducts in the process of detoxifying combustion products in smoke. Another gene induced only in smokers was glutathione reductase (2.1-fold induced, adjusted p < 0.01), which plays an important role in protection against oxidative stress in human alveolar macrophages (46
).We found that IGF-1 transcripts were decreased in smokers but increased in mouse models and confirmed these changes at the protein level in BALF. IGF-1 inhibits apoptosis of lung epithelial cells (47
), suggesting that decreases in IGF-1 could contribute to the increased epithelial cell apoptosis seen in human emphysema (48
). This pattern of expression also differentiates smoking-induced alveolar macrophage activation from findings in other lung diseases such as idiopathic pulmonary fibrosis, which are characterized by increased macrophage expression of IGF-1 (50
). Another distinctive gene expression change observed in smokers that contrasted sharply with mouse models was increased expression of osteopontin. BALF protein analysis confirmed increased expression of osteopontin in smokers and decreased expression in Itgb6−/−
mice. Osteopontin is a multifunctional protein that has been implicated in macrophage recruitment (53
) and plays a role in osteoclast differentiation and activation (55
), which suggests parallels between smoke-exposed alveolar macrophages and osteoclasts involved in degrading bone matrix. The role of each of these smoking-induced gene expression changes requires further study in humans and in suitable model systems.
We found that habitual cigarette smoking induces a remarkably consistent pattern of gene expression changes in all of the subjects that we studied, but epidemiologic data suggest that only a subset of smokers develop COPD (57
). We speculate that risk and susceptibility factors for COPD could function at three levels. First, the risk of COPD is related to the dose and duration of smoking exposure (58
), suggesting that COPD risk may relate to the cumulative effects of the consistently observed changes in macrophage gene expression in smokers. Second, there may be differences between individuals in the response of specific genes to smoking. For example, we found that osteopontin was increased in all smokers, but that the extent of this increase was greater in patients with more severe lung function impairment (). Further studies involving subjects with more extreme lung function abnormalities might allow for the identification of other macrophage gene expression changes that are selectively associated with the development of COPD in a subset of smokers. Third, it seems likely that susceptibility to COPD may relate to differences in the response of the lung to smoking-induced macrophage activation. For example, MMP-12 (which was elevated in smokers without apparent relation to lung function) might produce emphysema more readily in smokers with lower levels of antiproteinase activity. Our findings demonstrate that macrophage function is markedly altered by smoking and provide a useful framework for future studies that focus on contributions of specific macrophage gene expression changes to the pathogenesis of smoking-induced lung disease in humans.