In this study, we made a number of novel observations. First, we identified a common set of 101 genes that was differentially regulated in endotoxin exposed vs. control (PBS or air exposed) mice across all published models of recurrent endotoxin exposure. A number of genes previously identified as being important in either the pathogenesis or severity of COPD and asthma were present in the signature. In addition, there are a number of other genes identified not previously associated with obstructive lung disease, and may represent candidate genes for further investigation. Of particular interest is the ability of this 101 gene signature to accurately classify endotoxin, smoke, and endotoxin plus smoke exposed mice. While there appear to be some similarities in differential gene expression in response to endotoxin, to cigarette smoke, or to cigarette smoke in conjunction with endotoxin, whether this translates into a similar phenotype of obstructive lung disease cannot be concluded from this observation.
Comparison of these studies at the pathway level revealed additional similarities across these experiments. Notably, when looking within the same experimental model, longer duration of endotoxin exposure led to increased enrichment of inflammatory pathways, which is in contrast to prior studies suggesting that endotoxin tolerance develops in response to repeated endotoxin challenge 
. While development of tolerance may be related to a number of factors such as the method or route of exposure, it deserves further study as chronic inflammation is thought to play an important role in both asthma and COPD.
While the 101 genes identified represents a small number of genes that are differentially expressed across experimental models as compared to the number of genes differentially expressed within any of the identified experimental models, factors that have previously been identified as being important in the biological response to endotoxin and likely contributed to the heterogeneity in response include significant differences in strain of endotoxin, strain of mouse, exposure protocol, and local practices of each lab, and likely contributed to differentially expressed genes that were not conserved across experimental models 
. We evaluated the significance of these 101 genes on a number of levels. First, using hypergeometric tests, we demonstrated that this common intersect is highly unlikely due to chance alone. Second, examination of the 101 genes revealed the presence of lbp and cd14,
which are well described components of the endotoxin signaling pathway, thus affirming the biologic relevance of this gene signature. Third, this gene signature was able to accurately classify between endotoxin and phosphate buffered saline exposed mice in a distinct experiment not used to generate the gene signature. Finally, at the pathway level, we observed that increasing duration of endotoxin exposure led to increasing enrichment of inflammatory pathways; this was confirmed in a recent publication from our group 
, where increasing duration of inhaled endotoxin exposure was associated with increased IL-6 and decreased IL-10 concentrations in lung homogenate, consistent with a persistent pro-inflammatory profile. The increase in inflammation with prolonged endotoxin exposure was associated with and potentially mediated by an expansion of lung dendritic cells and a reduction in macrophages; thus it was not surprising to see that within the 101 gene signature, there were a number of genes important in antigen presentation that were differentially expressed (such as h2ab-1
which encodes for MHC class II, myeloid chemokines such as ccl6
cathepsins such as ctsz, ctss, ctsb,
and C-type leptin-like receptors such as clec4a2, clec4n,
). Recent human studies have demonstrated the accumulation of dendritic cells in COPD, with an association between disease severity and level of dendritic cell accumulation. 
On examination of the genes present in the 101 gene signature (Tables S1 and S2
), a number of asthma associated genes were present, including chi3l1
, which was identified in one of the first genome-wide association (GWAS) studies of asthma 
, and il33
, which has been shown in a number of experiments to be important in asthma development 
and asthma severity 
. Interestingly, interleukin-33 has also been found to enhance the endotoxin response of macrophages 
. Several COPD associated genes were present, including mmp-12,
which has been identified in both murine studies as being associated with the development of emphysema 
as well as in human studies as being associated with the risk of COPD development in smokers 
. Of further interest was the identification of fpr2 and saa3
. Serum amyloid protein (SAA) has previously been considered solely an acute phase reactant, and while saa1
are expressed primarily in liver and kidney 
is expressed in the lung and has only recently been identified as important in the pathogenesis of glucocorticoid refractory COPD by opposing organ protective signaling by lipoxins at the ALX/FPR2 receptors 
While inhaled endotoxin exposure as a model for de novo COPD development rather than COPD exacerbations has received little attention, from epidemiologic studies it is clear that between a quarter to a half of patients with COPD have never smoked 
. The phenotype of non-tobacco induced COPD as compared to tobacco-related COPD remains poorly studied. The third National Health and Nutrition Examination Survey (NHANES III) has suggested that non-smokers account for 24.9% of COPD cases in the United States 
; in this study many subjects with non-tobacco COPD previously had a physician diagnosis of asthma. The multi-center, international BOLD study 
confirms these findings, estimating that between a quarter to a fifth of all patients with COPD are nonsmokers. Indoor biomass fuel exposure and occupational exposure to biologic or organic dusts in the workplace, both of which has been associated with high levels of endotoxin exposure, were associated with non-tobacco COPD. As in the NHANES study, self-reported physician diagnosis of asthma was a strong predictor of non-tobacco COPD. While this may represent disease misclassification by physicians, as the existence of non-tobacco COPD is not widely appreciated, it is also possible that this may relate to the underlying phenotype of non-tobacco COPD that is different from tobacco-related COPD.
The ability of the 101 gene signature to accurately classify between endotoxin, smoke, and endotoxin plus smoke exposed mice, and not just mice exposed to air vs. mice exposed to any endotoxin (whether as endotoxin alone, tobacco smoke [which contains endotoxin], or endotoxin in addition to tobacco smoke) is intriguing. It suggests that genes selected for differential expression between endotoxin and control also play an important role in differential expression between various endotoxin containing exposures. Further examination of expression patterns of the 101 gene signature in this comparison indicate that (Figure S3
, which is associated with asthma, was upregulated only in endotoxin exposed mice. Conversely, mmp12
was significantly upregulated in all exposure groups although average log2
fold change for smoke vs. control was 4.76, for LPS vs. control was 2.05, and for smoke+LPS vs. control was 2.33. MMP-12 does not appear to have an important role in endotoxin induced inflammation 
, but the interaction between smoke and endotoxin has not been well studied. Of note mmp12
has been associated with the emphysema subtype of COPD 
. If these gene expression changes are reflected in human exposures and affect downstream clinical phenotypes, it is possible that endotoxin-related COPD has a phenotype more consistent with small airways disease rather than parenchymal disease as seen in tobacco-related emphysema. A further potential implication of the observed differences in mmp12
expression is that smokers who have recurrent bacterial infections (and thus are exposed recurrently to endotoxin) may be more likely to develop a predominantly airways disease subphenotype of COPD rather than a predominantly emphysema subphenotype of COPD 
Strengths of this work include the approach to identify the consistency of gene expression across experimental models, the use of consensus clustering to validate the importance of the identified gene signature against a randomly picked set of genes, and the potential biological applications of the gene signature. This is the first paper to assess the importance of the endotoxin component of cigarette smoke as an exposure using genomic techniques.
We acknowledge that there are several limitations to this work. These studies were performed in murine models and may not be translatable to human disease. Additionally, all of these studies were performed on lung homogenates, and it is difficult to distinguish which cell population contributed to the gene expression signature. Differences between duration of endotoxin exposure were not explicitly addressed as we were looking for agreement across studies; prior work has demonstrated that there are changes in short term vs. long term exposure. 
Finally, to verify the biologic importance of any one of the identified genes, additional functional work is needed.
To date all treatment trials of COPD have required prior significant tobacco use as an inclusion criteria, and thus we know little about the efficacy of COPD therapies in non-tobacco related COPD. While our work suggests that the endotoxin component of cigarette smoke may be important in disease development, the effect of additional endotoxin in conjunction with tobacco exposure leads to different gene expression changes compared to endotoxin alone. The differentially expressed genes in response to repeated endotoxin exposure that we identified have been implicated in both asthma and COPD, and based on our pathway analysis, chronic inflammation plays a significant role. There may be other biologically-targeted therapies that may have additional benefit in endotoxin-related obstructive lung disease. Prognosis and treatment implications of this disease may or may not differ from tobacco-related COPD, and deserves further study.