Comprehensive Landscape of Matrix Metalloproteinases in the Pathogenesis of Idiopathic Pulmonary Fibrosis

*3.1. MMPs Expression Significantly Altered in IPF* This study included four independent gene expression data (GSE32537, GSE47460, GSE150910, and GSE124695) from GEO database, including two sets of high-throughput sequencing data and two sets of microarray data, a total of 296 normal subjects and 431 IPF patients (Table S1). There were significant differences in MMPs expression between the healthy and IPF, while the MMPs expression profiles from the four datasets were sufficient to clearly distinguish between the normal and IPF in the PCA (Figure 1A–D). The majority of MMPs were obviously upregulated in IPF and the alterations of MMPs were much consistent across the microarray and high-throughput sequencing technologies. MMP1, MMP10, MMP11, MMP12, MMP13, MMP14, MMP16, MMP2, MMP21, MMP3 and MMP7 were consistently up-regulated in the four sets of data, while MMP28 was consistently down-regulated in the four sets of data. Some others, such as MMP15, MMP17, MMP19, MMP25, MMP8 and MMP9, were uncertain (Figure 1E). To further confirm whether the changes in the MMPs are an important feature that distinguishes IPF disease from other interstitial lung diseases, we introduced gene expression data from chronic hypersensitivity pneumonia (CHP) and acute lung injury (ALI). Even though persistent chronic hypersensitivity pneumonia and repetitive acute lung injury are the main initiative event of IPF, surprisingly, the expression of MMPs didn’t show significant change in CHP (Figure 1F) or ALI (Figure 1H) patients. Further, the empirical cumulative distribution curve analysis determined that the abnormal expression of MMPs is specifically present in IPF (Figure 1G,I). This suggests that the MMPs plays a key role in the pathogenesis of IPF. *3.2. MMPs Are Significantly Correlated with Clinical Features of IPF* To study the correlation between the expression of MMPs and the clinical features, further analysis was performed on lung function (diffusing capacity of the lungs for carbon monoxide, DLCO, forced vital capacity, FVC, etc.), susceptibility and alveolar surface density. Principal component analysis was performed to derive a single measure of meta lung function representing a combination of multiple pulmonary function parameters (Figure S1A). Next, to consider the potential confounding demographic variables, we performed linear multiple regression analysis to assess the correlated gene expression levels with meta lung function variables accounting for age and sex. Most of MMPs (MMP1, MMP7, MMP28, MMP11, MMP10, MMP14, MMP3, MMP12, MMP9, MMP15, MMP16) are closely related with meta lung function (Figure 2A). 13 of MMPs (MMP8, MMP9, MMP12, MMP16, MMP14, MMP3, MMP11, MMP10, MMP7, MMP1, MMP28, MMP15, MMP25) show a dramatically association with lung function and can be the potential diagnostic markers. MMP19, MMP26, MMP17 and MMP21 have no correlationship with lung function (Figure 2B). MMP1 and MMP7 have a significant positive correlation with meta lung function which are currently the most promising diagnostic markers (Figure 2C).Currently homozygous (TT) for the minor allele of the MUC5B polymorphism (rs35705950) possesses higher susceptibility and quicker progress of the disease than heterozygous (GT) [6]. The association between MUC5B mutation and the expression of MMPs by expression quantitative trait locus (eQTL) analysis in all samples demonstrated MMP3 and MMP7 expression were closely related to MUC5B rs35705950 (Figure 2D), indicating that the susceptibility of pulmonary fibrosis may also be mediated by the MMPs. In other words, the MUC5B polymorphism (rs35705950) homozygous (TT) is susceptible to fibrosis due to its ability to up-regulate the expression of MMPs (Mainly MMP7). Besides, MMP2 was strongly associated with the MUC5B polymorphism homozygous (TT) as well in IPF patients. The increase of alveolar surface density is often considered as the gold standard of pulmonary fibrosis. Further analysis of the relationship between MMPs and cellular lung density exhibited that most persistent upregulated MMPs in IPF were accompanied by the increase of alveolar surface density, and MMP1, MMP10, MMP11, MMP12, MMP13, MMP14, MMP16, MMP21, MMP3 and MMP7 increased significantly. There are also genes whose expression decreases as alveolar surface density increases, such as MMP24, and some MMPs were not associated with alveolar surface density, for example MMP25, MMP28, etc. (Figure 2E).*3.3. MMPs Predicted the Outcome of IPF* To explore whether the alteration of MMPs family in IPF could explain the prognosis of IPF, the erratic gene expression in MMPs within IPF was determined by standard deviation and average expression value (Figure 3A). The correlation between the expression of MMPs and clinical features only in IPF patients was performed. Among them, MMP19 had an incredible negative correlation with the DLCO and FVC only in IPF patients in comparison to the whole group (Figure 2B), suggesting that MMP19 expression was extremely critical to pulmonary fibrosis (Figure 3B). The heterogeneity of MMPs expressions predicted the survival and distinguish high-risk from low-risk of IPF patients. BALF microarray data including a total of 176 IPF patients from three regions were performed. Univariate COX regression analysis showed that several MMPs (MMP25, MMP9, MMP1, MMP7, MMP10, MMP24, MMP8, MMP19) were identified as important risk prognostic factors, and most of them were poor prognostic factors (HR > 1) (Figure 3C). Survival analysis further confirmed that MMP1 and MMP19 were prognostic indicators (Figure 3D). A gene set was optimized to construct a multivariate COX regression model including MMP1, MMP10, MMP8, MMP24 and MMP25 (Figure S1C). The risk score obtained by this model can distinguish the survival status of patients (Figure 3E). Time ROC analysis found that the AUC value was 0.81/0.80/0.82 at 1/2/3 years in Freiburg cohort (Figure 3F), 0.77/0.87/0.92 at 1/2/3 years in SIENA cohort (Figure 3G) and 0.75/0.70/0.68 at 1/2/3 years in LEUVEN cohort (Figure 3H) respectively.*3.4. MMPs Mediate a Variety of Biological Functional Pathways in IPF* MMPs plays a role in many biological processes in tissue injury and repair, such as tissue remodeling and regeneration, wound repair, tissue defense, and immune responses. To further understand the biological pathways which MMPs were involved in, gene set variation analysis (GSVA) was performed to analyse all MMPs based on four datasets (GSE32537, GSE47460, GSE150910, and GSE124695). We set the Pearson correlation coefficient greater than 0.6 and the adjusted P value less than 0.05 in at least three datasets. A total of 37 positively correlated pathways (Figure 4A) and 27 negatively correlated pathways (Figure 4B) were obtained (Table S2). Most of MMPs were involved in the production of extracellular matrix, which was the most fundamental role of MMP family. MMP10 was closely related to cytoskeleton formation and cell migration, and MMP11 was involved in the activation of other matrix metalloproteinases (Figure 4C). MMP2 reflects the classic functions of matrix metalloproteins, and MMP7-related pathways involve a wide biological process such as DNA replication, SUMOylating, apoptosis and cell-cell adhesion etc. (Figure 4D). Most of enriched signal pathways showed that MMP19 is involved in the TNF signaling pathways (Figure 4C, Table S2). 23 positively correlated signaling pathways was identified by correlation analysis between MMPs and signaling pathways in IPF patients (Figure S1D). Notably, MMP7 is no longer associated with biological processes including DNA replication, SUMOylating, apoptosis and cell-cell adhesion, etc. in IPF samples (Figure S1D,E), which are significantly associated in both normal samples and IPF samples (Figure 4A,B). The significance of MMP7 tends to be lower in pathogenesis of IPF, which may be the reason why MMP7 is not finally applied to biomarker in clinical practice by ATS/ERS.*3.5. The Roles of MMPs at the Single Cell Resolution* Many previous studies have demonstrated the important role of the MMPs, but most of them have been limited to the tissue sample and therefore may not explain the corresponding cell type specificity. In this study, we obtained 31 major lung cell populations from GSE135893 (Figure 5A) and screened only the single cell sequencing of normal and IPF patients (Figure 5B). MMP2 tends to be highly expressed in mesenchymal cells (Figure 5C). As shown in Figure 5D, MMP7 is specifically expressed in basal cells and secretory cells, and MMP19 highly expressed in macrophages, in line with our previous prediction based on TNF signaling pathway, and MMP28 expressed in epithelial cells and endothelial cells, which may directly affect the formation of epithelial barrier and the remodeling of alveolar structure. Subsequently, we focused on the abnormal expression of MMPs in the disease state. MMP7 has been recognized to be upregulated in the lung tissue of IPF patient. The profiling of MMPs in single cell demonstrated that MMP7 was indeed highly expressed in most cell types and MMP24 was much lower in all cell types of IPF lung tissues (Figure 5E), suggesting that more attention shoule be paid to the role of MMPs antisense RNA in the pathogenesis of IPF. The single cell sequencing data provide a rational interpretation of the unstable expression of MMP19 at the lung tissues, and MMP19 in myofibroblasts and macrophages of IPF patients is obviously upregulated, but the expression of MMP19 in PLIN2+ fibroblasts is downregulated.*3.6. Immunomodulatory Effects of MMP19+ Macrophages* The upregulated expression of MMP19 in macrophages prompted us to further explore its role in the immunomodulatory. So far, no studies have systematically elucidated the immunomodulatory function of MMP family in IPF. MMP9 and MMP19 in immune cells are upregulated in IPF, and hence it may be the main form of immunomodulatory function of MMPs. We found that MMP9 was specifically expressed in some macrophages, which is consistent with previous studies, showing that MMP9 secreted by macrophages mediates epithelial mesenchymal transition [21], but the immunomodulatory role of MMP19 has not been reported. We firstly divided the macrophages into subgroups according to the alteration of MMP19. We have MMP19+ macrophages and MMP19− macrophage group gene expression spectrum difference analysis to get 76 raised genes (|Log2Fold Change| > 0.5 and P value. Adj < 0.01). Functional enrichment analysis revealed 17 primary pathways. The differences are mainly reflected in Chylomicron remodeling, long-chain fatty acid transport, phospholipase inhibitor activity and chemokine activity (Figure 6A). To further explore how transcription factors were involved in these ways, we integrated the four-transcription database (JASPAR, CHEA, ENCODE, Motif Map) and found that MMP19 was regulated by the 79-transcription factor, at the same time these transcription factors regulating the differentially expressed genes (Figure 6B). Eight transcription factors with significant regulatory effects were selected and all were found to be highly expressed in MMP19+ macrophages (Figure 6C).