Meprin β contributes to collagen deposition in lung fibrosis

Lung fibrosis is a severe disease characterized by epithelial cell injury, inflammation and collagen deposition. The metalloproteases meprinα and meprinβ have been shown to enhance collagen maturation and inflammatory cell infiltration via cleavage of cell-cell contact molecules; therefore we hypothesized that meprins could play a role in lung fibrosis. An exhaustive characterization of bleomycin-treated meprinα, meprinβ and the double meprinsαβ knock-out (KO) with respective wt-littermates was performed by using several different methods. We observed no difference in lung function parameters and no change in inflammatory cells infiltrating the lung between wt and all meprins KO mice after 14 days bleomycin. No difference in epithelial integrity as assessed by e-cadherin protein level was detected in bleomycin-treated lungs. However, morphological analysis in the bleomycin-treated mice revealed decrease collagen deposition and tissue density in meprinβ KO, but not in meprinα and meprinαβ KO mice. This finding was accompanied by localization of meprinβ to epithelial cells in regions with immature collagen in mice. Similarly, in human IPF lungs meprinβ was mostly localized in epithelium. These findings suggest that local environment triggers meprinβ expression to support collagen maturation. In conclusion, our data demonstrate the in vivo relevance of meprinβ in collagen deposition in lung fibrosis.

These findings suggest that meprins can be important in the onset of fibrosis, contributing to epithelial layer disruption, inflammatory cell infiltration and collagen maturation. However, their role in lung fibrosis has not been investigated so far. Hence, the aim of the current study is to delineate the contribution of meprins to the onset of pulmonary fibrosis and to investigate potential underlying molecular mechanism.

Meprins are expressed in inflammatory and epithelial cells of human lungs.
We first assessed the localization of both meprins α and β by immunohistochemical staining in human donor and IPF lungs. We observed that meprins expression localized to epithelial and inflammatory cells (Fig. 1a). To confirm the epithelial cell localization of meprins we performed immunohistochemical staining on serial slides with pro-surfactant protein C (pro-SPC, marker for alveolar type II epithelial cells) and smooth muscle actin (SMA; marker for smooth muscle cells and myofibroblasts). The staining revealed that meprins are mostly localized to pro-SPC positive cells (Fig. 1a, arrow) and inflammatory cells (Fig. 1a, arrowhead), while myofibroblasts were negative for meprins (Fig. 1a). PCR analysis revealed that meprins are expressed at very low levels in human lungs; however meprin β level is higher compared to meprin α (as indicated by their Δ Ct values) (Fig. 1b). We did not detect any substantial change of meprins expression in lung homogenate of IPF in comparison to donor samples (Fig. 1b). All immunohistochemistry controls are shown in Supplementary Figure S1. Meprin β is regulated by TGF-β1. As analysis of total lung homogenate may mask cell specific changes in meprin expression, we examined how pro-fibrotic and pro-inflammatory factors could directly affect their expression in epithelial cells. Following stimulation of A549 cells with TGF-β 1, we observed that meprin β , but not meprin α , was upregulated upon 48 h (Fig. 2a,c), suggesting an indirect regulation upon activation of TGF-β 1 pathway. TNF-α and short time course with TGF-β stimulation did not influence the expression of either meprins (Fig. 2b,d and Supplementary Figure S2). Meprins are important for epithelial monolayer integrity. Meprins have previously been shown to cleave E-cadherin, a cell-cell contact molecule, in colon and kidney epithelial cells 17 . Disruption of the epithelial integrity is considered to be an early event in lung fibrosis 16 ; therefore we tested whether meprins can alter the A549 cell monolayer in vitro. Our permeability assay showed that meprin α and meprin β treatment resulted in a significant increase of cell permeability (Fig. 3a). Positive controls are depicted in Supplementary Figure S3. Immunofluorescence staining confirmed the weaker signal of E-cadherin (Fig. 3b), indicating that meprins can be involved in cell-cell contact protein degradation and thereby in the disruption of epithelial layer integrity.
Meprin β localisation changes upon bleomycin treatment. In order to assess meprins expression in the bleomycin model of lung fibrosis, we performed immunohistochemical staining. Similar to the human data, meprin α and meprin β were localized mostly to epithelial cells and to inflammatory cells ( Fig. 4a,b, arrows and arrowheads respectively). Interestingly, after 14 days bleomycin administration, the localization of meprin β was restricted to specific areas of the lung parenchyma (Fig. 4b, arrow). To verify the cellular localization of meprins, we performed serial slide staining with pro-SPC, α -SMA and von Willebrand Factor (marker for endothelial cells). In bleomycin treated mice, we observed that meprin β is mostly express in pro-SPC positive cells (Fig. 4c). Meprin α showed a ubiquitous expression in the lung, irrespectively of the treatment and a very weak staining indicating low expression (Fig. 4c).

Lack of meprins do not result in improved lung function in bleomycin treated mice. To test
whether meprins have a role in fibrosis development in vivo, we subjected wt, meprin α , β and α β KO mice to bleomycin. Histological analysis, revealed no apparent differences between wt mice and meprins KO mice (Fig. 5). Measurement of the lung function of the mice, revealed a decrease of compliance (Fig. 6a) and total lung capacity ( Fig. 6b) upon bleomycin treatment, however no differences were observed between bleomycin treated wt littermates and meprins KO mice irrespectively of genotype.
Absence of meprins does not prevent E-cadherin loss in bleomycin challenged mice. As it has been reported that epithelial cells downregulate E-cadherin expression in fibrosis 20,21 and meprins can influence its abundance, we analysed the expression of E-cadherin in in vivo settings upon saline and bleomycin treatment. No changes in E-cadherin mRNA level were observed between saline and bleomycin treated mice, irrespectively of the genotype (Fig. 7a). Western blot analysis revealed a decrease of E-cadherin in bleomycin wt mice in comparison to wt saline treated mice (Fig. 7b); however we did not observe any apparent difference of E-cadherin protein level between different genotypes ( Fig. 7b and Supplementary Figure S4).  23 . Hence, we analysed the inflammatory cells in the BALF of saline and bleomycin wt, meprin α , meprin β and meprin α β KO mice. As depicted in Fig. 8a, we observed an increase in inflammatory cells in the BALF of bleomycin treated mice in comparison to saline; however deletion of meprins did not affect this increase. To rule out the possibility that the inflammatory infiltrate composition is different, we performed FACS analysis from the BALF. Although we observed several changes such as increase of macrophages, T and B-cells ( Fig. 8b-i) due to bleomycin application in wt mice in comparison to wt saline mice; we could not detect any significant change among different genotypes ( Fig. 8b-i). These results suggest that meprins do not influence the inflammatory composition in the lung after 14 days of bleomycin treatment.
Meprin β KO mice have less collagen deposition in comparison to wt littermates. Meprins can also facilitate collagen maturation 3 ; we therefore investigated whether the collagen amount is altered in meprins KO mice in comparison to wt mice upon bleomycin treatment. Quantification of collagen deposition was performed on masson's trichrome stained slides. As expected, bleomycin treatment increased the amount of collagen in comparison to saline mice (Fig. 9a). Interestingly, we observed that meprin β KO mice upon bleomycin treatment had a lower amount of collagen in the lung in comparison to bleomycin treated wt littermates (Fig. 9a). The decreased collagen deposition in meprin β KO mice was concomitant with lower tissue density in comparison to wild type mice (Fig. 9b). Additionally, Sirius red stained slide subjected to polarized light confirmed a diminished birefringence signal in bleomycin treated meprin β KO mice compared to wt animals challenged with bleomycin suggesting a decrease collagen amount ( Fig. 9c and Supplementary Figure S5). Expression of collagen I and III was elevated upon bleomycin in comparison to saline, but no differences were observed between different genotypes, indicating that post-translational modification is responsible for the reduction in collagen content in meprin β KO mice ( Supplementary Fig. S6a,b). Another protease which belongs to meprin family, bone morphogenetic protein-1 (BMP-1), can also enhance maturation of collagens 24 . We therefore determined whether the expression of BMP-1 was enhanced in KO mice and could potentially compensate for the absence of meprins. However, we did not observe any difference in the expression of BMP-1 in saline or bleomycin treated meprins KO mice (Supplementary Figure S7).
Meprin β localizes in regions of the lung with immature collagen. To better assess the role of meprins with collagen maturation, we performed Alcian Blue combined with Elastica and van Gieson's staining in order to visualize the mature and immature collagen. Serial slides were stained for meprin β to assess possible co-localization of immature collagen with meprin β . Interestingly we observed that regions with mature

Discussion
In this study we have investigated the role of meprin α and meprin β proteases in development of lung fibrosis. We could show in vitro that meprin α and meprin β are involved in injury of the epithelial layer integrity and that they are mostly expressed by epithelial cells. Additionally, meprin β but not meprin α contributes to collagen deposition in vivo. Several other metalloproteinases (MMP2, MMP3, MMP7), have been shown to be elevated in BALF of IPF patients and to correlate with the decline of the lung function [25][26][27] . However, it is becoming clearer that their expression level has to be considered together with their cellular and tissue localization 28 . MMP-1 was found to be elevated in IPF patients, even though its main function is degradation of collagen 27,29 . Although it sounds paradoxical, further analysis revealed that MMP-1 is upregulated in reactive epithelium and not in the lung interstitium where the collagen is deposited 27 . In this study we could clearly show that meprins are expressed by epithelial and inflammatory cells of the mouse and human lung. Injury of the epithelial layer is an early event in lung fibrosis 10,16 , which is characterized by loss of cell-cell contact molecules, such as E-cadherin 30 . In line with previous studies 17, 18 , we could show that meprins are involved in disturbance of the lung epithelial cell layer integrity in vitro. In vivo, we did not observe any apparent differences in E-cadherin protein levels in meprins KO mice in comparison to wt. As E-cadherin is also degraded by other proteases 31,32 , it is likely that other molecular pathways take part in E-cadherin cleavage in bleomycin induced lung fibrosis. Degradation of epithelial cell-cell contact protein is followed by inflammatory cell recruitment in the tissue, which influences the progression of fibrosis 33 . Meprins KO animals have been shown to have deficient extravasation of inflammatory cells to the intestinal mucosa in inflammatory bowel disease mouse model 22 . In our study we could not observe any difference in number and type of cells infiltrating the lung, suggesting that meprins are not essential for the cell extravasation in the lung upon bleomycin treatment at the investigated time point. The infiltrated inflammatory cells release several pro-fibrotic chemokines which promote collagen production 28,34 .
Here we could show in vivo that bleomycin treated meprin β KO mice accumulate less collagen than bleomycin treated wt littermates. These findings suggest a potential role of the metalloprotease meprin β in collagen deposition in lung fibrosis. Interestingly, we did not observe any difference in lung function parameters in bleomycin treated meprin β KO mice in comparison to wt. As the decline of lung function in pulmonary fibrosis is a result of the epithelial injury, inflammatory influx and collagen deposition 35 , we could speculate that in meprin β KO mice change in collagen amount, but not in inflammation, was insufficient to improve the lung function parameters. Interestingly, in bleomycin treated mice and in human lung fibrosis, meprin β displayed a well localized and definite expression pattern to epithelial cells. The localization of meprin β on epithelial cells and the reduction of collagen in the meprin β KO lung, suggest that meprin β in pathological condition could take part to collagen maturation as a protection mechanism against the epithelial cell injury. This was further supported by the change of meprin β localization in saline and bleomycin condition, suggesting a switch on/off mechanism of meprin β expression depending on the need of producing collagen. This mechanism might fuel the perpetual injury and collagen maturation cycle which contribute to the progression of the disease 36 . To support this idea, we showed that immature collagen and pro-collagen I localized in close proximity to meprin β . However, in regions of mature collagen, meprin β was found to be very weakly or not expressed, indicating that meprin β expression is very dynamic and is regulated depending on the need of the surrounding environment. Interestingly, the two meprins have different localization which suggests different behaviour in pathological condition. Although in  human lung samples both meprins are expressed and in vitro studies have shown that both meprins are capable of maturating collagen I and III 2 , our animal experiments showed that meprin β , but not meprin α , has in vivo relevance and contributes significantly to collagen maturation in the lung. Broder et al. have previously reported that meprin α and meprin β KO mice have less collagen in the skin and therefore less tensile strength 1 . The different level of meprin α expression between skin and lung could explain the lack of phenotype in meprin α KO mice. Both meprins are weakly expressed in the mouse lungs; particularly with meprin α being often at the detection limit (data not shown). As meprin α is very low expressed in mouse lung, we would expect the α β KO mice to have a similar phenotype as meprins β KO mice in term of collagen amount. Interestingly, no difference in collagen amount has been noticed in meprins α β KO mice, which could be explained with other proteases being activated as a compensatory mechanism. Several proteases, such as BMP-1 24 , ADAMTSs 37 or furin like pro-protein convertase 38 are involved in pro-collagen processing. Even though we did not observe any difference in the mRNA expression level of BMP-1, we cannot exclude that BMP-1 or other protease´s activity is more pronounced in the α β KO mice.
Taken together our study assessed the role of meprins in lung fibrosis development. We showed that both meprin α and meprin β are involved in disturbance of epithelial layer integrity in vitro, which might represent one of the trigger mechanisms for the fibrosis development. Importantly, we could prove that meprin β is involved in collagen deposition in vivo in the lung of bleomycin treated mice and it localized in region with immature collagen. This suggests that meprin β can favour the progression of the fibrotic process enhancing collagen processing and deposition.  Lung function measurement. Analysis of the respiratory mechanics was performed using FlexiVent system (SCIREQ, Paris, France). Mice were intraperitoneal anesthetized with ketamine (15 mg/ml) and xylazyne (2 mg/ml). Tracheotomy was performed and mice were connected to the FlexiVent system to perform measurement of the respiratory mechanics. Mice were ventilated at a rate of 150 breaths/min with a positive end-expiratory pressure (PEEP) of 3 cmH 2 O. After stable ventilation was achieved three cycles of deep inflation, snapshot, prime-wave (quickprime-3) and pressure-volume loop (PVs-P) perturbations were applied for each mouse. All perturbations were automatically adjusted to the body weight of the mouse by the Flexiware software 7.6. Three measurements were averaged to give the represented the value of one biological sample. After measurement, the mice were euthanized and the blood, broncho-alveolar lavage fluid (BALF) and lungs of all animals were collected for molecular and histological analysis. FACS analysis. BALF samples were subjected to cytometric analysis. All inflammatory cells were first selected using a CD45 + gate. Staining for neutrophils (CD11b + , CD11c − , Gr-1 + ), alveolar macrophages (CD11b low , CD11c + , SiglecF + ), dendritic cells (CD11b+ , CD11c + , CD24 + MHC-II high ), T helper cells (CD3 + , CD4 + ), cytotoxic T cells (CD3 + , CD8 + ), B cells (CD19 + ), eosinophils (CD11b + , CD11c − , SiglecF + ), interstitial macrophages (CD11b + , CD64 + SiglecF − ) and Ly6C/G + monocyte macrophages (CD11b + , CD64 − , Ly6C/G + ) was performed and assessed by a LSRII flow cytometer 41 . Analysis was carried out with the FACSDiva software (both from BD Biosciences, Heidelberg, Germany).   (Sigma) was added to the upper compartment of the insert. The fluorescence from the bottom well was measured with a fluorescent plate reader (FLUOstar Omega, BMG LABTECH, Ortenberg, Germany) (Exc 544 nm, Em 590 nm). TRITC-dextran was added on a well without cells (called no monolayer, nML) to give the maximum fluorescent signal. This signal was used for gain adjustment of the detector sensitivity, in order to achieve the correct sensitivity.

Material and Methods
Immunohistochemistry and tissue staining. Formalin-fixed paraffin embedded lung tissues were cut to 2.5 μ m thick sections and stained with either Haemalaun-eosin, Masson's trichrome, used for immunohistochemistry, Sirus red or Alcian Blue combined with Elastica and van Gieson's staining.
To determine tissue density and collagen content Masson's trichrome slides were subjected to semi-automated image analysis software Visiopharm (Hoersholm, Denmark). Collagen content was calculated as a percentage of the total analysed area, as previously described 42 . Tissue density was calculated as a percentage of total tissue to the total analysed area. In both analysis the whole left and right lung were considered.
Sirius Red staining was performed according to manufactory's instruction. The stained slides were subjected to bright field and polarized light (Olympus) for qualitative analysis of the collagen using the microscope Olympus BX51.
Alcian Blue combined with Elastica and van Gieson's staining was performed on lung section from bleomycin treated wt mice. Previous studies have already shown the capability of identification of immature collagen with acid fuchsin 43,44 . This staining highlights: nuclei (dark brown), elastic fibres (blue-violet), mucus (blue), collagens (red), muscle (ochre) and immature collagen (greenish grey).
Immunofluorescence. For immunofluorescence analysis, A549 grown on 8-well chamber slides were fixed for 20 min in cold methanol at − 20 °C and rinsed five times in PBS. Afterwards, the cells were permeabilized for 5 min with 0.5% TritonX-100 in PBS at RT, washed twice with PBS and incubated overnight at 4 °C with anti-E-cadherin (1:500; U3254, Sigma). Cells were then washed three times with PBS and incubated with anti-rat conjugated with Alexa Fluor ® Dye 555 (Thermo Scientific) for 1 h at RT in the dark. Cells were washed three times in PBS and mounted with fluorescence vectashield mounting medium with 4′ ,6-diamidino-2-phenylindole dihydrochloride (DAPI) for the counterstaining (Vector). For microscopic inspection, an Olympus fluorescence microscope was used.
Statistical analysis. Data are presented as mean ± standard error of the mean (SEM) for all experiments.
Statistical analysis was performed in GraphPad Prism 5 software. For comparison of two groups, unpaired t-test was carried out. Differences between more than two groups were assessed by one-way ANOVA followed by Bonferroni post-hoc test. Differences in time course stimulation experiments were assessed with one-way ANOVA followed by Dunnet post hoc test. For the permeability assay, T-test with wilcoxon matched pair signed rank was used. Significance is indicated by an asterisk on the graphs. A p value less than 0.05 was considered significant for all analysis. All experiments were designed with matched control conditions to enable statistical comparison.