Paracrine signalling during ZEB1-mediated epithelial–mesenchymal transition augments local myofibroblast differentiation in lung fibrosis

The contribution of epithelial–mesenchymal transition (EMT) to human lung fibrogenesis is controversial. Here we provide evidence that ZEB1-mediated EMT in human alveolar epithelial type II (ATII) cells contributes to the development of lung fibrosis by paracrine signalling to underlying fibroblasts. Activation of EGFR–RAS–ERK signalling in ATII cells induced EMT via ZEB1. ATII cells had extremely low extracellular matrix gene expression even after induction of EMT, however conditioned media from ATII cells undergoing RAS-induced EMT augmented TGFβ-induced profibrogenic responses in lung fibroblasts. This epithelial–mesenchymal crosstalk was controlled by ZEB1 via the expression of tissue plasminogen activator (tPA). In human fibrotic lung tissue, nuclear ZEB1 expression was detected in alveolar epithelium adjacent to sites of extracellular matrix (ECM) deposition, suggesting that ZEB1-mediated paracrine signalling has the potential to contribute to early fibrotic changes in the lung interstitium. Targeting this novel ZEB1 regulatory axis may be a viable strategy for the treatment of pulmonary fibrosis.


Introduction
Epithelial-mesenchymal transition (EMT), a dynamic and reversible biological process by which epithelial cells lose their cell polarity and down-regulate cadherinmediated cell-cell adhesion to gain migratory properties, is involved in embryonic development, wound healing, fibrosis, and cancer metastasis 1 . EMT is executed in response to pleiotropic signalling factors, including the transforming growth factor β (TGFβ) superfamily, Sonic Hedgehog (Shh), Wnt/β-catenin, fibroblast growth factor (FGF) and epidermal growth factor (EGF). These factors regulate the expression of specific transcription factors (TFs) called EMT-TFs (e.g., Snail, ZEB, Twist, and others) that promote repression of epithelial features and induction of mesenchymal characteristics 2,3 . Unlike EMT in cancer, which is detrimental, wound-healing-driven EMT induced in response to injury is beneficial, but exaggerated healing responses can lead to fibrosis, or tissue scarring.
Fibrosis is a hallmark of many chronic degenerative disorders and is associated with reduced organ function and eventual organ failure. Fibrotic disease is on the increase; for example, idiopathic pulmonary fibrosis (IPF), the most common type of idiopathic interstitial pneumonia, occurs with similar frequency to that of stomach, brain, and testicular cancer 4 . IPF is now generally regarded as a consequence of multiple interacting genetic and environmental risk factors, with repetitive local micro-injuries to ageing alveolar epithelium playing a central role 5 . These microinjuries initiate the progressive accumulation of extracellular matrix (ECM) deposited by myofibroblasts. The origin of these myofibroblasts has been debated for many years, with EMT being considered as a potential source by driving the transformation of epithelial cells into ECM producing myofibroblasts 6,7,8,9,10 . However, lineage tracing in transgenic mice indicates that the contribution of those cells to the population of myofibroblasts is negligible 11,12,13,14 .
In this study, we identify a novel regulatory axis involved in lung fibrosis whereby EMT contributes to the fibrotic process via paracrine activation of fibroblasts. We demonstrate that epidermal growth factor receptor (EGFR)-RAS-extracellular signalregulated kinase (ERK) signalling induces the transcription factor ZEB1, which not only controls EMT but also regulates the production of locally-acting mediators.
Specifically we identified tissue plasminogen activator (tPA) as a downstream effector of ZEB1 transcriptional activity that contributes to paracrine signalling by enhancing TGFβ-induced profibrogenic responses in fibroblasts. Consistent with this, increased ZEB1 nuclear expression was detected in alveolar epithelium adjacent to sites of ECM deposition in IPF lung tissue. Thus, rather than contributing directly to the mesenchymal population, our data suggest that ZEB1-dependent EMT of ATII cells contributes to fibrosis via epithelial-fibroblast crosstalk. The occurrence of ZEB1 activation at sites of local ECM deposition in IPF lung tissue is consistent with the concept that ZEB1-regulated paracrine signalling contributes to development of a profibrogenic microenvironment leading to interstitial lung fibrosis.

Activation of EGFR signalling induces EMT in alveolar epithelial cells.
To investigate IPF associated signalling pathways, we analysed differentially expressed genes in IPF and control lung tissue from a publicly available microarray dataset (GSE24206) 15 . Using a false discovery rate (FDR) corrected P value of 0.05, we identified 7668 genes to be differentially expressed out of a total of 54675 probe sets. Gene network analysis using the Consensus Pathways Database 16  Based on the transcriptomic data, we hypothesised an important role of EGFR signalling in IPF. Identification of pathological mechanisms of IPF has been challenging; however, dysregulation of alveolar type 2 (ATII) epithelial cells is thought to be central 5 . We therefore treated a human ATII cell line (ATII ER:KRASV12 ) 17,18 with EGF ( Fig. 1b-d; Supplementary Fig. S1b) or transforming growth factor α (TGFα) ( Supplementary Fig. S1b) to activate EGFR signalling. The human ATII cell line grows in continuous culture and expresses the ATII cell marker, prosurfactant protein C (ProSP-C) ( Fig. 1a; Fig. 2f). Our results showed that treatment of ATII ER:KRASV12 cells with EGF for 24 hrs induced EMT, reflected by a change in their morphology from typical cuboidal epithelial cells to a more elongated mesenchymal cell phenotype with a reorganization of the actin cytoskeleton as demonstrated using Phalloidin staining of filamentous actin (F-actin) (Fig. 1b). This phenotypic switch was accompanied by a significant increase in mRNA expression of ZEB1 and VIM (Vimentin), and a reduction in CDH1 (E-cadherin); mRNA levels of other EMT-TFs, such as SNAI1, SNAI2, TWIST and ZEB2 were not increased by activation of EGFR 6 signalling (Fig. 1c). The changes in ZEB1 and E-cadherin were further confirmed by Western blot analysis ( Fig. 1d; Supplementary Fig. S1b).
Similar results were obtained using primary human ATII cells treated with EGF where an increase in ZEB1 expression was associated with down-regulation of Ecadherin (Fig. 1e). Under the same conditions, however, TGFβ was not able to induce EMT in the primary human ATII cells (Fig. 1e). Together, these results demonstrate that activation of EGFR signalling is able to activate the EMT programme in ATII cells, which is supported by a morphology change, the induction of the EMT-TF ZEB1 and a mesenchymal marker Vimentin as well as a reduction in E-cadherin expression.

Activation of the RAS pathway drives EMT via ERK-ZEB1 in ATII cells.
RAS signalling is one of the most important pathways downstream of EGFR activation and is involved in a variety of physiological and pathological responses, including EMT 19,20,21 . To investigate whether the RAS pathway is important for EMT in ATII cells, we utilised a RAS-inducible ATII cell model. KRASV12 (containing a single amino acid mutation in KRAS, glycine to valine at position 12) fused to the oestrogen receptor (ER) ligand-binding domain 22 was introduced into ATII cells to generate ATII ER:KRASV12 , in which KRASV12 expression is induced by 4-hydroxytamoxifen (4-OHT) 17,18 . Like EGF, direct activation of the RAS pathway in ATII ER:KRASV12 cells by treatment with 4-OHT induced EMT, reflected by a reduction in E-cadherin levels and an increase in ZEB1 and Vimentin expression ( Fig. 2a and   b). Time-course analysis further demonstrated that the induction of ZEB1 by RAS 7 activation preceded the down-regulation of E-cadherin (Fig. 2c). Consistently, an EMT morphology change with an increase in ZEB1 expression was observed upon RAS activation (Fig. 2d). When grown on a thick layer of Matrigel, ATII cells form spheres (a 3D culture model) 23 . We adopted this experimental system and used ATII ER:KRASV12 cells to investigate whether RAS activation induces EMT in 3D cultures. Control ATII ER:KRASV12 cells formed single round spheres. Induction of oncogenic KRAS by 4-OHT resulted in spheres invading into the Matrigel with protrusions ( Fig. 2e; Supplementary Fig. S2). We recovered these cells from the Matrigel, and examined the protein expression. We confirmed that RAS activation induced EMT in 3D cultures, demonstrated by a reduction in E-cadherin, and an increase in ZEB1 and Vimentin expression (Fig. 2f). These observations suggest that EGFR signalling and the downstream RAS pathway are able to induce EMT in ATII cells.
Since RAS activity regulates both the RAF-ERK and phosphoinositide 3-kinase (PI3K)-protein kinase B (AKT) signalling pathways, we next investigated which one is required for EMT in the ATII cells using inhibitors for these pathways. Treatment with the ERK inhibitor U0126 in ATII ER:KRASV12 cells was sufficient to inhibit RASinduced ZEB1 and Vimentin expression, as well as to restore the expression of Ecadherin and the epithelial morphology; in contrast, the AKT inhibitor AKT VIII failed to do so ( Fig. 3a and b; Supplementary Fig. S3a).
We next investigated which EMT-TFs are important for RAS-induced EMT in ATII cells. ZEB1 RNA interference (RNAi), but not Snail1 or Snail2 RNAi, was able to restore E-cadherin expression and the epithelial morphology in 4-OHT-treated ATII ER:KRASV12 cells ( Fig. 3c and d; Supplementary Fig. S3b), in line with the fact the ZEB1 was the only EMT-TF induced by EGFR-RAS signalling ( Fig. 1 and 2). Taken together, our results identify that RAS activation in human ATII cells drives EMT via ERK-ZEB1 pathway.
ZEB1 is highly expressed in IPF alveolar epithelium and is critical for transcriptional regulation of secreted factors that mediate crosstalk between ATII cells and fibroblasts.
Given our in vitro findings, we compared ZEB1 expression in IPF and control lung tissue. In IPF tissue, we detected strong nuclear expression of ZEB1 not only in fibroblastic foci ( Fig. 4a) but also in epithelial cells of thickened alveoli septae where collagen deposition in the interstitium was evident (Fig. 4b); in contrast, little ZEB1 staining or collagen deposition was observed in alveoli of control lung tissue (Fig.   4c). The presence of nuclear ZEB1 staining in alveolar epithelial cells within IPF lung tissue suggests that these cells are undergoing EMT; furthermore, the presence of ECM suggests induction of mesenchymal responses, either directly via the repairing epithelial cells undergoing EMT or by crosstalk with underlying fibroblasts.
Comparison of the relative expression of ECM components in RAS-activated ATII ER:KRASV12 cells and fibroblasts highlights that ATII cells produce extremely low levels of ECM genes even after the induction of EMT ( Supplementary Fig. S4a), suggesting that ECM production in fibrosis is more likely to be a consequence of fibroblast activation than direct deposition by epithelial cells undergoing EMT.
Therefore we investigated whether ATII cells undergoing RAS-induced EMT produce paracrine factors that activate fibroblasts. For these experiments, we took advantage of the ability of 4-OHT to induce RAS pathway activation in ATII ER:KRASV12 cells, as this was not dependent on exogenous growth factors that might directly affect fibroblast responses. We treated the MRC5 or primary human parenchymal lung fibroblasts with conditioned media (CM) from control or 4-OHT-treated ATII ER:KRASV12 cells in the absence or presence of TGFβ, and evaluated the fibroblast responses by measuring the expression of α-smooth muscle actin (α-SMA, a myofibroblast marker) and other ECM genes, including COL1A1, COL3A1 and FN1.
On its own, CM from RAS-activated ATII ER:KRASV12 cells (4-OHT-treated ATII CM) had little effect on the activation of fibroblasts (Fig. 5). However, 4-OHT-treated ATII CM together with TGFβ achieved a synergistic effect in activating fibroblasts, reflected by a larger increase in α-SMA (ACTA2), COL1A1 and FN1 levels ( Fig. 5a and b). Of note, 4-OHT-treated ATII CM did not augment Smad2 phosphorylation suggesting a Smad2-independent response ( Fig. 5a and c). Similar results were obtained using primary human lung fibroblasts from IPF patients (IPF fibroblasts, IPFFs) and control donors (normal human lung fibroblasts, NHLFs) (  By performing quantitative proteomic analysis of the CM from control or 4-OHTtreated ATII ER:KRASV12 cells, we identified ~ 430 secreted proteins whose levels changed during RAS-induced EMT. We then checked their expression in pulmonary epithelial cells from control and IPF lung tissue using a publicly available dataset 24 , and identified a total number of 25 genes/proteins that were elevated in IPF lung epithelial cells as well as in CM from 4-OHT-treated ATII ER:KRASV12 cells (Supplementary Table S1). Of these, PLAT, which encodes tissue plasminogen activator (tPA) was most up-regulated in IPF epithelial cells ( Fig. 7a; Supplementary Table S1) and we confirmed enhanced secretion of tPA in the CM from 4-OHTtreated ATII ER:KRASV12 cells by Western blotting (Fig. 7b). As we had identified ZEB1 as the key regulator of epithelial-mesenchymal crosstalk, we scanned the promoter of PLAT for the presence of ZEB1 binding motifs (5'-CANNTG-3') and found a ZEB1 binding site -419 bp upstream of the transcriptional start site (TSS) ( Supplementary   Fig. S6a). Further experiments showed that the mRNA expression of PLAT was increased upon RAS-activation in ATII cells and this was repressed by ZEB1 RNAi (Fig. 7c).
To validate the ZEB1 binding site in the PLAT promoter, we first performed a chromatin immunoprecipitation (ChIP) assay. An anti-ZEB1 antibody was used to precipitate formaldehyde cross-linked ZEB1-DNA complexes in ATII ER:KRASV12 cells treated without or with 4-OHT. The presence of PLAT promoter DNA sequences in the immunoprecipitate was verified by PCR using primers amplifying the region between -547 and -345 upstream of the TSS, and we found RAS activation in ATII cells increased ZEB1 occupancy on the PLAT promoter ( Fig. 7d; Supplementary Fig.   S6b). We next generated two PLAT promoter constructs (-689 to -1 upstream of the TSS) which were cloned into a pGL3 basic luciferase reporter plasmid and transfected into ATII cells; the pGL3 basic-PLAT (−689 to −1) construct contained the ZEB1 motif whereas this was deleted in the second construct (delta −419 to −414 upstream of the TSS) (pGL3 basic-Δ ZEB1 motif). RAS activation by 4-OHT in ATII ER:KRASV12 cells resulted in a significant increase in pGL3 basic-PLAT (−689 to −1) luciferase activity. Under the same conditions, luciferase activity was not increased using pGL3 basic-Δ ZEB1 motif (Fig. 7e). These data confirm that PLAT (tPA) is a transcriptional target of ZEB1 in response to RAS activation in ATII cells.
Consistent with a previous report 25 , tPA synergistically promoted TGFβ-induced α- Finally, in view of the requirement for exogenous TGFβ to demonstrate an effect of the 4-OHT-treated ATII CM on fibroblasts, we investigated whether ATII cells in fibrotic tissue in vivo or those undergoing injury/repair in vitro expressed endogenous TGFβ. Using a publicly available dataset 24 , we found that the major TGFB isoform expressed by alveolar epithelial cells in vivo was TGFB2 and that this was expressed at significantly higher levels in IPF compared with control lung tissue ( Supplementary   Fig. S7a). In contrast with the study in kidney 14 , the data also revealed that Snail2 is up-regulated in IPF vs. control lung epithelial cells, but not Snail1 or Twist ( Supplementary Fig. S7b). As we have previously shown that scrape-wounding of bronchial epithelial cells stimulates release of TGFβ2 independently of EGFR activation 26 , we examined whether damage of ATII cells similarly affected TGFB2 expression. This showed that scrape-wounded ATII cells expressed more TGFB2 and this increased in proportion to the extent of injury ( Supplementary Fig.   S7c). These data suggest that damaged ATII cells are a potential a source of TGFβ in vivo.

Discussion
Fibrotic diseases are a major cause of morbidity and mortality worldwide and their prevalence is increasing with an ageing population. Abnormal wound healing responses appear to make major contributions to the scarring process, but the underlying pathological mechanisms are unclear, especially the role of EMT. In this study, we have used a variety of approaches to show that activation of EGFR-RAS-ERK signalling in ATII cells induces EMT via the transcriptional regulator ZEB1.
Importantly, beyond its effects on the epithelial cell phenotype, we have identified that ZEB1 is a regulator of paracrine signalling between lung epithelial cells and fibroblasts, as ATII cells undergoing RAS-induced EMT secrete tPA to augment TGFβ-induced myofibroblast differentiation (Fig. 8). This may be an important profibrotic event as, relative to epithelial cells, the ability of fibroblasts to synthesise ECM is orders of magnitude greater.
Consistent with previous findings 27, 28 , we found strong expression of ZEB1 in the epithelium in proximity to fibroblastic foci in IPF lung tissue. However, we also found ZEB1 was expressed in epithelial cells of thickened alveolar septae where ECM deposition was evident. This suggests that ZEB1 is induced as an early response to alveolar epithelial injury and that, by regulating expression of factors involved in paracrine signalling, ZEB1 may promote TGFβ-induced fibroblast activation in IPF.
While this may be a normal physiological response to injury, persistent epithelial injury and/or failure to resolve the lesion may sensitise the underlying fibroblasts to drive a pathologic profibrogenic response. In line with this, exposure of human lung cells to nickel (Ni), an environmental and occupational pollutant linked to lung fibrosis 29 , caused ZEB1-dependent EMT, which was irreversible even after the termination of Ni exposure 30 . Thus, it is conceivable that repetitive environmental exposures to metals such as Ni could lead to deregulation of ZEB1 to cause persistent EMT and exaggerated profibrogenic crosstalk during the initiation of IPF.
EMT in the ATII cells was strongly induced by EGFR activation. The EGFR is a transmembrane receptor tyrosine kinase activated by members of the EGF family, including EGF and TGFα 31 . EGFR dimerisation activates one or more downstream effectors including the ERK, PI3K/AKT, STAT (signal transducer and activator of transcription), and mTOR (mammalian target of rapamycin) pathways through receptor autophosphorylation and cytoplasmic protein binding 32, 33, 34 . These in turn act as critical mediators of airway and alveolar homeostasis, with aberrant activation within one or more pathway components capable of driving a variety of respiratory pathologies, including lung fibrosis 33, 35 . The EGFR pathway has been implicated in lung fibrosis through studies in which transgenic mice that constitutively express TGFα in epithelial cells develop progressive lung fibrosis 36, 37 . Conversely, mice deficient in TGFα that lack normal EGFR signalling or that are treated with EGFR pathway inhibitors exhibit resistance to bleomycin-induced lung fibrosis 38 . In IPF patients, EGFR mutations 39 or increased expression of the EGFR 40 have been reported. Our evidence that an EGFR-RAS-ERK-ZEB1 axis may contribute to the early stages of lung fibrosis suggests that inhibiting EGFR signalling may be of clinical relevance for regulating human fibrotic lung disease.
A key finding of our study was identification that ZEB1 controls tPA expression and that this affects the sensitivity of fibroblast activation induced by TGFβ. While tPA is a key activator of fibrinolysis, it also has direct cellular effects by virtue of its ability to bind to the low density lipoprotein (LDL) receptor-related protein-1 (LRP-1), triggering LRP-1 tyrosine phosphorylation, and recruitment of β1-integrin signalling involving integrin-linked kinase (ILK) 25 . In this context, tPA acts as a However, as these paracrine effects required exogenous TGFβ, this raised the question of the source of TGFβ in lung fibrosis in vivo. While many cell types produce TGFβ isoforms which can also be stored as latent growth factor bound in the ECM 46, 47, 48 , we focussed on the epithelium and found increased expression of TGFB2, as well as SNAI2 in IPF epithelial cells using publicly available datasets, with scrape wounding of ATII cells also inducing TGFB2 expression in vitro. The increased epithelial TGFB2 signature highlights the potential for EGF (ZEB1) and TGFβ (Snail2) to synergize in paracrine activation of the underlying fibroblasts.
Together with previous findings in kidney fibrosis 14,43 , our study helps to provide a Samples were observed using a confocal microscope system (Leica SP8). Acquired images were analysed using Photoshop (Adobe Systems) according to the guidelines of the journal.

Immunohistochemistry, haematoxylin and eosin (H/E) and tinctorial stains.
Control or IPF lung tissues (n = 3 donors) were fixed and embedded in paraffin wax; tissue sections (4µm) were processed and stained as previously described 20  Raw data were processed and collated into a single .csv document. Values were then normalised to total fmol of each sample multiplied by 10,000. Pseudo-counts were applied to the normalised values to replace missing ones, to allow for full statistical analysis to be completed 53 . We first sorted the normalised values in each column in order of abundance, in ascending order, then the minimum value of each sample identified. This minimum was used to replace all missing values in the data set. A two-tailed, unpaired Student's t-test was used to compare two groups for independent samples. P < 0.05 was considered statistically significant.
In order to highlight their implications in IPF, differentially expressed proteins/genes identified in the quantitative secretome analysis were searched in LGEA web portal (https://research.cchmc.org/pbge/lunggens/mainportal.html) for their levels in pulmonary epithelial cells from control and IPF lung tissue.

Bioinformatics. IPF transcriptomic data was downloaded from the NCBI's Gene
Expression Omnibus (GEO). We used data from GSE24206 15 , a microarray study comparing samples from 11 IPF patients undergoing lung transplantation or diagnostic biopsy to 6 normal lung samples taken from lung transplantation donors.
Microarray series matrix files were imported into R, and differential expression 23 analysis comparing normal to IPF samples performed using the R package limma 54 .
Data were log-transformed before analysis. To correct for multiple testing, a Benjamini-Hochberg false discovery rate (FDR) of 5% was applied to the data, and a Q-value cut-off of 0.02 was used to determine significance. Differentially expressed gene lists were input into the human Consensus Pathways Database, which determined pathways with differentially expressed genes overrepresented in this database. A 5% FDR was used as above.

Statistical analysis and repeatability of experiments. Each experiment was
repeated at least twice. Unless otherwise noted, data are presented as mean and s.d., and a two-tailed, unpaired Student's t-test was used to compare two groups for independent samples. P < 0.05 was considered statistically significant.

Supplementary Methods
Quantitative proteomic analysis of the secretome.       Monolayers of ATII ER:KRASV12 cells were scrape wounded with 8 or 12 scratches and allowed to repair for 24 hrs. Data are mean ± s.d. n = 3 samples per group. ** P < 0.01. *** P < 0.001.