Bidirectional epithelial–mesenchymal crosstalk provides self-sustaining profibrotic signals in pulmonary fibrosis

Idiopathic pulmonary fibrosis (IPF) is the prototypic progressive fibrotic lung disease with a median survival of 2 to 4 years. Injury to and/or dysfunction of the alveolar epithelium is strongly implicated in IPF disease initiation, but the factors that determine whether fibrosis progresses rather than normal tissue repair occurs remain poorly understood. We previously demonstrated that zinc finger E-box-binding homeobox 1–mediated epithelial–mesenchymal transition in human alveolar epithelial type II (ATII) cells augments transforming growth factor-β–induced profibrogenic responses in underlying lung fibroblasts via paracrine signaling. Here, we investigated bidirectional epithelial–mesenchymal crosstalk and its potential to drive fibrosis progression. RNA-Seq of lung fibroblasts exposed to conditioned media from ATII cells undergoing RAS-induced epithelial–mesenchymal transition identified many differentially expressed genes including those involved in cell migration and extracellular matrix regulation. We confirmed that paracrine signaling between RAS-activated ATII cells and fibroblasts augmented fibroblast recruitment and demonstrated that this involved a zinc finger E-box-binding homeobox 1–tissue plasminogen activator axis. In a reciprocal fashion, paracrine signaling from transforming growth factor-β–activated lung fibroblasts or IPF fibroblasts induced RAS activation in ATII cells, at least partially through the secreted protein acidic and rich in cysteine, which may signal via the epithelial growth factor receptor via epithelial growth factor–like repeats. Together, these data identify that aberrant bidirectional epithelial–mesenchymal crosstalk in IPF drives a chronic feedback loop that maintains a wound-healing phenotype and provides self-sustaining profibrotic signals.

Idiopathic pulmonary fibrosis (IPF) is the prototypic progressive fibrotic lung disease with a median survival of 2 to 4 years. Injury to and/or dysfunction of the alveolar epithelium is strongly implicated in IPF disease initiation, but the factors that determine whether fibrosis progresses rather than normal tissue repair occurs remain poorly understood. We previously demonstrated that zinc finger E-box-binding homeobox 1-mediated epithelialmesenchymal transition in human alveolar epithelial type II (ATII) cells augments transforming growth factor-β-induced profibrogenic responses in underlying lung fibroblasts via paracrine signaling. Here, we investigated bidirectional epithelialmesenchymal crosstalk and its potential to drive fibrosis progression. RNA-Seq of lung fibroblasts exposed to conditioned media from ATII cells undergoing RAS-induced epithelialmesenchymal transition identified many differentially expressed genes including those involved in cell migration and extracellular matrix regulation. We confirmed that paracrine signaling between RAS-activated ATII cells and fibroblasts augmented fibroblast recruitment and demonstrated that this involved a zinc finger E-box-binding homeobox 1-tissue plasminogen activator axis. In a reciprocal fashion, paracrine signaling from transforming growth factor-β-activated lung fibroblasts or IPF fibroblasts induced RAS activation in ATII cells, at least partially through the secreted protein acidic and rich in cysteine, which may signal via the epithelial growth factor receptor via epithelial growth factorlike repeats. Together, these data identify that aberrant bidirectional epithelial-mesenchymal crosstalk in IPF drives a chronic feedback loop that maintains a wound-healing phenotype and provides self-sustaining profibrotic signals.
Fibrotic diseases are a major cause of morbidity and mortality worldwide, and their prevalence is increasing with an ageing population (1). Development of effective treatments for progressive fibrosis is considered one of the most challenging tasks in modern medicine.
Our studies focus within the lung upon fibrotic interstitial lung diseases, a diverse group of conditions that affect the space between the alveolar epithelium and the capillary endothelium to cause varying degrees of inflammation and fibrosis. Once considered rare, epidemiologic investigations have found them to be more common than previously recognized, with interstitial lung abnormalities detected in 7% of the general population (aged older than 50 years) resulting in an increase in all-cause mortality (hazard ratio = 2.7) (2). Idiopathic pulmonary fibrosis (IPF) is the prototypic chronic progressive fibrotic interstitial lung disease, affecting five million people worldwide (3). Like many fibrotic disorders, IPF is characterized by enhanced deposition and remodeling of the extracellular matrix (ECM). This leads to decreased lung compliance, disrupted gas exchange, and ultimately respiratory failure and death. Median survival from diagnosis is 2 to 4 years, and approved therapies only slow disease progression, so there is a significant unmet need.
An overexuberant wound healing response is regarded as a canonical cause of organ fibrosis (4). Abnormal wound healing may occur as a consequence of multiple stimuli, including infections, chemical exposures, and physical injuries (1,5). Whilst injury to and/or dysfunction of the alveolar epithelium is strongly implicated in IPF disease initiation, the factors that determine whether fibrosis progresses rather than normal tissue repair occurs remain poorly understood.
In this study, we identify that aberrant bidirectional epithelial-mesenchymal crosstalk provides self-sustaining activation signals driving disease progression in pulmonary fibrosis. Building on our previous report (6), we provide evidence that alveolar epithelial type II (ATII) cells undergoing RAS-induced epithelial-mesenchymal transition (EMT) utilize a zinc finger E-box-binding homeobox 1 (ZEB1)-tissue plasminogen activator (tPA) axis to provide paracrine signals that augment fibroblast recruitment and activation. In reciprocal paracrine signaling, transforming growth factor-β (TGF-β)activated fibroblasts or IPF fibroblasts (IPFFs) induce RAS activation in ATII cells, which is at least partially driven by secreted protein acidic and rich in cysteine (SPARC). These findings support the concept that aberrant bidirectional epithelial-mesenchymal crosstalk contributes to the development of a profibrogenic microenvironment in which a chronic wound healing response leads to interstitial lung fibrosis.

Results
Global transcriptomic changes in fibroblasts exposed to conditioned media from RAS-activated ATII cells identify multiple processes including cell migration We previously reported that RAS-induced EMT in the human ATII cell line, ATII ER:KRASV12 (in which RAS activation can be induced by 4-hydroxytamoxifen ) resulted in production of paracrine factors that augmented TGF-βinduced profibrogenic responses in lung fibroblasts (6). To determine if, and how, fibroblasts responded to epithelial signals in the absence of profibrogenic TGF-β signaling, we characterized the global transcriptomic changes in MRC5 lung fibroblasts exposed to conditioned media (CM) from control or RAS-induced ATII cells by performing RNA-Seq.
Genes with a false discovery rate (FDR)-adjusted p value less than 0.05 were considered as differentially expressed genes (DEGs). In total, 966 DEGs were identified, including 491 upregulated (Table S1) and 475 downregulated (Table S2) genes. We then performed Gene Ontology (GO) enrichment analysis of the identified DEGs using PANTHER (http:// pantherdb.org/) (7). The results were grouped into molecular function, biological process, and cellular component. Interestingly, several IPF-related pathological terms were identified, including ECM and collagen, cell migration, and response to cytokine growth factor inflammatory stimulus, which included TGF-β receptor signaling (FDR <0.05; Figs. 1A and S1 ;  Tables S3 and S4). Of these, cell migration and collagen containing ECM were among the top ten ranked GO terms from the biological process category (Fig. 1B). These results extend our previous report (6) by demonstrating that CM from RAS-activated ATII cells regulates the expression of many genes that augment fibroblast activation and migration independent of exogenous TGF-β. Of note, under these conditions, we did not see changes in expression of ACTA2, encoding α-smooth muscle actin (α-SMA) (adjusted p value equals 0.289), suggesting that the fibroblasts did not adopt a myofibroblast phenotype under the conditions tested.
The effect of CM from RAS-induced ATII cells on genes controlling cell migration was further demonstrated using gene set variation analysis (GSVA) with a gene list from GO:0010763 (positive regulation of fibroblast migration); the GSVA score calculated based on this gene list was higher in MRC5 cells incubated with CM from 4-OHT-treated versus control ATII ER:KRASV12 cells (p < 0.05; Fig. 2A). To validate this finding on cell migration, scratch wound assays were performed using MRC5 or IPFFs treated with CM from control or 4-OHT-activated ATII ER:KRASV12 cells. Twenty-four hours after the scratch wound, IPFFs treated with CM from 4-OHT-activated ATII cells almost filled all the gaps, whereas the wound in the control cells was still open (p < 0.01; Fig. 2B). Similar results were obtained using MRC5 lung fibroblasts (Fig. S2A). Under the same conditions, CM from RASactivated ATII cells did not affect the cell viability in MRC5 (Fig. S2B). In addition, in a transwell migration assay, a 3.5-fold increase in migration was detected using MRC5 lung fibroblasts treated with CM from 4-OHT-activated ATII ER:KRASV12 cells compared with those treated with CM from control cells over a 24-h period (p < 0.001; Fig. 2C). Together, these results demonstrate that RAS-activated ATII cells augment fibroblast migration via paracrine signaling.
RAS-activated ATII cells augment fibroblast migration via paracrine signaling, in which ZEB1 and tPA are key regulators As we previously demonstrated that ZEB1 and its transcriptional target, tPA (encoded by PLAT gene) are key regulators of epithelial-mesenchymal crosstalk (6), we hypothesized that these two proteins may also control the ability of RAS-activated ATII cells to promote migration and recruit fibroblasts by paracrine signaling. First, we depleted ZEB1 in ATII ER:KRASV12 cells via RNAi (Fig. S3A). We found that this completely abolished the promigratory effects of CM from RAS-activated ATII cells on lung fibroblasts in both wound healing assay (p > 0.05; Figs. 3A and S3B) and transwell migration assay (p > 0.05; Fig. 3B). Next, we determined whether tPA was a key paracrine regulator of fibroblast migration. Recombinant tPA treatment significantly enhanced fibroblast migration (p < 0.0001; Fig. S3C). Similar to ZEB1, CM was collected from control or PLAT-depleted ATII ER:K-RASV12 cells treated without or with 4-OHT (Fig. S3, D and E). As expected, RAS-activated ATII cells augmented fibroblast migration in both wound healing assay (p < 0.0001; Figs. 3A and S3F) and transwell migration assay (p < 0.0001; Fig. 3B), whereas under the same conditions, depletion of PLAT in ATII ER:KRASV12 cells completely abolished the promigratory effects of CM from RAS-activated ATII cells on lung fibroblasts (p > 0.05; Figs. 3, A and B and S3F). Taken together, these data highlight the importance of ZEB1 and tPA as key regulators of epithelial-mesenchymal crosstalk to promote fibroblast recruitment, as well as fibroblast activation, as demonstrated previously (6) (Fig. 3C).

TGF-β-activated fibroblasts or IPFFs induce RAS activation in ATII cells
Our current and previous (6) findings suggest that a ZEB1-tPA axis is involved in the paracrine signaling between RASactivated ATII cells and fibroblasts to augment not only their recruitment but also their activation in the presence of TGF-β. To explore whether this epithelial-mesenchymal crosstalk was bidirectional, we investigated the effect of the activated fibroblasts on the alveolar epithelium. First, MRC5 lung fibroblasts were left untreated or treated with TGF-β ( Fig. S4A) to induce myofibroblast differentiation, as measured using the myofibroblast marker, α-SMA. An overall increase in α-SMA protein expression manifested after 24 h of TGF-β treatment, with maximal induction being observed by Western blot at 48 h post-TGF-β treatment (Fig. S4B). To eliminate the direct effect of exogenous TGF-β in CM collected from fibroblasts on ATII cells, fibroblast cultures were placed in fresh medium without TGF-β for another 24 h before CM was harvested ( Fig. S4A; group "48 h+, 24 h−"). As shown in Figure S4B, fibroblast differentiation into myofibroblasts was maintained under these conditions, evidenced by a comparable level of α-SMA protein level. We then treated ATII cells with CM from control or TGF-β-activated MRC5 cells (Fig. S4C). Western blot analysis suggested that RAS signaling was activated upon treatment with CM from TGF-β-activated fibroblasts, reflected by significant increases in phosphorylation of AKT and extracellular-regulated kinase (ERK) (Fig. S4D). Of note, TGF-β-activated fibroblast CM also increased ZEB1 expression in ATII cells (Fig. S4D). Since it has been reported that IPFFs have a higher level of autocrine TGF-β activation (8), we next compared the effects of CM from normal human lung fibroblasts (NHLFs) or IPFFs on ATII cells. We found that RAS-ZEB1 signaling was significantly higher upon treatment with CM from IPFFs, reflected by significantly higher increases in phosphorylation of AKT and ERK levels (p < 0.05; Fig. 4, A and B) as well as ZEB1 (p < 0.01; Fig. 4C) and PLAT (tPA) expressions (p < 0.01; Fig. 4D). Together, these results suggest that signaling occurs between fibroblasts and ATII cells leading to RAS activation and induction of ZEB1, and that this effect is augmented by CM from fibroblasts that have been differentiated into myofibroblasts by TGF-β or by CM from IPFFs independently of exogenous TGF-β.

SPARC, a TGF-β-induced secreted protein, is highly expressed in IPF
As CM derived from TGF-β-activated fibroblasts or IPFFs induced RAS signaling in ATII cells, we next investigated which secreted factors from activated fibroblasts were responsible for this effect. By performing quantitative proteomic analysis of the CM from control or TGF-β-treated MRC5 cells, we identified 15 secreted proteins whose levels were upregulated upon TGF-β treatment (Table 1; Tables S5  and S6). Of these, SPARC was among the top candidates ( Fig. 5A; Table 1), suggesting a potential regulatory role of SPARC in mesenchymal-epithelial crosstalk driven by TGFβ. This was further validated in a publicly available dataset (GSE139963) (9), in which RNA-Seq was performed in human lung fibroblasts treated with or without TGF-β. Potential Epithelial-mesenchymal crosstalk in pulmonary fibrosis TGF-β-induced secreted proteins were identified by overlapping the upregulated genes with a list of secreted proteins (predicted) from Human Protein Atlas (https://www. proteinatlas.org/humanproteome/tissue/secretome). We identified 116 predicted secreted proteins that were significantly elevated (FDR-adjusted p value <0.05 and Log 2 [fold change] (Log 2 [fold change]) >1) in TGF-β-treated lung fibroblasts compared with controls ( Fig. S5; Table S7), and the   Fig. S5B; Table S7). The importance of SPARC in IPF pathogenesis was further confirmed in other datasets. In dataset GSE92592 (10), SPARC levels are increased in IPF tissues compared with control lungs (p < 0.01; Fig. 5C; Table 1), consistent with our recent report (11). Further analysis of GSE99621 (12) showed that SPARC expression is increased not only in scarred but also in those macroscopically unaffected (normal-appearing) regions of IPF lung (p < 0.05; Fig. 5D; Table 1). In addition, we confirmed a significantly enhanced secretion of SPARC in CM from IPFFs versus NHLFs by Western blotting (p < 0.001; Fig. 5E). Therefore, these data suggested that SPARC is induced as an early response to fibroblast activation and that it may be associated with IPF disease progression.
SPARC is a key fibroblast-derived paracrine regulator of RAS activation in ATII cells We recently reported that paracrine SPARC signaling from IPFFs dysregulates alveolar epithelial barrier integrity (11). We therefore hypothesized that SPARC might also mediate RAS activation in ATII cells during fibroblast-epithelial crosstalk. To test this hypothesis, ATII cells were exposed to CM harvested from control or SPARC-depleted NHLFs treated without or with TGF-β (Fig. S6A). As expected, we detected enhanced expression (Fig. S6B) and secretion (Fig. S6C) of SPARC from TGF-β-activated NHLFs compared with controls, and this was abolished using RNAi. When the CM was applied to ATII cells, CM from TGF-β-activated fibroblasts induced RAS activation, as evidenced by increases of phosphorylation in both AKT and ERK, whereas depletion of SPARC completely abolished the effects of CM from TGF-β-activated fibroblasts on ATII cells (Fig. S6C). We then compared CM from IPFFs: consistent with our previous results, an increased SPARC secretion was detected in CM from IPFFs versus NHLFs (Fig. S6D). Furthermore, while RAS signaling in ATII cells was activated by treatment with CM from IPFFs, these effects were completely abolished upon SPARC depletion in IPFFs ( Fig. 6A;  Fig. S6D), as evidenced by decreased phosphorylation of AKT and ERK levels ( Fig. 6B; Fig. S6D). Furthermore, human recombinant SPARC treatment alone was sufficient to induce epithelial growth factor receptor (EGFR) activation in ATII cells, demonstrated by increased phosphorylation levels of EGFR and its downstream AKT and ERK phosphorylation levels (Fig. 6C). EGFR activation was accompanied by a significant increase in mRNA expression of ZEB1 (Fig. 6C) and a reduction in CDH1 (E-cadherin) (Fig. S6E). Therefore, our results strongly suggest that SPARC is a critical paracrine regulator from activated lung fibroblasts that induces RAS activation in ATII cells (Fig. 6D).

Discussion
IPF is a chronic progressive lung disease with limited therapeutic options (1,13). Abnormal wound healing responses appear to make major contributions to the scarring process, but the underlying pathological mechanisms are unclear. Similar to studies in kidney fibrosis (14), our findings support the concept that aberrant epithelial-mesenchymal crosstalk contributes to the development of interstitial lung fibrosis. Our studies are consistent with the establishment of a bidirectional profibrogenic positive feedback loop, which maintains a chronic wound environment involving activated epithelial cells and fibroblasts that drive fibrosis progression rather than wound resolution (Fig. 7). Figure 5. SPARC, a TGF-β-induced secreted protein, is highly expressed in IPF. A, quantitative secretome analysis identifies an increased level of SPARC in the CM from TGF-β-treated MRC5 cells. Data are mean ± SD; n = 3. *p < 0.05. B, SPARC expression in human lung fibroblasts treated with or without TGF-β (GSE139963). ***p < 0.001. C, SPARC expression in lung tissues from healthy controls and IPF patients (GSE92592). **p < 0.01. D, SPARC expression in lung tissues from healthy controls and IPF samples with macroscopically normal-appearing (IPF-normal) or scarred (IPF-scarred) (GSE99621). *p < 0.05. E, protein expression of SPARC in CM from NHLFs or IPFFs. Ponceau S staining showing total protein levels. Graph showing SPARC protein levels in CM from NHLFs or IPFFs. Data are mean ± SD; n = 3. ***p < 0.001. CM, conditioned media; IPF, idiopathic pulmonary fibrosis; IPFFs, IPF fibroblasts; NHLFs, normal human lung fibroblast; SPARC, secreted protein acidic and rich in cysteine; TGF-β, transforming growth factor-β.
Epithelial cells undergo EMT in response to injury, allowing them to adopt a repair phenotype, which involves signaling to other cell types including inflammatory cells and resident mesenchymal cells. We previously reported that ATII cells undergoing RAS-induced EMT secrete tPA to augment TGF-β-induced myofibroblast differentiation and that ZEB1 is a key regulator of this paracrine signaling (6). To investigate further whether fibroblasts can respond to epithelial-derived signals in the absence of TGF-β, we used RNA-Seq to study the global transcriptomic changes in lung fibroblasts exposed to CM from ATII cells undergoing RAS-induced EMT. Using this unbiased approach, several potential mechanisms of relevance to IPF pathology were identified, including ECM and collagen, cell migration, cytokine binding, and inflammatory response, which included changes in expression of genes associated with TGF-β receptor signaling pathways. While we Figure 6. SPARC is a key fibroblast-derived paracrine regulator of RAS activation in ATII cells. A, protein expression of SPARC in CM from IPFFs with indicated treatment. Ponceau S staining showing total protein levels. B, protein expression of phospho-AKT (p-AKT), AKT, phospho-ERK (p-ERK), and ERK in ATII cells treated with CM from IPFFs with indicated treatment. β-tubulin was used as a loading control. Total AKT or ERK-normalized protein levels in ATII cells treated with CM from control IPFFs were used to set the baseline value at unity. C, protein expression of phospho-EGFR (p-EGFR), EGFR, p-AKT, AKT, p-ERK, and ERK in ATII cells treated with or without human recombinant SPARC (hrSPARC). β-tubulin was used as a loading control. Total EGFR, AKT, or ERKnormalized protein levels in control cells were used to set the baseline value at unity. Fold change in mRNA levels of ZEB1 in ATII cells treated with or without hrSPARC. β-actin-normalized mRNA levels in control cells were used to set the baseline value at unity. Data are mean ± SD; n = 3 samples per group. **p < 0.01. D, diagram summarizing SPARC as a key fibroblast-derived paracrine regulator of RAS activation in ATII cells. ATII, alveolar epithelial type II; CM, conditioned media; EGFR, epithelial growth factor receptor; ERK, extracellular-regulated kinase; IPFFs, IPF fibroblasts; SPARC, secreted protein acidic and rich in cysteine; ZEB1, zinc finger E-box-binding homeobox 1.
observed upregulation of ECM and collagen genes including COL3A1, COL4A6, COL14A1, and COL18A1 in the transcriptomic dataset, we did not observe changes in expression of either COL1A1, the major interstitial collagen produced in response to TGF-β stimulation, or ACTA2, which encodes alpha-smooth muscle actin, a marker of myofibroblast differentiation (Table S1; GSE163908). This suggests that the magnitude of TGF-β receptor signaling caused by the epithelial-derived CM was insufficient to drive differentiation of the fibroblasts into myofibroblasts. These findings are consistent with our previous studies (6) where exogenous TGF-β was required to drive myofibroblast differentiation. However, our observation of upregulation of genes associated with TGF-β receptor signaling and increased expression of many profibrotic genes in fibroblasts following exposure to CM from epithelial cells undergoing RAS-induced EMT may help explain why the CM augmented the effects of exogenous TGF-β that we observed previously (6).
Building on the strong gene expression signature for cell migration, we confirmed induction of a migratory phenotype in fibroblasts exposed to epithelial CM and demonstrated that this involved a ZEB1-tPA signaling axis, similar to that required for augmentation of TGF-β-induced myofibroblast differentiation (6). Fibroblasts from pulmonary fibrosis have altered pathological properties (15) including enhanced migration compared with fibroblasts from control lungs (16,17). In our studies, we found that CM produced by ATII cells undergoing RAS-induced EMT markedly induced fibroblast migration and recruitment. The importance of the ZEB1-tPA axis in this process was then confirmed, with their silencing abolishing the promigratory effects of CM from RASactivated ATII cells on lung fibroblasts. We previously identified that ZEB1 is expressed in epithelial cells of thickened alveolar septae where ECM deposition is evident (6). This suggests that ZEB1 is induced as an early response to alveolar epithelial injury and that, by regulating tPA expression, ZEB1 may promote fibroblast recruitment and thereby increase their numbers in the wound site. Consistent with our findings, it has been reported that neutralizing antibodies against tPA inhibit fibroblast migration (18). Mechanistically, tPA has direct cellular effects by virtue of its ability to bind to the low-density lipoprotein receptor-related protein-1, triggering low-density lipoprotein receptor-related protein-1 tyrosine phosphorylation and recruitment of β1-integrin signaling involving integrin-linked kinase (19,20).
While the activation of the ZEB-tPA axis may be a normal physiological response to injury, deregulation of this axis may sensitize the underlying fibroblasts to drive a pathological profibrogenic response (21). Consistent with this concept, in kidney studies, microinjuries on renal epithelium have been proposed to create a profibrotic microenvironment via TGF-β or connective tissue growth factor (14,22), promoting the aggregation of myofibroblasts, and in return, myofibroblasts may enhance the apoptosis of epithelial cells by secreting reactive oxygen species (23,24) or angiotensin II (25,26), so generating self-sustained pathological feedbacks. We used CM harvested from TGF-β-activated fibroblasts or IPFFs and found that both enhanced RAS activation and induction of ZEB1 in ATII cells via paracrine signaling. Moreover, we found that CM from IPFFs induced significantly more RAS activation and ZEB1 expression than normal fibroblasts in the absence of exogenous TGF-β, pointing to a pathological mechanism within the IPFF population. This finding may be explained by a previous report that autocrine TGF-β signaling is enhanced in IPFFs (8). Taken together, these data suggest that bidirectional epithelial-mesenchymal crosstalk may become dysregulated in IPF leading to fibrosis progression rather than wound resolution.
Using proteomics and bioinformatics, we were able to identify SPARC, secreted by TGF-β-activated fibroblasts, as a key paracrine regulator of RAS activation in ATII cells. SPARC, a cysteine-rich acidic matrix-associated protein, containing three epithelial growth factor-like repeats, is required for the collagen in bone to become calcified but is also involved in ECM synthesis and promotion of changes to cell shape (27)(28)(29)(30)(31)(32)(33)(34). Consistent with our recent findings (11), our analysis based on publicly available transcriptome datasets indicate that SPARC expression is increased not only in scarred regions but also in macroscopically unaffected (normal-appearing) regions of IPF lung tissue (10,12). This suggests that SPARC is induced as an early response to fibroblast activation. We found that while SPARC expression can be induced by TGF-β in normal lung fibroblasts, it is aberrantly expressed by IPFFs and that it acts via a paracrine mechanism to mediate RAS activation in ATII cells. SPARC has been shown by several studies to be able to regulate ERK signaling (35)(36)(37), although the upstream mechanism remains to be elucidated. In this study, our results demonstrated that SPARC alone is sufficient to induce EGFR activation in ATII cells, suggesting that SPARC might signal via the EGFR similar to tenascin C (38)(39)(40) via their epithelial growth factor-like repeats (41). Together, we have provided strong evidence suggesting that in both TGF-β-activated healthy lung Epithelial-mesenchymal crosstalk in pulmonary fibrosis fibroblasts and IPFFs, paracrine SPARC signaling dysregulates alveolar epithelial barrier integrity (11) and activates EGFR/ RAS signaling in ATII cells to maintain a chronic woundhealing phenotype.
In summary, we identify a mechanism whereby bidirectional epithelial-mesenchymal crosstalk results in self-sustaining activation of wound healing responses. This involves a ZEB1-tPA axis in ATII cells, which contributes to paracrine signaling that supports recruitment of fibroblasts and augments TGF-β-induced fibrotic responses; in turn, the aberrant behavior of IPFFs (or normal fibroblasts activated by high levels of TGF-β) promotes persistent alveolar epithelial activation via paracrine mediators including SPARC, which prevents resolution of normal epithelial repair responses and restoration of tissue homeostasis. Such a chronic wound environment may drive disease progression in pulmonary fibrosis (Fig. 7).

Lung tissue sampling
All human lung tissue samples for primary cell culture were approved by the Southampton and South West Hampshire and the Mid and South Buckinghamshire Local Research Ethics Committees, and all subjects gave written informed consent. The study abided by the Declaration of Helsinki principles. Clinically indicated IPF lung biopsy tissue samples and nonfibrotic control tissue samples (macroscopically normal lung sampled remote from a cancer site) were assessed as surplus to clinical diagnostic requirements. All IPF samples were from patients subsequently receiving a multidisciplinary diagnosis of IPF according to international consensus guidelines (42).

RNA isolation, library construction, and sequencing
To identify global transcriptomic changes in fibroblasts exposed to CM from RAS-activated ATII cells, RNA-Seq was performed in MRC5 cells treated with either CM from control or 4-OHT-activated ATII ER:KRASV12 cells. In brief, total RNA was isolated using RNeasy mini kit (Qiagen) according to the manufacturer's instructions and quantified using a Nanodrop Spectophotometer 2000c (Thermo Fisher Scientific). A total amount of 3 μg RNA per sample was used as input material for library construction. Sequencing libraries were generated using NEBNext UltraTM RNA Library Prep Kit for Illumina (NEB) following the manufacturer's instruction. Libraries were pooled in equimolar and sequenced using the paired-end strategy (2 × 150) on the Illumina NovaSeq 6000 platform following the standard protocols (Novogene).
Trim Galore (https://github.com/FelixKrueger/ TrimGalore) was used to trim adapters, reads with low quality (<30), and short length (<50 bp). RNA-Seq reads were mapped to Human genome Ensembl GRCh38 using Hisat2 (49) (version 2.1.0) with default codes. Sam files were transformed into bam files using samtools (50) (version 1.9). The read counts of each gene were summarized using featur-eCounts (51) (version 1.6.5). Raw read counts were imported into RStudio (version 3.6.1) and analyzed by using R package of DESeq2 (52) (version 1.26.0). Transcripts with low abundance (under ten counts across all samples) were removed. Genes with FDR p value less than 0.05 adjusted by using Benjamini-Hochberg method were considered as DEGs.

Pathway enrichment analysis
GO enrichment analysis was generated through PANTHER (7) Web site. REVIGO TreeMap was generated via an online tool available at http://revigo.irb.hr/ (53). Parameters were set as FDR <0.05. GSVA (54) was used to calculate pathway activity.

Wound-healing migration assay
The wound-healing migration assays were done in serumstarved fibroblasts. Confluent monolayers of cells were wounded with a p20 pipette tip (time 0). Phase-contrast images were taken using an Olympus inverted microscope at time 0 h or 24 h after the scratch wound. Wound areas were evaluated by ImageJ 1.42q software.

Transwell migration assay
For the transwell migration assay, transwell membranes (8-μm pore size, 6.5-mm diameter; Corning Costar; 3422) were used (45,55,56). The bottom chambers of the transwell were filled with indicated CM. The top chambers were seeded with 1.5 × 10 5 serum-starved fibroblasts per well. After 24 h, the filters were fixed with 4% paraformaldehyde for 10 min at room temperature; subsequently, the cells on the upper side of the membrane were scraped with a cotton swab. Filters were stained with crystal violet for light microscopy. Images were taken using an Olympus inverted microscope, and migratory cells were evaluated by ImageJ 1.42q software.
Quantitative proteomic analysis of the fibroblasts secretome and data processing Proteomic experiment was performed in three biological replicates. Serum-free CM from MRC5 cells treated with or without TGF-β1 (10 ng/ml; 48 h) were enriched based upon Strataclean resin (Agilent) and analyzed using LC-MSE (57), to provide estimates of absolute protein concentration with indepth coverage. Raw data were analyzed and exported into a csv document. Normalization method for each data is first dividing by the sum of each sample, using proteins containing no missing data, then multiplying by 10,000 (6). Minimum values of each column after normalization were used as pseudocounts to replace all the missing data, allowing for a complete statistical analysis (58). Differentially expressed proteins between two groups with p < 0.05 were identified as significant using a two-tailed and unpaired Student's t test (details described in supporting information section).

Statistical analysis and repeatability of experiments
Each experiment was repeated at least twice. Data are presented as mean and SD and compared with two-sample t test, Welch two-sample t test, Mann-Whitney U test, or one-way ANOVA test if appropriate. p < 0.05 was considered statistically significant. All data analyses and graphs were done in RStudio (version 3.6.1) or GraphPad Prism (version 8.2.1, GraphPad Software, Inc). Codes are available upon request.

Data availability
RNA-Seq data have been deposited in the Gene Expression Omnibus database (accession code: GSE163908). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE repository (59) with the dataset identifier PXD025821.
Supporting information-This article contains supporting information (57).