Lung fibroblasts express a miR-19a-19b-20a sub-cluster to suppress TGF-β-associated fibroblast activation in murine pulmonary fibrosis

Lung fibroblasts play a pivotal role in pulmonary fibrosis, a devastating lung disease, by producing extracellular matrix. MicroRNAs (miRNAs) suppress numerous genes post-transcriptionally; however, the roles of miRNAs in activated fibroblasts in fibrotic lungs remain poorly understood. To elucidate these roles, we performed global miRNA-expression profiling of fibroblasts from bleomycin- and silica-induced fibrotic lungs and investigated the functions of miRNAs in activated lung fibroblasts. Clustering analysis of global miRNA-expression data identified miRNA signatures exhibiting increased expression during fibrosis progression. Among these signatures, we found that a miR-19a-19b-20a sub-cluster suppressed TGF-β-induced activation of fibroblasts in vitro. Moreover, to elucidate whether fibroblast-specific intervention against the sub-cluster modulates pathogenic activation of fibroblasts in fibrotic lungs, we intratracheally transferred the sub-cluster-overexpressing fibroblasts into bleomycin-treated lungs. Global transcriptome analysis of the intratracheally transferred fibroblasts revealed that the sub-cluster not only downregulated expression of TGF-β-associated pro-fibrotic genes, including Acta2, Col1a1, Ctgf, and Serpine1, but also upregulated expression of the anti-fibrotic genes Dcn, Igfbp5, and Mmp3 in activated lung fibroblasts. Collectively, these findings indicated that upregulation of the miR-19a-19b-20a sub-cluster expression in lung fibroblasts counteracted TGF-β-associated pathogenic activation of fibroblasts in murine pulmonary fibrosis.

pivotal roles in fibroblast activation by stimulating TGF-β receptor, resulting in the expression of pro-fibrotic genes, such as connective-tissue growth factor (CTGF) 3 . Platelet-derived growth factor (PDGF) receptor signaling also contributes to fibroblast activation through cell migration and proliferation 4 . Accordingly, regulation of lung fibroblast activation by modulating these signaling pathways represents an attractive therapeutic strategy against pulmonary fibrosis. MicroRNAs (miRNAs) are a class of small endogenous noncoding RNAs of approximately 22 nucleotides in length. miRNAs negatively regulate the expression of target genes mainly by binding to the 3′ untranslated region (UTR) of the gene, thereby modulating various biological processes 5 . In humans, >45,000 miRNA-target sites have been identified, with miRNAs targeting >60% of protein-coding genes 6 . miRNAs have been implicated in a range of diseases, including cancer 7 , degenerative neurological disease 8 and inflammatory disease 9 .
A number of miRNAs have been implicated in IPF and pulmonary fibrosis models, including the miR-17~92 cluster 10 , miR-19a,b 11 , miR-21 12 , miR-26a,b 13,14 , miR-29 15 , miR-145 16 , miR-155 17 , and let-7d 18 . In rat models of pulmonary fibrosis, miRNA-expression profiling and proteomics studies suggest that miRNAs influence the proliferative, migratory, invasive, and survival properties of lung cells in the early phase of the disease 19 . However, because most previous studies analyzed miRNAs expressed in whole-lung tissue, the dynamics and roles of miR-NAs in fibroblasts of fibrotic lung remain elusive.
In this study, we performed global miRNA-expression profiling of lung fibroblasts during bleomycin-induced transient pulmonary fibrosis and silica-induced progressive pulmonary fibrosis using Col1(α)2-green fluorescent protein (Col1a2-GFP)-reporter mice to specifically isolate fibroblasts directly from mouse lungs. We identified the miR-19a-19b-20a sub-cluster, a member of fibrosis-associated miRNA signatures, as a suppressor of TGF-β-associated lung fibroblast activation in vitro. Pathway analysis of target genes associated with miRNA signatures also suggested a broad contribution of the miR-19a-19b-20a sub-cluster in TGF-β signaling.

Results
Identification of pulmonary fibrosis-associated miRNA signatures by global miRNA-expression profiling of lung fibroblasts. To measure global miRNA-expression patterns in activated lung fibroblasts, we performed global miRNA-expression profiling on fibroblasts isolated from untreated, bleomycin-treated, and silica-treated mice on days 7, 14, and 28 post-treatment. We identified 165 miRNAs that were differentially expressed with a fold-change ≥2 between at least two time points. These miRNAs were grouped based on their expression kinetics using the WGCNA and CLICK methods 20,21 (Fig. 1A; see Supplemental Table S3 for the full list of miRNAs in each group). Within the four miRNA groups identified, groups B1 and Y1 were transiently P-value (-log 10 Figure 1. Global expression profile of miRNAs in the activated lung fibroblasts of multiple lung fibrosis models. (A) Global miRNA-expression profile of fibroblasts from bleomycin-and silica-treated murine lungs on days 7, 14, and 28, as well as those of untreated mice on day 0. The heat map represents the group of differentially expressed miRNAs determined based on the WGCNA and CLICK methods. Each column represents the group and time point, and each row represents an individual miRNA. (B) KEGG pathway enrichment analysis of target genes associated with the miRNA signatures. The enrichment significance of the KEGG pathways was calculated using DIANA-miRPath version 3.0 with Tarbase version 7.0. P-values were calculated for each term using a two-sided hypergeometric test with Benjamini-Hochberg correction. See also Supplemental Tables S3 and S4. upregulated and groups B2 and Y2 were transiently downregulated on day 7 in bleomycin model (Fig. 1A). Notably, these expression changes were maintained in the silica-induced chronic pulmonary fibrosis model until at least day 28 post-administration, and no miRNA specifically expressed in the late phase of silica-induced chronic pulmonary fibrosis, was found (Fig. 1A). These results suggest that these miRNAs are associated with pulmonary fibrosis progression independent of the experimental model. To determine the biological functions associated with these miRNA signatures, we performed KEGG pathway analysis of predicted miRNA-target genes (see Supplemental Table S4 for the full list of significantly enriched KEGG pathways for each group). We found that genes associated with the TGF-β-signaling pathway were enriched in the predicted targets of miRNAs in groups B1 and Y1, but not in groups B2 and Y2 (Fig. 1B). By contrast, ECM-receptor interactions were only associated with group Y2. Because miRNAs are post-transcriptional gene repressors, these results indicated that the miRNAs in groups B1 and Y1 might suppress TGF-β-signaling-pathway related genes in fibrotic lung fibroblasts, and that the miRNAs in group Y2 might suppress ECM-receptor-interaction-related genes. Interestingly, fatty acid metabolism-related pathways (fatty acid metabolism, degradation, biosynthesis, and elongation) were highly enriched in all of the miRNA signatures (Fig. 1B). miR-19a-19b-20a sub-cluster is associated with TGF-β-signaling pathway. Within groups B1 and Y1, we selected miRNAs with mean tag count higher than 100 in the global miRNA-expression profile, and chose the guide strands of either 5p or 3p from these miRNAs. Next, we analyzed the predicted target genes, associated with the TGF-β-signaling pathway, and whose predicted scores had been described in TarBase version 7.0. We found 23 miRNAs with relatively high expression and predicted TGF-β-associated target genes with scores were more than 0.5 ( Fig. 2A). Among them, miR-19a, 19b, and 20a formed a sub-cluster, within 400 bp of genomic region. Consistent with the global miRNA-expression profile, qPCR revealed a significant increase of miR-19a, 19b, and 20a-expression in activated fibroblasts in the early phase of bleomycin and silica models (Fig. 2B). Interestingly, miR-19a-19b-20a sub-cluster, as a whole, had 12 predicted TGF-β-associated target genes, each miRNA predicted to target Tgfbr2, a cell surface receptor of TGF-β1. Therefore, we focused on miR-19a-19b-20a sub-cluster as a candidate that regulates fibroblast activation by suppressing TGF-β-signaling. miR-19a-19b-20a sub-cluster directly suppresses Tgfbr2 protein expression in primary lung fibroblasts. To investigate whether the miR-19a-19b-20a sub-cluster regulates endogenous Tgfbr2 protein levels in primary lung fibroblasts, we retrovirally transduced the sub-cluster and the reporter gene ∆hLNGFR (hCD271) into Col1a2-GFP fibroblasts (Supplemental Fig. S1A). Transduction efficiencies were similar between the control and the miR-19a-19b-20a sub-cluster vectors, and >60% in all samples (Supplemental Fig. S1C). We observed that miR-19a, 19b and 20a-expression levels were significantly higher in hCD271 + fibroblasts transduced with the miR-19a-19b-20a sub-cluster-vectors (sub-cluster fibroblasts) as compared with those transduced with a control vector (control fibroblasts) (Fig. 3A). Because the half-life of Tgfbr2 mRNA and Tgfbr2 protein was estimated to be between 1 and 5 h (average: 2-3 h) 22 , Tgfbr2 mRNA and Tgfbr2 protein expression were measured at 5-h post-infection. We found that Tgfbr2 mRNA expression did not change significantly (Fig. 3B); however, Tgfbr2 protein levels were significantly downregulated at 5-hour post-infection (Fig. 3C). To determine whether the miR-19a-19b-20a sub-cluster directly regulates Tgfbr2 protein levels, we assessed interactions between the miR-19a-19b-20a sub-cluster and the 3′ UTR of Tgfbr2 mRNA. We found that the 3′ UTR of Tgfbr2 mRNA contains 8-mer target sequences of each miRNA (Fig. 3D). A luciferase-reporter assay revealed that the miR-19a-19b-20a sub-cluster suppressed the expression of the wild-type Tgfbr2 3′ UTR reporter, but not that of the mutant Tgfbr2 3′ UTR reporters (Fig. 3E,F). These results suggested that the miR-19a-19b-20a sub-cluster suppressed Tgfbr2 protein levels by targeting the Tgfbr2 3′ UTR sequence in lung fibroblasts. miR-19a-19b-20a sub-cluster inhibits TGF-β1-mediated activation of primary lung fibroblasts. To assess the effect of the miR-19a-19b-20a sub-cluster on TGF-β1-mediated lung fibroblast activation, we measured Acta2 expression and Col1a2-promoter activity, both major downstream targets of TGF-β1, in primary lung fibroblasts in the presence and absence of TGF-β1 stimulation. Because the half-life of Acta2 mRNA is estimated at ~18 h, and the half-life of Acta2 protein is ~59 h 23 , the transduced fibroblasts were analyzed at 4-days post-infection. Flow-cytometric analysis revealed that the sub-cluster fibroblasts expressed significantly lower levels of α-SMA and Col1a2-GFP as compared with control fibroblasts (Supplemental Fig. S2A,B). When stimulated with TGF-β1 for 24 h after transduction, the sub-cluster fibroblasts expressed significantly lower levels of Acta2 mRNA as compared with control fibroblasts (Fig. 4A). Similar results were obtained using primary lung fibroblasts from Acta2-Kusabira Orange2 × Col1a2-GFP mice transduced with the miR-19a-19b-20a sub-cluster ( Fig. 4B,C).
(A) TGF-β-signaling-pathway associated miRNAs in groups B1 and Y1 identified using DIANA-miRPath version 3.0 with Tarbase version 7.0. The number of TGF-β-signaling genes associated with each mature miRNA is shown. (B) qPCR validations of miR-19a, 19b and 20a expression in fibroblasts from bleomycin-and silica-treated murine lungs on days 7 and 14, as well as untreated mice on day 0. The relative changes in gene expression were analyzed by the 2 −ΔΔCT method. qPCR and global miRNA-expression data showed similar trends related to expression kinetics. Graphs of the qPCR data show the mean ± SEM (n = 4). *P < 0.05; **P < 0.01; ***P < 0.001, one-way ANOVA with Tukey-Kramer's multiple-comparison post hoc test. Cohen's effect size d in these data was >0.7. (C) Details of the TGF-β signaling pathway-related genes potentially targeted with mmu-miR-19a-3p, 19b-3p, 20a-5p and 20a-3p.  5D). However, Acta2, Serpine1, Col1a1, and Col1a2 mRNA-expression levels in sorted donor fibroblasts were significantly lower in the sub-cluster fibroblasts as compared with control fibroblasts, suggesting that the miR-19a-19b-20a sub-cluster partially inhibited the upregulation of fibrosis-prone genes in fibroblasts of fibrotic lung (Fig. 5E).
miR-19a-19b-20a sub-cluster modulates the TGF-β-related changes in the global transcriptome profile of activated lung fibroblasts. To determine how the miR-19a-19b-20a sub-cluster regulates the global gene-expression profile of activated lung fibroblasts, we performed 3′ SAGE-seq analysis on intratracheally transferred control or the sub-cluster fibroblasts at 3-days post-transfer in the bleomycin model. We identified 160 genes that were differentially expressed in the sub-cluster fibroblasts relative to control fibroblasts (59 upregulated and 101 downregulated) and exhibiting a >1.5-fold-change in expression (P < 0.05) (Fig. 6A Table S5 for the full list of differentially expressed genes). The range of changes in gene expression was between 0.33-to 0.66-fold downregulation and 1.5-to 4.73-fold upregulation. The downregulated genes included TGF-β-associated pro-fibrotic genes, such as Acta2, Adam12 25 , Ctgf 3 , Has2 26 , Itga5 27 , Postn 28 , and Tgfbi 29 , and cell cycle-related genes, such as Ccnd1 and Mki67 (Fig. 6A). Additionally, Serpine1 expression was significantly downregulated (0.76-fold) in the sub-cluster fibroblasts relative to control fibroblasts (Supplemental Table S5). By contrast, upregulated genes included anti-fibrotic genes, such as Dcn 30 , Igfbp5 31 , and Mmp3 32 (Fig. 6B). Unexpectedly, some fibroblast-activation-associated genes such as Spp1 33 and Mmp2 34 , were also upregulated. qPCR analysis confirmed the changes in gene expression for Adam12, Dcn, Igfbp5, Mmp3 and Spp1 (Fig. 6C). To investigate the biological functions modulated by the miR-19a-19b-20a sub-cluster overexpression in activated lung fibroblasts, we performed GO 35 analysis on the 160 genes exhibiting the sub-cluster-mediated differential expression, with cell adhesion (GO:0007155) the only function significantly enriched among these genes (Fig. 6D). These results suggested that the miR-19a-19b-20a sub-cluster suppresses a part of the TGF-β-signaling-associated gene expression in the activated fibroblasts during bleomycin-induced lung fibrosis.

Discussion
In this study, we identified miRNA signatures in lung fibroblasts of transient/progressive murine pulmonary fibrosis models. Among these signatures, we found that the miR-19a-19b-20a sub-cluster suppressed TGF-β-associated gene expression and primary lung fibroblast activation. We used both bleomycin-induced transient and silica-induced progressive pulmonary fibrosis models to identify the fibrosis-associated common activation signature of miRNA expression in primary fibroblasts. Intratracheal instillation of bleomycin induces acute epithelial injury and transient fibrotic responses. Silica is ingested by macrophages, resulting in their activation and production of pro-fibrotic cytokines, such as tumor necrosis factor-α, PDGFs, and TGF-β, leading to progressive fibrotic responses. Notably, despite the differences in etiology between these two fibrosis models, most of the differentially expressed miRNAs (159 of 165 miR-NAs) were commonly upregulated or downregulated in line with fibrosis progression in activated fibroblasts from both models. Additionally, pathway analysis suggested possible contributions of these miRNAs in attenuating the TGF-β-signaling pathway and fatty acid metabolism-related pathways. A previous study reported that lung-resident fibroblasts contain lipid droplets, and that lipogenic genes contribute to lung fibroblast activation 36 . Moreover, a meta-analysis of global-expression profiles of mRNA and miRNA data from IPF patients revealed that the expression pattern of IPF-associated miRNA signatures is associated with that of TGF-β-signaling-related and lipid-metabolism-related genes 37 . Our results and previous reports suggested that the identified common miRNA-expression signatures in activated fibroblasts constitute a fibroblast-activation marker in a variety of fibrotic interstitial lung diseases, including IPF.
Here, we identified the miR-19a-19b-20a sub-cluster, a component of the miR-17~92 cluster, as fibroblast-activation-associated miRNAs. A previous study reported that expression of miR-17~92 cluster components, including the miR-19a-19b-20a sub-cluster, decreased in lung tissues from IPF patients and those from mice treated with bleomycin during late phase (day 28) 10 . Because fibroblasts constitute 15% to 20% of the whole-cell population in lungs, especially after induction of fibrotic disease in mice 33,38 , miRNA expression in whole lung tissues observed in the previous study might not reflect that observed in fibroblasts. The differences in the miR-19a-19b-20a sub-cluster expression between whole lung tissues and purified fibroblasts might be due to a specific regulatory mechanism associated with the sub-cluster expression in fibroblasts during fibrotic responses. miR-19a-19b target TGFBR2 in human cardiac fibroblasts 39 , and miR-20a inhibits TGF-β signaling by targeting TGFBR2 in non-fibroblastic cells, including cancer cells 40 , endothelial cells 41 . Consistent with previous studies, we observed that Tgfbr2 expression was suppressed by the miR-19a-19b-20a sub-cluster in primary lung fibroblasts and associated with suppression of TGF-β1-associated fibroblast activation. Intratracheal transfer of the sub-cluster fibroblasts revealed that inhibition of TGF-β1-associated fibroblast activation by the miR-19a-19b-20a sub-cluster also occurred in activated fibroblasts from a fibrotic lung. Moreover, the expression of Serpine1, a major TGF-β-responsive gene 42 , decreased in the sub-cluster-fibroblasts from a fibrotic lung, supporting speculation that the miR-19a-19b-20a sub-cluster suppresses TGF-β1 signaling in lung fibroblasts during pulmonary fibrosis. Our pathway analysis and previous report also suggested a possible contribution of miR-17~92 cluster components in TGF-β1 signaling in lung fibroblasts 10 . Overexpression of the miR-17~92 cluster decreases the amount of actin cytoskeleton and the expression of CTGF 10 , a downstream target of TGF-β1 signaling 3 . Overexpression of miR-19a and miR-19b reduces CTGF expression in lung fibroblast cell lines 11 . These reports and our results suggested that miR-17~92 cluster components, including the miR-19a-19b-20a sub-cluster, possibly cooperatively inhibit TGF-β1-mediated lung fibroblast activation.
Transcriptome analysis of adoptively transferred fibroblasts revealed decreases in expression of the pro-fibrotic genes Ctgf, Adam12, and Itga5 and increases in the expression of the anti-fibrotic genes Dcn (Decorin) and Mmp3 among multiple genes affected by the miR-19a-19b-20a sub-cluster overexpression. CTGF is a pro-fibrotic matricellular protein that contributes to pulmonary fibrosis through transcriptional activation cultured in medium containing TGF-β1 (10 ng/mL) and 1% FBS + DMEM. The surface area of the collagen gel was measured using ImageJ at 0-and 7-h post-stimulation, and the percentage of gel contraction in each cell group was calculated. (D) Gel images at 7-h post-stimulation. Dashed line shows the size of the gels at the stimulation-start point. Representative images (n = 3) are shown. (E) Percentage of collagen gel contraction at 7-h post-stimulation. Graphs show the mean ± SEM (n = 7; A), (n = 6; B), and (n = 3; E). *P < 0.05; **P < 0.01; ***P < 0.001, Student's ttest (A,B, and E) 3 . ADAM12 is a member of the ADAM protein family of disintegrins and metalloproteases, with Adam12 + cells reported as specific progenitors of collagen-overproducing cells generated following acute tissue injury. Genetic ablation of these Adam12 + cells limits generation of pro-fibrotic cells and interstitial collagen accumulation in muscle and dermis acute-injury models 45 . ITGA5 belongs to the integrin α-chain family, with ITGA5/ITGB1 expression facilitating lung stromal progenitor-cell differentiation into lung myofibroblasts 27 . Decorin is a proteoglycan that inhibits fibrotic responses by inhibiting TGF-β signaling 46 , and adenoviral Dcn transduction in lung tissues suppresses bleomycin-induced pulmonary fibrosis 30 . Collectively, the transcriptional changes associated with these genes support the anti-fibrotic roles related to the miR-19a-19b-20a sub-cluster expression in fibroblasts during pulmonary fibrosis. We also observed upregulated Spp1 expression in the sub-cluster fibroblasts. Spp1 encodes osteopontin, an ECM protein previously reported as a marker of activated fibroblasts present at the site of tissue remodeling in bleomycin-induced pulmonary fibrosis. Importantly, osteopontin + and α-SMA + fibroblasts show different distribution patterns in bleomycin-induced fibrotic lesions 33 . Additionally, osteopontin expression is induced by PDGF signaling in smooth-muscle cells 47 , which is another pro-fibrotic pathway distinct from TGF-β signaling. Therefore, it is possible that the sub-cluster suppresses only a subset of fibroblast-activation pathways, including that related to TGF-β signaling. Overall, we propose that upregulation of the miR-19a-19b-20a sub-cluster represents a compensatory response in activated fibroblasts that suppress detrimental TGF-β-associated fibrogenic activation of lung fibroblasts following tissue injury.
The progress and pathology of IPF differ from murine pulmonary fibrosis models; however, one commonality involves TGF-β activation of both human and murine lung fibroblasts. Considering that miR-19a-3p, 19b-3p and 20a-5p are conserved in mice and humans, the knowledge provided in this study will likely be applicable to IPF-patient-derived fibroblasts. Development of therapeutic strategies for the fibroblast-specific control of the miR-19a-19b-20a sub-cluster expression might lead to the identification of new therapeutic approaches for treating fibrotic diseases, including IPF.

Materials and Methods
Mice. C57BL/6J (B6) female mice were purchased from Japan SLC (Hamamatsu, Japan) or CLEA Japan (Tokyo, Japan). Col1a2-GFP female mice (C57BL/6 background for >10 generations) were generated in previous study 48 . Acta2-Kusabira Orange2 mice were generated as described in the Methods section of the online supplement. Acta2-Kusabira Orange2 × Col1a2-GFP male mice were generated by cross-breeding Col1a2-GFP mice and Acta2-Kusabira Orange2 mice. Experiments were initiated when mice were 6-to 12-weeks old. Mice were bred and maintained in specific pathogen-free facilities at the University of Tokyo, were group-housed with freely available food and water, and were under standard conditions with a light-dark cycle of 12 hours light and 12 hours dark. All methods were carried out in accordance with relevant guidelines of the University of Tokyo, and all experimental protocols were approved by the University of Tokyo.
Global miRNA-expression profiling. Small RNAs were isolated from lineage − GFP + lung fibroblasts using a mirVana miRNA-isolation kit (Thermo Fisher Scientific, Waltham, MA, USA). Small RNA libraries were constructed using an Ion Total RNA-seq kit v2 (Thermo Fisher Scientific) according to manufacturer instructions. The samples were sequenced twice on an Ion PGM system using an Ion PGM sequencing 200 kit (Thermo Fisher Scientific) according to manufacturer instructions. CLC Genomics Workbench software version 6.0.5 (CLC Bio, Aarhus, Denmark) was used to trim the sequence reads. miRNA expression analysis was performed by counting the resulting tags using miRBase database 49 (v19; http://www.mirbase.org/). Total tag counts were normalized to 80,000 tags, and the tag counts from two independent sequencing reads were averaged. Raw data from these experiments were deposited in the NCBI Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih. gov/geo; accession GSE100115).
Grouping of global miRNA-expression data. To group global miRNA-expression data, miRNAs with maximum tag counts (<10) across all experiments were filtered out. We then selected miRNAs exhibiting a fold-change of ≥2 between at least two samples. The total tag count for the 165 selected miRNAs was log 2 (X + 1) transformed, and miRNAs were grouped using R-3.3.1 software (https://cran.r-project.org/) and the WGCNA package 20 (power = 8, merge_thre = 0.25). Each detected miRNA group was further divided into positive-and negative-correlated groups using the CLICK method 21 . The expression level of each group was Z-scaled and visualized using R-3.3.1 software (https://cran.r-project.org/). Quantitative real-time polymerase chain reaction (qPCR). Total RNA from transduced lung fibroblasts was isolated using TRIzol reagent (Thermo Fisher Scientific). qPCR of mRNA was performed as described previously 38 . miRNA were reverse transcribed using TaqMan MicroRNA assays (Thermo Fisher Scientific) according to manufacturer instructions. qPCR analysis of miRNA was performed using THUNDERBIRD Probe qPCR mix (Toyobo) on an ABI 7500 real-time PCR system (Thermo Fisher Scientific

Retroviral miRNA transduction.
A 479 bp fragment containing miR-19a, miR-19b and miR-20a precursors was amplified by nested PCR by using PrimeSTAR Max DNA polymerase (Takara) and cloned into the downstream region of the CMV promoter in a modified pMY retroviral vector containing a ∆hLNGFR/hCD271 reporter gene (a truncated form of low-affinity nerve-growth factor receptor) (Supplemental Fig. S1A). Since the design of expression vector for individual miRNAs in miR-19a-19b-20a sub-clusters was technically difficult, we were unable to construct these vectors. Primer sequences for nested PCR are shown in Supplemental Fig. S1B. miRNA expression was confirmed using a sensor vector derived from the psiCHECK-2 vector (Promega, Madison, WI, USA). Retroviral vectors were produced in GP2-293T cells as described previously 51 . Viral supernatant was passed through 0.45-μm filters and concentrated using Amicon Ultra-15 centrifugal filters (50-kDa MWCO; Millipore, Bedford, MA, USA). The viral supernatant was concentrated to a pellet using a lentiX-concentrator (Takara, Tokyo, Japan) according to manufacturer instructions, and the pellet was dissolved with 10% FBS + DMEM. For miRNA transduction, 1 × 10 4 primary lung fibroblasts were transferred to 48-well plates. After a 12-h incubation, the cells were exposed to the viral supernatant (diluted to appropriate density) for 24 h.
Luciferase assay. Sensor and mutant vectors for Tgfbr2 were created from the psiCHECK-2 vector (Promega). Each of the control and sub-cluster vectors, and each of the sensor and mutant vectors were co-transfected into HEK293T cells using Lipofectamine LTX (Thermo Fisher Scientific) and cultured for 24 h. Renilla luciferase-firefly luciferase ratios in the cells were measured using a Dual-Glo reporter assay system (Promega) according to manufacturer instructions.
Cell culture under TGF-β1-stimulated conditions. Lung fibroblasts were cultured with 1% FBS + DMEM for 12 h, followed by a medium change to 1% FBS + DMEM in the presence or absence of TGF-β1 (10 ng/mL, R&D Systems). In some experiments, images of cultured fibroblasts were captured by a BZ-X700 fluorescence microscope (Keyence, Osaka, Japan).
Collagen gel contraction assay. Collagen gels were prepared according to manufacturer instructions to a final collagen concentration of 1.5 mg/mL (rat tail Col1; BD Biosciences). After transduction, 3 × 10 4 Col1a2-GFP lung fibroblasts were plated on the collagen gels in a 48-well plate. After cell attachment to the collagen gels (2 h), the gels were mechanically loosened from the sides of the wells, and the cells were cultured with 1% FBS + DMEM in the presence or absence of TGF-β1. The relative percentage of gel contraction represented the gel surface area covered each hour divided by the surface area just after mechanical loosening. The collagen gel images were quantified using ImageJ version 1.47t (http://imagej.nih.gov/ij; National Institutes of Health, Bethesda, MD).
Intratracheal transfer of lung fibroblasts. Intratracheal transfer was performed as described previously 24 . Details are provided in the Methods section of the online supplement. Since adoptively transferred fibroblasts were efficiently integrated into the fibrotic foci of bleomycin model, but not of silica model, we performed this experiment only in bleomycin model. Amplification of the whole transcript of fibroblasts. The whole transcript of intratracheally transferred fibroblasts was amplified according to a previous report, with some modifications 52 . Details are provided in the Methods section of the online supplement. Primer sequences are shown in Supplemental Table S1.

3′ serial analysis of gene expression (3′SAGE)-library generation and sequencing. Generation
of Na SAGE library was performed according to a previous report, with some modifications 53 . Details are provided in the Methods section of the online supplement. Raw data from these experiments were deposited in NCBI GEO (http://www.ncbi.nlm.nih.gov/geo; accession GSE100116). Primer and adapter sequences are shown in Supplemental Table S2.
Analysis of 3′SAGE-seq data. Analysis of 3′SAGE-seq data was performed according to a previous report, with some modifications [54][55][56][57] . Details are provided in the Methods section of the online supplement. Genes exhibiting differential-expression levels associated with an adjusted P-value < 0.05, a fold change >1.5, and a minimum expression level >30 were identified as showing statistically significant differential expression. Gene ontology (GO) analysis of the differentially expressed genes was performed by using DAVID version 6.8 58 .
Statistical analyses. Statistical comparisons were performed using unpaired Student's t-tests (two-tailed), one-way analysis of variance (ANOVA) with Tukey-Kramer's multiple-comparison post hoc tests, and a likelihood ratio test with Benjamini-Hockberg correction (transcriptome analysis). A P < 0.05 was considered statistically significant. Data are expressed as the mean ± standard error of the mean (SEM). Effect size was measured with Cohen's d. Statistical analysis was performed using Prism software version 5.01 (GraphPad Software, La Jolla, CA, USA) or R-3.3.1 software (https://cran.r-project.org/).