Identification of a fibronectin-binding protein signature associated with idiopathic pulmonary fibrosis

Abstract

The extracellular matrix (ECM) is a critical component of tissue where it provides structural and signaling support to cells. Its dysregulation and accumulation lead to fibrosis, a major clinical challenge underlying many diseases that currently has little effective treatment. An understanding of the key molecular initiators of fibrosis would be both diagnostically useful and provide potential targets for therapeutics. The ECM protein fibronectin (FN) is upregulated in fibrotic conditions and other ECM proteins depend on assembly of a FN foundational ECM for their matrix incorporation. We used cell culture and in vivo models to investigate the role of FN in the progression of lung fibrosis. We confirmed that normal human lung fibroblasts (NHLFs) treated with transforming growth factor-beta (TGF-β) to stimulate fibrotic gene expression significantly increased both FN expression and its assembly into a matrix. We found that levels of alternatively spliced EDA and EDB exons were proportional to the increase in total FN RNA and protein showing that inclusion of these exons is not enhanced by TGF-β stimulation. RNA-sequencing identified 43 core matrisome genes that were significantly up- or down-regulated by TGF-β treatment and a Luminex immunoassay demonstrated increased levels of ECM proteins in conditioned medium of TGF-β-treated NHLFs. Interestingly, among the regulated core matrisome genes, 16 encode known FN-binding proteins and, of these, insulin-like growth factor binding protein 3 (IGFBP3) was most highly up-regulated. To link the NHLF results with in vivo disease, we analyzed lung tissue and bronchoalveolar lavage fluid from bleomycin-treated mice and found dramatically higher levels of FN and the FN-binding proteins IGFBP3, tenascin-C, and type I collagen in fibrotic conditions compared to controls. Altogether, our data identify a set of FN-binding proteins whose upregulation is characteristic of IPF and suggest that FN provides the foundational matrix for deposition of these proteins as fibrosis develops.

1. Introduction

Fibrosis is the accumulation of extracellular matrix (ECM) proteins in insoluble and disorganized deposits within tissues that are distinct from the well-organized fibrillar networks and linear fibers that make up normal interstitial ECM [1,2]. Fibrotic deposits perturb tissue and organ functions changing cell growth and survival, adhesion, and gene expression of resident cells. The development of both normal tissues and fibrotic deposits depends on the assembly of ECM proteins into polymers. The ECM protein fibronectin (FN) is the foundational component of this assembly process and FN matrix fibrils are required to provide a scaffold for assembly of other ECM proteins, in particular, collagens [3-6]. FN polymers are key components in numerous biological functions including blood clot and provisional matrix formation in wound healing, cell invasion in pathogenesis, cell adhesion and migration required for organization of cells into tissues and organs, and by providing environmental cues for tissue resident cells as well as immune and inflammatory cells [7].

Many events that occur in fibrosis have been attributed to increased signaling by TGF-β family members [8]. FN, collagens, and their receptors along with other molecules involved in ECM formation and cell contractility are well-known to be up-regulated by TGF-β. Other mechanisms have been linked to excess ECM deposition including activation of intracellular pathways that stimulate receptors to assemble more ECM [9,10]. Because FN fibrils are central to the initiation of matrix assembly, FN levels and the formation of FN fibrils may have a critical role in the development of fibrosis and progression of disease. In that regard, degradation products of FN have been proposed as a biomarker of pulmonary fibrosis [11]. Pharmacological inhibition of FN fibrillogenesis has been reported to mitigate tissue fibrosis in an animal model [12] and FN1 does show up in comprehensive proteomic profiling of lung injury [13]. However, it is still poorly understood how FN is dysregulated and whether measurements of total FN levels or of FN matrix are useful for disease prognosis and/or diagnosis. To investigate FN’s role in idiopathic pulmonary fibrosis (IPF), we used both cell culture and in vivo models, namely, TGF-β-treated normal and diseased human lung fibroblasts and a bleomycin-treated mouse model [14], respectively. We report a list of proteins that are reproducibly and significantly increased with TGF-β treatment of fibroblasts and are also found in lavage fluid of bleomycin mouse lungs. Notably, a significant fraction of these proteins is known to bind to FN, identifying a novel “FN-binding protein signature” of IPF. We suggest that this signature may be an important tool for diagnosing and understanding fibrosis in the lung and other tissues.

2. Materials and methods

2.1. Cell culture, reagents, and antibodies

Normal human lung fibroblasts (NHLF, Lonza CC-2512) and diseased human lung fibroblasts from idiopathic pulmonary fibrosis patients (DHLF, Lonza CC-7231) were cultured in complete medium, FGM-2 Fibroblast Growth Medium-2 BulletKit (Lonza CC-3132, FBM Basal Medium and FGM-2 supplements). NHLFs and DHLFs were used for experiments through Passage 6. For TGF-β treatment, fibroblasts were seeded in complete medium at a density of 3*10⁴/cm² and allowed to attach for 6 h followed by overnight starvation in FBM Basal Medium supplemented with 0.1% FBS (Hyclone). Cells were stimulated with TGF-β1 (5 ng/ml) (Peprotech, 100-21) for 6, 24, and 48 h in starvation medium for gene and protein expression studies and in complete medium for matrix assembly experiments.

Antibodies used in this study were mouse monoclonal HFN 7.1 anti-human FN antibody from concentrated supernatant (Developmental Studies Hybridoma Bank, Univ. Iowa), rabbit polyclonal anti-FN antiserum (R184) and rabbit polyclonal anti-TN-C antiserum (R759) [15] (both prepared in-house), rabbit polyclonal anti-FN (Abcam, ab32419), rabbit polyclonal isotype control (Abcam, ab37415), rabbit anti-GAPDH monoclonal antibody (Cell Signaling, 2118), mouse anti-EDA monoclonal antibody (AdipoGen, AG-20B-6001PF-C100), rabbit anti-IGFBP3 monoclonal antibody (Cell Signaling, 13216), and goat anti-IGFBP3 polyclonal antibody (R&D systems, AF775).

2.2. Animals

Male C57BL/6 mice with an average weight of 25-30g were obtained from Jackson Laboratory. Mice were single housed in a temperature-controlled environment (20–24 °C) on 12 h light–dark cycle and provided food and water ad libitum. All animal husbandry and procedures were approved by the Bristol Myers Squibb Company Animal Care and Use Committee and were conducted in conformity with the Public Health Service Policy on Humane Care and Use of Laboratory Animals.

2.3. Intratracheal Bleomycin Mouse Model

Mice were placed in an induction chamber and anesthetized by a mixture of isoflurane delivered in oxygen until a surgical plane of anesthesia was reached which was determined by respiratory rate and toe pinch. The animals were then suspended on an induction board by their incisors and an otoscope and lead wire were used to guide an indwelling catheter (20g x 1in) into the trachea. Once the catheter was in place, the lead wire was removed and a single treatment (PBS or Bleomycin 1 U/kg Sigma lot # 3670656) was administered into the hub of the catheter and passively inspired during normal respiration. Once the total volume (4 ml/kg) was inhaled the catheter was removed and the animals were allowed to recover from anesthesia prior to being returned to the housing room.

Bronchoalveolar lavage fluid was collected 21 d after bleomycin treatment. Mice were anesthetized with an injectable anesthesia mixture (10 mg/ml Ketamine and 0.7 mg/ml Xylazine) administered via IP injection (12.5 ml/kg) until a surgical plane of anesthesia was reached as determined by respiratory rate and toe pinch. A small skin excision was made on the ventral neck region and blunt dissection was used to expose the trachea. The trachea was then catheterized with a luer hub (20g x 0.5in) and secured with suture. A 1ml syringe pre-filled with cold 5 mM EDTA in HBSS was secured to the luer hub and a single wash was performed to collect bronchoalveolar lavage (BAL) fluid. BAL fluid was centrifuged (500xg for 8 min in 4 °C) to pellet cells and supernatant was collected for analysis. Mice were then exsanguinated via cardiac puncture. The tracheobronchial tree and lungs were excised with luer hub still sutured in place. A 3 ml syringe pre-filled with 10% neutral buffered formalin (NBF) was attached to luer hub and a volume was delivered until lungs were fully inflated without overdistension. A suture was used to tie off trachea below luer hub and tissues were placed in individual jars pre-filled with 10% NBF.

2.4. Histochemistry

5 µm tissue sections from paraffin embedded mouse lung were processed on the Sakura Tissue Tek Prisma automated tissue stainer using manufacturer recommended protocols for hematoxylin and eosin (H&E) staining. In brief, xylene was used to dewax the tissue, followed by rehydration with alcohol and water. Hematoxylin was applied to stain nuclei, followed by differentiation reagent. Next, eosin counterstain was used to stain the cytoplasm and extracellular components of the tissue. Alcohol rinses and xylene clearing made the tissue transparent before mounting with a coverslip and mounting media. Slides were left at room temperature to cure before imaging on the Aperio AT2 (Leica Biosystems, Deer Park, IL, USA) whole slide scanner.

For immunohistochemistry, the Bond Polymer Refine detection system (Leica Biosystems, DS9800) was used in the automated staining protocol with the polyclonal rabbit anti-FN antibody. Slides were rinsed in ddH2O prior to mounting with a glass coverslip and brightfield imaging on the Phenoimager HT. Images were exported into HALO v3.4.2986.166 (Indica Labs, Albuquerque, New Mexico, USA) and analyzed using the CytoNuclear and Area algorithms. Single cell positivity and area analysis was graphed using GraphPad Prism v9.5.

Assessment of fibrotic regions was performed using a HALO AI-based classifier called Densenet V2. Alveolar regions of interest were selected in the distal portion of healthy lung sections stained with H&E; these images were annotated as normal and used to train the classifier. Alveolar regions with higher cell density and strong eosin staining outside of the nuclei were selected on lung sections from bleomycin-treated mice and used as fibrotic for classifier training. The classifier was utilized to analyze all of the H&E images from the two groups (2 images per animal). The % of the tissue that was classified as fibrotic or normal is reported.

2.5. Immunoblotting and protein analyses

Conditioned medium (CM) was collected from NHLFs, cells were lysed in deoxycholate (DOC) buffer, DOC-soluble and DOC-insoluble fractions were isolated by centrifugation, and the DOC-insoluble pellet was solubilized in SDS buffer by boiling following the procedures as described [10]. To analyze intracellular FN in sub-confluent cells, NHLFs were stimulated with TGF-β for 48 h, then re-plated at a sub-confluent density for 4 h before lysing in SDS lysis buffer (2% SDS, 20 mM Tris-HCl, pH 8.8, 2 mM EDTA) with protease inhibitor. Total protein concentrations of DOC-soluble and sub-confluent SDS lysates were determined by BCA assay (Pierce). Equal volumes of CM, equal amounts of total protein from SDS lysates or DOC-soluble fractions and proportional volumes of DOC-insoluble samples were separated by SDS-PAGE on 6% (for FN, EDA-FN, TN-C) and 10% (for IGFBP3 and GAPDH) polyacrylamide gels. Gels were transferred to a nitrocellulose membrane and immunoblotted as described [10]. Concentrations of BALF samples were determined by spectrophotometry using A₂₈₀ (1Abs = 1mg/ml, a general reference setting based on a 0.1% protein solution producing an Absorbance of 1.0 A). BALF samples in control and in fibrotic groups were analyzed by SDS-PAGE and silver staining [16]. The fibrotic BALF samples, diluted 1:10, were analyzed using R184 anti-FN and R759 anti-TN-C antisera at 1:10,000 and 1:2000 dilutions, respectively. Other antibody dilutions were: HFN 7.1 (1:10,000), anti-EDA-FN (1:1000), and both anti-IGFBP3 antibodies (1:1000). Secondary HRP-conjugated antibodies (ThermoFisher Scientific) were used at 1:10,000 dilution. SuperSignal West Pico Plus ECL reagent (Pierce) was used for developing blots. Azure imaging system was used for imaging blots and AzureSpot Pro was used for band density analysis. Band intensities within the linear range were used for quantification.

2.6. Immunofluorescence microscopy

For matrix staining, NHLFs were seeded on glass coverslips and stimulated with TGF-β as described above. After 48 h, NHLFs were fixed with 3.7% formaldehyde in PBS and incubated with anti-FN (R184 at 1:100) or anti-EDA antibody (1:100) followed by Alexa-Fluor-488- or 568-conjugated goat anti-rabbit or goat anti-mouse IgG (Invitrogen,1:500). For intracellular FN staining after TGF-β stimulation, NHLFs were re-seeded at sub-confluence at a density of 0.75*10⁴cells/cm². After 4-6 h attachment, NHLFs were fixed and permeabilized with 0.5% Triton X-100 in PBS for 15 min at room temperature, followed by staining as above. Samples were mounted in VECTASHIELD Vibrance antifade mounting medium (H-1700). A Nikon Eclipse Ti microscope with a Hamamatsu C10600 ORCA-R2 digital camera was used to capture images. Mean fluorescence measurements were performed using ImageJ on 7 randomly selected fields per condition. Representative fields are shown. Images were adjusted equivalently using ImageJ.

2.7. Quantitative PCR

NHLFs were lysed with TRIzol reagent (Invitrogen), lysates were homogenized with a 20G needle, chloroform was added, the aqueous layer containing RNA was separated by centrifugation and RNA was isolated from the aqueous layer using a RNeasy Mini Kit (Qiagen, 74104) as described [17]. RNA concentrations were determined with a Nanodrop spectrophotometer. RNA (1 μg) was reversed transcribed using random hexamer primers and qPCR reactions were performed with a Mx3000P QPCR system (Agilent Technologies) using primers listed in table and conditions as described [17]. The delta Rn value is calculated by normalizing the CT value of a target gene to the CT value of the reference gene (GAPDH) in each sample.

RT-PCR primers

Primer Pair	Forward 5'-3'	Reverse 5'-3'
FN	ACAACACCGAGGTGACTGAGAC	GGACACAACGATGCTTCCTGAG
EDA-FN	TGCTCAGAATCCAAGCGGAG	GACTGGGTTCCAATCAGGGG
EDB-FN	TGGTCCATGCTGATCAGAGC	CCAGCCCTGTGACTGTGTAG
GAPDH	GTCTCCTCTGACTTCAACAGCG	ACCACCCTGTTGCTGTAGCCAA
THBS1	GCTGGAAATGTGGTGCTTGTCC	CTCCATTGTGGTTGAAGCAGGC
TNC	GGAAGAGACGGTCAACCTGG	TGCATTTCCAGACCACCCTC
POSTN	CAGCAAACCACCTTCACGGATC	TTAAGGAGGCGCTGAACCATGC
IGFBP3	CGCTACAAAGTTGACTACGAGTC	GTCTTCCATTTCTCTACGGCAGG
THBS2	CAGTCTGAGCAAGTGTGACACC	TTGCAGAGACGGATGCGTGTGA

2.8. RNA-sequencing and analysis

Duplicated RNA samples in each condition (+/− TGF-β) from each donor were submitted to the Genomics Core Facility at the Lewis Sigler Institute for Integrative Genomics (Princeton University). RNA samples of normalized concentrations were used as input for reverse transcription using barcoded oligo-dT primers to index each cDNA sample following the drop-seq method [18]. The pooled barcoded cDNA samples were purified and amplified by PCR, then turned into sequencing libraries using the Nextera kit (Illumina, CA) to include only the poly-A tail adjacent 3’ ends of the transcripts. The cDNA samples and sequencing libraries were examined on the Bioanalyzer (Agilent, CA) DNA HS chips for size distribution, and quantified by Qubit fluorometer (Invitrogen, CA). The purified library pool was sequenced on Illumina NovaSeq 6000 using the SP 100 cycle kit v1.5. Raw sequencing reads were filtered by Illumina NovaSeq Control Software and only the Pass-Filter (PF) reads were used for further analysis.

FASTQ files generated from above were used to identify the differentially expressed genes. The reads were aligned using STAR [19] on the Human Reference Genome (GRCh38). The transcript abundance estimation was performed using RSEM [20]. The quality of the sequence data was assessed using multiQC v1.12 [21]. Count and TPM data were generated from the pipeline for further analysis and visualizations to identify differentially expressed genes in NHLF cells treated with TGF-β at 48 h. After filtering non-expressed or low-expressed genes based on counts per million (CPM), 24,489 genes with CPM values above 1 remained. Differential Expression analysis was carried out using PyDEseq2 python package [22]. The Matrisome Project website (https://sites.google.com/uic.edu/matrisome/home) contains a tool, Matrisome AnalyzeR (https://matrinet.shinyapps.io/MatrisomeAnalyzer/), that was used to classify our RNA-sequencing gene lists into three groups: non-matrisome, matrisome-associated and core matrisome.

2.9. Luminex analysis

NHLFs were cultured and stimulated as described above in starvation medium. The CM was collected at 48 h and stored with 10 mM EDTA, 2 mM PMSF at −20°C until analyzed. The concentration of soluble protein analytes in CM was measured using a Luminex^™ Bead-Based multiplexed assay with commercially available soluble protein analyte panels from Millipore-Sigma (analyte information is listed in Supplementary Table 4). Data were collected using a FLEXMAP 3D^™ instrument. Standard curves were derived, and analyte concentrations were calculated with Bio-Plex Manager^™ (Bio-Rad) software using phycoerythrin median fluorescence intensity (MFI) data. Ranges of reliable analyte detection are presented in Supplementary Table 4. For FN analyses, cell supernatants were diluted 1:1000 and patient plasma samples were diluted 1:400,000 to bring FN into the range of the assay.

For each biomarker, the concentration values in each group were checked to ensure they were within the lower limit of quantification (LLOQ) and upper limit of quantification (ULOQ). The values that fell outside this range were excluded from further analysis (with a few exceptions for manual check as their values before or after TGF-β treatment drastically changed and fell outside of the thresholds). The remaining values were log2 transformed and normality was assessed using the Shapiro-Wilk test [23] and visual inspection of raw data. Unpaired t-tests were performed to compare the mean concentration values of each biomarker between the comparison groups (such as +/− TGF β). The log2 fold change was calculated between the mean values of each group for each biomarker. All statistical analyses were performed using Python and data visualizations were created using Spotfire (v12) software.

2.10. Statistical analyses

Immunoblot and immunofluorescence results are reported as the mean ± standard error of multiple independent experiments as noted in figure legends. Statistical analyses for experiments with stimulation at a single time point were performed using an unpaired, two tailed Student’s t-test. With stimulation at multiple time points, two-way ANOVA was applied, with p < 0.05 considered statistically significant. RNA-seq and Luminex data statistics were calculated as described above.

3. Results

3.1. Effects of TGF-β on fibronectin expression

Normal human lung fibroblasts (NHLFs) derived from donors were shown to respond to TGF-β treatment as expected at both the mRNA and protein levels. TGF-β-dependent changes in FN expression were confirmed by comparing NHLFs either without or treated with TGF-β for 24 and 48 h after serum starvation. Using qPCR, we found that total FN mRNA increased more than 4-fold at 24 h but continued to increase to > 5-fold by 48 h of treatment (Figure 1A, Supplementary Table 1). Results were equivalent with NHLFs from two different donors. Because alternative splicing at the extra domain A (EDA) and extra domain B (EDB) exons of the FN transcript has been suggested to associate with fibrosis [24,25], we used qPCR to measure levels of EDA+ and EDB+ FN transcripts relative to total FN. Our results show higher levels of these splice variants with TGF-β treatment (Figure 1B, C), but the change in each was similar to or less than the change in total FN transcripts (Supplementary Table 1) indicating that these isoforms are not enriched compared to total FN with TGF-β stimulation. We observed a similar pattern of change in total and spliced FNs in our analyses of diseased cells (DHLFs) from two donors (Supplementary Figure 1). Direct comparison shows that total FN mRNA levels expressed by untreated DHLFs were 2-3-fold higher than in untreated NHLFs.

Figure 1. Transcription of FN is increased by TGF-β stimulation.

Total cellular RNAs from control (Ctrl, blue) and TGF-β-treated (red) NHLFs from two donors (NHLF 1 and NHLF 2) were isolated after treatment for 24 h (light colors) and 48 hr (dark colors). Expression was quantified by qPCR for total FN (A) and splice isoforms EDA (B) and EDB (C). Means ± SEM of dRn for 3 independent experiments are reported for each condition of each donor. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Increased mRNA resulted in higher production of FN protein. Subconfluent NHLF cells were lysed after 48 h with or without TGF-β in starvation medium. When subconfluent, the majority of FN is intracellular and thus a measure of protein production. Quantification of immunoblots showed > 7-fold increases in FN protein for both donors when treated under standard conditions on an uncoated culture surface (Figure 2A). Significantly more FN was also detected intracellularly by immunofluorescence of control and treated cells (Figure 2B). Cell-ECM interactions can affect gene expression and TGF-β activity [8,26] so we also tested a collagen-coated substrate, as type I collagen is known to be upregulated under these culture conditions. Comparison of cells cultured on a collagen coat with an uncoated surface showed a similar level of increase (Figure 2A) indicating that increased collagen is not enhancing the effects of TGF-β.

Figure 2. TGF-β increases FN protein expression.

NHLFs were treated with TGF-β or untreated (Ctrl) for 48 h and then FN levels were analyzed. A) SDS lysates were prepared from subconfluent NHLFs from two donors grown on tissue culture plastic (Non-Coat) or on type I collagen-coated plates (Collagen-Coat). Equal amounts of total protein were immunoblotted with anti-human FN antibody. Numbers below the blots are average fold-changes of band intensities with TGF-β relative to control from three independent experiments. Representative blots are shown. Locations of FN and MW marker are indicated on left and right, respectively. B) NHLFs were fixed, permeabilized and stained with DAPI and anti-FN antibody to visualize intracellular FN in control (Ctrl) and TGF-β-treated cells. Images are representative of three independent experiments. Scale bar = 100 μm.

3.2. Fibronectin matrix assembly in response to TGF-β stimulation

A primary function of FN is to assemble into a fibrillar matrix and excess matrix fibrils are a hallmark of fibrosis. To determine how FN matrix levels are affected by TGF-β treatment, NHLFs were incubated with TGF-β in complete medium with serum, which is needed for efficient matrix assembly. Conditioned medium (CM) from confluent cells collected at 48 h of TGF-β treatment contained > 5-fold more FN than CM from untreated cells (Figure 3A). A deoxycholate (DOC) detergent insolubility assay was used to determine levels of FN fibril formation with and without TGF-β treatment. The DOC-soluble material contains intracellular and cell-associated proteins as well as ECM fibrils that are initially soluble in DOC buffer. Over time, fibrils are irreversibly converted into a DOC-insoluble form; the quantity of these insoluble fibrils is a measure of the level of stable FN matrix. Results show dramatic and significant increases in both DOC-soluble and DOC-insoluble FN matrix (Figure 3A, Supplementary Figure 2A). Immunostaining showed that fibril densities and mean fluorescence intensities are also much higher with TGF-β (Figure 3B, Supplementary Figure 2B). Type I collagen fibrillogenesis was increased in treated cells (Supplementary Figure 3) as expected, as collagens depend on FN fibrils to initiate fibrillogenesis.

Figure 3. FN matrix assembly is stimulated by TGF-β.

NHLFs were grown to confluence with or without TGF-β treatment and analyzed at 48 h. A) Conditioned media (CM) were collected, and cells were lysed with buffered DOC. DOC-soluble cell-associated material was separated from DOC-insoluble matrix. Equal volumes of CM and equal amounts of total protein in DOC samples normalized to GAPDH in the DOC-soluble fraction were electrophoresed and immunoblotted with anti-FN antibodies. Average fold-changes of FN in TGF-β vs. control were 5.7 ± 0.18* for CM, 2.6 ± 0.1* for DOC-soluble and 5.1 ± 0.1** for DOC-insoluble fractions. n = 3; *p < 0.05, **p < 0.01. Locations of FN and MW markers are indicated on left and right, respectively. B) In parallel, cells were fixed and immunostained for human FN and DAPI. Mean fluorescence intensity of FN in TGF-β vs. control condition is 1.7 ± 0.1*, n = 3. *p < 0.05. Scale bar = 100 μm.

EDA+ FN production and incorporation into the matrix were analyzed by immunofluorescence. Mean fluorescence intensities of total matrix FN and EDA+ matrix were 1.5 and ~1.2 fold higher, respectively, in TGF-β treated cultures (Supplementary Figure 4A). Band intensities on immunoblots of DOC-soluble and DOC-insoluble fractions increased with TGF-β treatment for both total FN and EDA+ FN (Supplementary Figure 4C, D), while the accumulation of EDA FN in CM was slightly less than that detected for total FN (see Figure 3) (Supplementary Figure 4B). While EDA+ FN protein levels increase with TGF-β stimulation in secreted and matrix fractions, the EDA+ isoform does not show a disproportionately higher increase than total FN. Therefore, contributions to the fibrotic process of the EDA+ isoform cannot be separated from effects of increases in total FN. EDB+ FN protein levels were too low to be detected in our samples.

3.3. Effects of TGF-β on core matrisome gene expression

To identify global ECM gene expression patterns as well as ECM changes with TGF-β stimulation in NHLFs, RNA-seq analyses were performed. The effects of TGF-β were confirmed by expression of the well-established biomarker of fibrosis, smooth muscle actin (ACTA2), which increased nearly 20-fold (log2FC = 4.31, padj = 2.66E-13). Analysis of the RNA-seq data in Matrisome AnalyzeR [27] identified 206 core matrisome genes. The 80 most highly expressed of the 206 ECM genes, with normalized transcripts per million (TPM) > 1, were further analyzed. Expression of 43 of these 80 genes was significantly changed between control and TGF-β conditions with adjusted p-values < 0.05; expression of 34 genes increased and 9 genes decreased (Figure 4A, Supplementary Table 2). IGFBP3, which is known to be up-regulated by TGF-β and several other soluble factors [28], showed the largest Log2FC of 7.31. RNA-seq expression differences with TGF-β were confirmed using two genes with significant changes (THBS1 and THBS2) and each increased at least two-fold by qPCR (Supplementary Figure 5C, D). Periostin gene (POSTN) expression was also analyzed because it showed a Log2FC of 1.16 with a p-value < 0.05 but padj > 0.05. Our qPCR results show a level of change (~3-fold) similar to the RNA-seq results (2.2-fold) (Supplementary Figure 5E).

Figure 4. Changes in core matrisome gene and protein expression with TGF-β stimulation.

Analysis of RNA-sequencing data for NHLFs from two donors shows significant regulation of 43 ECM genes with TGF-β compared to control treatment. A) The heatmap illustrates log2FC for these 43 genes. Expression levels, log2FC, and adjP values are listed in Supplementary Table 2. B) A Luminex bioassay was used to analyze protein levels in CM from TGF-β-treated and control NHLFs at 48 h. 10 core matrisome proteins were detected and 7 of them changed in response to TGF-β with ∣Log2FC∣ > 1 or with P < 0.05. The graph shows these proteins ranked by Log2FC. See Supplementary Table 3 for complete dataset. C) CM and DOC-soluble fractions were electrophoresed as in Figure 3 and immunoblotted for IGFBP3. IGFBP3 was detected in the TGF-β-treated but not control-treated samples. MW markers are indicated on the right.

In parallel, we tested the expression of a panel of 124 soluble protein analytes using a Luminex immunoassay. Antibodies were included that recognize matrisome proteins and soluble factors (Supplementary Table 4) and CM was analyzed from NHLF cells treated with TGF-β or untreated for 48 h. We found that 38 of the 124 proteins were detected in the NHLF medium, 10 of which are core matrisome proteins (Supplementary Table 3). The levels of 7 of the 10 core matrisome proteins (70%) showed log2FC > 1.0 or < −1.0 and/or p-values < 0.05 with TGF-β treatment (Figure 4B). Six of the proteins increased while TN-C levels decreased compared to the control. Immunoblotting of CM and DOC lysates for IGFBP3 protein showed significant accumulation of protein with TGF-β treatment compared to CM and DOC lysates from untreated cells where IGFBP3 was not detectable (Figure 4C; donor 2 data not shown). The fold-change in IGFBP3 from RNA-seq could not be quantified by qPCR because, as with the protein expression, IGFBP3 mRNA was undetectable in untreated cells. TN-C was also analyzed because this protein has been implicated in the development and progression of fibrosis in a number of tissues, including in lung [29], but showed little change with TGF-β in RNA-seq data and a decrease in CM in the Luminex analysis. We confirmed by immunoblotting that TN-C levels were similar in CM with and without treatment (Supplementary Figure 5A). On the other hand, qPCR results (Supplementary Figure 5B) and analysis of TN-C levels in DOC lysates showed higher levels with TGF-β treatment for NHLFs from both donors (Supplementary Figure 5A; donor 2 data not shown). This difference between CM and the cell layer suggests that the lack of a protein change in CM may be due to excess TN-C being incorporated into the FN matrix. Also of note, the DOC-insoluble matrix contained multiple TN-C bands, especially in the TGF-β treated sample (Supplementary Figure 5A, right), suggesting that DOC-soluble and DOC-insoluble fractions may contain different splice variants of TN-C.

Interestingly, of the 80 core matrisome proteins that increased with TGF-β stimulation in our RNA-seq dataset, 25 (30%) have been shown to bind to FN (Supplementary Table 2). Of these, expression of 14 genes was increased, including POSTN, TNC, and FN1 itself, and 2 were decreased (DCN and FBLN1) with TGF-β (Table 1). Of the 7 core matrisome proteins that changed in the Luminex analysis, only IGFBP-7 has not been shown to bind FN (Supplementary Table 3), while the other 6 are in our FN-binding list from the RNA-seq analysis (Table 1). Other than core matrisome proteins, two of the identified soluble factors in the Luminex assay bind to FN (VEGF [30] and HGF [31]) and two extracellular proteases are known to cleave FN (MMP-9 [32] and cathepsin D [33]). Together, our results identify a set of ECM proteins that are both up-regulated with TGF-β stimulation and are documented binding partners for FN, suggesting the existence of a “FN-binding protein signature” that defines IPF.

Table 1. FN-binding core matrisome proteins identified by RNA-seq.

16 core matrisome genes that were significantly changed with TGF-β treatment encode FN-binding proteins. Genes are ranked by log2FC from RNA-seq data. Gene products that were also changed in the Luminex analysis are indicated. References cite publications that show FN-binding activity for each protein.

Gene name	Luminex assay	#Reference
IGFBP3	*	[49]
COL7A1		[50]
TNFAIP6		[51]
COL1A1		[7]
DPT		[52]
COL4A2	*	[7]
FN1	*	[5]
THBS2	*	[53]
FBN1		[44]
COL16A1		[54]
THBS1	*	[7]
COL1A2		[7]
POSTN	*	[55]
TNC	*	[56]
DCN		[57]
FBLN1		[46]

^#References show FN-binding activities.

^*Included in Luminex antibody panel.

For FN binding to promote development of fibrosis, one would expect FN levels to be high prior to up-regulation of its binding partners. To test this, we compared changes in the mRNA levels of FN and several other proteins over time. Our results show that FN levels are high at 6 h and increase significantly at subsequent time points (24 and 48 h) (Figure 5). IGFBP3, THBS1 and 2, POSTN, and TNC RNAs increased over time but never to the level of FN. FN protein was also significantly higher than other proteins in the Luminex assay (data not shown). These time course results indicate that FN matrix levels increase early after TGF-β stimulation, supporting the suggestion that sufficient FN matrix would be present to support deposition of other up-regulated core matrisome proteins. Thus, coordinated up-regulation of FN along with a set of FN-binding proteins could potentially increase the rate of progression and the extent of fibrosis in the lung.

Figure 5. Time course analysis of changes in core matrisome gene expression.

RNAs from control (Ctrl, blue) and TGFβ-treated (red) NHLFs from two donors (NHLF 1 and NHLF 2) were isolated after treatment for 6 h, 24 h and 48 h. Expression was quantified by qPCR for FN1, THBS1, IGFBP3, POSTN, TNC and THBS2. n=3, **p < 0.01, ***p < 0.001.

3.4. In vivo changes in ECM protein levels with IPF

To determine the relationship between ECM changes in NHLFs treated with TGF-β and in lung disease, we used a mouse model of IPF in which lung fibrosis is induced by treatment with bleomycin [34]. As fibrosis develops, changes in protein expression and deposition will differ between bleomycin- and control-treated mouse lungs. Excess ECM deposition is clearly detectable in bleomycin lung sections with H&E-positive staining area measurements showing a significant fibrotic region (Figure 6A, B). Immunostaining with anti-FN antibody compared to an isotype-matched control antibody demonstrates a dramatic increase in FN deposition with bleomycin treatment (Figure 6C). Both cell positivity and FN staining area were significantly higher (Figure 6D, E). If increases in protein expression are sufficiently high in the tissue, then secreted proteins may be detectable in bronchoalveolar lavage fluid (BALF). We analyzed protein levels in BALF collected 21 days after treatment and found up to 15-fold more total protein mass in BALF after bleomycin treatment (Supplementary Figure 6A). SDS-PAGE followed by silver staining for total protein detection showed dramatically higher levels of proteins in the 50 – 75 kDa range (Supplementary Figure 6B), most likely serum albumin and immunoglobulins from blood leakage. In addition, many higher molecular mass bands were present in the bleomycin-treated samples that were not detected in the saline controls. Excess FN in BALF could be explained by leakage of blood plasma. In fact, Luminex analyses of FN levels in plasma from healthy individuals compared to individuals with interstitial lung disease (ILD) showed that the concentration of FN was 2-fold higher on average in the ILD plasma (Supplementary Figure 6C), similar to a previous report [35]. To confirm that BALF contains not only FN from the blood but also tissue-produced proteins, we examined it for the presence of three ECM proteins that are not normally present in blood (TN-C and type I collagen) or are present at very low levels (IGFBP3). All three of these proteins were very high in BALF from treated mice while little if any protein was detected in control mouse BALF (Figure 6F). Anti-IGFBP3 staining of control and bleomycin-treated lung sections show dramatically higher IGFBP3 in the fibrotic lung ECM (Supplementary Figure 7). We conclude that increased IGFBP3 is deposited in the FN matrix of diseased lungs and that BALF contains tissue-derived proteins that have leaked from fibrotic regions of the lung. Taken together, these results show an overlap between proteins that are increased in the cell culture model with those that accumulate in vivo with lung fibrosis.

Figure 6. FN and FN-binding proteins are increased in bleomycin-treated mouse lung tissue.

Tissue sections of saline-treated and bleomycin-treated mouse lungs were stained with (A) H & E or with (C) hematoxylin and either anti-FN or isotype control antibodies as indicated. B) Staining intensities in (A) were quantified to identify the fibrotic areas which were significantly increased in the bleomycin lungs. D, E) Graphs show quantification of signals in (C) for cells positive for FN stain and the percent of tissue area stained for FN. No FN signal was detected in the isotype control sections. *p < 0.05, ***p < 0.001. N = 10 for A - E. F) BALF samples from 4 control and 4 bleomycin-treated mice were analyzed by immunoblotting for levels of FN, TN-C, type I collagen (Col I) and IGFBP3 as indicated. For Col I and IGFBP3 analyses, equal amounts of BALF total protein from saline-treated mice and from bleomycin-treated mice were analyzed. For FN and TN-C analyses, saline samples were electrophoresed without dilution, but bleomycin-treated samples were diluted 1:10 for immunoblotting. Locations of proteins and MW markers are indicated.

4. Discussion

As a major contributor to fibrosis, TGF-β enhances production of FN and other ECM molecules. FN forms a foundational ECM, and therefore excess deposition of FN could drive fibrosis through accumulation of other ECM proteins on the FN fibrils. To determine whether FN has a central role in progression of fibrosis, we analyzed FN levels in a mouse model of lung fibrosis and TGF-β stimulated NHLFs in culture. In vivo samples showed elevated FN protein in both soluble and matrix forms and TGF-β-stimulated NHLFs expressed significantly higher amounts of FN mRNA, secreted FN protein, and FN matrix. We also found that FN was increased in IPF patient plasma compared to healthy donors. Together, these findings show that up-regulation of FN is a common theme across these models of lung fibrosis. Gene and protein expression profiling by RNA-seq and Luminex assays, respectively, identified core matrisome and secreted factors that changed with TGF-β treatment of NHLFs, with IGFBP3 being the most highly up-regulated. Several of these ECM proteins were examined in BALF from mouse diseased lungs and found to be dramatically increased. Interestingly, of the NHLF core matrisome genes significantly up-regulated by TGF-β treatment, over 40% (14 out of 34) of the encoded proteins are known to bind to FN suggesting the idea that FN may direct the accumulation of these other core matrisome proteins. We propose that fibrosis progresses through expression of a “FN-binding protein signature” that initiates and propagates excess ECM deposition.

The accumulation of disorganized and insoluble matrix fibrils is the basis for fibrosis. Fibrillogenesis of ECM proteins is a cell-mediated process and a number of molecular mechanisms have been linked to increased assembly, besides up-regulation of expression of major ECM components like FN or collagen and their receptors. For example, chronically high glucose levels activate PKC signaling in kidney cells resulting in increased activity of α5β1 integrin, the receptor responsible for FN matrix assembly [9]. FN assembly is also modulated by lysine acetylation [10], the presence of certain ECM molecules such as TN-C or proteoglycans [17,36,37], changes in ERK and Rho GTPase signaling [38-40], and activity levels of focal adhesion proteins such as FAK and Src [41,42]. Thus, there are numerous mechanisms that can contribute to fibrosis. However, even with activation of intracellular pathways, how excess ECM proteins accumulate rather than turnover is not clear. Some evidence has been provided to suggest that the presence of EDA+ and EDB+ alternatively spliced isoforms of FN have fibrosis-specific roles [25]. However, we found that they increased only in proportion to total FN, but not in excess, making their contributions difficult to parse from the simple increase in FN itself.

In its role as a foundational matrix, increased numbers of FN fibrils would be able to promote excessive ECM deposition of a variety of ECM proteins. In fact, our results show not only increased expression and secretion of FN, but also dramatically more fibrillar FN matrix assembled by cells treated with TGF-β and also in bleomycin-treated mouse lungs. RNA and protein analyses show higher FN production prior to increases in other FN-binding proteins. Relevant to fibrosis, higher levels of FN matrix would then provide more binding sites for the FN-binding core matrisome proteins that are also increased in fibrotic conditions. An example of this role was provided by our analysis of type I collagen which was significantly increased in our RNA-seq dataset and also accumulated in association with FN fibrils in TGF-β-treated NHLFs. In addition to type I collagen, other up-regulated FN-binding proteins are known to take on a fibrillar organization in association with FN matrix, such as thrombospondin-1 [43], fibrillin-1 [44] and TN-C [45].

Unlike FN, collagens, and other fibrillar ECM proteins, IGFBP3 is a small protein and is not a structural component of matrix fibrils, and as expected, we detected it in the DOC-soluble matrix, but not in the DOC-insoluble FN fraction (Y.S, unpublished observations). IGFBP3 does have heparin-binding activity [28] and could bind to heparan sulfate proteoglycans and FN fibrils at the same time, perhaps providing a linkage that might organize fibrils in specific geometries. IGFBP3 was undetectable in both gene and protein expression levels in NHLFs without stimulation, but dramatically increased in response to TGF-β and was also abundant in BALF. IGFBP3 forms a complex with IGF1 [28] and therefore, increased production of both IGFBP3 and FN might have an impact on fibrosis development through increased incorporation of IGF1 into matrix.

Investigation of how up-regulated FN-binding proteins contribute to the progression of fibrosis may elucidate novel targets for diagnosis and treatment. But it is also possible that down-regulation of certain core matrisome proteins might exacerbate fibrosis. Fibulin-1 binds to FN’s heparin-binding domain [46] and we showed that binding to this domain of FN modulates cell-ECM interactions by blocking the binding of syndecan receptors [45,47]. Reducing syndecan-FN interactions affects activation of the intracellular signaling proteins FAK and Rho GTPase [47,48] resulting in decreased FN matrix assembly [47]. Perhaps down-regulation of fibulin-1, or other modulatory ECM proteins, might relieve this inhibitory activity such that cell-mediated assembly can proceed in excess.

5. Conclusions

Our studies, using a TGF-β-treated NHLF cell culture model and a bleomycin-treated mouse lung fibrosis model, identify a “FN-binding protein signature” linked to excess ECM deposition in IPF. Both FN expression and FN fibril formation were increased by TGF-β stimulation of NHLFs and these increases were confirmed in tissue sections and BALF samples from fibrotic mouse lungs and blood plasma from IPF patients. RNA-seq and Luminex immunoassay experiments were used to identify up-regulated gene expression and protein secretion, respectively, using stimulated NHLFs. Data analyses identified the core matrisome proteins expressed by these cells and, of these, 14 TGF-β-up-regulated proteins are known to bind to FN. Several of the FN-binding proteins were tested and found to be elevated in BALF. The most highly expressed gene and protein was IGFBP3 which may contribute to fibrosis by localizing IGF1 to the FN matrix. Other proteins in the FN-binding signature have roles in adhesion modulation, matrix structure and stability, and cell-ECM interactions suggesting that this signature participates in multiple processes involved in the progression of fibrosis.

Abbreviations

ECM

extracellular matrix

FN

fibronectin

NHLF

normal human lung fibroblasts

DHLF

diseased human lung fibroblasts

BALF

bronchoalveolar lavage fluid

IPF

idiopathic pulmonary fibrosis

CM

conditioned medium

TGF-β

transforming growth factor-β

DOC

deoxycholate