1. Introduction
Idiopathic pulmonary fibrosis (IPF) is a chronic and fatal interstitial lung disease, characterized by rapid progression of lung fibrosis with usual interstitial pneumonia (UIP) as the typical histologic pattern [
1,
2]. The prognosis of IPF is dismal. The median survival of IPF is around 2–3 years from diagnosis [
3,
4]. The cause of IPF remains unclear. It is believed that gene-environmental interactions involve in the pathogenesis of IPF. Repetitive environmental stimuli in susceptible individuals lead to aberrant injury-repairing process. Lung epithelial cells are activated and produce pro-fibrotic growth factors and other mediators, which promote epithelial-mesenchymal transition, and activation and differentiation of fibroblasts and myofibroblasts. The activated fibroblasts and myofibroblasts secrete excessive extracellular matrix proteins, resulting in destruction and scaring of lung architecture [
5,
6,
7,
8].
To date, therapeutic options for IPF remain limited. Nintedanib and pirfenidone are the only two drugs demonstrated to be effective for IPF [
9]. Nintedanib, approved in 2014 in the United States and in 2015 in Europe for IPF treatment, is a multi-target tyrosine kinase inhibitor. The main molecular targets of nintedanib are the fibroblast growth factor receptor (FGFR), the vascular endothelial growth factor receptor (VEGFR), and the platelet-derived growth factor receptor (PDGFR) [
10,
11]. By competitively and reversibly blocking the receptor tyrosine kinases FGFR, VEGFR, and PDGFR, nintedanib inhibits fibroblasts proliferation, migration and transformation to myofibroblasts [
12,
13].
IPF is considered as a complex genetic disorder [
14]. Several genetic studies has demonstrated the possible dysregulated genes associated with the development of IFP [
15,
16,
17]. Plantier and colleagues used individual data from four microarray studies to compare the transcriptome of IPF versus normal lung fibroblasts. They found downregulated genes in IPF were highly related to inflammation and immunity, and upregulated
CTGF and
SRF may play key roles in lung fibrogenesis [
18]. How anti-fibrotic agents change gene regulations in IPF, however, has never been systematically investigated. Recently, Kwapiszewska and colleagues compared transcriptomic profiles in IPF lungs obtained from patients treated with or without pirfenidone [
19]. They found CEMIP (cell migration-inducing and hyaluronan-binding protein), which has been previously associated with ECM production, was strongly downregulated by pirfenidone treatment.
The rapid evolution of next-generation sequencing (NGS) technology provides a powerful tool for systematic analysis of transcriptomic profiles [
20,
21]. After creating initial testing strategy and generating sequencing data, investigator could survey for the differential gene expression in the target cells with or without treatment. In order to understand the changes of gene regulation associated with nintedanib treatment in IPF, we conducted a study using the NGS technology and various bioinformatic tools to systematically evaluate the changes of mRNA and miRNA (microRNA) profiles in IPF fibroblasts treated with 2 µM and 4 µM nintedanib, and without treatment. This study not only enhanced current understanding of nintedanib molecular mechanisms, but also provided useful information for future research on pharmacogenomics.
2. Experimental Section
2.1. Study Design
The flowchart of study design is illustrated in
Figure 1. IPF fibroblasts (the second passage) were cultured with 0.1% DMSO (control), 2 µM and 4 µM nintedanib for 24 h, and harvested for RNA sequencing and expression profiling using the NGS platform. The significantly upregulated and downregulated mRNAs were analyzed with bioinformatic tools including Search Tool for the Retrieval of Interacting Genes (STRING) database and the Database for Annotation, Visualization and Integrated Discovery (DAVID) database to investigate functions and pathways associated with nintedanib treatment. In addition, the upregulated and downregulated miRNAs were analyzed with miRmap for target prediction, and went through Venn diagram to determine genes with potential miRNA–mRNA interactions. These potential miRNA–mRNA interactions were further validated by another prediction database TargetScan and miRDB. Finally, the dysregulations of miRNAs and mRNAs were validated using reverse transcription quantitative polymerase chain reaction (RT-qPCR).
2.2. Cultures of IPF Lung Fibroblasts
Human IPF lung fibroblasts (Disease Human Lung Fibroblasts, Idiopathic Pulmonary Fibrosis. Catalog No. CC-7231), purchased from Lonza Inc. (Walkersville, MD, USA), were incubated at 37 °C in a 5% CO2-containing incubator in FGM™-2 Fibroblast Growth Medium-2 Bulletkit™ (Lonza, Catalog No. CC-3132) containing 0.5 mL human fibroblast growth factor-basic (hFGF-B), 0.5 mL insulin, 10 mL fetal bovine serum, and 0.5 mL GA-1000. The medium was changed every 2–3 days and the cells were passaged at 80–90% confluence for the following experiments.
2.3. Nintedanib Treatment
Nintedanib (BIBF1120, Catalog No. S1010), purchased from Selleckchem (Houston, TX, USA), was dissolved in dimethyl sulfoxide (DMSO) (Sigma Chemical Co., St. Louis, MO, USA) to obtain various concentrations. The reagents were stored at −20 °C until use in the experiments. In nintedanib-treated cultures, cells were treated with 1 µM, 2 µM, and 4 µM nintedanib for 24 h, 48 h, and 72 h. In control cultures, cells were treated with the carrier solvent (0.1% DMSO). All experiments were performed in triplicate. The selection of nintedanib concentrations of 1–4 µM in this study was mainly based on previous study evaluating the effects of nintedanib on IPF fibroblasts [
22].
2.4. Cell Morphological Observation
Human IPF lung fibroblasts were incubated in 6 cm plates (1 × 105 cells/plate). When the confluence reached 80%, carrier solvent (0.1% DMSO), 1 µM, 2 µM, or 4 µM nintedanib were added into plates. Cell morphological changes were observed and photographed with an inverted-phase contrast microscope at 24 h, 48 h, and 72 h after nintedanib treatment.
2.5. Cell Proliferation Assay
Cell proliferation and viability were assessed using the Premix WST-1 Cell Proliferation Assay System (Takara Bio Inc., Shiga, Japan. Catalog No. MK400), according to the manufacturer’s instructions. In brief, the Premix WST-1 assay measures cell proliferation and cell viability with a colorimetric assay, based on cleavage of tetrazolium salts by mitochondrial dehydrogenase in viable cells. Human IPF lung fibroblasts seeded in the 96-well plates (5 × 103 cells/well) were incubated to confluence of 80% and treated with carrier solvent (0.1% DMSO), 1 µM, 2 µM, or 4 µM nintedanib for 24–72 h. Then 10 µL Premixed WST-1 reagent was added into each well, followed by incubation at 37 °C for 30 to 150 min. Quantification of cell proliferation and viability was measured by detecting the absorbance at 450 nm using a multi-well plate reader (FLX80, BioTek, Winooski, VT, USA). Cell proliferation was expressed as percentage, compared with controls (proliferation % = absorbance in the experimental group/absorbance in the control group × 100%).
2.6. Cell Apoptosis Assay
Cell apoptosis was assessed using the Annexin V-FITC Early Apoptosis Detection Kit (Cell signaling Technology, Beverly, MA, USA. Catalog No. 6592S), according to the manufacturer’s instructions. In brief, human IPF lung fibroblasts seeded in the 6 cm plates (1 × 105 cells/well) were incubated to confluence of 80% and treated with carrier solvent (0.1% DMSO), 1 µM, 2 µM, or 4 µM nintedanib for 24–72 h. Harvested cells were double stained with Annexin V and propidium iodide (PI), then analyzed by the BD Accuri™ C6 Plus personal flow cytometer (Becton Dickinson, San Josè, CA, USA). Cell apoptosis was expressed as the percentage of cells with Annexin-V and PI double staining in total cells.
2.7. Next-Generation Sequencing (NGS) for miRNA and mRNA Expression Profiling
The expression profiles of miRNAs and mRNAs were examined by using an NGS platform. Total RNA was extracted by TRIzol
® Reagent (Thermo Fisher Scientific, Waltham, MA, USA. Catalog No. 15596018) according to the instruction manual. Purified RNA was quantified at OD
260 nm by using a ND-1000 spectrophotometer (Nanodrop Technologies, Wilmington, DE, USA) and qualitated by using a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA) with RNA 6000 LabChip
® kit (Agilent Technologies, Santa Clara, CA, USA). The small RNA library construction and deep sequencing were carried out at the biotechnology company Welgene (Taipei, Taiwan). Samples were prepared using Illumina sample preparation kit according to the TruSeq Small RNA Sample Preparation Guide. The 3′ and 5′ adaptors were ligated to total RNA and reverse transcription followed by PCR amplification. The enriched cDNA constructs were size-fractionated and purified on a 6% polyacrylamide gel electrophoresis and the bands containing the 18–40 nucleotide RNA fragments (140–155 nucleotide in length with both adapters). Libraries were sequenced on an Illumina instrument (75 single-end cycles). Sequencing data was processed with the Illumina software. Small RNA sequencing data was also obtained and analyzed. Initially, the sequences generated went through a filtering process to obtain qualified reads. Trimmomatic (version 0.36) [
23] was implemented to trim or remove the reads according to the quality score. Qualified reads after filtering low-quality data were analyzed using miRDeep2 [
24] to clip the 3′ adapter sequence and discarding reads shorter than 18 nucleotides, before aligning reads to the human genome (version: GRCh38.p10) from UCSC (University of California, Santa Cruz, CA, USA). Only reads that mapped perfectly to the genome five or less times were used for miRNA detection, since miRNAs usually map to a limited number of genomic locations. MiRDeep2 estimates expression levels of known miRNAs, and also identifies novel miRNAs. The miRNAs with low levels (<1 normalized read per million (rpm)) were excluded, and those with >2-fold changes (rpm in 2 µM nintedanib-treated IPF fibroblasts/rpm in controls) in a dose-dependent manner (fold change in 4 µM nintedanib-treated IPF fibroblasts > fold change in 2 µM nintedanib-treated IPF fibroblasts) were considered dysregulated.
For transcriptome sequencing, the Agilent’s SureSelect Strand Specific RNA Library Preparation Kit was used for preparing libraries. Then the libraries were sequenced on a Solexa platform (150 paired-end cycles), using the TruSeq SBS Kit (Illumina, Inc. San Diego, CA, USA). Similar with small RNA sequencing data, reads with low quality score were trimmed and removed using the Trimmomatic (version 0.36). Qualified reads were analyzed using HISAT2 (version 2.1.0) [
25], and genes with low expression levels (<0.3 fragment per kilobase of transcript per million mapped reads (FPKM)) in any groups were excluded.
p values were calculated by cuffdiff (Cufflinks version 2.2.1) with non-grouped samples using the “blind mode”, in which all samples were treated as replicates of a single global "condition" and used to build one model for statistical test [
26,
27]. To correct for multiple testing, False Discovery Rate (FDR)
p values were calculated. Genes with FDR
p values <0.05 (i.e., −log10 (FDR
p) > 1.3) in 4 µM nintedanib-treated IPF fibroblasts and >2-fold changes (FPKM in 4 µM nintedanib-treated IPF fibroblasts / FPKM in controls) in a dose-dependent manner (fold change in 4 µM nintedanib-treated IPF fibroblasts > fold change in 2 µM nintedanib-treated IPF fibroblasts) were considered as significantly dysregulated genes. For analysis of functions and pathways associated with nintedanib treatment, we uploaded all significantly dysregulated genes into the STRING and DAVID database.
2.8. miRmap Database Analysis
miRNA have the ability to repress the expression of protein-coding genes. miRmap is an open-source software library which can provide the prediction of comprehensive miRNA targets (
http://mirmap.ezlab.org/) [
28]. By calculating the complementary ability of miRNA–mRNA interactions, the putative target genes could be identified. miRmap uses high-throughput experimental data from immunopurification, transcriptomics, proteomics, and polysome fractionation experiments to examine feature correlations and also compare the predictive power. The prediction results provide a list of putative target genes with miRmap score, which is a predictive reference value. In this study, the criteria for selection of putative miRNA targets were miRmap score ≥99.0.
2.9. TargetScan Database Analysis
TargetScan predicts the target of miRNA by searching for the presence of conserved 8mer, 7mer, and 6mer sites matching the seed region of each miRNA (
http://www.targetscan.org). This is an online database. The results of predictions are ranked by the predicted efficacy of targeting or by their probability of conserved targeting [
29]. TargetScan could provide a valuable resource for investigating the role of miRNAs in gene-regulatory networks.
2.10. miRDB Database Analysis
miRDB provides web-based microRNA-target prediction and functional annotations (
http://mirdb.org) [
30,
31]. In miRDB, all the targets were predicted by MirTarget. MirTarget was developed by analyzing microRNA-target interactions from high-throughput sequencing experiments. miRDB can predict microRNA targets in five species, including human, mouse, rat, dog, and chicken.
2.11. STRING Database Analysis
The functional interactions between expressed proteins in cells are very important and complicated. The Search Tool for the Retrieval of Interacting Genes (STRING) database (
https://string-db.org/) has collected and integrated this information, by consolidating known and predicted protein–protein association data for many different organisms. STRING offers the information of the associations include direct (physical) interactions and indirect (functional) interactions. Interactions in STRING are derived from five main sources: Conserved Co-Expressions, High-throughput Lab Experiment, Genomic Context Predictions, Automated Text-mining, and Previous Knowledge in Database [
32].
2.12. DAVID Database Analysis
The Database for Annotation, Visualization and Integrated Discovery (DAVID) is a powerful tool for gene functional classification (
https://david.ncifcrf.gov/) [
33]. It integrates Gene Ontology (GO), biological process and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway. In DAVID database, a list of interesting genes can be classified into clusters of related biological functions, signaling pathways, or diseases by calculating the similarity of global annotation profiles with agglomeration algorithm method. The criteria of Expression Analysis Systematic Explorer (EASE) score in DAVID database is a modified Fisher’s Exact
p value. The reference score represents how specifically the user genes are involved in the category (for example: signaling pathways). In this study, we selected EASE score = 0.1 as default and define pathways with a Bonferroni
p value < 0.05 as significant.
2.13. RT-qPCR
The expression levels of mRNAs and miRNAs in IPF fibroblasts treated with 0.1% DMSO (control) and 4 µM nintedanib for 24 h were assessed with RT-qPCR. In brief, total RNA was extracted using TRIzol
® Reagent (Thermo Fisher Scientific). The mRNAs and miRNAs were reverse transcribed using the PrimeScript RT reagent Kit (Perfect Real Time) (Takara Bio USA, Mountain View, CA, USA) and the Mir-X miRNA First-Strand Synthesis Kit (Takara Bio, USA. Catalog No. 638315), respectively. The mRNA and miRNA expressions were assessed using real-time analysis with Fast SYBR Green Master Mix (Thermo Fisher Scientific. Catalog No. 4385612) catalog number: 4385612)) on the QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific). The primers used are listed in
Supplementary Table S1. The relative expression levels of the mRNAs and miRNAs were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and U6 small nuclear RNA (snRNA), respectively. The relative standard curve method (2
−ΔΔCt) was used to determine relative mRNA and miRNA expression levels.
4. Discussion
IPF is the most common among idiopathic interstitial pneumonias. The incidence of IPF ranges between 0.22 and 17.4 per 100,000 population and the patients with IPF are usually in the sixth or seventh decade of life [
12]. Because of the increasing incidence and the lethal nature of this disease, more attention has been drawn to develop effective treatment for IPF. Nintedanib, a multi-target tyrosine kinase inhibitor, has been demonstrated to retardate the disease progression in IPF patients [
34]. Nintedanib is able to inhibit TGF-β signaling, downregulate extracellular matrix (ECM) gene/protein expression and promote noncanonical autophagy [
22]. To further investigate the molecular mechanisms associated with nintedanib treatment in IPF, we used NGS to scan mRNA and miRNA changes in primary human IPF fibroblasts treated with 2 µM and 4 µM nintedanib, compared to those without treatment. We identified 151 significantly upregulated genes and 157 significantly downregulated genes associated with nintedanib treatment in a dose-dependent manner. These 308 dysregulated genes were then analyzed in different bioinformatic tools for their protein–protein interactions, biological pathways, and molecular functions.
The STRING analysis showed strong protein–protein interactions within the top dysregulated genes. Interestingly, most genes in the PPI network were related to biological pathways of cell cycle, cell cycle process, mitotic cell cycle process, and mitotic cell cycle. The DAVID analysis revealed that the role of nintedanib in treating IPF fibroblast might most possibly relate to the biological process of cell division, mitotic nuclear division, sister chromatid cohesion, and DNA replication. We also identified 7 upregulated genes and 5 downregulated genes with potential miRNA–mRNA interactions. After validation in TargetScan and miRDB, upregulated SLC25A23 (might be de-repressed by downregulated hsa-miR-92a-1-5p) and downregulated DDX11, E2F1, NPTX1, and PLXNA4 (might be repressed by the upregulated hsa-miR-486-3p) were validated as the dysregulated genes with potential miRNA–mRNA interactions associated with nintedanib treatment. The dysregulations of these mRNAs and miRNAs were further validated with RT-qPCR, except for hsa-miR-92a-1-5p.
Our data showed that
DDX11 was downregulated after nintedanib treatment.
DDX11 is the predicted target of hsa-miR-486-3p, which is upregulated in nintedanib-treated IPF fibroblasts.
DDX11, also known as CHL-1, is able to promote cell proliferation, metastasis, and migration in human glioma cells [
35]. Although no previous study investigating the role of
DDX11 in IPF fibroblasts, our data suggest that nintedanib may suppress
DDX11 in IPF fibroblasts and therefore suppress the proliferation of IPF fibroblasts. According to our NGS data, the expression change of
DDX1 in nintedanib-treated IPF fibroblasts was in a dose-dependent manner. Therefore, we hypothesized that nintedanib might decreased proliferation of IPF fibroblasts through regulation of hsa-miR-486-3p-
DDX11.
We found that nintedanib treatment was associated with downregulation of
PLXNA4.
PLXNA4 is also the predicted target of the upregulated hsa-miR-486-3p in nintedanib-treated IPF fibroblasts. PLXNA4 may promote progression and angiogenesis of tumor, possibly by enhancement of VEGF and bFGF signaling pathways [
36]. Downregulation of
PLXNA4 may be associated with decreased angiogenesis. Therefore, our data suggest the changed expressions of hsa-miR-486-3p-
PLXNA4 associated with nintedanib treatment may contribute to decreased angiogenesis in IPF.
Our data also showed that
E2F1 was downregulated after nintedanib treatment. Similar to
DDX11 and
PLXNA4,
E2F1 is the predicted target of hsa-miR-486-3p. E2F1 protein is a member of the E2F family of transcription factors. The E2F1 transcription factor play critical roles in the regulation of cell cycle and apoptosis [
37]. It is considered as one of the most important protein required for entering into the S phase during cell cycle progression [
38]. Several growth factors including epidermal growth factor (EGF), keratinocyte growth factor (KGF), and VEGF exert their proliferating biological activity by increasing expression of
E2F1 [
37,
39,
40,
41]. It is not surprising to know that treatment with nintedanib, which targets many growth factor receptors, was associated with downregulation of
E2F1. Our interaction analysis suggests that upregulation of hsa-miR-486-3p might have an interaction with downregulation of
E2F1 in nintedanib-treated IPF fibroblasts. However, this hsa-miR-486-3p-
E2F1 interaction was predicted from the bioinformatic database. How hsa-miR-486-3p, E2F1 transcription factors, and growth factors interact with each other in nintedanib treatment need further studies to confirm and elucidate.
Another downregulated gene
NPTX1 was also the predicted target of hsa-miR-486-3p. The protein encoded by
NPTX1 is a member of the neuronal pentraxin family. The expression of this gene is mainly localized to the nervous system [
42]. The only upregulated gene with miRNA–mRNA interaction found in this study was
SLC25A23, which might have an interaction with the downregulated miRNA hsa-miR-92a-1-5p based on our bioinformatic prediction.
SLC25A23 is a member of the mitochondrial transporter family SLC25 embedded in the inner mitochondrial membrane [
43].
SLC25A23 has been shown to augment mitochondrial matrix Ca
2+ influx, and may induce oxidative stress-mediated cell death [
44]. Although the function of this gene remains largely unknown, it is possible that nintedanib treatment downregulated hsa-miR-92a-1-5p which in turn de-repressed
SLC25A23 and enhance cell death of fibroblasts. However, the downregulation of hsa-miR-92a-1-5p in nintedanib-treated IPF fibroblasts was not confirmed by the RT-qPCR method.
To the best of our knowledge, this is the first study using NGS and bioinformatics to scan mRNA and miRNA changes following nintedanib treatment. We identified 308 significantly dysregulated genes associated with nintedanib treatment in IPF fibroblasts. We found strong protein–protein interactions within these dysregulated genes, mostly involved in the pathways of cell cycle and mitotic cell cycle. We also identified 1 upregulated and 4 downregulated genes with potential miRNA–mRNA interactions. According to our data from nintedanib-treated IPF fibroblasts and the literatures, we proposed that upregulation of hsa-miR-486-3p following nintedanib treatment might be associated with decreased expression of
DDX11, E2F1, and
PLXNA4. These gene expression changes together might contribute to decreased proliferation of fibroblasts and decreased angiogenesis in the microenvironment of IPF (
Figure 8). Our findings improve current understandings of molecular mechanisms of nintedanib treatment in IPF. Our approach and results may also provide useful information for future research on nintedanib pharmacogenomics. Nevertheless, there remain several limitations in this study. Firstly, as nintedanib is known to interfere with VEGFR, FGFR, and PDGFR, and potentially also with TGF-β signaling, its therapeutic effects will likely be observed when cells are stimulated with VEGF, FGF, PDGF, and/or TGF-β. The culture medium used in this study contains only FGF, but VEGF and PDGF might also be present in the fetal bovine serum in unknown quantities. The FGF-containing culture condition might somewhat explain our findings that nintedanib-induced gene expression changes were mostly related to pathways of cell cycle, instead of pathways of TGF-β signaling, ECM, and autophagy. Considering that results would be different when fibroblasts were stimulated with different growth factors in the medium, plus there is no perfect culture condition mimicking the true in vivo microenvironment of IPF lung, we decided to cultivate the IPF lung fibroblasts using the standard condition as suggested by Lonza. Secondly, the dysregulated miRNAs, mRNAs and their interactions found in this study were based on primary IPF lung fibroblasts, RNA deep sequencing, bioinformatics, and RT-qPCR. The dysregulations of these genes need to be confirmed by other methods and in more clinical samples. The relationships between miRNAs and gene expressions found in this study also need be confirmed in future cell-based assays.