Idiopathic pulmonary fibrosis (IPF) is a type of chronic, progressive, fibrosing interstitial pneumonia, and its etiology and pathogenesis are not yet fully understood (Richeldi et al., 2017). Its key pathological characteristics include alveolar epithelial cell (AEC) injury, inflammatory cell infiltration, massive extracellular matrix accumulation, epithelial–mesenchymal transition, and fibroblast transformation to myofibroblasts. These detrimental pathological effects eventually result in an irreversible and progressive respiratory insufficiency (Reddy et al., 2016; Colunga Biancatelli et al., 2020). Some of the risk factors related to the incidence of IPF include smoking and old age (Hill et al., 2019; Rahaghi et al., 2020), and more men are affected by IPF than women. The occurrence rate of IPF and its related mortality rate increase substantially with age (Ebner et al., 2020; Phan et al., 2021); two-thirds of patients with IPF are aged 60 years or more at the onset of the disease and the average age at the time of diagnosis is 66 years; the estimated prevalence of IPF among individuals over 65 years of age may be as high as 94 per 100,000 people (Hecker et al., 2014; Ryerson et al., 2016; López-Ramírez et al., 2018; Kishaba, 2019; Lee et al., 2020). At present, the key treatment approaches for IPF are based mainly on the symptoms and have low therapeutic effects; the 5 year survival rate after diagnosis is less than 50%, and the median survival time is only 2–3 years (Kistler et al., 2014; Yoshihara et al., 2020). Increasing evidence shows that different biological functions such as cell proliferation, apoptosis, senescence, and autophagy play a role in IPF pathogenesis (Predescu et al., 2017; Cong et al., 2020; Krempaska et al., 2020). Among these, cell senescence plays a major role.

Cell senescence is defined as an irreversible inhibition of cell proliferation wherein the cell cycle is usually arrested permanently in the G0 or G1 phase. Changes in cell morphology include cell flattening, nuclear enlargement, and chromatin aggregation (Anwar et al., 2016). Another prominent feature of senescent cells is that they secrete a variety of cytokines, chemokines, growth factors, and matrix metalloproteinases, thereby constituting the senescence-associated secretory phenotype (SASP) (Huda et al., 2019). Increasing evidence supports a correlation between cellular senescence and IPF pathogenesis. For example, Hecker et al. (2014) showed that persistent fibrosis in the lung tissues of aged mice was characterized by an accumulation of senescent and apoptosis-resistant myofibroblasts, and these mice demonstrated an impaired capacity for fibrosis resolution. Yanai et al. (2015) suggested that cellular senescence could serve as a bridge connecting lung aging and pulmonary fibrosis, and is a crucial factor in disease progression. Álvarez et al. (2017) found that IPF human lung fibroblasts developed senescence leading to decreased apoptosis, and the development of the SASP might be a critical contributor to the fibrotic process observed in IPF. Additionally, some studies have highlighted that a variety of aging-related pathways are activated in epithelial cells, fibroblasts, and progenitor cells in the lungs of patients with IPF (Mora et al., 2017; Chen et al., 2019); these activated pathways promote abnormal secretory phenotype in lung epithelial cells, augment the resistance of myofibroblasts to apoptosis, and accelerate IPF progression. In the lung biopsies of patients with IPF, the level of SA-β-gal, a specific cellular senescence marker, was increased compared to that in patients with chronic obstructive pulmonary disease or hypersensitivity pneumonitis (Kellogg et al., 2021). Kim et al. (2021) evaluated the genomic profile of fibrotic and normal lung tissues and found that the core molecular network of IPF featured p53 signaling pathway and cellular senescence. Nevertheless, it remains unknown which aging-related genes are critical for the development of IPF and thus, the correlation between IPF and aging-related genes is yet to be understood. Further studies are needed to determine new biomarkers for the treatment of IPF based on potential aging-related genes involved in IPF.

GSE150910 is an IPF-related data set with a large sample size (103 IPF lung and 103 normal lung tissues). Furusawa et al. (2020) completed a sequencing analysis of GSE150910 and revealed 1,183 differentially expressed mRNAs between IPF and normal lung tissues. In this study, we aimed to analyze the GSE150910 data set from different perspectives. The differential expression of aging-related IPF genes was determined by bioinformatic methods using limma test, protein-protein interaction (PPI) analysis, correlation analysis, gene ontology (GO) enrichment analysis, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. Then, machine learning approaches were used for extensive filtration and diagnostic IPF molecular marker identification. Finally, the expression levels of aging-related genes screened by machine learning approaches was validated using other Gene Expression Omnibus (GEO) datasets and clinical samples.

IPF is a progressive aging-related disease characterized by the replicative senescence of lung epithelial cells, with lung epithelial II cell senescence being the key manifestation. Abnormally activated AECs induce the expansion and activation of fibroblast populations, causing deterioration of the lung structure (Jiang et al., 2017). Accumulating evidence indicates that senescent cells are metabolically active and secrete a large number of leukocyte mediators, including interleukin (IL)-1β, IL-6, and IL-8, which induce the differentiation of lung fibroblasts into myofibroblasts, thereby promoting IPF onset (Sugihara et al., 2020; Xu et al., 2020). Nevertheless, extensive validation is essential to better understand the role of cellular senescence in IPF pathogenesis.

Recent studies have explored the correlation between PBMCs, lung tissue RNAs, and IPF. For example, Herazo-Maya et al. (2013) performed a microarray expression profile analysis of RNAs expressed in PBMCs obtained from healthy individuals and patients with IPF, and identified some of the differentially expressed mRNAs that included a series of aging-related mRNAs, such as CD28, ICOS, LCK, and ITK. Moreover, Huang et al. (2014) revealed that the expression level of LYCAT was greatly reduced in the PBMCs and lung tissues of patients with IPF, as well as in a mouse model of bleomycin- and radiation-induced IPF. Hence, the PBMCs of patients with IPF were also used to verify the results of the bioinformatic analysis in this study. Gao et al. (2017) reported that the expression of the aging-related gene telomeric repeat-containing RNA (TERRA) in the PBMCs of patients with IPF was significantly enhanced and showed a substantial negative correlation with the percentage of predicted forced vital capacity. This finding might be attributed to the ability of TERRA to regulate the expression of telomeres and mitochondria, suggesting that aging plays an important role in IPF. Nevertheless, very few studies have explored this field, and further investigations are needed to better understand the role played by cellular senescence in IPF.

A few cancer-based studies have analyzed aging-related genes (Jia et al., 2018; Avelar et al., 2020; Song et al., 2020; Yue et al., 2021). For example, a recent study highlighted the role of 15 aging-related genes in colorectal cancer, among which some were associated with disease prognosis (Yue et al., 2021). However, a bioinformatic analysis of aging-related genes in IPF has not yet been performed. In the present study, 16 potential aging-related genes associated with IPF were identified using bioinformatic analysis. Of these, some genes have been studied previously. For instance, Hernandez et al. (2020) demonstrated that IGF1 expression was upregulated in myofibroblasts and the fibrotic lung tissue by transforming growth factor β, whose expression was positively correlated with deteriorated lung function in advanced IPF; Guiot et al. (2016) found that the expression of IGFBP2 was substantially higher in the serum of patients with IPF than that in healthy participants, but was reduced in the serum of patients receiving specific anti-fibrosis therapy. These results support the findings of our bioinformatic analysis and suggest that IGFBP2 may be a promising biomarker of IPF. Moreover, Toren et al. (2021) investigated the relationship between lung fibrosis-related and longevity-associated genes in a mouse model of bleomycin-induced fibrosis; they highlighted age as an important risk factor for pulmonary fibrosis and found that pro-longevity genes tended to be related to anti-fibrosis. Our results are partially consistent with these findings. Notably, KL (Klotho) gene, which is a pro-longevity gene, was found to be overexpressed in healthy control lung tissues in our study. Consistently, (Toren et al., 2021) also demonstrated that KL is a pro-longevity and anti-fibrotic gene. However, other genes screened in our bioinformatic analysis were not mentioned in the study by Toren et al. This discrepancy could be due to the differences in the study design and aims of the two studies. We aim to explore more aging-associated genes that could potentially be related to IPF in the future.

The GO and KEGG enrichment analyses were also performed in this study to investigate the potential biological functions of these differentially expressed aging-related genes. Our results highlight the involvement of these genes in senescence and mitochondrial apoptosis. Consistently, accumulating evidence supports that IPF progression is affected by cellular senescence. For example, extracellular vesicles containing microRNA-23b-3p and microRNA-495-3p derived from lung fibroblasts increase reactive oxygen species levels in mitochondria and cause mitochondria-related damage in lung epithelial cells, resulting in DNA damage and subsequently, epithelial cell senescence (Kadota et al., 2020); another recent study reported that the PTEN/NF-κB pathway in senescent AECs facilitated the accumulation of collagen in fibroblasts, leading to lung fibrosis (Tian et al., 2019). Moreover, the KEGG pathway enrichment analysis suggested that these aging-related genes were mainly involved in endocrine resistance and the MAPK pathway. Endocrine resistance is a common diabetes-related phenomenon (Ahmad et al., 2014); diabetic lung damage is a part of the multi-system disease, and the lungs are one of the target organs of diabetic damage. Additionally, lung tissue abnormalities and pathophysiological changes caused by diabetes result in tissue fibrosis, such as diabetic nephropathy and diabetic cardiomyopathy. Notably, diabetes can greatly enhance the risk of pulmonary fibrosis, indicating that pulmonary fibrosis may be a complication of diabetes (Irfan et al., 2011; V. ; Kumar et al., 2020; Wang et al., 2020). Furthermore, Fang et al. (2018) demonstrated that p38 MAPK is essential for the epithelial–mesenchymal transition of AECs induced by Wnt1 and lithium chloride, and Antoniou et al. (2010) proposed the involvement of the MAPK signaling pathway in IPF pathogenesis. These findings are consistent with the results of our KEGG enrichment pathway analysis. However, further studies are needed to explore the biological functions of differentially expressed aging-related genes.

SVM-RFE is a machine learning technique based on an SVM, which searches for the best variable by subtracting the feature vector generated by the SVM (Wan et al., 2019). The advantage of the SVM is that an SVM classifier depends only on the support vectors, and the classifier function is not influenced by the entire dataset. LASSO regression is another machine learning technique that identifies the variable by finding the value of λ when the classification error is the least (Zhao et al., 2020). These two algorithms are primarily used to select characteristic variables and create the best classification model. Unlike the conventional univariate analysis, the LASSO regression used in this study aimed to select variables for logistic regression to avoid overfitting. After screening based on these two methods, six aging-related genes (IGF1, RET, IGFBP2, CDKN2A, JUN, and TFAP2A) were identified. The random forest algorithm provided an important metric for these aging-related hub genes. These genes could be used for improved IPF diagnosis; the ROC curves of GSE150910 and GSE32537 datasets had higher AUC values, suggesting that these six genes are potential diagnostic biomarkers for IPF.

To verify the results of the bioinformatic analysis, blood samples were collected from 20 patients with IPF and 20 healthy individuals, and qRT-PCR was performed to identify the six differentially expressed aging-related genes screened by LASSO regression and SVM-RFE. We observed that the expression levels of IGF1, RET, and IGFBP2 in IPF blood samples were considerably enhanced compared to those in the blood samples of healthy individuals, while expression levels of CDKN2A, JUN, and TFAP2A were comparable between the two groups. Notably, studies have suggested that IGF1, RET, and IGFBP2 are closely related to IPF. For example, Sun et al. (2021) reported that PI3K/AKT signal activation induced by IGF1 was involved in the aging of type II AECs and IPF progression by inducing the release of connective tissue growth factor, transforming growth factor-beta 1, and matrix metalloproteinase 9; Tran et al. (2014) showed that continuous IGF1 treatment inhibited the biological activity of SIRT1 deacetylase, which induces p53 acetylation and increases the stability of p53 and biological activity, thereby enhancing senescence in immature cells; Guiot et al. (2017) reported that the concentration of insulin-like growth factor binding protein (IGFBP)-2 in the sputum supernatant of patients with IPF was greatly increased compared to that in healthy participants. The mechanism may be related to cell senescence with SASP, which involves the secretion of soluble factors, such as IGFBPs that perform extracellular and intracellular functions in an IGF-dependent or -independent manner. Interestingly, while extracellular IGFBP2 counter-regulates IGF-induced cell hyperproliferation, apoptosis is inhibited by intracellular IGFBP2 via its interaction with p21 to protect itself from ubiquitin-dependent degradation (Mercurio et al., 2020).

To date, there are no reports on the role of RET in IPF. RET is a new driver gene discovered after EGFR and ALK in non-small cell lung cancer. RET fusion may be responsible for the pathogenesis in patients with lung cancer without mutations in EGFR, ALK, or ROS1, which may be related to the RET-induced promotion of apoptosis resistance (Drilon et al., 2020). Additionally, the experimental verification results of JUN and TFAP2A differed from the results of biological information analysis, probably due to the small sample size in this study. Moreover, no major difference was observed in CDKN2A mRNA expression between patients with IPF and healthy participants in this study. However, Du et al. (2018) showed that the CDKN2A expression level was lower in the peripheral blood of patients with IPF, whereas it was higher in healthy controls. This finding is different from the results of our present study; this discrepancy could be due to the differences in the sample source and the composition of study participants. The study by Du et al. included only male patients, whereas both male and female patients were included in our study. In addition, IPF severity differed in patients of both the studies. Furthermore, this study had a smaller sample size of only 20 cases and different experimental operators and varied experimental conditions in the two studies might have caused a variation in the results. Hence, studies with larger sample sizes should be designed in the future to further confirm the findings of our study.

This study has some limitations. First, the results of bioinformatic analysis were based on data obtained from IPF and normal lung tissues; however, due to technical reasons, we could not obtain the lung tissue of patients with IPF in our hospital, and the results could only be verified experimentally through blood samples. Moreover, we were unable to analyze differential gene expression between old and young patients with IPF because most of the patients included in this study were middle-aged or old. In the future, we intend to study the differentially expressed genes between old and young patients with IPF. Second, the sample size was small because fewer clinical samples were included in this study. Hence, our conclusions need to be confirmed using a larger IPF cohort. Third, only the expression levels of differentially expressed aging-related genes were verified in the clinical samples, and the mechanisms underlying the functions of these genes were not explored in IPF cells or mouse models. Thus, a more detailed investigation is required in the future.

In conclusion, six potential aging-related genes associated with IPF were identified in this study using bioinformatic analysis and machine learning methods. The prognostic role of key genes IGF1, RET, and IGFBP2 was verified using clinical samples. These genes may affect the occurrence and prognosis of IPF by regulating senescence. These present findings improve our knowledge regarding IPF and may help design treatment strategies for this disease in the future.