Background

Idiopathic pulmonary fibrosis (IPF) is the most common type of interstitial pneumonia [1]. It is a fatal, invariably progressive disease. Lung transplantation is the most effective treatment [2]. The incidence of IPF appears to be increasing, so early recognition and intervention with supportive pharmacologic agents are needed to prevent its progression. Lung cancer (LC) is one of the most common complications in IPF patients, and mortality among patients with IPF and LC is highest [3]. Likewise, IPF is an independent risk factor for LC [4]. Nintedanib, used as an anti-fibrotic drug for IPF, is also approved for non-small cell lung cancer (NSCLC) [5], which further suggests the close relationship between IPF and NSCLC. Thus, the pathogenic overlap of IPF and LC may help clinicians to better understand the molecular mechanisms involved in both diseases, and may contribute to provide therapeutic strategies.

C1q protein was first identified in 1961 and was described as a “11s thermolabile serum protein” that participated in immune hemolysis [6]. It is the first subcomponent of the C1 complex, which activates the classical pathway of the complement [7]. C1q is composed of three polypeptide chains (C1qa, C1qb and C1qc) and regulates immune and non-immunes responses. The complement system appears to accelerate the pathogenesis of IPF [8, 9], while studies about the relationship between IPF and C1q are limited. C1q can act as a promoting factor of cancer in the tumor microenvironment [10]. However, the role of C1q in IPF and LC has not been fully explored.

In our study, three mRNA microarray datasets were downloaded from the Gene Expression Omnibus (GEO) and 37 differentially expressed genes (DEGs) were obtained comparing lung tissue with pulmonary fibrosis and normal lung tissues. Next, we explored common DEGs in both IPF and NSCLC. Gene ontology (GO) terms and KEGG pathway enrichment analysis were performed using DAVID. Protein–protein interaction (PPI) network analysis was performed and hub genes were screened to aid better understand the relationships between DEGs. C1q was finally identified as a co-hub gene. We verified the role of C1q in IPF using another four datasets from the GEO and establish mouse model of IPF. Besides, we analyzed multiple databases including those of Kaplan–Meier Plotter, Tumor Immune Estimation Resource (TIMER2.0) and the Human Protein Atlas (HPA) to reveal the clinical significance and function of C1q in NSCLC especially in lung adenocarcinoma. Finally, SurvivalMeth platforms and GSE63704 was used to detecte methylation levels of C1q. The flow diagram of this study is shown in Fig. 1.

Fig. 1
figure 1

The flow diagram of this study

Full size image

Methods

Mice

All experimental procedures involving animals were performed in accordance with the guidelines of the National Institutes of Health for the care and use of laboratory animals (NIH Publication, 8th Edition, 2011) and approved by the Animal Care and Use Committees of Shanghai Tenth People’s Hospital (Shanghai, China; permit number: SHDSYY-2019-2149). 12–16-week-old female mice were purchased from Charles River Laboratories. The murine pulmonary fibrosis model was induced with intratracheal injection of 1.5 U/kg BLM for 4 weeks, and control mice were given normal saline in same volume. After the study, all the mice were first anesthetized by allowing them to inhale 2.0% isoflurane and subjected to cervical dislocation. The study received approval by the Animal Care and Use Committees of Shanghai Tenth People’s Hospital for animal welfare.

Real-time RT-PCR

Total RNA from lung tissue was extracted by Trizol reagent (Life Technologies), and then reserve transcribed to generate cDNA by the PrimeScript reverse transcription reagent kit (TaKaRa Bio, Shiga, Japan). Quantitative real-time PCR was carried out using SYBR Green (Roche). The gene Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal control. Primers were listed in Supplemental Table 1.

Histology and Immunohistology

Lung tissue samples were fixed by 4% of paraformaldehyde, dehydrated, and paraffin embedded for sectioning into 5-μm-thick sections and then used for Masson staining and immunohistochemical staining. The results of Masson staining were used to evaluate the cross-sectional area and the collagen volume respectively by light microscopy. Immunostainings for C1QA (A13659; ABclonal; 1:100), C1QB (A5339; ABclonal; 1:100) and C1QC (A9227; ABclonal; 1:100) was performed on paraffin embedded sections and incubated with goat anti-rabbit secondary antibodies to examine positive cells in lung tissue. All histopathological sections were measured with Image-Pro Plus 6.0 image analysis system.

Microarray data

Data were retrieved from the GEO database, including the GSE485 series on the GPL81 platform (Affymetrix Murine Genome U74A Version 2 Array), the GSE97546 series on the GPL6887 platform (Illumina MouseWG-6 v2.0 expression beadchip), the GSE37635 [11] series on the GPL6885 platform (Illumina MouseRef-8 v2.0 expression beadchip), and the GSE31013 [12] series on the GPL1261 platform (Affymetrix Mouse Genome 430 2.0 Array). We chose samples from established animal models that required more than 1 week for fibrosis to be established to ensure the IPF model was successful (Table 1). Datasets including GSE98468, GSE102751 and GSE124685 [13, 14] were used to confirme the role of C1q in IPF. GSE63704 was used to detecte methylation levels of C1q in IPF. Datasets including GSE135304 and GSE76033 were used to confirme the role of C1q in NSCLC (Table 2).

Table 1 Characteristics of datasets for analysis
Full size table
Table 2 Characteristics of the datasets for validation
Full size table

Data processing of differentially expressed genes

GEO2R is an online tool that allows the comparison of groups in the GEO series to identify DEGs [15]. We identified DEGs comparing healthy mice and mice with lung fibrosis using GEO2R https://www.ncbi.nlm.nih.gov/geo/geo2r/ ). Genes without corresponding gene symbols were removed. Only one probe was kept for genes with different probes, and p < 0.05 was considered 143 for statistical significance. Co-differentially expressed p < 0.05 was considered for statistical significance. Co-differentially expressed genes were identified and Venn maps were drawn via the Venn online tool (http://bioinformatics.psb.ugent.be/webtools/Venn/ ).

Functional enrichment analysis of differentially expressed genes

DAVID ( https://david.abcc.ncifcrf.gov/ ) is an online bio-informatics tool composed of a comprehensive biological knowledge base and analytic tools. It can extract biological meaning of genes or proteins systematically [16]. We used DAVID to perform a functional enrichment analysis, including biological processes (BP), molecular functions (MF), and cellular components (CC), and KEGG pathway analysis of the overlapping DEGs( p < 0.05). The bubble chart was plotted by http://www.bioinformatics.com.cn , an online platform for data analysis and visualization.

Construction of the protein interaction network and modules selection

The String online database (https://string-db.org) was used to construct a protein–protein interaction (PPI) network for DEGs and a combined-score > 0.4 was considered a significant difference. The PPI network was visualized with Cytoscape (www.cytoscape.org). Cytoscape has been used for integrated models of biomolecular interaction networks [17]. The essential proteins were predicted using the Molecular Complex Detection (MCODE) within the PPI network as follows: MCODE score > 5, degree cutoff = 2, node density cutoff = 0.1, node score cutoff = 0.2, k-score = 2, and maximum depth = 100.

Survival analysis and validation of the hub genes

Kaplan–Meier Plotter (http://kmplot.com/analysis/index.php?p=service) was used to assess the prognostic value of hub genes in non-small-cell LC [18], which includes 1925 patients. We scanned the immunohistochemical images of lung adenocarcinoma and squamous cell carcinoma from The HPA (https://www.proteinatlas.org/), has open access allows free access to data for exploration of the human proteome [19]. TIMER2.0 [20] (http://timer.cistrome.org/) was used to analyze the correlation of C1q with the fibrosis indexes. SurvivalMeth (http://bio-bigdata.hrbmu.edu.cn/survivalmeth/) was used to analyze the effect of DNA methylation of C1q on LUAD prognosis [21].

Statistical analysis

Microarray data was analysed using Statistical Package for Social Sciences (SPSS, Version 20.0). Heatmap, box plot and violin plot were drawn by Graphpad prism 8. Continuous variables were presented as mean ± SEM. Comparisons between groups were made using unpaired two-tailed t test or one-way ANOVA when appropriate and p < 0.05 was considered to be significant.

Results

Identification and analysis of differentially expressed genes in idiopathic pulmonary fibrosis

Three microarray datasets were chosen to identify DEGs in pulmonary fibrosis from the GEO. The GSE485 dataset contained 2 control samples and 2 pulmonary fibrosis samples. The GSE97546 dataset contained 3 lung samples of healthy mice and 3 lung samples of mice with lung fibrosis. The GSE37635 included 6 timepoints: the control (n = 7), 1 week (n = 7), 2 weeks (n = 6), 3 weeks (n = 6), 4 weeks (n = 6), and 5 weeks (n = 6) after bleomycin-treatment to induce lung fibrosis. We first extracted gene expressions changes at each time points in the GSE37635 dataset (Fig. S1), then the results were intersected with the other two datasets.

Sixty one genes overlapped across the three datasets as shown in the Venn map. Genes with opposite trends were then excluded. Finally, we identified the 37 overlapping DEGs in the 3 datasets, consisting of 32 genes that were up-regulated and 5 genes that were down-regulated (Fig. 2A).

Fig. 2
figure 2

Analysis of DEGs in pulmonary fibrosis. (A) Venn diagram of DEGs in mRNA expression profiling datasets GSE485, GSE37635, and GSE97546. (B-D) GO and KEGG enrichment analysis of the DEGs. (F) PPI network of DEGs. (G) A significant module selected from PPI network, all of them were upregulated genes

Full size image

Differentially expressed gene ontology and KEGG pathway analysis in pulmonary fibrosis

DAVID was used to perform GO term and KEGG pathway enrichment analysis for further insight into the function of the identified DEGs. We selected the top 5 terms of biological processes and cell components according to p value. We found cell components of DEGs mainly enriched in extracellular regions, extracellular exosome and extracellular space (Fig. 2B). The biological processes were mainly related to inflammation and immune pathways (Fig. 2C). Further, molecular function mainly involved chemokine activity, cytokine activity and protein homodimerization activity (Fig. 2D). Moreover, eight KEGG pathways were overrepresented in DEGs, including staphylococcus aureus infection, prion diseases, systemic lupus erythematosus, chemokine signaling pathway, pertussis, complement and coagulation cascades, cytokine-cytokine receptor interaction and chagas disease (Fig. 2E).

Protein-protein interaction network construction and hub genes selection

The PPI network, which was consisted of 25 nodes and 74 edges, was constructed using STRING and was visualized by Cytoscape software to explore the association between the DEGs. The MCODE plugin was used to identify hub genes. C1qa, Fcgr1, C1qb, C1qc, Ccr5, Slc11a1, Aif1, Emr1 and Cxcl10 were identified as the most closely connected module which were highlighted in yellow, including 9 nodes and 32 edges (Fig. 2F), and the genes in this region were upregulated in pulmonary fibrosis.

Screening and analysis of genes related to hub co-expressed genes in lung cancer

Considering IPF increased the risk of LC development especially NSCLC [22], we selected the GSE31013 dataset containing 6 samples of lung tumors and 6 samples of age-matched normal lung tissue. After the DEGs between lung tumors and control samples were identified via GEO2R online tools with |logFC| > 1 and p value < 0.05, it was surprising to find that nearly half of the DEGs in pulmonary fibrosis also exhibited had significant difference in spontaneous lung tumors formations (Fig. 3A-B, Table S2).

Fig. 3
figure 3

Analysis of commin DEGs in pulmonary fibrosis and non-small cell lung cancer. (A) Venn diagram of common DEGs in DEGs of IPF and GSE31013. (B) GO enrichment analysis of the common DEGs. (C) KEGG pathways of the common DEGs. (D) PPI network of common DEGs. (E) The most significant module related to common DEGs

Full size image

The DEGs gene ontology and KEGG pathway of these genes were re-analyzed via DAVID to better understand their functions. The results showed that genes of the cellular component of this module were also mainly enriched in the extracellular region and the molecular function mainly enriched in protein homodimerization activity. The biological process analysis results suggested complement activation and cellular potassium ion homeostasis together with immune and inflammatory pathways might contribute to the occurrence and development of pulmonary fibrosis and LC (Fig. 3C-D). The top hub (C1qa, C1qb, C1qc and Ccr5) genes were also identified by the MCODE plugin, including 4 nodes and 6 edges (Fig. 3E-F).

Validation of C1q in IPF and lung cancer

As C1q ranked high among differentially expressed genes (Table S3), we focused on the role of C1q in IPF and LC. First, pulmonary fibrosis was induced by intratracheal instillation of bleomycin, while the control group was given normal saline in the same condition. The levels of fibrosis were measured by QPCR (Fig. 4A) and Masson staining (Fig. 4C-D) in IPF group increased, indicating the accomplishment of modeling. We further detected the levels of C1q by QPCR and immunohistochemistry, and found that the expression of C1q increased in IPF (Fig. 4B-C, E), consistent with previous analysis. Besides, we found the concentrations of C1q in the blood and balf of IPF increased as well (Fig. S2, Fig. S3).

Fig. 4
figure 4

The expression of C1q (C1qa, C1qb, C1qc) in IPF. (A) Relative mRNA level analysis of fibrosis markers. (B) Relative mRNA level analysis of C1q. (C) Masson staining and immunostaining of C1q in IPF. *p < 0.05 versus Control group (n = 4 for each experimental group). All data are presented by mean ± SEM

Full size image

The immunohistochemical stainings from the HPA showed C1q expression of lung adenocarcinoma was higher than that of normal lung, while squamous cell carcinoma showed no significance up-regulated trend (Fig. 5A, Fig. S4). Furthermore, data in GSE135304 revealed that the levels of C1q also increased in blood of patients with malignant (MN) pulmonary (Fig. 5B). To confirm the result in blood, we tested the expression of C1q in blood monocytes by GSE76033 as C1q mainly expresses in macrophages (Fig. S5, Fig. S6).

Fig. 5
figure 5

The expression of C1q (C1qa, C1qb, C1qc) in NSCLC including LUAD and LUSC. (A) Immunohistochemistry data of C1q (C1qa, C1qb and C1qc) in lung tissue of patients of control, LUAD and LUSC from The Human Protein Atlas. (B) The expression of C1q in the blood of patients with malignant (MN) pulmonary and patients with no nodules

Full size image

As methylation induced aberrant epigenetic regulation is an important common pathogenic mechanism of IPF and lung cancer [23], we explored DNA methylation levels of C1q in NSCLC (LUAD and LUSC) and IPF separately using the SurvivalMeth database and GSE63704. The results were shown in the box plot. We found most probes of hub genes decreased significantly in LUAD (n = 438) compared to normal samples (n = 32) (p < 0.05) (Fig. S7A-D). In LUSC, decreased probe of C1q also showed significant difference (Fig. S8A-D). Moreover, 4 probes of C1qa, 5 probes of C1qb, and 6 probes of C1qc were significantly decreased in IPF group (Fig. S9A-D). Common significant probes were listed in Table 3 and Table 4.

Table 3 Summary of significant methylation probes of hub genes
Full size table
Table 4 Characteristics of significant methylation probes of hub genes
Full size table

The prognostic value of C1q in IPF and NSCLC

Expression of C1qa, C1qb and C1qc genes were examined and their clinical prognostic significance was investigated within the Kaplan–Meier Plotter database including 1925 cases of NSCLC. We noted that NSCLC patients with elevated C1q (C1qa, C1qb, C1qc) levels had lower OS (p < 0.05) (Fig. 6A). Patients with higher concentrations of Ccr5 showed higher survival probability while it was positively correlated with collagen fiber related indexes, which was inconsistent with previous results, so we didn’t focus on the role of Ccr5(Fig. S10, Fig. S11). Furthermore, we investigated the Kaplan–Meier curve in 672 patients with lung adenocarcinoma (LUAD) and 271 patients with lung squamous cell carcinoma (LUSC). The results demonstrated that high levels of C1q expression revealed worse prognosis of adenocarcinoma (p < 0.05) (Fig. 6B) compared with squamous cell carcinoma (p > 0.05) (Fig. 6C).

Fig. 6
figure 6

Overall survival analyses of C1q in NSCLC. (A) NSCLC patients with elevated C1q (C1qa, C1qb, C1qc) levels had lower OS (p < 0.05). (B) LUAD patients with elevated C1q levels had lower OS (p < 0.05). (C) The levels of C1q had no correlation with prognosis of LUSC. (D) C1q including C1qa, C1qb and C1qc were positive correlative with α-SMA, COL1A1, CTGF and TGF-β1 in lung adenocarcinoma at TIMER2.0 database

Full size image

We detected a relationship between C1q and TGF-β1 as well as other mesenchymal markers (ACTA2, COL1A1, and CTGF) in lung adenocarcinoma using the online tool TIMER2.0. These factors involved in pulmonary fibrosis were positively related to the level of C1q expression (Fig. 6D).

GSE37635 mentioned before was used to verify the role of C1q in IPF. It was found that elastin and collagen fiber related markers including elastin (Eln), Fgfr1 (FR1) and TGFBI peaked at first week and decreased with time, in line with the expression of C1q (Fig. 7A, B). We further investigated it in IPF human patients by GSE124685 which had 84 samples including control samples (n = 35) and IPF samples (n = 49) from 10 control patients and 6 IPF patients (Table S4). The results were consistent with previous studies. The FPKM of C1q was also significantly increased in IPF patients (Fig. 7C). And then we divided samples IPF into two groups according to alveolar surface density (ASD) which was negatively correlated with degree of fibrosis. We identified samples with ASD > 6/μm as early or progressive fibrosis and grouped them to IPF1. Other samples (ASD < 6/μm) with end-stage fibrosis were grouped to IPF2. Although the difference between the IPF1 and IPF2 was not significant, the average expression levels of C1q in IPF2 group had an uptrend compared to IPF1 group (Fig. 7D, Table S5). These results suggested that C1q might also have prognostic value for IPF progression.

Fig. 7
figure 7

Overall survival analyses of C1q in IPF. (A) The variations of the expression of collagen fiber related markers including elastin (Eln), Fgfr1 (FR1) and TGFBI with time in IPF mice. (B) The variations of the expression of C1q with time in IPF mice. (C) The violin plot in expression of C1q between control and IPF. (D) The box plot in expression of C1q among control, IPF1 and IPF2

Full size image

Discussion

IPF is the most common of the idiopathic interstitial pulmonary disease. It is a fatal disease with a median survival of 4 years after diagnosis [24]. However, the underlying mechanisms and the treatment of IPF still remain unknown. Furthermore, growing epidemiologic evidence supports the association between IPF and LC. IPF can increase the risk of the development of LC by 7 to 20% [22]. Lung tumors in patients with IPF first occur adjacent to fibrotic areas. Compared with non-IPF-associated lung tumors, they have different histologic distribution and immunohistochemical features [23]. NSCLC is the predominant type in LC-IPF patients, of which adenocarcinoma (ADC) is the most common subtype [25]. Studies show that there are many mechanistic similarities between IPF and LC especially in NSCLC diseases, of which methylation is an important common mechanism [23, 26]. Myofibroblasts of IPF and cancer cells of LC are characterized by altered intercellular communication, enhanced migration and mobility, and increased invasion through the extracellular matrix. The fibroblasts in LC also exhibited mesenchymal-like features and have heterogeneous phenotypes [27].

In our study, a series of bioinformatics analyses were performed on three mRNA microarray datasets, and 37 common DEGs were identified including 32 up-regulated genes and 5 down-regulated genes. The cell component of DEGs was mainly enriched in the extracellular region and biological process were enriched in the inflammatory and immune responses, consistent with the previous study of IPF [28, 29]. It was found that among the DEGs, nine potential hub genes (C1qa, Fcgr1, C1qb, C1qc, Ccr5, Slc11a1, Emr1, Aif1 and Cxcl10) were obtained using the MCODE plug-in of Cytoscape.

To explore the deep correlation between pulmonary fibrosis and NSCLC, we screened co-expressed genes in the DEGs between pulmonary fibrosis and LC, and surprisingly found 16 out of the 37 initial DEGs were also significantly more expressed in LC. Four hub genes for both IPF and NSCLC were finally identified (C1qa, C1qb, C1qc, Ccr5). C1qa, C1qb, and C1qc are three polypeptide chains the composed C1q which is the first subcomponent of the classical pathway of the complement system [30]. Ccr5 is a chemokine receptor. It can mediate receptor activation by binding several endogenous chemokines [31].

We explored the methylation levels of hub genes, and found most of probes in IPF and NSCLC decreased which could give a good interpretation to the up regulated expression of C1q. In order to evaluate the diagnostic and prognostic value of the hub genes, we also determined the correlations between key genes and clinical data. The results demonstrated that the OS rate was significantly lower among NSCLC patients with high levels of C1q, while patients with high concentrations of Ccr5 showed higher survival probability. We further investigated the role C1q played in the survival rate of lung adenocarcinoma and squamous cell carcinoma respectively. There was a significant difference in lung adenocarcinoma but no significant difference in lung squamous cell carcinoma, consistent with the immunohistochemical images analyzed from the PHA. Kaplan–Meier curves and immunohistochemical and stained pictures also indicated that C1q was closely related to the prognosis of NSCLC particularly lung adenocarcinoma. TGF-β is the most important mediator of the pathogenesis and carcinogenesis of IPF [32]. C1q showed a positive correlation with LUAD as well as other fibrosis markers like a-SMA, CTGF and collagen-1. These findings suggested that representative molecules contributing to IPF might also be regulated by C1q in lung adenocarcinoma. We further investigated the expression of C1q in IPF human patients, and found C1q might also have prognostic value for IPF progression. Besides, we comfirmed the role of C1q in the blood and balf of IPF.

C1q is synthesized primarily by macrophages and dendritic cells [33]. The complement component C1q plays an important role in host immune responses. Previous studies mainly focused on its role in systemic lupus erythematosus. Deficiency of C1q leads to high susceptibility to systemic lupus erythematosus-like symptoms, and complications with skin and renal diseases [34]. However, the role and mechanism of C1q in IPF and LC still remain unclear.

Correlation studies on the contribution to IPF are lacking. Sakkas et al. found that high levels of anti-C1q autoantibodies could predict the occurrence of pulmonary fibrosis patients with SSc [35]. Besides, C1q was positive in 73% of IPF cases by investigating bronchoalveolar lavage fluids, higher than in hypersensitivity pneumonitis (40%) or in sarcoidosis (31%) [36]. These result from another standpoint indicate that C1q participates in the occurrence and development of pulmonary fibrosis.

Studies accumulated over the past few years suggest that C1q can promote tumor formation by facilitating the proliferation and migration of cancer cells along with angiogenesis and metastasis [37, 38]. Bulla et al. reported that C1q promoted the growth of malignant pleural mesothelioma. C1q could induce adhesion and proliferation of mesothelioma cells and bind to hyaluronic acid in the microenvironment of MPM [39]. Tedesco et al. discovered that C1q-deficient (C1qa−/−) mice exhibited inhibited melanoma tumor growth which led to a better prognosis. C1q is mainly expressed in mesenchymal elements and promotes tumor growth cooperating with fibronectin [10]. Besides, C1q was also found to stimulate the progression of hepatocellular tumor. It enhanced migratory and invasive phenotypes of liver cancer cells [40]. Furthermore, C1q had a negative prognostic effect on kidney carcinomas [41].

C1q has pro-fibrotic and immunosuppressive properties, which are two contributing factors to both of IPF and NSCLC. C1q binding in human lung fibroblasts was reported to be heterogeneous [42], and able to interact with fibronectin to mediate adhesion of fibroblasts to immune complexes, resulting in stimulation in collagen synthesis [43, 44]. Low circulating levels of C1q prevented proliferation of fibroblast cells via down regulation of Wnt/β-catenin signaling [45, 46]. In addition, C1q plays a part in immune dysregulation. C1q inhibited proliferation of the Th1 and Th17 T cell subset mediated by human macrophage and dendritic cells [47]. It also suppressed the activation, proliferation, and cytotoxic functions of CD8+ T cells in vitro [48]. Meanwhile, tumor cells could directly assemble with C1q, enabling the formation of complement activation and further promoting tumor growth [49]. Other complement components like C4d and C5a showed similar phenomena in LC [50, 51]. Considering complement components have been demonstrated to promote tumor growth, the inhibition of complement activation and binding have been proposed as effective treatment for cancer [52, 53].

Our study revealed that the concentrations of C1q were increased and took part in many physiological and pathological processes in IPF and lung cancer. The results of bioinformatics analysis highlighted that high concentrations of C1q have important clinical prognostic value in NSCLC, especially in adenocarcinoma. The levels of C1q in severe pulmonary fibrosis also have an uptrend. It’s the first time that we find methylation status of C1q decreases and the expression of C1q increases in IPF and NSCLC, which is positively correlated with pulmonary fibrosis. This may also be a connection between the two diseases. In conclusion, our findings can provide the basis of new targets for diagnosis, therapy, and prognosis of IPF and LC.