Systematic analysis of functional implications of fibrosis in pan-cancer

doi:10.21203/rs.3.rs-4938310/v1

Download PDF

Research Article

Systematic analysis of functional implications of fibrosis in pan-cancer

https://doi.org/10.21203/rs.3.rs-4938310/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The common pathogenic mechanisms and molecular pathways of fibrosis and tumors remain unclear. We aimed to conduct in-depth analysis of fibrosis feature genes role from a pan-cancer perspective and identify potential therapeutic targets for idiopathic pulmonary fibrosis and cancers. We downloaded mRNA expression, copy number alterations, and DNA methylation data of 33 cancers from The Cancer Genome Atlas (TCGA). Clinical and mutation data were obtained from the UCSC Xena database. The mutation frequencies of fibrosis-feature-related genes (FRGs) in the TCGA database were examined. Protein expression levels were analysed using the Clinical Proteomics Tumor Analysis Consortium. Gene Set Variation Analysis and Gene Set Enrichment Analysis algorithms were used. Most FRGs were differentially expressed in tumors owing to somatic cell copy number alterations and DNA methylation. We established a fibrosis potential index (FPI), and in most cancers, the FPI was lower than that in normal tissues and correlated with subtypes and clinical features. The FPI correlated negatively with multiple metabolic pathways and immune function but positively with several important tumor features or pathways. The FPI correlated with prognosis in different tumors, despite finding heterogeneity. Fibrotic features have excellent diagnostic and prognostic capabilities for various cancers. This may help predict responses to immunotherapy.

fibrosis features

diagnosis

immunotherapy

pan-cancer

prognosis

Idiopathic pulmonary fibrosis (IPF) is a chronic, progressive interstitial lung disease of unknown aetiology that has high incidence and continues to rise [1]. Patients with IPF have a significantly reduced quality of life and poor prognosis, with a median survival time of 3−5 years [2, 3]. Typical clinicopathological features of IPF include the activation of lung fibroblasts and excessive accumulation of the extracellular matrix (ECM) [4], leading to irreversible damage to the lung structure and, ultimately, respiratory failure due to impaired gas exchange.

Of note, IPF is associated with an increased risk of lung cancer, which is five times higher in patients with IPF compared with that in the normal population [5]. Epidemiological data [6] show that the incidence of lung cancer in patients with IPF ranges between 3% and 22%; in extreme cases, this incidence exceeds 50%. Notably, lung tumors in patients with IPF tend to appear initially in fibrotic areas; it is speculated that there may be common pathogenic mechanisms and molecular pathways between these two diseases [7]. Compared with non-IPF-associated lung cancers, IPF-associated lung cancers show significantly different histological distributions and immunohistochemical features [8]. Increasing evidence suggests that IPF is strongly associated with cancer [9, 10]. However, little is known about this relationship. Thus, exploring potential target molecules may better explain the molecular mechanisms by which fibrosis features increase the risk of cancer, the knowledge about which is important for developing new treatments of cancer based on fibrosis features and personalised and precise treatment strategies. Therefore, we aimed to perform an in-depth analysis of the role of fibrosis feature genes from a pan-cancer perspective to explore the common pathogenesis of fibrosis and cancer and identify potential therapeutic targets for IPF and cancer.

Datasets and source

To understand the complex microenvironment of IPF and its possible induction of malignant features at the single-cell level, we downloaded the GSE132771 dataset containing single-cell RNA sequencing data of three patients with IPF. Gene expression, copy number variants, methylation, and clinical data were downloaded from the UCSC Xena database (https://xenabrowser.net/). Protein profiling data were collected from the Clinical Proteomics Tumor Analysis Consortium (CPTAC) database (https://proteomics.cancer.gov/programs/cptac). Immune infiltration was evaluated using the TIMER 2.0 database (http://timer.cistrome.org/) containing multiple immune infiltration algorithms. Since these public databases are freely accessible, this study was granted an exemption by the Ethics Committee of the First Affiliated Hospital of Anhui Medical University.

Identification of biomarkers based on single-cell datasets

The ‘Seurat’ R software package was used for analysis and quality control (minimum gene count 500, maximum gene count 4000, mitochondrial gene percentage <5%, erythrocyte gene percentage <1%) to filter low-quality cells and to avoid zero- or multi-cell capture. The ‘SCTransform’ function was used to normalise the data, which can correct or remove the batch effect caused by different batches (e.g., different time points, different samples.) in the experiments, making the data more comparable. The correction effect on the sequencing depth was better than that of log normalisation. We used the ‘RunHarmony’ function to integrate multiple batches of single-cell data to remove the batch effect. Compared with other algorithms, the ‘harmony’ algorithm integrates data while still being sensitive to rare cells, making it suitable for more complex single-cell analysis experiments that compare cells from different donors, tissues, and technology platforms. The top 3000 highly variable features were subjected to principal component analysis downscaling using the ‘RunPCA’ function to select the top 20 principal components. Using the ‘clustertree’ R With the ‘clustertree’ R package, we chose 0.4 as the best clustering resolution. The ‘FindClusters’ function was used to identify cell subgroups, and the ‘plotUMAP’ function was used for visualisation. Subsequently, the cell clusters were manually annotated based on cell lineage-specific genes. Based on the ‘FindAllMarkers’ function, genes that were highly expressed in each cell subpopulation were identified as markers for each cell cluster. Cell–cell interactions form the basis for information exchange between cells and are essential for various biological processes. Using the ‘cellchat’ R package, we established the probability of cell–cell communication from the perspective of pattern recognition and multiple learning. We then made predictions about cell–cell communication and provided multiple visualisation methods. We analysed and visualised ligand-receptor-mediated cellular interactions in the 14 cell types. The analysis focused on molecules that interact with endothelial cells and fibroblasts with other cells and their key signalling pathways. We identified IPF microenvironmental communication patterns using non-negative matrix decomposition. Pseudo-trajectory analysis was performed using the Monocle2 algorithm. The first differentiation node genes during fibroblast differentiation intersected with highly expressed fibrosis feature genes, which were considered intersections and defined as fibrosis feature genes.

Multi-omics differential expression analysis

To determine the expression of fibrosis feature genes in tumors and normal tissues, we performed a three-dimensional differential analysis. First, we compared the mRNA expression differences in TCGA using the Wilcoxon rank sum and signed rank tests. Additionally, we performed the above tests on paired cancer samples grouped by TCGA. Since the TCGA data contained fewer normal tissues, we utilised the Genotype-Tissue Expression (GTEx) database and used the tests to test the significance of the differences. Additionally, we validated the protein expression of fibrosis feature genes using the CPTAC proteomics data. We performed somatic copy number alteration (SCNA) and mutation frequency analyses, where more than 5% were considered high-frequency SCNAs. The association between the SCNAs and gene expression was assessed using Spearman's correlation.

Survival analysis, single nucleotide variants, and DNA methylation analysis

The statistical significance of fibrosis feature genes and pan-cancer survival risk was assessed using Cox univariate analysis. Single nucleotide variants (SNVs) contribute to the genetic variability of an organism, influence the course of biological processes, and may play a role in the genetic factors that determine disease susceptibility. Thirty-three pan-cancer SNV datasets were downloaded from the TCGA database (n = 8,663), and an SNV oncoplot plot was generated using maftools. The methylation probes for each gene promoter were annotated using the R package ‘Illumina Human Methylation’. Differences in methylation were detected using the Wilcoxon signed rank Test.

Establishing the fibrosis potential index

To understand the role of fibrosis features in oncogenesis, the fibrosis potential index (FPI) was established using the Z-score parameter in the R package ‘GSVA’ to calculate the level of fibrosis features in the tissue samples. We compared the difference in FPI between tumor and normal tissues using the Wilcoxon analysis and visualised the distribution of FPI expression in human organs using the human anatomical ‘gganatogram’ package. We also performed Wilcoxon tests on paired samples using TCGA cancer. The importance of FPI for pan-cancer diagnosis was assessed based on the pROC package evaluating the receiver operating characteristic (ROC) curve, with area under the curve (AUC) values ranging between 0.5 and 1. An AUC value closer to 1 indicated a better diagnostic performance. Generally, AUC values of 0.5−0.7, 0.7−0.9, and >0.9 indicated low, average, and high precision, respectively.

Clinical relevance of the FPI

We downloaded survival data from the TCGA database and analysed the correlation between FPI level and prognostic indicators, including overall survival (OS), disease-specific survival, progression-free interval, and disease-free interval using the ‘survival’ and ‘survminer’ R packages. Kaplan−Meier and Cox univariate analyses were used to investigate the role of FPI and draw an FPI survival map.

Analysis of the tumor microenvironment and immune infiltration

The tumor microenvironment plays a crucial role in oncogenesis. TIMER 2.0 [11] (http://timer.cistrome.org/) uses seven algorithms to reliably estimate the level of immune infiltration. Next, we conducted a tumor immunophenotype (TIP) analysis [12] using seven stages of the cancer immune cycle. The anti-cancer immune status at each step was analysed and scored; the difference between the high and low FPI level groups was calculated.

Pathways and functional enrichment analysis

Based on the FPI level, each tumor sample was divided into either high or low FPI group to identify the pathways associated with fibrotic features. Gene Set Enrichment Analysis (GSEA) was used to investigate metabolic and other related pathways in different tumors. The CancerSEA database comprises 72 datasets from the SRA, GEO, and ArrayExpress websites, with a total of 41,900 tumor cells from 25 cancers. Functional analysis results were obtained from datasets redefining a total of 14 functional states, calculated using the R-package GSVA algorithm. The value of each gene set was calculated using the Z-score [13] algorithm to generate the GSVA. In each tumor, we classified FPI into high and low levels. If a gene was highly expressed in more than five tumors, we considered it related to FPI, defined it as an FPI-related gene, and performed a Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. Thus, the pan-cancer function of FPI and related biological pathways were identified.

Identification of FPI-related genes associated with chemotherapy and immunotherapy

The GSCA database [14] was used to analyse drug sensitivities. We calculated the correlation between the FPI-related genes and IC50 values. DTP NCI-60 data were acquired from CellMinerDB [15]. We assessed the drug sensitivity Z-score and calculated Spearman correlation coefficients. To identify differentially expressed genes in the high and low FPI groups across cancer types, the 150 most upregulated or downregulated genes were identified as FPI-associated markers. We used Connectivity Map (Cmap) analysis (https://www.pmgenomics.ca/bhklab/sites/default/files/downloads), containing 1288 compound-associated signatures, to calculate match scores. The analysis followed that of previous studies [16, 17]. The top five results for the 33 cancer types were summarised and graphically presented using the R language. Additionally, we used the FPI to predict survival and response in an immunotherapy cohort (IMvigor210/urinary tract epithelioma and GSE78220 cutaneous melanoma).

Identification and Validation of A2M in Lung Squamous Cell Carcinoma

A2M was identified as a critical gene in lung squamous cell carcinoma through the application of multiple machine learning algorithms, with the results of the analysis being illustrated by residual analysis. The expression levels of A2M were subsequently validated using a multi-omics approach, including transcriptomic profiling, spatial omic analysis, and pathomic characterization. To further reinforce these findings, the expression pattern of A2M was corroborated within single-cell RNA sequencing datasets.

Statistical analysis

All data were statistically analysed using WebTools and R software (version 4.3.0) (http://www. r-project. org/)). Pearson's correlation analysis was used to analyse normally distributed data; otherwise, Spearman's correlation analysis was used. The Kruskal−Wallis rank sum test, Wilcoxon rank sum test, and signed rank test were used to detect the significance of multiple and two variables, respectively. Cox and Kaplan–Meier Survival analyses were performed using the ‘survival’ package. Relative risks were described using hazard ratios and 95% confidence intervals. ROC curve analyses were performed using the ‘pROC’ software package. All statistical tests were two-sided. A p-value <0.05 was considered statistically significant.

Identification of fibrosis feature genes

We identified 15 single-cell clusters (Fig. 1a), i.e., clusters 0−14 (Fig. 1b). A large number of differentially expressed genes existed in each cell cluster relative to the other clusters (Fig. 1c), including a large number of lineage-specific and functional genes. We annotated 15 cell-type clusters based on their marker genes, yielding 14 cell types (Fig. 1d). The proportion of each cell type in the three samples was calculated; the results showed that myeloid and endothelial cells were dominant (Fig. 1e). Specific annotations were based on markers [18], as shown in Fig. 1f: T cells (marked with CD3D,CD3E, and CD3G), CD8T cells if both CD8A/CD8B were expressed, CD4T cells if both CD4 were expressed, NK T cells if both NK cell markers were expressed, such as KLRD1 and GNLY, plasma cell-like dendritic cells (marked with GZMB, SOX4, IRF7, LILRA4, and TCF4), endothelial cells (marked with RAMP2, PLVAP, PECAM1, FLT1, ENG, CDH5, or VWF), fibroblasts (labelled with COL1A1, COL1A2, COL3A1, DCN, or PDGFRA), VSMC (labelled with ACTA2, MYH11, TAGLN, CDKN1A, and PDGFRB), myeloid cells (marked with CD14, FCGR3A, LYZ, CD68, and C1QC), B-cells (CD79A, CD79B, CD19, and MS4A1), ciliated airway epithelial cells (marked with EPCAM and TPPP3) and secretory club cells (marked with SCGB1A1). We performed and visualised the Cellchat analysis (Fig. 2a–b). Incoming and outgoing signals were mediated by different ligands and receptors and differed between cell types (Fig. 2c). Notably, many signals were transmitted to endothelial cells, and many signalling pathways emanated from secretory club cells and fibroblasts (Fig. 2d). Endothelial cells received signals mainly through the chemokine pathway (Fig. 2e), while fibroblasts sent signals mainly through the MK and MIF-related pathways (Fig. 2f). Notably, MK- and MIF-related pathways are important for IPF cell chat because they are involved in many cell-to-cell communication processes (Fig. 2g–h). The NMF algorithm identified incoming versus outgoing communication patterns with k=6, as determined by the cophenetic coefficient and silhouette score. We observed certain commonalities in the communication patterns across different cells. For example, in outgoing signalling, T, CD4+T, B, and ciliated airway epithelial cells shared a communication mode (Fig. 2i–n). Pseudotime analysis identified six key differentiation nodes of fibroblasts using the Monocle2 algorithm (Fig. 3a). Branches in single-cell trajectories were generated according to cell fate, and the branched expression analysis modelling function was used to identify genes with branch-dependent expression (Fig. 3b). Genes that were highly expressed in fibroblasts intersected with first differentiation node branch-dependent genes, resulting in 30 fibrosis feature genes that may be involved in key roles in fibrosis (Fig. 3c).

Expression and mutation of fibrosis feature genes in pan-cancer

To determine the expression patterns of fibrosis-related genes in cancers, we determined their expression levels. We found that fibrosis feature genes were differentially expressed in most types of cancers; that is, COL1A1, COL1A2, CTHRC1, and TNFRSF12A were predominantly highly expressed in pan-cancer, and A2M, ADH1B, C7, CFD, CYR61, and PTGDS were predominantly expressed at low levels in pan-cancer (Fig. 4a). Subsequently, the above results were fully validated based on paired differential analyses of cancer versus paracancerous samples and differential analyses of TCGA samples combined with GTEx (Fig. 4b–c). The CPTAC protein data behaved similarly at the TCGA transcriptional level (Fig. 4d). The reproducibility and consistency of such differential results have been demonstrated across multiple databases, tumors, methods, and histologies, suggesting that differences in the expression of fibrosis-related genes may play a role in different cancers. We examined SCNAs in tumor and normal tissues. We examined the percentage of SCNAs that occurred at a high frequency in most cancer types and at a low frequency in only a small number of tumors (Fig. 4e). Spearman's correlation coefficient between gene expression and SCNAs was calculated. The results showed that, in most tumors, fibrosis feature genes were significantly related to SCNAs (Fig. 4f). This suggests that copy number abnormalities in fibrosis feature genes could be common in cancers.

Survival and genetic alterations of fibrosis feature genes in cancers

Cox univariate analysis indicated that fibrosis-related gene expression was strongly associated with patient survival (Fig. 5a). Additionally, we analysed the SNV data of fibrosis feature genes to detect the frequency and type of variants in each cancer subtype. As shown in Fig. 5b, SKCM contained the highest number of mutations, and SNV percentage analysis showed that FN1, COL3A1, CFH, and COL1A2 had the highest mutation frequencies (Fig. 5c). Promoter methylation regulates the expression of genes associated with tumorigenesis. We found that fibrosis feature genes showed complex methylation patterns (Fig. 5d–e). In addition, although most tumor tissues showed insignificant methylation differences compared with normal tissues, the negative correlation between the expression of fibrosis genes such as CCL2, POSTN, CFH, and CTSK and DNA methylation suggests that these genes may be regulated by methylation.

Identification of the FPI

To explore the factors related to fibrosis feature genes, we calculated gene enrichment scores using the GSVA algorithm [19], which was used to establish the FPI model. The FPI was significantly downregulated in most cancers compared with the normal group (Fig. 6a). The distribution pattern of FPI in each organ was visualised using the ‘gganatogram’ package (Fig. 6b). Differential analysis of TCGA pairwise samples was consistent with these results (Fig. 6c). Notably, the FPI levels correlated with eight tumor stages (Fig. 6d–k), BLCA, BRCA, ESCA, KIRC, SKCM, STAD, TGCT, and THCA, with higher expression at higher stages.

Clinical significance of FPI

The role of the FPI in cancer survival was analysed to demonstrate its clinical significance. The pan-cancer survival panorama showed that FPI was associated with multiple survival in a wide range of cancer types (Fig. 7a). Moreover, the correlations were relatively heterogeneous, as, in some cases, fibrosis could act as a risk factor for several cancers. There are also some tumors in which FPI was a protective factor; for example, patients with higher FPI expression show worse survival in BLCA, COAD, ESCA, GBM, KIRC, KIRP, LGG, and STAD, and better survival in DLBC, LIHC, UCES, and UCS. This suggests that the features of fibrosis could play different roles in different cancer types. Additionally, we used forest plots to show the four survival endpoints using Cox survival analysis (Fig. 7b–e) and Kaplan−Meier curves through the log-rank test (p<0.001) (Fig. 7f).

FPI correlates with immune infiltration in cancer

Next, we explored whether fibrotic features were involved in the process of immune infiltration in pan-cancer. In the pan-cancer cohort, the FPI showed significant positive correlations with MHC molecules, immunosuppressive and immune-activating genes, chemokine receptors, and ligands (Fig. 8a). TIP analysis suggested that the high FPI level group was characterised by priming and activating cancer cell antigens (Fig. 8b). To elucidate the specific cell types regulated by FPI in the TME, we explored the correlation between FPI and immune-infiltrating cells using the TIMER2.0 database. Overall, various immune-infiltrating cells were associated with FPI, particularly tumor-associated fibroblasts, endothelial cells, T cells, and macrophages (M2 type) (Fig. 8c).

Association between FPI and cancer pathways

Many tumor-related features and pathways were associated with FPI levels, including angiogenesis, apoptosis, cell cycle, differentiation, epithelial-mesenchymal transition (EMT), invasion, proliferation, metastasis, and stemness (Fig. 9a). We obtained highly expressed genes in the high FPI group for KEGG enrichment analysis. FPI was associated with signal transduction, particularly with the cAMP, calcium, MAPK, and PI3K-AKT signalling pathways (Fig. 9b). Additionally, based on the transcriptomes of the two groups (top and bottom 30% of the fibrosis feature genes), we performed GSEA for the relevant cell signalling in cancer. Many pathways were activated, specifically the EMT, inflammatory response, IL-2, STAT5, and TNF-α signalling via the NF-kB and K-Ras signalling pathways (Fig. 9c). This partly explains the survival heterogeneity, suggesting that fibrosis may be a double-edged sword in tumors. Both induce malignant features and promote the immune reserve.

Potential of FPI in chemotherapy, immunotherapy, and small molecule drug therapy

We used the FPI levels to predict immunotherapy response in patients with urothelial carcinoma of the bladder (Fig. 10a–b); that is, the complete response/partial response (CR/PR) group had lower FPI, and more CR/PR patients were in the low FPI group, indicating that the low FPI group was more suitable for undergoing immunotherapy. Consistent results were observed in the immunotherapy cohort for cutaneous melanoma (Fig. 10c-d). In the Cellminer data, the FPI was positively correlated with the sensitivity Z-score for a large number of drugs (bleomycin, temsirolimus, staurosporine, rapamycin, and irofulven) and negatively correlated with some drugs (CUDC 305, palbociclib, docetaxel, and At-13378) (Fig. 10e). In the gene and chemotherapy analyses, we investigated the potential correlation between drug sensitivity and fibrosis feature genes using two different databases (CTRP and GDSC). Fibrosis feature genes showed different sensitivities to different drugs (Fig. 10f–g). To explore potential therapeutic options that could counteract FPI-mediated tumor promotion, we performed a CMap analysis. We constructed a signature of FPI-associated genes, including the 150 most significantly upregulated and 150 most significantly downregulated genes identified by comparing patients with high and low FPI levels in each cancer type. The FPI-associated signature was compared with the CMap gene signature using the eXtreme sum algorithm to obtain similarity scores for the 1288 compounds. W.13, Stock1n.35874, and exisulind exhibited significantly lower scores for most cancer types, suggesting that they could inhibit FPI-mediated anti-cancer effects (Fig. 10h).

Identification and expression of A2M in Lung Squamous Cell Carcinoma

Further examination of lung squamous cell carcinoma data through machine learning analysis revealed that A2M expression was consistently prominent across various models, as evidenced by residual analysis (Fig. 11a, b). Notably, A2M ranked highly in terms of expression in these computational models (Fig. 11c). Pan-cancer cohort analysis indicated that A2M was underexpressed in tumor tissues (Fig. 12a), with a consistent expression pattern observed in paired samples across different cancers (Fig. 12b). Spatial omics analysis provided insights into the spatial distribution of A2M, showing a gradient decrease in expression from normal to tumor regions in head and neck tumors, lung squamous cell carcinoma (LUSC), and ovarian cancer (OV), which aligns with the whole transcriptome data (Fig. 12c-e).

Immunohistochemical analysis of A2M, accessed through the Human Protein Atlas (HPA) database, suggested that A2M was either undetected or expressed at low levels in the majority of tumors, corroborating the transcriptional findings (Fig. 13a, b). Proteomic mass spectrometry data from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) further substantiated these observations, with A2M protein levels found to be consistently underexpressed across various tumor types (Fig. 13c).

Single-cell RNA sequencing data highlighted that A2M was predominantly expressed in normal cell types, with its tissue structure absent in tumor tissues, elucidating the reason for its low expression in tumor tissues (Fig. 14a-c). Moreover, A2M emerged as a prognostic factor for overall survival in several cancers, including lung squamous cell carcinoma (Fig. 14d, e). Finally, the gene expression levels of A2M and Gene Set Variation Analysis (GSVA) scores for 14 tumor statuses were presented (Fig. 14f), demonstrating a significant positive correlation between A2M expression and numerous malignant tumor characteristics. It is plausible to hypothesize that higher A2M expression by tumor cells may be associated with a poorer prognosis for patients with this subtype of lung squamous cell carcinoma.

In this study, we performed an in-depth analysis of the role of fibrosis feature genes from a pan-cancer perspective to explore the common pathogenesis of fibrosis and cancer. Fibrotic features were related to the immune microenvironment and drug resistance. These findings emphasise the pivotal role of fibrotic features and contribute to further research on molecular mechanisms and therapeutic developments related to fibrotic features.

Tumor immunotherapy has achieved remarkable success in the last decade. However, high response rates have only been observed in haematologic tumors, and poorly responding solid tumors show significant heterogeneity in tumor cells and the surrounding TME. The fibrotic stroma within the TME produces hard ECM proteins that transmit potent signals, encouraging tumor cells to evade immunotherapeutic attacks. Cancers with inflammation-derived chronic fibrosis as their susceptibility factor, such as hepatocellular carcinoma and breast cancer, are less responsive to immunotherapy. Tumor-associated fibrosis involves ECM remodelling mediated by cancer-associated fibroblasts, which results in complex and dynamic stromal changes. Excessive deposition of fibrotic matrices, such as collagen I and IV, fibronectin, laminin, hyaluronic acid, and elastin, and nonstructural matrix cellular proteins, such as the cell communication network (CCN) family members, CCN factor 1 (CCN1), CCN2, and tenosynthesis glycoprotein C, leads to matrix remodelling and the subsequent release of proteases, such as metalloproteinases and tissue proteases. Additionally, tumor-associated immune cells are involved in the formation of fibrotic stroma, mainly through the synthesis of proteases and cytokines and, to some extent, ECM proteins. Fibrotic matrices impede immunity through various mechanisms. Accordingly, we speculated whether the fibrotic features of IPF, a benign disease, were similar to those of cancer-associated fibroblasts, with corresponding effects on solid tumors.

Researchers have found that fibrosis features are closely associated with tumorigenesis and play an important role in cancer treatment [20, 21]. However, studies on the features of fibrosis and their relationship with different cancers are lacking. In this study, we utilised single-cell data from GEO and clinical pan-cancer data from TCGA to reveal global alterations in fibrosis feature genes at the genetic, epigenetic, and transcriptional levels. We also used GSVA to establish an FPI to characterise fibrotic features and explore immune infiltration, pathways, and drug therapy about the FPI.

The KEGG and GSEA results indicated that FPI was associated with the PI3K-AKT, JAK-STAT, and EMT pathways. The PI3K⁃AKT signalling pathway is one of the core signalling pathways in cells and is involved in cell growth, proliferation, metabolism, and other biological functions. Activating the PI3K-AKT signalling pathway increases fibroblast proliferation [22]. PI3K is a group of plasma membrane-associated lipid kinases that can be categorised into three classes: class I, class II, and class III. Class I PI3K exists in four isoforms, including class IA (PI3Kα, PI3Kβ, and PI3Kδ) and class IB (PI3Kγ). Among them, PI3Kα and PI3Kγ are usually overexpressed in IPF. AKT also comprises three members: AKT1, AKT2, and AKT3. AKT1 enhances the resistance of alveolar macrophages to apoptosis by regulating mitochondria in pulmonary fibrosis, and AKT2 deficiency inhibits bleomycin-induced pulmonary fibrosis and inflammation [23]. The PI3K-AKT pathway can promote angiogenesis, inhibit apoptosis, and increase tissue invasion and metastasis, among other processes involved in the development of non-small cell lung cancer [24]. Many specific drugs targeting the PI3K-AKT signalling pathway are currently under development and in various stages of preclinical studies and early clinical trials.

The JAK-STAT signalling pathway is an important cellular signalling pathway that includes various cytokines and receptors involved in cell growth, proliferation, differentiation, apoptosis, and other processes [25]. Studies have shown that activation of the JAK-STAT signalling pathway induces the transformation of alveolar epithelial cells into fibroblasts and promotes their proliferation and invasion. Milara [26] et al. found that JAK2 and STAT3 were elevated and activated in the lungs of patients with IPF. Phosphorylation of JAK2 and STAT3 induces the mesenchymal transformation of ATII and the transformation of fibroblasts to myofibroblasts in patients with IPF.

We used Cellminer data to explore drug sensitivity, and the results showed that the FPI was positively correlated with the sensitivity Z-score for many drugs and negatively correlated with some drugs.

Limitations

There were some limitations to our study. First, the datasets from TCGA were a relatively small sample size. Second, our study was based on TCGA public database, and the result should be further validated using randomized controlled data. Third, further experiments are required to explore the underlying mechanisms. Therefore, in our future studies, we plan to explore the interactions of these genes in vitro and in vivo.

In this study, we systematically analysed FPI and their regulatory genes in different cancers. We established the FPI to assess the level of fibrosis and found that in most cancer types, the FPI was lower in tumors and associated with prognosis as well as drug sensitivity. Our findings highlight the potential of FPI-based cancer therapies.

We would like to thank TCGA project organisers as well as all study participants.

Funding

This work was supported by the Anhui Provincial Education Department, Major Project of Natural Science Research Project of Anhui Universities [grant number KJ2019ZD22].

Competing interests

The authors have no relevant financial or non-financial interests to disclose.

Authors’ contributions

R.Q.Z. designed and supervised the study. L.Y., L.L.L., and H.L.W. performed the research, analysed the data, and wrote the paper. Y.L.H. and K.M.W. validated the research findings with software. X.H. and J.R.X. reviewed and edited the paper.

Ethics approval and consent to participate

TCGA and GEO are public databases. The patients involved in the database have provided ethical approval. Users can download relevant data for free for research and publish relevant articles. Our study is based on open-source data, and therefore did not require Ethics Committee approval.

Consent for publication

Not applicable.

Data availability statement

The raw data required to generate the findings are available for download from the UCSC Xena (https://xena.ucsc.edu/).

Richeldi L, Collard HR, Jones MG: Idiopathic pulmonary fibrosis. Lancet. 2017;389(10082):1941-1952. doi: 10.1016/S0140-6736(17)30866-8.
Wang J, Jiang M, Xiong A, Zhang L, Luo L, Liu Y, Liu S, Ran Q, Wu D, Xiong Y et al: Integrated analysis of single-cell and bulk RNA sequencing reveals pro-fibrotic PLA2G7^high macrophages in pulmonary fibrosis. Pharmacol Res. 2022;182:106286. doi: 10.1016/j.phrs.2022.106286.
Raghu G, Remy-Jardin M, Richeldi L, Thomson CC, Inoue Y, Johkoh T, Kreuter M, Lynch DA, Maher TM, Martinez FJ, Molina-Molina M: Idiopathic Pulmonary Fibrosis (an Update) and Progressive Pulmonary Fibrosis in Adults: An Official ATS/ERS/JRS/ALAT Clinical Practice Guideline. Am J Respir Crit Care Med. 2022;205(9):e18-e47. doi: 10.1164/rccm.202202-0399ST.
Weng T, Ko J, Masamha CP, Xia Z, Xiang Y, Chen NY, Molina JG, Collum S, Mertens TC, Luo F et al: Cleavage factor 25 deregulation contributes to pulmonary fibrosis through alternative polyadenylation. J Clin Invest. 2019;129(5):1984-1999. doi:10.1172/JCI122106
Tzouvelekis A, Gomatou G, Bouros E, Trigidou R, Tzilas V, Bouros D: Common Pathogenic Mechanisms Between Idiopathic Pulmonary Fibrosis and Lung Cancer. Chest. 2019;156(2):383-391. doi: 10.1016/j.chest.2019.04.114.
Samarelli AV, Masciale V, Aramini B, Coló GP, Tonelli R, Marchioni A, Bruzzi G, Gozzi F, Andrisani D, Castaniere I, et al: Molecular Mechanisms and Cellular Contribution from Lung Fibrosis to Lung Cancer Development. Int J Mol Sci. 2021;22(22):12179. doi: 10.3390/ijms222212179
Chen T, Guo J, Ai L, Wang Y, Wang Y, Chen B, Liu M, Zhuang S, Liu K, Zhao Z et al: Up-regulated SPP1 increases the risk from IPF to lung cancer via activating the pro-tumor macrophages.. Comput Struct Biotechnol J. 2023;21:5751-5764. doi: 10.1016/j.csbj.2023.11.018.
Barravecchia I, Lee JM, Manassa J, Magnuson B, Ferris SF, Cavanaugh S, Steele NG, Espinoza CE, Galban CJ, Ramnath N, et al: Modeling Molecular Pathogenesis of Idiopathic Pulmonary Fibrosis-Associated Lung Cancer in Mice. Mol Cancer Res. 2024;22(3):295-307. doi: 10.1158/1541-7786.MCR-23-0480
Ballester B, Milara J, Cortijo J: Idiopathic Pulmonary Fibrosis and Lung Cancer: Mechanisms and Molecular Targets. Int J Mol Sci. 2019;20(3):593. doi: 10.3390/ijms20030593
Yu R, Gao D, Bao J, Sun R, Cui M, Mao Y, Li K, Hu E, Zhai Y, Liu Y et al: Exogenous Thymosin Beta 4 Suppresses IPF-Lung Cancer in Mice: Possibly Associated with Its Inhibitory Effect on the JAK2/STAT3 Signaling Pathway. Int J Mol Sci. 2023;24(4):3818. doi: 10.3390/ijms24043818
Li T, Fu J, Zeng Z, Cohen D, Li J, Chen Q, Li B, Liu XS. TIMER2.0 for analysis of tumor-infiltrating immune cells. Nucleic Acids Res. 2020;48(W1):W509-W514. doi: 10.1093/nar/gkaa407.
Xu L, Deng C, Pang B, Zhang X, Liu W, Liao G, Yuan H, Cheng P, Li F, Long Z et al: TIP: A Web Server for Resolving Tumor Immunophenotype Profiling. Cancer Res. 2018;78(23):6575-6580. doi: 10.1158/0008-5472.CAN-18-0689.
Lee E, Chuang HY, Kim JW, Ideker T, Lee D. Inferring pathway activity toward precise disease classification. PLoS Comput Biol. 2008; 4(11):e1000217. doi: 10.1371/journal.pcbi.1000217.
Liu CJ, Hu FF, Xie GY, Miao YR, Li XW, Zeng Y, Guo AY: GSCA. an integrated platform for gene set cancer analysis at genomic, pharmacogenomic and immunogenomic levels. Brief Bioinform. 2023; 24(1):bbac558. doi: 10.1093/bib/bbac558.
Reinhold WC, Wilson K, Elloumi F, Bradwell K, Ceribelli M, Varma S, Wang Y, Duveau D, Menon N, Trepel J et al: CellMinerCDB: NCATS Is a Web-Based Portal Integrating Public Cancer Cell Line Databases for Pharmacogenomic Explorations. Cancer Res. 2023; 83(12): 1941–1952. doi.org/10.1158/0008-5472.CAN-22-2996.
Malta TM, Sokolov A, Gentles AJ, Burzykowski T, Poisson L, Weinstein JN, Kamińska B, Huelsken J, Omberg L, Gevaert O et al: Machine Learning Identifies Stemness Features Associated with Oncogenic Dedifferentiation. Cell. 2018;173(2):338-354.e15. doi: 10.1016/j.cell.2018.03.034.
Yang C, Zhang H, Chen M, Wang S, Qian R, Zhang L, Huang X, Wang J, Liu Z, Qin W et al: A survey of optimal strategy for signature-based drug repositioning and an application to liver cancer. Elife. 2022;11:e71880. doi:10.7554/eLife.71880.
Xing X, Yang F, Huang Q, Guo H, Li J, Qiu M, Bai F, Wang J: Decoding the multicellular ecosystem of lung adenocarcinoma manifested as pulmonary subsolid nodules by single-cell RNA sequencing. Sci Adv. 2021;7(5):eabd9738. doi:10.1126/sciadv.abd9738
Hänzelmann S, Castelo R, Guinney J: GSVA: gene set variation analysis for microarray and RNA-Seq data. Bmc Bioinformatics. 2013; 14:7. doi: 10.1186/1471-2105-14-7.
Idilman R. Low liver fat in non-alcoholic steatohepatitis-related significant fibrosis and cirrhosis is associated with hepatocellular carcinoma, decompensation and mortality. Aliment Pharm Ther. 2024; 59:80
Chen JY, Yiu WH, Tang PM, Tang SC: New insights into fibrotic signaling in renal cell carcinoma. Front Cell Dev Biol. 2023; 11:1056964. doi: 10.3389/fcell.2023.1056964.
Dai J, Sun Y, Chen D, Zhang Y, Yan L, Li X, Wang J: Negative regulation of PI3K/AKT/mTOR axis regulates fibroblast proliferation, apoptosis and autophagy play a vital role in triptolide-induced epidural fibrosis reduction. Eur J Pharmacol. 2019;864:172724. doi: 10.1016/j.ejphar.2019.172724.
Wang J, Hu K, Cai X, Yang B, He Q, Wang J, Weng Q: Targeting PI3K/AKT signaling for treatment of idiopathic pulmonary fibrosis. Acta Pharm Sin B. 2022;12(1):18-32. doi: 10.1016/j.apsb.2021.07.023.
Tan AC. Targeting the PI3K/Akt/mTOR pathway in non-small cell lung cancer (NSCLC). Thorac Cancer. 2020;11(3):511-518. doi: 10.1111/1759-7714.13328.
Jaiswal J, Egert J, Engesser R, Peyrotón AA, Nogay L, Weichselberger V, Crucianelli C, Grass I, Kreutz C, Timmer J et al: Mutual repression between JNK/AP-1 and JAK/STAT stratifies senescent and proliferative cell behaviors during tissue regeneration. PLoS Biol. 2023 May 30;21(5):e3001665. doi: 10.1371/journal.pbio.3001665.
Milara J, Hernandez G, Ballester B, Morell A, Roger I, Montero P, Escrivá J, Lloris JM, Molina-Molina M, Morcillo E et al: The JAK2 pathway is activated in idiopathic pulmonary fibrosis. Respir Res. 2018;19(1):24. doi: 10.1186/s12931-018-0728-9.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Systematic analysis of functional implications of fibrosis in pan-cancer

Status:

Version 1

Abstract

Figures

Background

Materials and methods

Results

Discussion

Conclusions

Declarations

References

Additional Declarations

Status:

Version 1