A model of coagulation-related genes for prediction of prognosis in lung adenocarcinoma and verification in vitro

Abstract

Objective

Given the close link between abnormal coagulation function and lung adenocarcinoma (LUAD) progression, this study aims to identify coagulation-related biomarkers that influence LUAD prognosis and to validate the biological functions of key genes through in vitro experiments.

Methods

We obtained coagulation-related genes (CRGs) from the TCGA-LUAD and Genecards databases. To identify key CRGs, Cox and lasso analyses were conducted, leading to the establishment of a predictive model for the coagulation risk score (CRRS). The model's predictive accuracy was examined using Kaplan-Meier and receiver operating characteristic (ROC) curve, followed by external validation to ensure robustness. A nomogram was then created by integrating risk scores with clinical factors, enabling the quantification of survival probabilities for individual patients. Additional analyses examined the relationships between risk scores and functional pathways, tumor immune features, and chemotherapy drug sensitivity. Immunohistochemistry (IHC) was performed to validate ASPH expression. Finally, ASPH expression was silenced in A549 and NCI-H1975 cell lines, and the impact on tumor cell behavior was evaluated.

Results

The model constructed in this article can effectively predict the prognosis of LUAD patients. CRRS is an independent prognostic factor for LUAD, and CRRS was significantly correlated with multiple clinical indicators. Additionally, immune checkpoint expression was meaningfully higher in high-risk patients compared to low-risk patients. IHC analysis revealed an increase in ASPH expression in LUAD specimens. Downregulation of ASPH impairs the malignant biological behavior of LUAD cells.

Conclusion

The CRRS model provides reliable prognostic value for LUAD patients. ASPH could function as both a molecular marker and a potential treatment target for patients.

Introduction

Lung cancer is one of the most prevalent and fatal malignancies, with both its frequency and death rates on the rise. According to the Global Cancer Statistics 2022, lung cancer represents 18.7% of global cancer and continues to be the main contributor to cancer deaths worldwide. ¹ The early symptoms of lung cancer are not easy to detect, which leads to many patients being diagnosed at an advanced stage. ² However, with the increased awareness of health check-ups in recent years, the availability of convenient CT scans and screening programs has also improved survival rates. ³ Although targeted therapy and immunotherapy have made significant progress in the field of lung cancer treatment in recent years, their overall efficacy remains limited. ⁴ Identifying key molecules involved in lung cancer and unraveling the molecular mechanisms driving its progression are crucial steps toward improving early diagnosis, treatment and prognosis.

Recently, the function of blood coagulation system in tumor progression has caused wide public concern. Clinical observations have found that patients with malignant tumors commonly exhibit a hypercoagulable state. The most frequent clinical manifestation of cancer-associated hypercoagulability is venous thromboembolism (VTE), which is related to the deterioration of the patient's condition and potential delays or interruptions in systemic cancer treatment, ⁵ and VTE remains a major cause of mortality among lung cancer patients. ⁶ Tumors can disrupt the balance between the coagulation system and fibrinolytic system, and tumor cells activate the coagulation system to form a hypercoagulable state by producing tissue factor (TF) and cancer procoagulants, leading to increased blood coagulability. ⁷ In the tumor microenvironment (TME), coagulation factors interact with tumors and stromal cells. Thrombin interacts with protease activated receptors on tumor cell surfaces, triggering cellular signaling, enhancing the tumor cells’ ability to penetrate the vascular wall and invade surrounding tissues.^8,9 Moreover, thrombin's direct interaction with tumor cells can stimulate the formation of new blood vessels. ¹⁰ Furthermore, the coagulation system is also interconnected with immune regulation. Tumor associated macrophages autonomously produce FXa, leading to protease activated receptor 2 signaling and PD-L1 expression, aiding in tumor immune evasion and metastasis. ¹¹ Studies have found that the thrombin-platelet GARP–LTGF-β1 axis is a pathway in cancer immunotherapy. Thrombin cleaves glycoprotein A repetitions predominant of the platelet surface, leading to the release of active transforming growth factor–β1, allowing tumor cells to evade the attack of immune system. ¹² Inhibiting certain key factors in the coagulation pathway has become one of the potential strategies for anti-cancer therapy.

The ASPH (aspartate β-hydroxylase) gene is located on chromosome 8. It adds hydroxyl groups to β carbons of particular aspartate or asparagine residues of proteins.^13–15 ASPH is rarely expressed in most normal tissues but is frequently dysregulated in various cancers. ¹⁶ ASPH induces a malignant phenotype in tumor cells by enhancing the proliferation, motility, and anti-apoptotic capabilities of cancer cells.^17,18 Furthermore, ASPH is engaged in regulating multiple signaling pathways. ¹⁹ Inhibiting the catalytic site of the C-terminal region of ASPH using the MO-I-1100 has shown promising anti-tumor effects. It reduces tumor growth and distant metastasis in hepatocellular carcinoma (HCC) through suppressing the NOTCH signaling pathway. ²⁰ ASPH activates the SRC signaling pathway through interactions with ADAM12/ADAM15 and initiates MMP-mediated extracellular matrix (ECM) degradation/remodeling, guiding the migration, invasion, and distant metastasis of pancreatic tumor cells. ²¹ Studies have found that knocking down ASPH expression inhibits the development and growth of cholangiocarcinoma by reducing RB1 phosphorylation. ²² Although some studies have explored the multiple mechanisms of ASPH in tumor progression, the specific mechanisms of ASPH as a potential oncogene in LUAD have not been systematically revealed. Therefore, further investigation of the comprehensive role of ASPH in LUAD could offer valuable insights.

The aim of this study was to explore the relationship between CRGs and the prognosis of LUAD patients, providing new candidate tools for the prognostic prediction of LUAD. In this research, we utilized data from the GEO and TCGA databases to construct a CRRS prognostic model aimed at predicting individualized outcomes for LUAD patients and validated its reliability. Subsequently, we integrated the risk scores with clinical information to create a nomogram, finding that the risk score is a prognostic factor associated with adverse clinical indicators. Additionally, we also analyzed the relationships between the model and immune microenvironment, as well as between the model and drug sensitivity. Finally, we examined the transcriptional regulation of genes related to the model. The specific mechanism of ASPH in lung cancer has yet to be determined, and we identified ASPH for further analysis. The results indicated that ASPH expression differs between LUAD and normal lung tissues. ASPH is a risk factor for LUAD and has prognostic significance. Moreover, in vitro experiments demonstrated that inhibiting ASPH could reduce the proliferation ability of LUAD cells. This research offers a molecular basis for understanding the pathogenesis of LUAD and provides therapeutic targets for lung cancer therapy.

Materials and methods

Data acquisition

We collected the processed raw mRNA expression data of LUAD from the TCGA (https://portal.gdc.cancer.gov/), including the normal group (n = 59) and the tumor group (n = 541). Only patients with complete clinical information were included in the subsequent analysis. Download the Series Matrix File data file of GSE68465 ²³ from the GEO public database, annotated on GPL96 platform, and extract a total of 442 patient data with complete expression profiles and survival information. Download the Series Matrix File data file of GSE72094, ²⁴ annotated on GPL15048 platform, and extract a total of 398 patient data with complete expression profiles and survival information. For the TCGA dataset, FPKM values were transformed using log2 (FPKM +1), genes with low expression were filtered out. For GEO data, probe ID was mapped to the corresponding gene symbol, and multiple probes corresponding to the same gene were averaged to obtain a single value. The subcellular localization information of ASPH was obtained from the HPA website(https://www.proteinatlas.org/) through online searching.

Obtaining CRGs

We obtained 1192 coagulation genes (Relevance score>1) from the GeneCards website (https://www.genecards.org/). LUAD mRNA expression data was downloaded from the TCGA data portal, and Venny analysis was performed using the VENNY 2.1 (http://bioinfogp.cnb.csic.es/tools/venny/index.html). Overlapping genes were screened using Cox univariate analysis (p < 0.001).

Functional enrichment analysis of prognostic genes

To investigate the biological functions associated with disease progression, we utilized the online tool Metascape for comprehensive enrichment analysis. Prognosis-related genes were analyzed using Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway tools. Statistically significant results were defined by a min overlap ≥ 3 & p ≤ 0.01.

Model construction and prognosis

Select CRGs and further construct a prognosis related model using lasso regression. Using the expression values of each gene we screened, a risk score formula was constructed for each patient, with weights assigned based on regression coefficients estimated in lasso regression analysis. Patients were categorized into high-risk and low-risk groups based on the median risk score as the cut-off point. The Kaplan-Meier analysis was applied to assess survival differences between these groups, while comparisons were conducted using log-rank tests. ROC was conducted to assess the predictive accuracy of the model.

Nomogram model construction

According to regression analysis, a nomogram was constructed using gene expression levels and clinical symptom data, with scaled line segments on the same plane representing the relationships between the variables in the predictive model. Thereby calculating the predictive value of the individual outcome event. We constructed a nomogram integrating the risk scores and clinical information using the ‘rms’ package. At the same time, decision curve analysis (DCA) was performed using the “ggDCA” package to assess the model performance by calculating the net benefit. In addition, calibration curves and ROC curves were constructed to assess the predictive performance of the nomogram. Cox regression analyses were conducted to assess whether the risk score could serve as an independent prognostic factor.

Analysis of immune cell infiltration

The CIBERSORT analysis tool is often used to assess immune cell types within the microenvironment. It provides 547 biomarkers and 22 different human immune cells. In this research, the CIBERSORT method was used to analyze the relative proportions of 22 types of immune infiltrating cells.

Drug sensitivity analysis

We utilized the R software package ‘pRRophetic’ to predict chemotherapy sensitivity for tumor samples based on the Pharmacogenomics database (GDSC cancer drug sensitivity Genomics database, https://www.cancerrxgene.org/). Obtain the estimated IC50 values for each particular chemotherapy drug treatment through regression method, and conduct 10-fold cross validation tests on the GDSC training set to test the prediction accuracy.

Gene set variation analysis (GSVA)

GSVA can convert the gene matrix into a gene-set matrix, and convert the changes at the gene level into changes at the pathway level, so as to evaluate whether different pathways are enriched between various samples. In this analysis, gene sets from the Molecular Signatures Database (MSigDB, version 7.0) will be downloaded and analyzed using the Gene Set Variation Analysis (GSVA) algorithm.

Gene set enrichment analysis (GSEA)

GSEA was used to analyze the differences in signaling pathways between high-risk and low-risk groups. Background gene set, as the annotation gene set of subtype pathway, is a group of annotation genes downloaded from the MsigDB database. We performed differential expression analysis targeting different pathways among subtypes, and significantly enriched gene sets (p < 0.05) were ranked according to the concordance score.

Analysis of model genes regulation

We analyzed transcriptional regulation of key genes. Motifs are conserved sequences that play a role in gene regulation analysis. We perform normalized enrichment score (NES) on motifs. The initial step in evaluating overexpression of each motif for a gene set is calculating the area under the curve (AUC). The NES was calculated for each motif based on its AUC distribution. In this research, R-package “RcisTarget” was used to analyze transcription factors. Rcistarget.hG19.motifdb.cisbpont.500 bp was used for gene-motif ranking database.

Cell culture and transfection

The LUAD cell lines A549 (RRID: CVCL_0023) and NCI-H1975 (RRID: CVCL_1511) were obtained from the ATCC. The cells were carefully cultured in F-12 medium and RPMI-1640 medium supplemented with 10% fetal bovine serum and a constant-temperature incubator (37°C, 5% CO2). Lung cancer cells in the logarithmic growth phase were seeded in six-well plates, and the ASPH knockdown plasmid and negative control plasmid were transfected into A549 and NCI-H1975 cells using the transfection reagent Lipo3000. After 48 h, the medium was replaced with fetal bovine serum-containing culture medium.

IHC staining

Five pairs of tumor tissues were obtained from the Affiliated Hospital of Nantong University on February 2025. After pathological sectioning, they were sequentially baked, dewaxed with xylene, and hydrated with gradient ethanol. Antigen retrieval was performed using a microwave, after which the primary antibody ASPH was added. ASPH (14116-1-AP, Proteintech, RRID:AB_2060148) was purchased from Wuhan Sanying Biotechnology Company. The antibody was diluted at 1:500. Goat anti-Rabbit IgG (H + L) Secondary Antibody (HRP) (G-21234, Invitrogen, RRID:AB_2536530) was purchased from thermo Fisher Scientific Inc. and diluted at 1:500. The sections were then placed in a 4°C refrigerator for approximately 14–16 h. On the next day, the secondary antibody was added to sections after 3 × 2 min of cleaning with Tris-Buffered Saline with Tween (TBST) solution. After 30 min of incubation, the secondary antibody was removed after TBST cleaning and DAB staining was performed for 5 min. Subsequently, hematoxylin counterstaining and dehydration in alcohols, neutral gum was used for mounting. The positive staining area and intensity were observed under the microscope and scored (0 points: 0-5%, 1 points: 6%-30%, 2 points: 31%-70%, 3 points: > 71%). The total score ≤7 was classified as negative expression and > 7 was classified as positive expression.

Western blotting

ASPH (14116-1-AP, Proteintech, RRID:AB_2060148) and GAPDH (10494-1-AP, Proteintech, RRID AB_2107436) were purchased from Wuhan Sanying Biotechnology Company. The ASPH and GADPH were diluted at 1:1000 and 1:10000. HRP-conjugated Goat Anti-Rabbit IgG (H + L) (SA00001-2, Proteintech, RRID:AB_2722564) and HRP-conjugated Goat Anti-Mouse IgG (H + L) (SA00001-1, Proteintech, RRID:AB_2722565) were purchased from Wuhan Sanying Biotechnology Company. The antibodies were diluted at 1:10000. Samples from each experimental group were collected, and equal proportions of RIPA lysis buffer were added. The protein concentrations of each group were measured using the BCA method, and the loading amounts were adjusted according to the measured concentration data. After gel electrophoresis, the protein sample was transferred to a PVDF membrane, 5% skim milk was used to block the membranes for 1 h. The membranes and the first antibody were incubated in refrigerator at 4°C for 12 h, then washed three times with TBST for 5 min each, incubated with secondary antibodies for 90 min at room temperature. Finally, enhanced chemiluminescence (ECL) reagent was used for exposure in the gel imaging system.

Cell proliferation assay

Cells from the negative control group and the ASPH knockdown group were seeded in a 96-well plates at a density of 5000 cells per well. At the designated times of 24, 48, and 72 h, 10 µL of the Cell Counting Kit-8 (CCK8) reagent was added. The plate was kept in the dark and placed in an incubator for 3 h. Finally, absorbance was recorded using a microplate reader.

Transwell assay

Cells were resuspended in 200 µL of serum-free medium at a density of 2 × 10^4 cells and then added to the upper chamber. The lower chamber was filled with 600 µL of medium containing 10% fetal bovine serum. After incubating for 48 h in the incubator, the chambers were removed, and the cells were fixed with formaldehyde. Subsequently, the cells were stained with crystal violet.

Scratch migration test

Cells were seeded in a 6-well plate, and when they reached complete confluence, a straight wound was made using a 1000 µL pipette tip. Fresh complete culture medium was added and cultured after washing the cells with PBS solution. And the cells were allowed to continue to culture. Images of the wound area were taken at 0 h and 48 h to analyze differences between the groups.

Statistical analysis

Survival curves were constructed using the Kaplan-Meier method, and group differences were evaluated with the log-rank test. Statistical analysis was performed using Student's t-test, the Wilcoxon rank-sum test, and one-way ANOVA. All statistical analysis was performed using R software (version 4.2), SPSS (version 20), GraphPad Prism (version 8.0), and ImageJ (version 1.5). P < 0.05 indicates statistical significance. All experiments were performed with three independent biological replicates.

Ethics statement

This study was conducted in accordance with the Helsinki Declaration of 1975 as revised in 2024. Written informed consent was obtained from all patients included in this study. This investigation was approved by Ethics Committee of the Affiliated Hospital of Nantong University (October 9, 2024; 2024-L226).

Results

Construction of prognostic models

First, we obtained 1192 Coagulation genes (Relevance score>1) through the GeneCards (https://www.genecards.org/), and then downloaded the LUAD mRNA expression data from the TCGA. After intersection, 1137 genes were overlapping in total. (Figure 1A). Overlapping genes were further screened using single-factor Cox regression analysis. The results Indicated that 22 genes were prognostic genes related to LUAD (p < 0.001) (Figure 1B). Subsequently, we used the Metascape database to analyze the pathways of these 22 prognostic related genes, and the analysis revealed that the enriched pathways were mainly regulation of phospholipid transport, hexose metabolic process, and cell population proliferation (Figure S1).

Figure 1.

Identification of prognostic genes. (A) The Venn diagram. (B) Univariate Cox regression analysis.

We used lasso regression analysis to further search for prognostic genes., and we finally obtained 12 coagulation-related prognostic genes (Figure 2A-C). Figure 2D-E showed the expression heat map of 12 CRGs between different groups. Patients were randomly assigned to the training set and the validation set in a ratio of 4:1. Calculate the optimal risk score value for each sample (Risk Score = FUT1 × (−0.108) + PRKCD × (−0.101) + SFTPB × (−0.0316) + ABCC2 × 0.001 + RPE65 × 0.028 + ASPH × 0.041 + KLK8 × 0.063 + ASPM × 0.070 + F2 × 0.093 + GALNT2 × 0.102 + B4GALT1 × 0.141 + ENOX1 × 0.144). Based on the risk scores, patients were categorized into high-risk and low-risk groups, and prognostic analysis was performed using the Kaplan-Meier curve. The results indicated that the overall survival (OS) of the high-risk group was significantly lower than that of the low-risk group, both in the training set and the testing set (Figure 2F-G). The ROC curve showed that the AUC for the 1-year, 2-year, and 3-year training and testing sets exceeded 0.70, suggesting strong validation performance of the model (Figure 2H-I).

Figure 2.

Establish a prognostic model. (A, B) LASSO regression analysis. (C) LASSO coefficients of the 12 predictor genes for constructing the prognostic model. (D, E) Gene expression heatmap. (F) Kaplan–Meier survival curve analysis in the high-risk and low-risk groups in the training set. (G) Kaplan–Meier survival curve analysis in the high-risk and low-risk groups in the testing set. (H) ROC curve in the training set. (I) ROC curve in the testing set.

In addition, we download data of LUAD patients from the GEO (GSE68465, GSE72094) to verify the robustness of the model using external datasets. The outcome indicated that in the external validation set, the OS of patients in the high-risk group was significantly lower than that in the low-risk group (Figure S2A-B). The ROC curve also demonstrated that the model demonstrated good efficacy in the prognosis of patients at 1 year, 2 years, and 3 years (Figure S2C-D).

Establishment of the nomogram for survival prediction

For the purpose of analyzing the independent prognostic factors of LUAD patients, we constructed nomogram plots according to age, sex, TNM, and risk score, providing a visual method to predict the survival rate of LUAD patients (Figure 3A). The calibration curve for 3-year survival indicated that the nomogram demonstrated good predictive performance (Figure 3B). The ROC curve results showed reliable prediction performance, with a 3-year AUC of 0.7312(Figure 3C). The decision curve analysis also shows that risk score has a good net benefit (Figure 3D). In addition, Cox regression analysis indicated that N and risk score were independent factors for the prognosis of LUAD patients (Figure 3E-F). These results suggest that the nomogram developed in this research has strong prognostic potential for LUAD patients. The results showed that risk score was related to Fustat and N (Figure 3G).

Figure 3.

Establishment of the nomogram for survival prediction. (A) The nomogram for predicting the 1- and 3- years OS of LUAD patients. (B) The calibration curve of the nomogram for predicting 1- and 3-years OS of LUAD patients. (C) Time-dependent ROC curve for 1-year, 2-years, and 3-years prediction. (D) Decision curve analysis. (E) Univariate Cox regression analysis. (F) Multifactorial Cox regression analysis. (G) Correlation analysis of risk scores with clinical characteristics.

Relationship between prognostic models and the immune landscape

We investigated the association between risk scores and tumor immune infiltration. The figure illustrated the differences in immune cell proportions between the high-risk and low-risk groups (Figure 4A). Immune cell infiltration analysis indicated that T cells regulatory, Macrophages M0 and Macrophages M2 were significantly increased in the high-risk group (Figure 4B). In addition, immune-related chemokines, immune checkpoints, immunostimulatory factors, and immune receptors were analyzed, and the differences in expression between the high and low risk groups are shown in the figure (Figure 4C). Subsequently, we predicted the sensitivity of risk groups to anti-tumor immunotherapy. The findings indicated that the high-risk group had a worse response to immunotherapy (Figure 4D). We also observed that risk scores were closely correlated with patient sensitivity to chemotherapeutic agents, including paclitaxel, docetaxel, gemcitabine, cisplatin, doxorubicin, and the targeted therapy erlotinib (Figure 4E). These results do not represent clinical evidence and require further experimental and clinical validation.

Figure 4.

Prognostic models and immune microenvironment analysis. (A) Proportions of immune cells in low-risk and high-risk subgroups. (B) Box plot of the differentiation of immune cells in low-risk and high-risk subgroups. (C) The distinct expressions of immune-related chemokines, immunosuppressants, immunostimulatory factors, immune receptor genes between two risk groups. (D) The response to immunotherapy. (E) The sensitivity to chemotherapy drugs between high and low risk groups.

Potential signaling pathways associated with LUAD

Subsequently, we investigated the signaling pathways associated with the model. Significant differences of signaling pathways were observed between the low-risk and high-risk groups across various biological processes, the results of GSVA revealed that EPITHELIAL MESENCHYMAL TRANSITION, G2 M CHECKPOINT, E2F TARGETS and other signaling pathways were mainly enriched in the high-risk group, FATTY ACID METABOLISM, ESTROGEN RESPONSE LATE, BILE ACID METABOLISM and other signaling pathways were mainly enriched in another group (Figure 5A). The results of GSEA indicated that Central carbon metabolism in cancer, HIF−1 signaling pathway, and p53 signaling pathway was up-regulated in the high-risk group (Figure 5B). The enrichment of these pathways suggests that the poor prognosis in the high-risk group may be associated with dysregulated energy metabolism, neovascularization, cell proliferation, and tumor drug resistance. However, further experimental validation is needed to confirm these mechanistic links.

Figure 5.

Potential signaling pathways associated with LUAD. (A) Differences in pathway activities scored by GSVA between high-risk and low-risk groups. The blue column indicates activated pathways in the high-risk group, and the green column indicates activated pathways in the low-risk group. (B) GSEA analysis based on the risk grouping.

Transcriptional regulation of model genes

We found that 12 model genes were regulated by various mechanisms including transcription factors. For this reason, the enrichment analysis of these transcription factors was conducted using the cumulative recovery curve (Figure 6A). The results indicated that cisbp__M5975 was the motif with the highest standardized enrichment score (NES: 5.09), and four model genes (B4GALT1, ENOX1, FUT1, KLK8) were enriched in this motif. The enriched motifs and transcription factors of the model gene were demonstrated (Figure 6B).

Figure 6.

Analysis of transcriptional regulation of model genes. (A) Enrichment analysis of these transcription factors was performed using cumulative recovery curves. (B) All the enriched motifs and corresponding transcription factors of the model genes.

Expression of ASPH in LUAD

Among the 12 coagulation-related prognostic genes, ASPH piqued our interest. We displayed the subcellular localization of ASPH through the HPA database. The results indicated that based on the A549 cell lines, ASPH was mainly localized in the endoplasmic reticulum (Figure 7A). We analyzed the expression of ASPH, the analyses indicated that the expression level of ASPH was higher in tumor tissues compared to normal tissues (Figure 7B). In this study, we sliced the pathological specimens of patients who underwent radical resection of lung cancer in thoracic surgery, and detected the ASPH in tumor tissues and matched normal lung tissues through IHC staining. ASPH expression is notably higher in LUAD tissues compared to normal tissues (Figure 7C). We used Kaplan Meier analysis to confirm the relationship between ASPH expression and OS. The analysis indicated that the OS of patients with high ASPH was lower than that of patients with low ASPH (Figure 7D).

Figure 7.

Clinical role of ASPH in LUAD. (A)Subcellular localization of ASPH. (B) Analysis of the expression level of ASPH. (C) IHC image of ASPH in tumor and normal tissue. Images are shown at 10× magnification. (D) The OS difference in high- and low-ASPH expression patients.

Silencing of ASPH inhibits the malignant biological phenotype of LUAD cells

For the purpose of investigating the influence of ASPH on the biological function of LUAD cells, we silenced the ASPH gene in LUAD cell lines and detected the knockdown efficiency through the Wb experiment (Figure 8A). Finally, we selected the two fragments with the highest knockdown efficiency for subsequent experiments. Firstly, CCK8 experiments showed that knocking down ASPH significantly reduced the proliferation ability of A549 and NCI-H1975 (Figure 8B). Subsequently, we conducted experiments related to invasion and migration. Transwell experiments showed that knocking down ASPH significantly reduced the invasive ability of LUAD cells (Figure 8C). Finally, the cell scratch healing experiment showed that the in vitro migration ability of tumor cells after knocking down for 48 h decreased compared to the control group (Figure 8D). In light of the aforementioned experiments, knocking down ASPH can significantly inhibit the malignant biological phenotype of LUAD cells.

Figure 8.

Effect of ASPH expression on the biological behavior of LUAD cells. (A) Western-blot detects knockdown efficiency of ASPH. (B) CCK8 assay to detect the proliferative capacity of A549 and NCI-H1975 cells. (C) Transwell assay to examine the invasion ability of A549 and NCI-H1975 cells. (D) Scratch healing assay examining the migration ability of A549 and NCI-H1975 cells. All images are shown at 10× magnification.

Discussion

Tumor development and progression involve intricate and multifaceted biological processes. Exploring the pathogenesis of LUAD at the molecular level may aid in identifying new therapeutic targets for this cancer. Numerous clinical studies have shown that enhanced coagulation is often observed in cancer patients, a phenomenon known as cancer-associated thrombosis, suggesting a potential link between tumor progression and the coagulation system.^5,7 Tumor cells can release various soluble proinflammatory factors and proangiogenic factors. In endothelial cells, these molecules promote TF expression, stimulate the production of plasminogen activator inhibitor-1 (PAI-1), downregulate thrombomodulin and promote the upregulation of cell adhesion molecules, favoring localized coagulation activation and thrombosis. ²⁵ The functions of the coagulation system are multifaceted, involving blood coagulation and hemostasis, maintenance of blood fluidity, and prevention of thrombosis. ²⁶ Recent studies indicate that the coagulation system also participates in tumorigenesis, including angiogenesis, tumor proliferation, and immune response regulation.^27,28 On one side, various components of the coagulation cascade (such as TF, tumor procoagulant proteins, microparticles, proangiogenic factors, and cytokines) can promote tumorigenesis, contributing to a persistent hypercoagulable state, significantly raising the risk of thrombotic events. ²⁹ A meta-analysis has shown that various coagulation abnormalities are closely related to lung cancer, with coagulation indicators may offer valuable insights for the timely diagnosis of lung cancer patients. ³⁰ Elevated plasma fibrinogen and high levels of D-dimer in LUAD patients have been associated with reduced survival rates.^31,32 On the other side, coagulation factors in TME can influence the activity of immune cells, potentially allowing cancer cells to evade immune surveillance. ³³ We built a prognosis risk model based on CRGs to predict patient survival, facilitating clinical treatment management.

Currently, prognostic models based on CRGs have proven effective in various cancers, such as glioma, breast cancer, and HCC.^34–36 Our findings are consistent with previous reports showing that coagulation-related gene signatures are closely associated with prognosis and immune microenvironment in lung adenocarcinoma. ³⁷ Coagulation genes have shown great potential in predicting the prognosis of LUAD patients, helping to guide clinical treatment decisions. In this research, we performed differential analysis of coagulation-related mRNA expression matrices for LUAD in the TCGA database and identified 22 prognostic-related genes. We then employed lasso regression analysis to identify the characteristic genes of LUAD and established a new prognostic model based on 12 CRGs. Predictive model identified ENOX1, B4GALT1, GALNT2, F2, ASPM, KLK8, ASPH, RPE65, and ABCC2 as risk factors, while SFTPB, PRKCD, and FUT1 were protective factors. B4GALT1 facilitates the proliferation and invasion of cancer cells, leading to immune evasion in lung cancer and is involved in maintaining and proliferating lung cancer stem cells.^38,39 GALNT2 has been recognized as a possible target for treatment and prognostic marker for LUAD. ⁴⁰ ASPH may promote tumor growth by stimulating angiogenesis and immunosuppression, ¹⁹ while overexpression of ABCC2 can increase the resistance of LUAD cells to cisplatin, ⁴¹ and ASPM has been related with poor prognosis in lung cancer patients. ⁴² Subsequently, the optimal risk score values for each sample were collected by lasso regression analysis, and the patients were assigned to two groups based on the median risk score. Patients in the high-risk group had worse OS compared to the low-risk group. ROC curve analysis indicated good validation performance of this model. We combined the risk scores and clinical indicators to create a nomogram plot, which allows for a more personalized prediction of survival probabilities in LUAD patients. The univariate and multivariate Cox analysis indicated that the risk score could be considered an independent prognostic factor.

The TME is a cellular environment consisting of tumor cells, immune cells, ECM, fibroblasts, and cytokines. The microenvironment not only provides a site for tumor cells to survive and proliferate, but also assists tumors to escape immune surveillance and develop distant metastasis. Tumor-infiltrating immune cells can effect the efficacy of immunotherapy.^43,44 CD8+ T cells directly kill tumor cells by releasing cytotoxins, and CD4+ T cells can secrete cytokines to regulate immune responses, while regulatory T cells and M2 macrophages play a negative immunomodulatory role in the microenvironment.^44,45 We analyzed immune cell infiltration in patients of different risk groups. We observed significantly decreased levels of CD8+ T cells, resting memory CD4+ T cells, activated NK cells, monocytes, M1 macrophages, resting dendritic cells, and resting mast cells in the high-risk group. Conversely, activated memory CD4+ T cells, regulatory T cells, resting NK cells, M0 macrophages, and M2 macrophages were elevated, indicating an immunosuppressive state. Tumor cells exploit abnormally expressed immune checkpoint molecules to evade immune surveillance and attack. ⁴⁶ For example, tumor cells overexpress PD-L1, which binds to PD-1 on the surface of T cells and inhibits T cell activity. Immune checkpoint inhibitors restore the recognition and killing capacity of T cells against tumor cells by blocking PD-1 binding to PD-L1.^47–49 Therefore, we explored the association between risk scores and immune checkpoint expression and observed that the expression levels of LAG3, CTLA4, PD-1, and PD-L1 were significantly higher in the high-risk group compared to the low-risk group. This observation implies that an immunosuppressive microenvironment might be partially responsible for the negative prognosis seen in high-risk patients, reflecting a potential biological link between coagulation and immune evasion. Interestingly, our study displayed that high-risk patients have a lower response to immunotherapy. This observation reflects the complexity of the TME, which may be related to T cell exhaustion, tumor mutation burden, and microsatellite instability.^50–52 Although our drug sensitivity analysis suggests differential responses between risk groups, these results are predictive in nature and do not establish clinical efficacy. Future experimental studies, including in vitro drug assays and patient-derived xenograft models, are required to validate these findings.

The differential pathways identified by GSVA enrichment analysis primarily include Epithelial-Mesenchymal Transition, G2 M Checkpoint, and E2F Targets pathways. GSEA results indicated that the principal pathways were Central carbon metabolism in cancer, HIF−1 signaling pathway, and p53 signaling pathway. These findings indicate a correlation between the high-risk score and aggressive tumor hallmarks, although the specific regulatory mechanisms remain to be elucidated. Finally, ASPH (aspartate beta-hydroxylase) particularly attracted our attention among the 12 genes in our model. ASPH is a transmembrane protein and expresses in 70–90% of human solid tumors, plays a significant role in cell proliferation, invasion, and metastasis.^14,18 Current research has indicated that ASPH is highly expressed in LUAD, consistent with our IHC findings. ⁵³ However, the connection between ASPH and the malignant biological behavior of LUAD cells is not clear. We knocked down the expression level of ASPH in A549 and NCI-H1975, and examined the growth, invasion and migration of cells by CCK8 assay, transwell and cell scratch assay. The findings indicated that knockdown of ASPH substantially inhibited the malignant behaviors of both A549 and NCI-H1975 cell lines, suggesting that ASPH may function as an oncogene promoting advancement of LUAD.

Despite the promising findings, our study has several limitations that should be acknowledged. First, in terms of experimental verification, the clinical cohort used for IHC staining was small (n = 5), serving primarily as a preliminary verification of protein expression trends. Second, while we validated the function of ASPH in two representative LUAD cell lines (A549 and NCI-H1975) with distinct genetic backgrounds, we did not employ in vivo models (such as xenografts) in this study. Future investigations incorporating animal models and larger prospective clinical cohorts are necessary to confirm the therapeutic potential of ASPH and the robustness of the prognostic model.

Conclusions

We determined 12 key CRGs associated with LUAD. The CRRS prognostic model can effectively predict the prognosis of LUAD patients and has clinical application potential. In addition, ASPH was both highly expressed in LUAD and associated with its prognosis. ASPH may function as a biomarker and therapeutic target for LUAD.