DKK1 Overexpression in Lung Adenocarcinoma: Prognostic Implications, Immune Microenvironment Correlates, and Regulatory Network Characterization

Abstract

Background

Lung adenocarcinoma (LUAD) is a prevalent malignancy with poor survival outcomes, underscoring the need for better biomarkers and therapeutic targets. Dickkopf-1 (DKK1), a secreted Wnt pathway inhibitor, is dysregulated in many cancers and can promote tumor progression and immune evasion. This study aimed to investigate DKK1 expression and its clinical, immunological, and molecular significance in LUAD.

Methods

Transcriptomic and clinical data from TCGA and GTEx were analyzed to evaluate DKK1 expression, prognostic relevance, immune infiltration, and associated pathways. Functional enrichment and co-expression networks were constructed in silico. Key findings were validated in vitro using siRNA-mediated DKK1 knockdown in lung cancer cells, followed by assays of proliferation, migration/invasion, apoptosis, cell cycle, and protein expression.

Results

DKK1 was significantly overexpressed in LUAD and correlated with advanced stage and poor survival. Enrichment analyses suggested roles in extracellular matrix remodeling and invasion. High DKK1 expression was also associated with increased infiltration of innate immune cells and elevated PD-L1 expression, indicating a potential role in shaping an immunosuppressive microenvironment. Functional experiments further confirmed that DKK1 knockdown reduced proliferation, impaired migration and invasion, induced cell-cycle arrest, and promoted apoptosis in LUAD cells.

Conclusions

DKK1 serves as a strong prognostic biomarker in LUAD, linking high expression to aggressive clinicopathologic features and immunosuppressive microenvironments. Its knockdown reversed malignant phenotypes in vitro, supporting DKK1 as a potential therapeutic target and a predictor of immunotherapy resistance.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12575-025-00319-9.

Introduction

Lung cancer remains one of the most frequently diagnosed and lethal malignancies worldwide [1]. Non-small cell lung cancer (NSCLC) accounts for approximately 85% of all lung cancer cases, with lung adenocarcinoma (LUAD) representing its most common histologic subtype [2]. Despite notable advancements in the surgical, chemotherapeutic, targeted, and immunotherapeutic management of LUAD, long-term prognosis for patients remains poor, with less than 20% achieving five-year survival [3, 4]. A key challenge contributing to this outcome is the scarcity of reliable biomarkers capable of accurately predicting prognosis and guiding treatment selection [5]. There is an urgent need to better understand the molecular drivers of LUAD and the composition of tis tumor microenvironment (TME) in order to identify novel prognostic indicators and therapeutic targets [6–8].

The TME plays a pivotal role in LUAD progression and treatment response [9]. In particular, dynamic interactions between tumor cells and various immune cell populations significantly influence tumor growth, metastatic potential, and sensitivity to therapy [10]. Although immune checkpoint inhibitor‑based immunotherapy has improved outcomes for a subset of NSCLC patients, treatment responses are highly variable, and acquired resistance remains common [11]. Thus, identifying robust biomarkers that reflect the complex interplay between tumor and immune cells is essential for selecting patients most likely to benefit from immunotherapy and for guiding strategies to overcome immune resistance [5].

Dickkopf-1 (DKK1), a secreted antagonist of the Wnt/β-catenin signaling pathway, has emerged as a potentially important regulator in cancer biology [12–14]. DKK1 expression is frequently dysregulated in multiple malignancies; it is significantly elevated in tumors including breast cancer and NSCLC [15–17]. Clinically, high levels of DKK1 in patient tumors tissue or serum have been associated with advanced disease stage and unfavorable prognosis [18, 19]. In lung cancer, DKK1 has previously been proposed as a promising serologic and prognostic biomarker. Functionally, DKK1 can promote tumor progression through diverse mechanisms. It actively shapes the TME by modulating stromal and immune cells. For instance, DKK1 can recruit or enhance immunosuppressive myeloid-derived suppressor cells (MDSCs) while impairing cytotoxic T-cell activity [20]. Elevated DKK1 expression has also been linked to immune-evasion phenotypes, including upregulation of programmed death‑ligand 1 (PD‑L1) and suppression of natural killer cell-mediated tumor clearance [21, 22]. These immunomodulatory properties position DKK1 as an attractive target for cancer immunotherapy [23–26]. Indeed, preclinical studies demonstrate that DKK1 blockade can inhibit tumor growth and remodel the TME toward a more immunopermissive state. Reflecting this therapeutic potential, an anti-DKK1 monoclonal antibody (DKN-01) has entered clinical trials in combination with other agents, underscoring the growing interest in DKK1 as a target in solid tumors [27].

Given the accumulating evidence implicating DKK1 in LUAD pathogenesis and immune regulation [28, 29], we performed a comprehensive, multi‑modal investigation to systematically characterize its expression, clinical relevance, immunologic correlates, and underlying regulatory networks. In this study, we integrated transcriptomic data from The Cancer Genome Atlas (TCGA), single-cell RNA sequencing resources, and in silico predictions with experimental validation in vitro. Our specific aims were to: (1) evaluate DKK1 expression levels in LUAD versus normal lung tissues; (2) assess the association of DKK1 expression with clinicopathological characteristics and patient survival; (3) examine correlations between DKK1 and tumor-infiltrating immune cells as well as immune checkpoint molecules; (4) explore the DKK1 co-expression landscape and perform functional enrichment and protein-protein interaction network analysis; (5) investigate DKK1 expression patterns at single-cell resolution; (6) analyze the relationship between DKK1 expression and drug sensitivity; (7) construct a putative long non-coding RNA (lncRNA)–microRNA (miRNA)–DKK1 competing endogenous RNA (ceRNA) regulatory axis; and (8) experimentally validate the functional impact of DKK1 knockdown on proliferation, apoptosis, and invasion in LUAD cell models. By illuminating the multifaceted role of DKK1 through both computational and experimental approaches, our findings may provide insights into new prognostic biomarkers and opportunities for targeted therapy or immunotherapy in this disease.

Materials and Methods

Public Data Acquisition and Preprocessing

Gene expression profiles and clinical annotations for LUAD were downloaded from TCGA (access date: 05 January 2025) [30]. RNA-seq expression values (FPKM) were transformed as log₂ (value + 1) and subsequently converted to TPM for downstream analyses. To increase the number of non-tumour controls for expression comparisons, normal lung tissues from GTEx were integrated through GEPIA2, which harmonizes TCGA and GTEx using a unified processing pipeline [31]. Where indicated by the analysis workflow, UALCAN was used for secondary validation of tumour-versus-normal expression and clinicopathological stratifications based on TCGA LUAD clinical metadata. All TCGA data were restricted to primary LUAD samples after removal of duplicates and entries lacking essential clinical variables.

Differential Expression Analysis

Differential expression of DKK1 between LUAD tumour and normal lung tissues was evaluated in R using limma (v3.38.3) and edgeR (v3.32.1). For group contrasts, samples were filtered for adequate library size and low-abundance features were excluded using the edgeR filterByExpr default. Results were reported as log₂ fold change (log₂FC) with Benjamini–Hochberg FDR correction; unless otherwise specified, significance thresholds were set at |log₂FC| ≥ 1 and FDR < 0.05. Visualization (volcano, box/violin) was performed with ggplot2 (v3.4.4).

Clinicopathological Correlation

Associations between DKK1 TPM and clinicopathological variables were tested using Wilcoxon rank-sum tests for two-group comparisons or Kruskal–Wallis tests for ≥ 3 groups. Where multiple categories existed, post-hoc pairwise Wilcoxon tests with FDR correction were performed. Continuous variables were assessed by Spearman’s rank correlation. Effect sizes and adjusted P values were reported. Plots were generated in ggplot2 with compact letter displays for multiple comparisons when applicable.

Diagnostic and Prognostic Evaluation

Diagnostic performance was evaluated with receiver-operating-characteristic (ROC) curves using pROC (v1.18.5). The area under the ROC curve (AUC) and 95% CIs were computed via DeLong’s method; AUC > 0.90 was interpreted as excellent discrimination. Prognostic value was assessed by Kaplan–Meier (KM) survival analysis for overall survival (OS) using survival (v3.5-7) and survminer (v0.4.9). Patients were dichotomized by the median DKK1 expression unless otherwise stated. Hazard ratios (HRs) were estimated with Cox proportional hazards models; proportional hazards assumptions were checked using Schoenfeld residuals. Log-rank P < 0.05 denoted statistical significance.

Co-expression Analysis, Functional Enrichment, and PPI Network Construction

Pairwise Spearman correlations between DKK1 and all protein-coding genes were computed from TCGA-LUAD TPM. Genes were ranked by correlation coefficient (ρ), and the top positively and negatively correlated sets were retained after multiple testing correction (FDR < 0.01). Functional annotation of the top positively correlated genes (default n = 100 unless otherwise stated) was conducted in Metascape for Gene Ontology (GO) biological processes and KEGG pathways, using default background and enrichment settings with BH correction. Protein–protein interaction (PPI) networks were constructed in STRING v11.5 with the “medium confidence” interaction score cutoff (0.400). Networks were visualized in Cytoscape v3.10.0; highly connected subnetworks were extracted using MCODE (degree cutoff = 2, node score cutoff = 0.2, k-core = 2, max depth = 100) [32]. Enrichment for MCODE clusters was computed within Metascape.

Single-cell Transcriptomic Analysis and Functional State Inference

To investigate the expression pattern and potential biological functions of DKK1 at single-cell resolution in LUAD, we employed multiple publicly available databases. First, the TISCH database (Tumor Immune Single-cell Hub, http://tisch.comp-genomics.org/) was queried using the dataset NSCLC_GSE127465, which provides annotated single-cell RNA sequencing data from non-small cell lung cancer [33, 34]. DKK1 expression was examined across diverse cell populations within LUAD tissues, and distribution was visualized at the cluster level. Second, the Human Protein Atlas (HPA, http://www.proteinatlas.org/) database was used to validate the expression of DKK1 across different single-cell lineages of lung tissue (e.g., alveolar cells, fibroblasts, endothelial cells, macrophages), providing additional context regarding its localization and tissue specificity [35]. Finally, the CancerSEA database (http://biocc.hrbmu.edu.cn/CancerSEA/) was utilized to explore the functional state associations of DKK1 at the single-cell level [36]. CancerSEA integrates 41,900 cancer single cells from 25 human cancers with 14 curated functional modules, enabling systematic analysis of single-cell functional states. Spearman correlation analysis was applied to assess the relationship between DKK1 expression and cellular processes such as proliferation, cell cycle, DNA damage response, and invasion/metastasis, with P values adjusted for multiple testing using the false discovery rate (FDR).

Immune Infiltration Estimation and Immune-checkpoint Correlation

Immune infiltration was estimated by single-sample gene set enrichment analysis (ssGSEA) implemented in the GSVA package (v1.48.3), using the 24 immune-cell signature gene sets described by Bindea et al. Enrichment scores were z-scaled across samples for visualization. Associations between DKK1 expression and immune-cell scores were tested by Spearman correlation with BH FDR control. Correlations between DKK1 and major immune-checkpoint genes frequently evaluated in LUAD (e.g., PDCD1, CD274, CTLA4, LAG3, TIGIT, HAVCR2, SIGLEC15) were also computed using Spearman tests. For all correlation analyses, 95% CIs were generated by bootstrap (1,000 resamples) when stated.

Construction of the lncRNA–miRNA–DKK1 Regulatory Axis

Candidate DKK1-targeting miRNAs were predicted using DIANA-microT-CDS (v5.0), miRWalk (v3.0), TargetScan (v8.0), and ENCORI (starBase). Only miRNAs present in at least two databases were retained. Using ENCORI, lncRNAs harbouring validated or predicted binding sites for these miRNAs were retrieved. A competing ceRNA model was then constructed by applying reciprocal correlation filters in TCGA-LUAD: lncRNA–DKK1 (positive; Spearman ρ > 0.25, P < 0.05) and lncRNA–miRNA (negative; ρ < −0.25, P < 0.05), alongside miRNA–DKK1 (negative; ρ < −0.25, P < 0.05). The resulting tripartite lncRNA–miRNA–DKK1 interactions were visualized as Sankey diagrams using SangerBox (v3.0).

Ethics Statement

Analyses involving TCGA/GTEx and other de-identified public resources complied with their respective data-use policies and did not require additional IRB review. No human intervention or patient-identifiable information was collected by the authors for this computational component.

Reagents and Antibodies

Primary antibodies against DKK1 (Abcam, Cambridge, UK; Cat# ab93017; 1:1000 dilution for WB) and GAPDH (Abcam, Cambridge, UK; Cat# ab8245; 1:2000 dilution for WB) were used. Secondary horseradish peroxidase (HRP)-conjugated anti-rabbit and anti-mouse IgG antibodies were obtained from Cell Signaling Technology (Danvers, MA, USA; Cat# 7074 and 7076; 1:5000 dilution). For cell culture, RPMI-1640 medium and fetal bovine serum (FBS) were purchased from Gibco (Thermo Fisher Scientific, Waltham, MA, USA). Penicillin–streptomycin solution (100 U/mL and 100 µg/mL) was from HyClone (Logan, UT, USA). For protein detection, the BCA Protein Assay Kit (Thermo Fisher, Cat# 23227), and RIPA lysis buffer (Beyotime, Shanghai, China; Cat# P0013B) were used.

Cell Culture and Transfection

H1975 lung adenocarcinoma cells, A549 cells, 95-D cells,16HBE cells, and BEAS-2B normal lung epithelial cells were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). The human circulating lung tumor cell line CTC-TJH-01 was previously established and characterized by our laboratory Briefly, CTC-TJH-01 cells were cultured in F12K medium supplemented with 10% foetal bovine serum (Biological Industries, Israel) and penicillin– streptomycin (Gibco Life Technologies, Carlsbad, CA, USA).H1975, 95-D cells,16HBE and BEAS-2B cells were cultured in RPMI-1640 medium (Corning Cellgro, USA) containing 10% FBS and penicillin– streptomycin.A549 cells, were cultured in DMEM (Corning Cellgro, USA) containing 10% FBS and penicillin– streptomycin. All cells were maintained at 37 °C in a humidified atmosphere with 5% CO₂.

In brief, the cells were seeded in ultralow adsorption 24-well plates. After 4 h, the cells were transfected with 50 nM DKK1 (#SIGS 0010021-4) siRNA via a riboFECT CP transfection kit (RiboBio, China) according to the manufacturer’s guidelines. An unrelated, scrambled siRNA was used as a negative control. For the DKK1 plasmid, as we previously reported, 2 µg of the DKK1 plasmid was transfected via a riboFECT CP transfection kit, as instructed by the manufacturer. Downstream experimental analysis was carried out 48 h after transfection.

Cell Activity Detection

In brief, Cells subjected to siRNA-DKK1 or control treatment were assessed for viability using the CCK-8 assay according to the manufacturer’s instructions. Absorbance at 450 nm was measured, and survival curves were plotted using GraphPad Prism.

Migration and Invasion Assays

Migration and invasion assays were performed as previously described. Briefly, 8 × 10⁴ cells with knocking down DKK1 suspended in 500 µl of serum-free medium were seeded into the upper chamber of a Transwell insert, while the lower chamber was filled with 750 µl of medium containing 20% FBS. After a 16-h incubation period, non-migrated cells were gently removed using a cotton swab, and the migrated cells in the Transwell chamber were fixed and stained with Giemsa solution. Subsequently, under an inverted microscope, five random fields were photographed and counted at a magnification of 100×. For the cell invasion assay, Transwell inserts coated with Matrigel were used.

Cell Clone Formation Assay

Clone formation assays were conducted following a previously established protocol. Briefly, 500 cells were seeded into individual wells of a 6-well plate and subsequently treated with knocking down DKK1. The cells were cultured for a period of 12 days under suitable conditions. After the incubation period, colonies were stained with Giemsa and counted.

Cell Apoptosis Analysis

H1975 cells were treated with knocking down DKK1 and incubated for 24 h. Cell apoptosis was assessed using the Annexin V-FITC/PI apoptosis detection kit (BD Biosciences, CA, USA) and analyzed using a FACSVerse flow cytometer (BD Biosciences, CA, USA).

Western Blot Analysis

Western blotting was performed following a previously described method. Briefly, 3 × 10⁵ cells were seeded into individual wells of a 6-well plate and treated with knocking down DKK1. The cells were then lysed and centrifuged at 12,000 rpm for 30 min at 4 °C. The protein content in the lysates was extracted and quantified using the BCA assay. Subsequently, 40 µg of the proteins were loaded onto 5% to 12.5% SDS polyacrylamide gels and subjected to electrophoresis. The separated proteins were then transferred to a polyvinylidene difluoride membrane. The membranes were blocked using skim milk and subsequently incubated with specific primary antibodies, followed by appropriate secondary antibodies. Immunoreactive bands were visualized using an ECL kit (Bio-Rad, CA, USA). The density of the bands was analyzed using ImageJ software.

Statistical Analysis

All statistical analyses were performed using R (v4.3.1) and GraphPad Prism 8.0 (GraphPad Software, San Diego, CA), unless otherwise specified. For in vitro and in vivo experimental data, results are presented as mean ± SD from at least three independent experiments. Normality of data distribution was tested using the Shapiro–Wilk test; for normally distributed variables, two-tailed Student’s t-tests (two groups) or one-way ANOVA with Tukey’s post hoc test (≥ 3 groups) were applied. For non-normally distributed or ordinal data, the Wilcoxon rank-sum test (two groups) or Kruskal–Wallis test with Dunn’s post hoc correction (≥ 3 groups) was used. Correlation analyses were performed using Pearson or Spearman methods as appropriate. Multiple testing was adjusted by the Benjamini–Hochberg FDR method. Unless otherwise indicated, P < 0.05 (two-tailed) was considered statistically significant. Data visualization was conducted using ggplot2 in R, while experimental plots were prepared in GraphPad Prism 8.0. Final figure assembly was performed in Adobe Illustrator 2024. To ensure reproducibility, all analyses were executed with fixed random seeds, and key parameters (software versions, thresholds, and database access dates) are reported in the Methods.

Result

DKK1 is Up-regulated and Correlates with Clinical Features and Prognosis in LUAD

Pan-cancer screening (TCGA) showed broad up-regulation of DKK1, including in LUAD (Fig. S1A). In LUAD specifically, DKK1 mRNA was significantly higher in tumor tissues (T) than in normal lung tissues (N) (Fig. 1A).TP53-mutant LUAD showed slightly higher DKK1 expression than non-mutant cases (Fig. S1B). Receiver-operating-characteristic analysis supported diagnostic utility (Fig. 1B). KM analysis indicated poorer OS in patients with high DKK1 expression Fig. 1C). Clinically, higher DKK1 was associated with advanced pathological stage, nodal status, and tumor (T) classification (Fig. 1D–F). Notably, the adverse prognostic impact persisted in early-stage subsets, including Stage I, N0, and T1 cohorts (Fig. 1G–I). Together, these data highlight DKK1 as a candidate diagnostic and prognostic biomarker in LUAD.

Fig. 1.

DKK1 is up-regulated in LUAD and associates with clinicopathological features and prognosis. A Boxplot showing higher DKK1 expression in LUAD tumor tissues (T, n = 483) compared with normal lung tissues (N, n = 347). B ROC curve assessing diagnostic performance (AUC = 0.816; 95% CI: 0.780–0.853). C KM overall-survival curve comparing high versus low DKK1 expression (HR = 2.3; log-rank p = 4.9 × 10⁻⁸). D–F Violin plots showing higher DKK1 expression with advanced pathological stage (D: Stage I–II vs. III–IV), nodal stage (E: N0–N1 vs. N2–N3), and tumor stage (F: T1–T2 vs. T3–T4). G–B KM curves demonstrating worse OS with high DKK1 in early-stage subgroups: Stage I (G), N0 (H), and T1 (I).*Significance: P < 0.05, *P < 0.01, *P < 0.001

The Co-expression Pattern of DKK1 in LUAD

To elucidate the molecular interactions and potential biological functions associated with DKK1 in LUAD, we identified co-expressed genes using TCGA datasets and constructed a volcano plot to visualize significantly correlated genes (Fig. 2A). Heatmaps were generated to highlight the most significant co-expressed genes. The top 20 positively correlated genes included CREG2, SH2D5, TNS4, ITGB1-DT, LAMC2, CAMK2N1, ITGA6, NTSR1, and CASC8 (Fig. 2B). Conversely, the top 20 genes negatively correlated with DKK1 expression comprised NKX2-1, SFTA3, IRX5, CRNDE, MBIP, ZDHHC21, CADM1, MTURN, CNGA3, and UMODL1 (Fig. 2C). These genes may interact with DKK1 and provide a foundation for further investigations into the molecular pathways associated with DKK1 in LUAD.

Fig. 2.

Results of DKK1 co-expression analysis in LUAD. A Volcano map illustrating genes significantly correlated with DKK1 expression. Red dots represent positively correlated genes, and blue dots indicate negatively correlated genes. B Heatmap displaying the top 20 genes positively correlated with DKK1 expression. C Heatmap illustrating the top 20 genes negatively correlated with DKK1 expression

Functional Enrichment and PPI of DKK1 Co-expressed Genes in LUAD

To better understand the potential molecular mechanisms mediated by DKK1 in LUAD, we performed functional pathway enrichment analysis for the top 100 genes co-expressed with DKK1 using the Metascape tool. The co-expressed genes primarily clustered into pathways associated with cornified envelope formation, tissue morphogenesis, type I hemidesmosome assembly, pancreatic cancer subtypes, supramolecular fiber organization, and cell-extracellular matrix interactions (Fig. 3A, B). Furthermore, GO enrichment and KEGG analyses indicated that these genes significantly participated in critical biological pathways, including the PID P53 downstream pathway, ERK1 and ERK2 signaling cascades, gap junction assembly, and glucocorticoid receptor pathways (Fig. 3C).

Fig. 3.

Pathway enrichment and PPI analysis of DKK1 co-expressed genes in LUAD. A-B Network clustering analysis illustrating pathways enriched by DKK1 co-expressed genes. Nodes represent pathways, and connections represent shared genes between pathways. C Bar chart showing significantly enriched pathways based on GO and KEGG analyses of DKK1 co-expressed genes. D-E PPI network of DKK1 and its co-expressed genes. F-G Key modules identified by MCODE analysis from the PPI network, indicating highly interconnected regions and related pathways

Additionally, we constructed a PPI network to illustrate interactions between DKK1 and its co-expressed genes. Analysis of the network highlighted the involvement of DKK1-related genes in pathways crucial to tumor progression, such as cellular adhesion, matrix remodeling, and morphogenesis (Fig. 3D, E). Moreover, MCODE analysis identified key functional modules within the network. The most significant modules were enriched in laminin interactions, type I hemidesmosome assembly, gap junction trafficking, and integrin signaling pathways (Fig. 3F, G). Taken together, these findings indicate that DKK1 and its co-expressed genes are involved in diverse biological processes central to cancer progression, suggesting that DKK1 might promote LUAD development and metastasis through regulation of cellular adhesion, extracellular matrix interactions, and signaling pathways associated with cellular morphogenesis and tumorigenesis.

Analysis of DKK1 Expression in LUAD Tissues and at the Single-cell Level

To further validate and contextualize DKK1 expression, we examined both immunohistochemistry (IHC) staining and single-cell transcriptomic data. IHC analysis showed weak or absent DKK1 staining in normal lung tissues, whereas adenocarcinoma tissues displayed markedly stronger cytoplasmic expression, confirming DKK1 overexpression at the protein level (Fig. S2A–B). Complementary analyses using the TISCH single-cell RNA sequencing dataset (NSCLC_GSE127465) revealed that DKK1 expression was enriched in malignant epithelial clusters and alveolar epithelial subtypes (AT1/AT2), with additional signals detected in fibroblasts (Fig. S2C–D). Functional correlation analysis via CancerSEA suggested modest but significant negative associations between DKK1 expression and cell-cycle activity as well as DNA damage responses, indicating functional heterogeneity across distinct tumor cell states (Fig. S2E). Although not the main focus of this study, these findings provide additional evidence that DKK1 is differentially expressed in LUAD tissues and may influence cellular behavior through cell-state–specific mechanisms.

The Correlation between DKK1 Expression and Immune Infiltration in LUAD

To investigate the potential immunomodulatory role of DKK1 in LUAD, we first analyzed the relationship between DKK1 expression and the infiltration levels of 24 immune cell types using Spearman correlation analysis. The results demonstrated that DKK1 expression was significantly positively correlated with the infiltration of mast cells, neutrophils, macrophages, iDC, DC, eosinophils, and pDC (Fig. 4A–F). These findings suggest that DKK1 is involved in shaping the tumor immune microenvironment, especially through interactions with innate immune cells.

Fig. 4.

DKK1 expression is associated with immune cell infiltration in LUAD. A Spearman correlation analysis between DKK1 expression and the infiltration of 24 immune cell types; circle size denotes correlation magnitude and color encodes P value. B Scatter plot showing the positive correlation between DKK1 expression and mast cell infiltration. C Scatter plot showing the positive correlation between DKK1 expression and neutrophil infiltration. D Scatter plot showing the positive correlation between DKK1 expression and macrophage infiltration. E Scatter plot showing the positive correlation between DKK1 expression and immature dendritic cell (iDC) infiltration. F Scatter plot showing the positive correlation between DKK1 expression and dendritic cell (DC) infiltration. *P<0.05, **P<0.01, ***p<0.001

Furthermore, we explored the association between DKK1 expression and immune checkpoint molecules. As illustrated in the heatmap, the expression of DKK1 was positively correlated with several immune checkpoint genes, including CD274 (PD-L1), HAVCR2, and SIGLEC15 (Fig. 5A-D). These results imply that DKK1 may participate in immune evasion by regulating checkpoint pathways in LUAD. Collectively, these data highlight that DKK1 plays a critical role in modulating immune cell infiltration and may serve as a promising immunotherapeutic target in LUAD.

Fig. 5.

DKK1 expression is associated with immune checkpoint gene expression in LUAD. A Heatmap showing correlations between DKK1 and eight immune checkpoint genes. B Scatter plot of DKK1 expression versus CD274 (PD-L1). C Scatter plot of DKK1 expression versus HAVCR2 (TIM-3). D Scatter plot of DKK1 expression versus SIGLEC15. *P<0.05, **P<0.01, ***p<0.001

Construction of the lncRNA–miRNA–DKK1 Regulatory Axis

To delineate post-transcriptional regulators of DKK1, we queried four prediction engines (DIANA-microT, miRWalk v3, TargetScan v8 and ENCORI). After de-duplication, 38 miRNAs were recognised by ≥ 2 tools, six appeared in ≥ 3 lists and one (hsa-miR-302a-3p) was unanimously detected by all four (Fig. S3A). Correlation analysis across TCGA-LUAD samples demonstrated a robust positive correlation between miR-302a-3p and DKK1 mRNA abundance (Fig. S3B) and a mild, but significant, relationship for miR-34b-5p (Fig. S3C). Consistently, both miRNAs were expressed at higher levels in tumour tissues than in non-malignant counterparts (Fig. S3D–E). These observations nominate miR-302a-3p and miR-34b-5p as key components of the DKK1 post-transcriptional network.ENCORI cross-linking immunoprecipitation data were next interrogated to uncover lncRNAs predicted to bind the two miRNAs. Seven lncRNAs—FGD5-AS1, GAS5, ZNF667-AS1, OIP5-AS1, AC006511.5, SNHG3 and KCNQ1OT1—showed significant binding evidence and satisfied reciprocal expression criteria (Fig. S3F). Expression profiling confirmed that FGD5-AS1, ZNF667-AS1 and GAS5 were markedly up-regulated in lung-cancer samples (Fig. S3G–I), supporting a potential sponge function. Collectively, these findings support a model in which over-expressed lncRNAs (FGD5-AS1/ZNF667-AS1/GAS5) may titrate miR-302a-3p and/or miR-34b-5p away from DKK1 transcripts, culminating in elevated DKK1 protein levels and potentiation of Wnt signalling. The outlined axis offers new mechanistic insight and suggests multiple therapeutic entry points (e.g., miRNA mimics or lncRNA knock-down) to attenuate DKK1-driven oncogenesis.

Knocking down DKK1 Inhibits Lung Cancer Cell Malignant Phenotypes

Given the strong clinical association between high DKK1 expression and poor prognosis as well as immune evasion in LUAD, we next performed functional experiments to directly examine its biological role in lung cancer cells. We selected the H1975 cell line, which exhibited high endogenous DKK1 expression, for loss-of-function studies using siRNA-mediated knockdown. Western blot analysis confirmed that DKK1 was highly expressed in H1975 and other lung cancer cell lines compared to normal lung epithelial cells (Fig. 6A). Efficient knockdown of DKK1 in H1975 cells was validated by Western blot following siRNA transfection (Fig. 6B). Functionally, CCK-8 assays showed that DKK1 silencing significantly reduced cell proliferation relative to the control group (Fig. 6C). Cell cycle analysis by flow cytometry revealed that DKK1 knockdown caused S/G₂ transition delay, with an increase in the proportion of cells in S phase, a modest rise in G₂/M phase, and a corresponding decrease in G₀/G₁ phase cells (Fig. 6D–E). Furthermore, Annexin V/PI staining demonstrated a significant elevation of apoptotic cells following DKK1 knockdown (Fig. 6F–G). In terms of cellular motility, wound-healing assays indicated a slight reduction in migration capacity after DKK1 silencing, although this differences did not reach statistical significance (Fig. 6H). In contrast, Transwell migration and invasion assays clearly showed that DKK1 depletion, particularly with siDKK1-3, markedly decreased both migratory and invasive abilities of H1975 cells (Fig. 6I–K).

Fig. 6.

Functional validation of DKK1 knockdown in LUAD cells. A Western blot analysis of DKK1 expression across multiple lung cancer (CTC-TJH-01, H1975, A549, 95-D) and normal lung epithelial (16HBE, BEAS-2B) cell lines. B Validation of DKK1 knockdown efficiency in H1975 cells by western blotting. C CCK-8 assays revealed reduced viability in H1975 cells following DKK1 knockdown. D–E PI staining and flow cytometry showed that DKK1 silencing induced S-phase accumulation together with an increase in G2/M and a reduction in G0/G1. F–G Annexin V-FITC/PI double staining confirmed that DKK1 knockdown significantly increased apoptosis. H Wound-healing assays showed a modest decrease in migration after DKK1 knockdown. Scale bar, 250 μm. I–K Transwell migration and invasion assays demonstrated that DKK1 knockdown significantly suppressed the invasive behavior of H1975 cells. L Western blot detection of apoptosis-related proteins. DKK1 knockdown increased cleaved PARP (89 kDa), BAX (21 kDa) and Cleaved Caspase-3 (17/19 kDa). M Western blot detection of cell-cycle-related proteins showing reduced CDK1 (34 kDa) and Cyclin B1 (55 kDa) after DKK1 silencing. GAPDH served as loading control.Scale bar, 250 μm. Data are presented as the mean ± SD of three independent experiments. Statistical significance: *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant

To explore the molecular mechanisms underlying these phenotypes, we examined key apoptosis- and cell cycle-related proteins. DKK1 knockdown increased the expression of apoptotic markers, including PARP cleavage, BAX, and Cleaved Caspase-3 (Fig. 6L), while decreasing levels of cell cycle regulators CDK1 and Cyclin B1 (Fig. 6M). These results corroborate the flow cytometry findings of enhanced apoptosis and S–G2/M arrest. Collectively, these data provide functional evidence that DKK1 contributes to the malignant behavior of LUAD cells by promoting proliferation, migration, and invasion, while concurrently inhibiting apoptosis. Silencing DKK1 effectively reverses these oncogenic traits, supporting its potential as a therapeutic target in LUAD.

Discussion

In this study, we showed that DKK1 is markedly overexpressed in LUAD and associated with advanced stage, lymph node metastasis, and poor survival, supporting its role as a marker of aggressive disease. These results align with earlier studies identifying DKK1 as a serological and prognostic biomarker in lung cancer [37, 38]. Importantly, we demonstrate that its prognostic value extends to early-stage LUAD, suggesting utility for risk stratification—such as identifying Stage I patients who might benefit from adjuvant therapy or closer surveillance. This addresses a major challenge in LUAD management: distinguishing low-risk patients from those with hidden molecular drivers of recurrence.

The overexpression of DKK1 in LUAD raises questions about its functional role. Although classically described as a Wnt inhibitor, our results and prior studies suggest that DKK1 can act as an oncogenic driver [39]. Although classically a Wnt inhibitor, our results and prior studies suggest that DKK1 can act as an oncogenic driver. One explanation is context: by antagonizing Wnt signaling in immune or stromal compartments, DKK1 may indirectly promote tumor growth and invasion. In NSCLC, DKK1 mediates cancer–fibroblast crosstalk, and blocking DKK1 in co-culture partly reverses EGFR-inhibitor resistance [23]. Similar findings highlight its pro-tumorigenic role within the microenvironment [40]. Our co-expression and enrichment analyses further indicate that DKK1-high tumors are enriched for extracellular matrix remodeling, adhesion, and invasion programs, with integrins and laminins among top correlated genes. Conversely, negatively correlated markers such as NKX2-1 and surfactant proteins point to reduced epithelial differentiation [41]. Together, these findings support a model in which DKK1 drives a shift toward a de-differentiated, pro-metastatic phenotype that contributes to LUAD progression.

Beyond bioinformatics, our functional assays provided direct evidence that DKK1 sustains malignant behaviors in LUAD. Knockdown in high-expressing H1975 cells suppressed proliferation, migration, and invasion, while inducing cell-cycle arrest and apoptosis. These findings mirror our transcriptomic analyses, confirming that DKK1 not only associates with aggressive molecular signatures but also functionally drives tumor progression, reinforcing its value as both a prognostic biomarker and a therapeutic target.

A key novelty of our study is linking DKK1 to immune evasion in LUAD. High DKK1 expression correlated with enrichment of innate immune cells—neutrophils, macrophages, dendritic cells, and mast cells—often linked to tumor-promoting inflammation, while T-cell signatures were reduced. This mirrors prior findings that DKK1 associates with MDSCs and reduced CD8⁺ T-cell infiltration [42]. We also observed modest positive correlations with PD-L1, consistent with reports in hepatocellular carcinoma where DKK1 upregulated PD-L1 [21]. Together, these results suggest a dual immunosuppressive role for DKK1: recruiting suppressive cells and enhancing checkpoint signaling. Clinically, this may explain why patients with DKK1-high tumors respond poorly to PD-1/PD-L1 inhibitors, as supported by immunogenomic models incorporating DKK1 as a risk factor. Importantly, neutralizing antibodies such as DKN-01 are in clinical testing. Preclinical studies show that DKK1 blockade alleviates immune suppression, and in a Phase 1/2 mCRPC trial, DKN-01 plus docetaxel yielded partial responses in most patients, whereas monotherapy showed only modest activity [43]. These findings highlight DKK1 as both a predictor of immunotherapy resistance and a promising immunomodulatory target.

Single-cell RNA sequencing further clarified the cellular sources of DKK1 in LUAD. Malignant epithelial cells, especially AT1/AT2-like populations, were the main producers, with lower expression in cancer-associated fibroblasts. The enrichment of DKK1 in AT1-like subpopulations suggests that specific tumor cells may act as “leader cells,” interacting with stroma and immune components to establish metastatic niches. Functionally, CancerSEA analysis showed that DKK1-high cells had lower proliferation and DNA-repair signatures, which, although counterintuitive, fits the concept that metastatic or therapy-resistant cells often cycle slowly but possess stress tolerance and enhanced migratory capacity (e.g., partial EMT). Such quasi-dormant cells may evade chemotherapy and immune attack, later seeding recurrence. These findings indicate that DKK1 marks context-dependent states favoring invasiveness and immune evasion rather than direct proliferation. Validation through co-culture systems—combining DKK1-high LUAD cells or fibroblasts with immune cells—will be essential to confirm causality and further dissect how DKK1 shapes the tumor ecosystem.

We also observed a correlation between DKK1 expression and drug sensitivity. CellMiner analysis showed that LUAD cells with high DKK1 were more sensitive to inhibitors of MAPK/ERK, RAF, and PI3K/AKT signaling (e.g., SB-203580, ZM-336372) as well as to HIF-1α (IDF-11774) and MMP (RG-6016) inhibitors. Although counterintuitive for a pro-tumor factor, this likely reflects molecular dependencies: by inhibiting canonical Wnt/β-catenin, DKK1 may force reliance on compensatory pathways, creating vulnerabilities exploitable by targeted agents. Clinically, this suggests DKK1 could serve as a biomarker to guide therapy, with DKK1-high tumors potentially benefiting from combinations that include MAPK or HIF-1α inhibitors [44]. Moreover, blocking DKK1 itself can disrupt stromal crosstalk and re-sensitize tumors to EGFR inhibitors [45]. These results highlight DKK1 as both a predictor of therapeutic vulnerabilities and a candidate for combination strategies aimed at overcoming resistance.

Beyond coding genes, we also proposed a ceRNA network regulating DKK1 in LUAD. Two microRNAs, miR-302a-3p and miR-34b-5p, emerged as candidate regulators. Unexpectedly, both showed positive rather than inverse correlations with DKK1, suggesting indirect regulation. We therefore hypothesized that long non-coding RNAs (lncRNAs) act as sponges for these miRNAs. Integrative analysis identified several upregulated lncRNAs with predicted binding sites, including GAS5, KCNQ1OT1, FGD5-AS1, and ZNF667-AS1, which may contribute to DKK1 overexpression in LUAD.

Several of these lncRNAs have recognized roles in cancer. For instance, GAS5 is usually described as a tumor suppressor, often downregulated and linked to apoptosis and proliferation [46], whereas KCNQ1OT1 has been reported to modulate Wnt/β-catenin signaling [47]. Interestingly, in LUAD these lncRNAs were aberrantly overexpressed, suggesting a context-specific function. We propose that they may act as ceRNAs, sequestering miR-302a-3p and miR-34b-5p, thereby indirectly upregulating DKK1. This ceRNA circuitry offers a potential explanation for DKK1 overexpression independent of genetic alterations. Therapeutically, targeting RNAs rather than the protein itself could provide new entry points—for example, restoring miRNA activity with mimics or suppressing lncRNAs with antisense oligonucleotides. Such interventions could ultimately lower DKK1 levels and reduce its tumor-promoting effects. We acknowledge that this model is based on predictions and correlations; rigorous validation with assays such as luciferase reporters or RNA pulldown is required. Nevertheless, this network highlights the complexity of post-transcriptional regulation and resonates with systems biology approaches to reveal novel therapeutic targets.

Despite its breadth, our study has limitations. First, most analyses are in silico, and the causal role of DKK1 requires further mechanistic validation. Although published studies and our own functional assays show that DKK1 knockdown reduces proliferation, migration, and invasion while promoting apoptosis, additional experiments—such as assessing downstream Wnt/β-catenin signaling, evaluating EMT marker expression, or examining immune interactions in co-culture models—are needed to clarify the underlying pathways. Second, the proposed ceRNA regulatory axis remains speculative, as it was inferred from sequence-based predictions and expression correlations. Additional microRNAs or regulatory layers—including epigenetic modifications, TP53 status, or Wnt/β-catenin feedback—may also influence DKK1 expression. Our data hint at potential interactions with TP53 alterations, consistent with the known induction of DKK1 within Wnt signaling feedback loops. Overall, while our findings provide a multifaceted characterization of DKK1 in LUAD, they warrant confirmation through focused experiments and validation in independent clinical cohorts.

Conclusions

In conclusion, we provide a comprehensive profile of DKK1 in LUAD, integrating clinical outcomes, tumor biology, immune context, and regulatory mechanisms. We confirm DKK1 as a poor-prognosis biomarker linked to an immunosuppressive microenvironment and uncover novel regulatory layers involving specific miRNAs and lncRNAs. These insights clarify how a Wnt pathway modulator can paradoxically promote tumor progression through de-differentiation, immune evasion, and metastatic niche formation. Importantly, they highlight actionable opportunities: DKK1 may serve as both a therapeutic target and a biomarker to guide treatment. Future directions include evaluating whether DKK1 levels predict immunotherapy response, and testing whether targeting the DKK1 axis—via genetic knockdown, antibodies such as DKN-01, or RNA-based interventions—can enhance anti-tumor immunity. Success in these areas would establish DKK1 as both a prognostic marker and a contextual therapeutic target, exemplifying the challenges and opportunities in modern cancer therapy and immunotherapy.

Abbreviations

LUAD

Lung adenocarcinoma

NSCLC

Non-small cell lung cancer

DKK1

Dickkopf-1

TME

Tumor microenvironment

TCGA

The Cancer Genome Atlas

GTEx

Genotype-Tissue Expression project

TPM

Transcripts per million

FPKM

Fragments per kilobase of transcript per million mapped reads

FDR

False discovery rate

ROC

Receiver operating characteristic

AUC

Area under the ROC curve

OS

Overall survival

HR

Hazard ratio

MDSC

Myeloid-derived suppressor cell

ssGSEA

Single-sample gene set enrichment analysis

PD-1

Programmed cell death protein 1 (PDCD1 gene)

PD-L1

Programmed death-ligand 1 (CD274 gene)

CTLA4

Cytotoxic T-lymphocyte associated protein 4

TIM-3

T-cell immunoglobulin and mucin-domain containing-3 (HAVCR2 gene)

NK cells

Natural killer cells

PPI

Protein–protein interaction

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

UMAP

Uniform manifold approximation and projection (dimension reduction)

AT1/AT2

Alveolar type 1 / type 2 (cells)

EMT

Epithelial–mesenchymal transition

IC₅₀

Half maximal inhibitory concentration

ceRNA

Competing endogenous RNA

miRNA

microRNA

lncRNA

long non-coding RNA

Authors’ Contributions

Y. Wang, responsible for data analysis, manuscript writing, and experimental validation; J. Li, responsible for data analysis, figure preparation, and manuscript drafting; Y. Wang responsible for bioinformatics analysis and manuscript writing; J. Tian, responsible for study design, supervision, and funding acquisition as corresponding author; Z. Liu and Y. Wang, responsible for data collection and collation; J. Liu and Z. Que, responsible for clinical data support and sample resources; Z. Que, responsible for project supervision and funding acquisition as corresponding author; W. Shangguan, responsible for data interpretation, manuscript revision, and funding acquisition as corresponding author.

Funding

This study was supported by Shanghai Frontier Research Base of Disease and Syndrome Biology of Inflammatory Cancer Transformation (2021KJ03-12), National TCM Advantageous Specialty Construction Project of the National Administration of Traditional Chinese Medicine (Oncology Department-2024-510), Clinical Evaluation of the Tiaoshen Anti-Cancer Regimen in Treating Psycho-Neurological Symptom Cluster in Ovarian Cancer [the National Health Commission of the People’s Republic of China (formerly the Health and Family Planning Commission) of China (No. 2024ZD0521402)], and the Key Specialty Development Project of Baoshan Branch, Renji Hospital Affiliated to Shanghai Jiao Tong University School of Medicine (rbzdzk-2023-002).

Data Availability

All data analyzed in this study were obtained from public databases. The TCGA-LUAD gene expression and clinical data are available from the Genomic Data Commons (https://portal.gdc.cancer.gov/). GTEx normal lung data can be accessed via the GTEx Portal (https://gtexportal.org/). Single-cell RNA-seq data (GSE127465) were obtained from the TISCH database. Further details on data and custom analysis code are available from the corresponding author on reasonable request.

Declarations

Ethics Approval and Consent to Participate

This study primarily used publicly available, de-identified datasets (TCGA, GTEx, etc.), and therefore no additional ethics approval or patient consent was required. The lung nodule and lung cancer specimens used in this work were provided by Shanghai Pulmonary Hospital and collected in accordance with institutional ethical guidelines.

Consent for Publication

Not applicable. This manuscript does not contain individual person’s data in any form.

Competing Interests

The authors declare no competing interests.