Loss of the RAGE function in hypoxic conditions exacerbates the malignant progression of lung adenocarcinoma via damage-associated molecular pattern signaling

Highlights

• •RAGE levels were reduced and linked to poor survival in lung adenocarcinoma.
• •Tumors showed low oxygen and increased hypoxia markers.
• •Blocking RAGE slowed cell growth and migration in vitro.
• •RAGE inhibition accelerated tumor growth in vivo.
• •Tumors activated alternative pathways to sustain progression.

Abstract

Despite therapeutic advances, lung adenocarcinoma (LUAD) remains difficult to treat, and the mechanisms by which hypoxia promotes tumor progression are still incompletely understood. Because the receptor for advanced glycation end products (RAGE) has been implicated in hypoxia-related signaling, we investigated the interaction between damage-associated molecular patterns (DAMPs) and RAGE in LUAD under hypoxic conditions using the RAGE inhibitor FPS-ZM1 in vitro and in vivo. RAGE expression was significantly reduced in LUAD samples and was associated with poor survival. LUAD tissues also showed evidence of hypoxia, including elevated hypoxia-inducible factor-1α (HIF-1α). Under hypoxia, S100A8/A9 redistributed from the nucleus to the cytoplasm, whereas intracellular HMGB1 levels decreased. In vitro, FPS-ZM1 suppressed NF-κB phosphorylation and inhibited LLC cell viability and migration. In contrast, in vivo FPS-ZM1 treatment was associated with enhanced tumor progression, increased HIF-1α and S100A8/A9 expression, and reactivation of NF-κB signaling. These findings suggest that RAGE blockade may trigger context-dependent compensatory responses within the tumor microenvironment. We therefore present a working model in which a hypoxia-associated HIF-1α/S100A8/A9 axis may bypass RAGE inhibition and restore pro-tumor signaling; however, this mechanism requires direct experimental validation. Overall, our data highlight the importance of tumor-microenvironment context when interpreting RAGE-targeted interventions in LUAD.

Introduction

Lung cancer remains the leading cause of cancer-related death worldwide. Although mortality has gradually declined, the absolute number of deaths remains higher than that for any other cancer type [1]. Lung cancer is broadly classified into small cell lung cancer (10–15%) and non-small cell lung cancer (NSCLC; 80–85%) [2,3]. Lung adenocarcinoma (LUAD), the major histological subtype of NSCLC, accounts for approximately 40% of all lung cancer cases [4]. Current treatment strategies for LUAD include surgery, chemotherapy, radiotherapy, immunotherapy, and supportive care. However, recurrence, treatment resistance, and late-stage diagnosis continue to limit survival gains [5]. A better understanding of LUAD biology is therefore needed to identify more effective therapeutic targets.

Tumor growth and metastasis are strongly shaped by the tumor microenvironment (TME). Among TME features, hypoxia is particularly important because it regulates gene expression through hypoxia-inducible factors (HIFs) and promotes angiogenesis, metabolic reprogramming, immune evasion, treatment resistance, and stem-like phenotypes [6,7]. Hypoxia can also induce cell damage, membrane disruption, and metabolic stress, thereby promoting the release of damage-associated molecular patterns (DAMPs), including high-mobility group box 1 (HMGB1) and S100A8/A9. These molecules act as alarm signals and can stimulate tumor-promoting inflammation. HMGB1 has been linked to NSCLC progression and may recruit inflammatory cells into tumor tissue [8]. S100A8 and S100A9 are likewise overexpressed in several malignancies and are associated with tumorigenesis and metastasis [9]. A major receptor for these ligands is the receptor for advanced glycation end products (RAGE), a multi-ligand pattern-recognition receptor that activates diverse downstream pathways [10]. RAGE signaling has been implicated in cellular adaptation to hypoxia [11]. However, unlike many other cancers, normal lung tissue expresses abundant RAGE, whereas NSCLC often shows RAGE downregulation during malignant transformation [12,13]. At the same time, RAGE ligands such as HMGB1 and S100 proteins remain enriched in lung tumors and may signal through RAGE-dependent and RAGE-independent pathways [14]. These observations suggest that the role of RAGE in lung cancer is context dependent and incompletely understood. Accordingly, we investigated the interaction between DAMPs and RAGE in LUAD under hypoxia as an exploratory study designed to generate mechanistic hypotheses for future validation.

Methods

Human studies

Clinical specimens, including LUAD tissues (n = 8), corresponding paracancerous tissues (n = 4), and peripheral blood samples from patients with LUAD (n = 19) and healthy controls (n = 19), were collected from Qinghai Provincial People’s Hospital. The study protocol was approved by the Ethics Committee of the Clinical Medical College of Qinghai University (approval number: P-SL-2024–010), and all procedures were performed in accordance with the principles of the Declaration of Helsinki. All individuals involved in the study were apprised of its details, and written informed consent was procured from each participant before sample collection.

Animals

Male wild-type (C57BL/6) mice aged 8 weeks were purchased from Jiangsu Huachuang Xinuo Pharmaceutical Co., Ltd. (Suzhou, China). All animals were maintained under specific pathogen-free conditions at the Research Center for High Altitude Medicine, which is licensed for laboratory animal use (SYXK [QING] 2020–0001). The lung cancer animal model was established via tail vein injection of Lewis lung carcinoma (LLC) cells (1 × 10⁶ cells per mouse; n = 10 per group). Mice inoculated with LLC cells were randomly divided into three groups. One week after inoculation, FPS-ZM1 (1 mg/kg, twice weekly) [15] was administered via intraperitoneal injection. Four weeks later, the mice were weighed and placed in a sealed induction chamber. According to animal ethics regulations, a small animal gas anesthesia system (RWD Life Science, Shenzhen, China) was used to deliver 3% isoflurane in oxygen (flow rate: 2–3 L/min) for induction anesthesia. Anesthesia depth was confirmed by loss of the corneal reflex and the absence of a response to a hindlimb pinch.

Under deep anesthesia, blood was collected by cardiac puncture, followed by euthanasia. Whole blood was transferred to EDTA-coated collection tubes to prevent coagulation. Finally, lung tissues were harvested. Tumor nodules on the lung surface were assessed macroscopically, weighed, photographed, and subjected to further analyses. All animal procedures received formal approval (approval number: P-SL-2024–010) from the Ethics Committee of the Clinical Medical College of Qinghai University, operating in full compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (8th ed.). The experimental design, execution, and reporting strictly adhered to the ARRIVE 2.0 guidelines to guarantee methodological robustness and ethical integrity.

Tumor cell line

The mouse Lewis lung carcinoma (LLC) cell line (Pricella Biotechnology, Wuhan, China) was cultured in high-glucose Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin. Cells were maintained at 37 °C in a humidified incubator containing 5% CO₂ under normoxic (21% O₂) or hypoxic (1% O₂) conditions. LLC cells were exposed to 1% O₂ for 12, 24, or 48 h; these time points were selected on the basis of a preliminary time-course screen ranging from 6 to 72 h. For RAGE inhibition under hypoxia, cells were treated with 5 or 10 µM FPS-ZM1 (MedChemExpress, HY-19,370), and the 0 µM group received an equivalent volume of dimethyl sulfoxide (DMSO) as vehicle control. To modulate HIF-1α, cells were treated with 2 mM dimethyloxalylglycine (DMOG; MedChemExpress, HY-15,893) under normoxia for 48 h or with 20 µM LW-6 (MedChemExpress, HY-13,671) under hypoxia for 24 h. The corresponding 0 µM/0 mM groups received DMSO vehicle only.

Cell viability assays

Cells in logarithmic growth phase were harvested, counted, and seeded into 96-well plates at 1.2 × 10^4 cells per well in 100 µL medium. After 24 h of pre-incubation (37 °C, 21% O₂, 5% CO₂), cells were treated with FPS-ZM1 at 1, 5, 10, or 50 µM. These concentrations were selected after preliminary dose-response screening over a broader range. After 24 h, 100 µL fresh medium and 10 µL CCK-8 reagent were added to each well. Plates were incubated for 1–4 h, and absorbance at 450 nm was measured hourly to determine the optimal readout time. Each experiment was performed in triplicate.

$$\text{Cell}\text{viability}(\%)=\left[\mathrm{A}\left(\text{with}\text{drug}-\text{applied}\right)--\mathrm{A}\left(\text{blank}\right)\right]/\left[\mathrm{A}\left(0-\text{drug}\right)--\mathrm{A}\left(\text{blank}\right)\right]\times 100$$

A(drug-applied) denotes absorbance in wells containing cells, CCK-8 solution, and drug; A(blank) denotes absorbance in wells containing complete medium and CCK-8 solution without cells; and A(0-drug) denotes absorbance in wells containing cells and CCK-8 solution without drug (vehicle control).

Transwell cell migration in vitro

For the in vitro migration assays, 5 × 10⁴ cells suspended in serum-free medium were seeded into 8.0-µm pore-sized hanging cell culture plates (Corning Inc., Corning, NY, USA). The lower chamber contained medium supplemented with 10% FBS that acted as a chemical attractant. Following 24 h incubation, the culture medium was aspirated, and the membrane was subjected to sequential processing: phosphate-buffered saline (PBS) washing, methanol fixation, and Giemsa staining. Migrated cells were counted in three randomly selected microscopic fields(Research-grade fluorescence inverted microscope, IX71). Each experiment was performed in triplicate, and the results are presented as the mean ± standard deviation.

Western blotting

Cell lysates and murine lung tissue lysates were prepared on ice in RIPA buffer supplemented with protease and phosphatase inhibitors (150 mM NaCl, 50 mM Tris–HCl, pH 8.0, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% SDS). After centrifugation (12,000 × g, 10 min, 4 °C), supernatants were quantified by BCA assay. Equal amounts of protein were separated by SDS-PAGE and transferred to PVDF membranes. Membranes were blocked with 5% skim milk for 1 h at room temperature and incubated with primary antibodies overnight at 4 °C, followed by secondary antibodies for 1 h at room temperature. Signals were visualized using ECL substrate (Amersham Imager 800) and quantified with ImageJ 1.53e. Target-protein signals were normalized to the corresponding β-actin band from the same lane. Quantification was based on at least three independent biological replicates. The antibodies used were β-actin (Proteintech, 60,008-1-Ig), HMGB1 (Abcam, ab79823), HIF-1α (Cell Signaling, 14179S), S100A8/A9 (Abcam, ab288715), RAGE (Abcam, ab216329), p65 (Abcam, ab16502), phospho-p65 (ABclonal, TP56372), BCL-2 (Abcam, ab182858), p21 (Abcam, ab109199), VEGF (Abcam, ab46154), and CA9 (Proteintech, 66,243-1-Ig). Uncropped membrane images are provided in the Supplementary Materials.

Immunohistochemistry and immunofluorescent staining

Paraffin-embedded tissue sections containing normal lung and LUAD tissue were deparaffinized, rehydrated, and subjected to antigen retrieval, followed by overnight incubation with primary antibodies at 4 °C. After PBS washing, sections were incubated with polymer enhancer for 20 min at 37 °C and then with peroxidase-labeled mouse/rabbit IgG for 30 min at 37 °C. Nuclei were counterstained with hematoxylin, and images were analyzed with ImageJ. For immunofluorescence, specimens were fixed in 4% formaldehyde for 20 min, permeabilized in 0.5% Triton X-100 for 20 min, blocked with serum for 1 h, and incubated with primary antibodies overnight at 4 °C followed by fluorescent secondary antibodies for 1 h at room temperature. Nuclei were stained with DAPI, and slides were scanned using the TissueFAXS imaging system. Double-positive cells were defined as cells with spatially overlapping signals in both fluorescence channels above the predefined positivity thresholds. Quantification was performed with StrataQuest software, and the percentage of double-positive cells was calculated relative to all DAPI-positive nuclei in three randomly selected high-power fields per section. The antibodies used were HIF-1α (Elabscience, E-AB-52,265), HMGB1 (Abcam, ab79823), S100A8/A9 (Abcam, ab288715), RAGE (Abcam, ab216329), and CA9 (Proteintech, 66,243-1-Ig).

Enzyme-linked immunosorbent assay (ELISA)

Blood samples from patients with LUAD, healthy volunteers (n = 19 per group), and mice (n = 5 per group) were used for ELISA. Following the manufacturer’s instructions (Jianglai Biology, Shanghai, China), the human/mouse HIF-1α, HMGB1, S100A8/A9, and RAGE ELISA kits were used to quantify these biomarkers in plasma. Absorbance was measured at 450 nm, and all experiments were performed in triplicate.

Bioinformatics analysis

Gene expression validation was conducted using the GEPIA2 web tool (http://gepia2.cancer-pku.cn), a publicly accessible database for cancer transcriptome analysis [16]. Analyses included box plots, clinical-stage plots, survival analyses, and correlation analyses. The gene symbol or Ensembl ID was entered in the “Gene” field for all analyses. “Lung Adenocarcinoma (LUAD)” was selected as the cancer type. Parameters were set as follows:

• •Box Plot Module: The P-value threshold was set to 0.05, with “Multiple Datasets” and “Match TCGA normal and GTEx data” options enabled. The log scale was set as log₂ (TPM + 1), and data from the GTEx database were used to integrate additional lung tissue samples.
• •Stage Plot Module: The log₂ (TPM + 1) transformation was applied. This was used as the logarithmic scale. The primary pathological stage grouping was visualized.
• •Survival Analysis Module: “Overall survival (OS)” was selected as the analysis metric, with the “median expression value” designated as the group cutoff, while retaining all other default parameters.
• •Correlation Analysis Module: Pairwise gene symbols (e.g., Gene A and Gene B) were input, followed by the selection of Pearson correlation analysis, with “Lung Adenocarcinoma (LUAD)” specified as the malignancy type.

Statistical analysis

All data are presented as mean ± standard error of the mean unless otherwise indicated. Statistical analyses were performed using GraphPad Prism 9.0 (GraphPad Software, San Diego, CA, USA). Comparisons among multiple groups were performed by one-way analysis of variance followed by the least significant difference test or Dunnett’s T3 test, as appropriate. Kaplan–Meier survival curves were compared by the log-rank test, and Pearson correlation analysis was used for GEPIA2 correlation analyses. A two-sided p value < 0.05 was considered statistically significant.

Results

Low expression of RAGE in LUAD was associated with poor survival

To investigate the role of RAGE in LUAD, we first analyzed GEPIA2, which integrates transcriptomic data from TCGA and GTEx. RAGE expression was significantly lower in LUAD than in normal lung tissue and was especially reduced in stages II–IV (Fig. 1A). Lower RAGE expression was also associated with worse overall survival (Fig. 1B). In contrast, HIF-1α expression did not differ significantly between tumor and normal tissues in GEPIA2, but higher HIF-1α expression was associated with poorer survival (Fig. 1C). CA9, a canonical hypoxia marker, was significantly upregulated in LUAD and was likewise associated with unfavorable outcome (Fig. 1D). GEPIA2 further showed a significant negative correlation between RAGE and both HIF-1α and CA9 expression (Fig. 1E). Together, these data support an association between reduced RAGE, increased hypoxia-related signaling, and poor prognosis in LUAD.

Fig. 1.

Low RAGE expression in LUAD is associated with poor survival. (A, B) RAGE mRNA expression and overall survival (OS) in normal lung tissue (n = 347) and LUAD (n = 483) samples obtained from the GEPIA2 database. (C, D) HIF-1α and CA9 mRNA expression and OS in normal lung and LUAD tissue samples from the GEPIA2 database. (E) Correlation analysis among the gene expression levels of HIF-1α, CA9, and RAGE according to the GEPIA2 database. Differential gene expression was analyzed using one-way analysis of variance; survival analysis employed the Kaplan–Meier method with the log-rank test; and correlation was assessed through Pearson correlation analysis.Kaplan-Meier survival curves include the Hazard Ratio (HR) with 95% confidence intervals and Log-rank P-values as indicated. Abbreviations: HIF-1α, hypoxia-inducible factor-1α; LUAD, lung adenocarcinoma; RAGE, receptor for advanced glycation end products; mRNA, messenger RNA; GEPIA2, Gene Expression Profiling Interactive Analysis 2.

Low levels of RAGE expression in the tumor tissues of hypoxic patients with LUAD

To further examine hypoxia and RAGE expression in LUAD, we performed immunofluorescence staining for RAGE and HIF-1α in tumor and adjacent non-tumor lung tissues(5–8 cm from tumor margins with no pathological evidence of cancer cell infiltration). Normal lung tissue showed strong RAGE staining, whereas LUAD tissue showed markedly reduced RAGE expression. In contrast, HIF-1α staining was increased in LUAD samples (Fig. 2A, B). We then measured circulating RAGE and HIF-1α in serum samples. Compared with healthy controls, patients with LUAD showed lower serum RAGE levels and higher HIF-1α levels (Fig. 2C). Although the tissue cohort was small, these findings are consistent with the GEPIA2 analysis and support the presence of a hypoxic, low-RAGE state in LUAD.

Fig. 2.

Low RAGE expression in the tumor tissues of patients with hypoxic LUAD. (A, B) Immunofluorescence staining of RAGE (purple) and HIF-1α (red) in human LUAD samples, with pancreatic acinar protein (PAP) as an epithelial cell marker (blue). (A) Representative image (scale bar: 50 μm). (B) Quantitative analysis of immune reactivity in RAGE- and HIF-1α–positive cells within LUAD samples. Data are presented as mean ± standard deviation (normal group: n = 4; LUAD group: n = 8). Three random fields were selected from each specimen for averaging. (C) Analysis of RAGE and HIF-1α levels in the serum of healthy controls and patients with LUAD using ELISA. The results are expressed as mean ± standard deviation (normal group: n = 19; LUAD group: n = 19). *indicates p < 0.05; **indicates p < 0.01. Abbreviations: RAGE, receptor for advanced glycation end products; HIF-1α, hypoxia-inducible factor-1α; LUAD, lung adenocarcinoma; ELISA, enzyme-linked immunosorbent assay.

Hypoxia regulated DAMP expression in LUAD

GEPIA2 analysis showed significant positive correlations between HIF-1α and the mRNA levels of HMGB1, S100A8, and S100A9 in LUAD (Fig. 3A). In human LUAD specimens, immunofluorescence staining showed increased S100A8/A9 immunoreactivity in the cytoplasm and nucleus, whereas intracellular HMGB1 staining was reduced and showed diminished nuclear localization (Fig. 3B, C). Serum analysis revealed significantly increased levels of HMGB1 and S100A8/A9 in patients with LUAD (Fig. 3D). Multicolor immunofluorescence further showed more frequent colocalization of residual HMGB1 and S100A8/A9 with HIF-1α-positive cells in LUAD tissue (Fig. 3E, F), consistent with a relationship between hypoxia and DAMP redistribution. In parallel, the proportion of cells co-expressing HMGB1 or S100A8/A9 with RAGE was reduced (Fig. 3G, H). These findings suggest that hypoxia is associated with altered DAMP localization and with disruption of canonical RAGE-associated signaling in LUAD.

Fig. 3.

Hypoxia regulates DAMP expression in LUAD. (A) Correlation analysis between HIF-1α and the expression levels of HMGB1 and S100A8/A9 using the GEPIA2 database. Pearson correlation analysis was performed. (B, C) Immunofluorescence staining of HMGB1 (green) and S100A8/A9 (yellow) in human LUAD samples. (B) Representative image with a scale bar of 50 μm. (C) Quantitative analysis of immune reactivity in HMGB1- and S100A8/A9-positive (+) cells in LUAD samples. Data are expressed as mean ± standard deviation (normal group: n = 4 and LUAD group: n = 8; three random fields per specimen were selected for averaging). (D) ELISA analysis of HMGB1 and S100A8/A9 expression in the serum of healthy controls and patients with LUAD. Data are expressed as mean ± standard deviation (normal group: n = 19; LUAD group: n = 19). Individual data points are shown to reflect biological variability(Mann-Whitney U test). (E, F) Dual immunofluorescence staining of HMGB1, S100A8/A9, and HIF-1α in human LUAD samples. DAPI staining of nuclei (blue). (E) Representative image, with a scale bar of 50 μm; magnified view on the right side, with a scale bar of 5 μm. (F) Quantitative analysis of double-positive cells for HMGB1, S100A8/A9, and HIF-1α. Results are expressed as mean ± standard deviation (normal group: n = 4; LUAD group: n = 8; three random fields per specimen were selected for averaging). (G, H) Immunofluorescence double staining of HMGB1, S100A8/A9, and RAGE in human LUAD samples. (G) Representative images, with a scale bar of 50 μm; magnified view on the right side, with a scale bar of 5 μm. (H) Quantitative analysis of cells dual-positive for HMGB1, S100A8/A9, and RAGE. Results are presented as mean ± standard deviation (normal group: n = 4; LUAD group n = 8). Three random fields were selected from each specimen and averaged. *indicates p < 0.05; **indicates p < 0.01. Abbreviations: RAGE, receptor for advanced glycation end products; HIF-1α, hypoxia-inducible factor-1α; LUAD, lung adenocarcinoma; ELISA, enzyme-linked immunosorbent assay; DAMPs, damage-associated molecular patterns; DAPI, 4′,6-diamidino-2-phenylindole; HMGB1, high-mobility group box 1.

Hypoxia inhibited RAGE expression and modulated the proliferation of lung cancer cells

To examine how hypoxia affects DAMP and RAGE expression, LLC cells were cultured under hypoxic conditions for different durations. After 24 h of hypoxia, HIF-1α, HMGB1, and S100A8/A9 levels increased, whereas RAGE showed a modest decrease compared with normoxic controls (Fig. 4A). After 48 h, all four proteins were reduced. These results suggest that acute hypoxia induces a transient DAMP response, whereas prolonged hypoxia may lead to adaptation, reduced protein synthesis, or cell injury, thereby attenuating the response.

Fig. 4.

Hypoxic conditions inhibit RAGE expression and modulate lung cancer cell proliferation. (A) Western blot and quantitative analysis of HIF-1α, HMGB1, S100A8/A9, and RAGE in LLC cells under hypoxic culture at different time points(0h, 12h, 24h, 48h). Each test was performed in triplicate. (B) Western blot and quantitative analysis of HIF-1α, HMGB1, S100A8/A9, and RAGE in LLC cells treated with LW-6(20 μM) or an equivalent volume of DMSO (0 μM, vehicle control) for 24 h under hypoxia and DMOG(DMOG (2 mM) or DMSO (0 mM, vehicle control) for 48 h under normoxic conditions. Each test was performed in triplicate. (C) Cell viability of LLC cells treated with FPS-ZM1 under hypoxia. Here, 0 μM FPS-ZM1 (DMSO treatment) served as the control group. Each test was performed in triplicate. (D) Transwell migration capability assessment of LLC cells after FPS-ZM1 treatment. (E) Western blot and quantitative analysis of HIF-1α, HMGB1, S100A8/A9, RAGE, and NF-κB expression in LLC cells treated with different concentrations(0uM, 5uM, 10uM) of FPS-ZM1 for 48 h under hypoxic conditions. Data are normalized to the 0-μM FPS-ZM1 (DMSO treatment) group. Each test was performed in triplicate. *indicates p < 0.05; **indicates p < 0.01; ***indicates p < 0.001; **** indicates p < 0.0001. Abbreviations: RAGE, receptor for advanced glycation end products; HIF-1α, hypoxia-inducible factor-1α; LUAD, lung adenocarcinoma; LLC, Lewis lung carcinoma; DMOG, dimethyloxycarbamate; DMSO, dimethyl sulfoxide; NF-κB, nuclear factor kappa beta; HMGB1, high-mobility group box 1.

We next examined whether HIF-1α modulation altered the expression of HMGB1, S100A8/A9, and RAGE. Under hypoxia, treatment with the HIF-1α inhibitor LW-6 reduced HIF-1α, HMGB1, and S100A8/A9 expression, whereas RAGE expression changed little (Fig. 4B). Under normoxia, treatment with the HIF-1α stabilizer DMOG increased HIF-1α and S100A8/A9 but decreased intracellular HMGB1 and RAGE (Fig. 4B). We interpret the HMGB1 decrease cautiously: it may reflect altered intracellular distribution or secretion rather than simple transcriptional repression. Overall, these data support HIF-1α as an upstream modulator of DAMP expression, while the mechanism linking hypoxia to reduced RAGE remains to be defined.

To assess the functional consequences of RAGE inhibition, we treated LLC cells with FPS-ZM1. FPS-ZM1 reduced LLC cell viability (Fig. 4C) and decreased migration in the Transwell assay in a dose-dependent manner (Fig. 4D). Western blot analysis showed decreased RAGE and phospho-p65 levels, increased HIF-1α and S100A8/A9 expression, and reduced intracellular HMGB1 (Fig. 4E). Thus, in isolated tumor cells, pharmacological RAGE inhibition suppressed NF-κB activation and was associated with reduced malignant behavior, despite evidence of compensatory stress-related changes in HIF-1α and S100A8/A9.

Systemic inhibition of RAGE promotes tumor progression in vivo via compensatory activation of the DAMPs–HIF-1α axis

Although FPS-ZM1 showed inhibitory effects in vitro, the in vivo setting yielded a different result. In the LLC lung metastasis model, mice treated with FPS-ZM1 developed larger pulmonary tumor nodules than untreated tumor-bearing mice (Fig. 5B). Body weight decreased and lung weight and lung index increased in tumor-bearing groups, with the most pronounced changes in the FPS-ZM1 group (Fig. 5C). Histology confirmed metastatic nodules in all tumor-bearing groups, and immunohistochemistry/Western blot analyses showed further reduction of RAGE in the FPS-ZM1 group (Fig. 5D,E). Notably, unlike the in vitro experiments, FPS-ZM1 treatment in vivo was associated with increased phospho-p65, together with increased HIF-1α and S100A8/A9 expression (Fig. 5E). VEGF and BCL-2 were upregulated and p21 was reduced, consistent with a more aggressive tumor phenotype. Serum HIF-1α, HMGB1, and S100A8/A9 levels were also elevated in tumor-bearing mice and were highest in the FPS-ZM1 group, whereas serum RAGE levels were reduced (Fig. 5F). These results indicate that systemic RAGE inhibition may have context-dependent effects and, within the in vivo tumor microenvironment, may be associated with compensatory signaling that overrides the direct inhibitory effect seen in vitro.

Fig. 5.

RAGE inhibition promotes tumor growth in vivo. (A) LLC cells were intravenously injected into mice via the tail vein to establish a lung metastasis mouse model. One week after modeling, FPS-ZM1 (1 mg/kg, twice weekly) was administered via intraperitoneal injection, and a solvent control group was established. The mice were euthanized, and lung tissues were collected after 4 weeks. (B) Gross observation of pulmonary tumor lesions and photographic documentation. (C) The body and lung weights of the mice were recorded, and the lung index was calculated (n = 10 per group). (D) Hematoxylin and eosin staining was used to visualize the lung tumor morphology, whereas immunohistochemical staining was used to detect HIF-1α, CA9, S100A8/A9, HMGB1, and RAGE expression (scale bar: 2.5 mm). (E) Western blot analysis was performed to evaluate the protein levels and quantify HIF-1α, HMGB1, S100A8/A9, RAGE, phosphorylated p65, p65, VEGF, BCL-2, and P21 levels in the normal control (N), tumor group (T), solvent control (O), and inhibitor group (F), with three replicates per group. (F) Analysis of serum HIF-1α, HMGB1, S100A8/A9, and RAGE levels in the N, T, and F groups using ELISA. Data are presented as mean ± standard deviation. n = 5. Individual data points are shown to reflect biological variability (Mann-Whitney U test). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Abbreviations: RAGE, receptor for advanced glycation end products; HIF-1α, hypoxia-inducible factor-1α; LUAD, lung adenocarcinoma; LLC, Lewis lung carcinoma; VEGF, vascular endothelial growth factor; ELISA, enzyme-linked immunosorbent assay; HMGB1, high-mobility group box 1.

Discussion

Analysis of GEPIA2 data and clinical specimens consistently showed that RAGE expression is reduced in LUAD and that a low-RAGE state is associated with poor outcome. This pattern agrees with previous reports indicating that, unlike many other solid tumors, normal lung tissue expresses abundant RAGE whereas NSCLC often shows RAGE loss during malignant transformation [[17], [18], [19], [20], [21]]. One explanation is tissue context: RAGE is highly expressed in type I alveolar epithelial cells, whereas many NSCLC cells arise from compartments with lower basal RAGE expression [22]. Our data therefore support the idea that RAGE has a context-dependent role in lung cancer biology and that its loss may accompany, rather than simply oppose, malignant progression.

Our results also suggest a time-dependent effect of hypoxia on DAMP regulation. During shorter hypoxic exposure, HIF-1α accumulation was accompanied by increased HMGB1 and S100A8/A9, whereas prolonged hypoxia was associated with reduction of these proteins. This pattern may reflect an early stress response followed by adaptation or cell injury. The divergent behavior of HMGB1 and S100A8/A9 in human tissues and pharmacologic experiments further suggests that these DAMPs are regulated through partially distinct mechanisms [23,24]. In particular, the increase in S100A8/A9 after DMOG treatment is consistent with prior work supporting HIF-1-dependent regulation of these genes [25]. By contrast, the decrease in intracellular HMGB1 after DMOG treatment may reflect redistribution, secretion, or post-transcriptional regulation rather than direct transcriptional suppression. Importantly, our conclusion that HIF-1α lies upstream of DAMP changes is based on pharmacologic modulation and correlative analysis; direct transcriptional regulation will require ChIP, promoter-reporter assays, or genetic perturbation.

The central observation of this study is the discrepancy between the in vitro and in vivo effects of FPS-ZM1. In vitro, FPS-ZM1 suppressed LLC viability, migration, and NF-κB phosphorylation. In vivo, however, FPS-ZM1 was associated with greater tumor burden and increased phospho-p65. At present, our data do not establish the exact mechanism of this reversal. A plausible explanation is that within the complex tumor microenvironment, increased DAMP availability and hypoxia-related stress activate compensatory pathways that bypass RAGE blockade. S100A8/A9 can signal through receptors such as TLR4, and these alternative pathways may converge on NF-κB and promote angiogenesis and survival [[26], [27], [28], [29]]. Supporting this interpretation, we observed increased S100A8/A9, HIF-1α, VEGF, and BCL-2 in the FPS-ZM1-treated tumors, and our supplementary ELISA data showed higher circulating TLR4 levels in patients with LUAD and tumor-bearing mice. However, these findings should be interpreted as supportive rather than definitive, because receptor-specific loss-of-function experiments were not performed.

This study has several limitations. First, the mechanistic links among HIF-1α, DAMP regulation, alternative receptors, and NF-κB reactivation are inferred from convergent observations rather than demonstrated directly. Second, our study relies mainly on pharmacological inhibition with FPS-ZM1; pharmacological blockade may not fully recapitulate genetic loss of RAGE and may include off-target effects. Third, the human tissue cohort was small, although the tissue findings were directionally consistent with public datasets and our serum analyses. Fourth, the use of murine LLC cells and mouse models limits direct extrapolation to human LUAD biology. Accordingly, Fig. 6 should be regarded as a working model that summarizes our observations and generates testable hypotheses for future studies.

Fig. 6.

Proposed working model of context-dependent RAGE signaling in LUAD. Under hypoxic conditions, S100A8/A9 undergoes nuclear-to-cytoplasmic redistribution. In isolated cell culture, FPS-ZM1 inhibits the canonical RAGE/NF-κB axis and is associated with reduced viability and migration. In the in vivo tumor microenvironment, however, RAGE inhibition is accompanied by increased HIF-1α and S100A8/A9, reactivation of NF-κB signaling, and upregulation of VEGF and BCL-2, resulting in paradoxical tumor acceleration. Solid lines indicate experimentally observed associations in the present study, whereas dashed lines indicate proposed compensatory pathways that require further validation. Abbreviations: RAGE, receptor for advanced glycation end products; HIF-1α, hypoxia-inducible factor-1α; NF-κB, nuclear factor kappa beta; HMGB1, high-mobility group box 1.

In conclusion, our data indicate that RAGE inhibition has context-dependent effects in LUAD. In isolated tumor cells, pharmacological RAGE blockade suppressed malignant phenotypes, whereas in vivo it was associated with enhanced tumor progression and reactivation of NF-κB-related signaling. These findings support a model in which hypoxia-associated compensatory pathways may bypass RAGE inhibition in the tumor microenvironment, but the specific molecular route remains to be established. Further studies using genetic models, receptor-specific pathway interrogation, and validation in human LUAD systems are needed before therapeutic implications can be drawn.

Funding

The author(s) declare that financial support was received for the research and/or publication of this article. This study was supported by the Qinghai Province Science and Technology Plan Project (2025-ZJ-748)

Data availability

The data underlying the conclusions of this article are included in the article. Further inquiries can be directed to the corresponding authors.

CRediT authorship contribution statement

Han Xiao: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Data curation, Conceptualization. Xiaoyan Song: Writing – review & editing, Resources, Data curation. Tana Wuren: Writing – review & editing, Supervision, Resources, Funding acquisition. Ri-li Ge: Writing – review & editing, Supervision, Resources, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.