Pan-Cancer Landscape of the Novel Oxygen Sensor ADO and Its Potential Role in Hepatocellular Carcinoma

Abstract

Background

Hypoxia is a key driver of tumor progression across cancers, yet oxygen-sensing mechanisms beyond HIFs remain underexplored. 2-Aminoethanethiol dioxygenase (ADO) has recently been identified as an oxygen sensor, but its role in malignancy is poorly defined. We conducted a pan-cancer analysis of ADO with a special focus on hepatocellular carcinoma (HCC), to assess its oncogenic significance and clinical potential.

Methods

A multi-omics pan-cancer analysis of ADO expression and survival was performed using TCGA and GTEx, with validation in HCC across ICGC, GEO, and CNHPP proteomic cohorts. Correlations with genetic, epigenetic, immune, and pathways were evaluated. Drug sensitivity was predicted. Functional validation was conducted in HCC cells through proliferation, colony formation, Western blotting, and xenograft assays.

Results

ADO was aberrantly expressed across cancers and showed cancer type–specific survival associations. Integrative analyses revealed links with tumor mutation burden, microsatellite instability, chromatin regulator methylation, RNA modification, proliferative signaling (G2M checkpoint, MYC, TGF-β), an immunosuppressive microenvironment, and negative correlations with ROS-responsive genes. In HCC, ADO was consistently overexpressed, associated with advanced stage, poor differentiation, residual disease, and unfavorable survival across independent cohorts. ADO-high HCC showed reduced predicted responsiveness to checkpoint blockade but increased sensitivity to sorafenib and fluorouracil. Experimentally, ADO overexpression activated ERK signaling, upregulated CD276 and HMGB1, and promoted HCC cell proliferation, while ADO depletion suppressed tumor growth in vitro and in vivo, reversible upon re-expression.

Conclusion

ADO plays oncogenic and immunomodulatory roles in HCC, and may serve as a potential prognostic biomarker and therapeutic target in liver cancer.

Introduction

Hepatocellular carcinoma (HCC) is the most common primary liver malignancy and an inevitable cause of cancer-related mortality worldwide.¹ Despite progress in early detection and therapeutic strategies, outcomes for patients with advanced HCC remain poor due to high recurrence rates and limited responsiveness to systemic therapies.² The unique hepatic microenvironment,³ often shaped by chronic liver disease, viral hepatitis, and cirrhosis, further complicates disease management and underscores the need for novel molecular biomarkers and therapeutic targets. Consistent with the need for better risk stratification, candidate biomarkers associated with HCC outcome continue to be identified, including Vav1 protein expression and mitochondrial lipid-metabolism gene signatures.^4–6

Hypoxia is a hallmark of solid tumors and exerts a profound influence on cancer biology. Oxygen deprivation drives stabilization of hypoxia-inducible factors (HIFs), metabolic reprogramming, angiogenesis, epithelial-mesenchymal transition, immune suppression, and resistance to systemic therapies. These adaptive responses enhance tumor aggressiveness and contribute to poor patient outcomes across malignancies.^7,8 In hepatocellular carcinoma (HCC), hypoxia is especially relevant. HCC frequently develops in cirrhotic livers, where fibrotic scarring and vascular distortion compromise oxygen delivery.⁹ Within the tumor, abnormal vasculature and high metabolic demand create steep oxygen gradients, leading to regions of severe hypoxia. This microenvironment fuels tumor progression, recurrence, and therapeutic resistance. Clinical interventions such as transarterial chemoembolization (TACE) and anti-angiogenic therapies may further intensify hypoxia by restricting perfusion, inadvertently promoting angiogenesis, invasion, and immune evasion.^10,11 Collectively, these observations highlight the critical role of oxygen-sensing enzymes in hepatocarcinogenesis and therapeutic response.

2-Aminoethanethiol dioxygenase (ADO) is a thiol dioxygenase first characterized for catalyzing the oxidation of cysteamine to hypotaurine in taurine metabolism.¹² More recently, ADO has emerged as an enzymatic oxygen sensor.¹³ By hydroxylating N-terminal cysteine residues, ADO regulates protein stability via the Arg/Cys branch of the N-degron pathway, directly coupling oxygen availability to proteasomal degradation. ADO substrates include signaling proteins such as RGS4/5/16 and interleukin-32, underscoring its regulatory functions beyond transcriptional control.^13,14 This pathway operates in parallel with the canonical HIF axis and expands the repertoire of cellular oxygen-sensing mechanisms. Despite these insights, the expression, clinical significance, and mechanistic role of ADO in cancer, particularly in HCC, remain poorly defined.

To address these gaps, first, we performed a pan-cancer analysis to define its expression patterns, molecular associations, and prognostic value across human malignancies. Second, we focused on HCC, to evaluate its tumor-promoting role, potential as a prognostic biomarker and relevant mechanisms. This study provides a systematic view of ADO in human cancers and clarifies its significance in the pathogenesis of HCC.

Materials and Methods

Expression and Survival Analyses

The Cancer Genome Atlas (TCGA) Pan-Cancer (PANCAN, N = 10,535, G = 60,499) dataset and the Genotype-Tissue Expression (GTEx) databases were obtained via the USC Xena database (https://xenabrowser.net/). Cancer types with fewer than 10 samples were removed, yielding a final dataset comprising 39 cancer types. Samples with a follow-up time shorter than 30 days were excluded. Additional 14 GEO datasets (GSE124535, GSE135631, GSE144269, GSE14520, GSE169289, GSE184733, GSE214846, GSE29721, GSE39791, GSE45267, GSE45436, GSE57957, GSE67764, GSE95698) were retrieved from GEO (https://www.ncbi.nlm.nih.gov/geo/). ICGC-LIRI-JP cohort was downloaded from the ICGC data portal (https://dcc.icgc.org/). The proteomic expression data of ADO in HCC were obtained from the CNHPP (Chinese human proteome program) liver data portal (http://liver.cnhpp.ncpsb.org/). The Wilcoxon rank-sum test was applied to compare ADO expression between tumor and normal tissues, as well as among clinicopathological subgroups in the TCGA-LIHC cohort. This nonparametric, rank-based test was selected because expression values were not assumed to follow a normal distribution and because the compared subgroups were frequently small and unbalanced; rank-based inference is robust to skewed distributions and outlying values. The Human Protein Atlas (HPA, https://www.proteinatlas.org/) was used to examine ADO protein expression in liver cancer and normal liver tissues. The coxph function was applied to construct Cox proportional hazards regression models, evaluating the association between ADO expression and patient survival across individual cancer types and the log-rank test was employed to assess statistical significance. Cox regression is a semi-parametric model that leaves the baseline hazard unspecified; it therefore estimates hazard ratios with 95% confidence intervals while appropriately handling censored survival data. Results were presented as forest plots. Kaplan-Meier plots were also analyzed using the TCGA-LIHC cohort, GSE14520, and the CNHPP cohort.

TMB and MSI Analysis

Somatic mutation data (level 4 Simple Nucleotide Variation) processed by MuTect2 (DOI: 10.1038/nature08822) were downloaded from the GDC portal (https://portal.gdc.cancer.gov/). Tumor mutation burden (TMB) was calculated using the tmb function of the maftools R package (v2.8.05). Microsatellite instability (MSI) scores for each tumor type were obtained from a previous study by Bonneville et al.¹⁵ ADO expression values were extracted across all samples from the TCGA Pan-Cancer dataset and log2-transformed (log2[x + 0.001]). Cancer types with fewer than three samples were excluded, yielding 37 cancer types for downstream analysis. Then the correlations between ADO and TMB and MSI data were compared.

Methylation Analysis of Chromatin Regulators

A total of 902 chromatin regulators were collected through database mining and literature retrieval,^16–18 including DNA methylation modifiers (eg, DNA methyltransferases and demethylases), histone modification regulators, and chromatin remodelers. Using the 450K methylation data from TCGA across multiple cancer types, the correlation between ADO expression and the methylation levels of these chromatin regulators was analyzed. For each cancer type, the proportion of regulators showing significant correlation with ADO expression was calculated.

RNA Modification Analysis

From the TCGA Pan-Cancer dataset, we extracted the expression profiles of ADO and 44 marker genes corresponding to three categories of RNA modification genes (m1A: 10 genes, m5C: 13 genes, m6A: 21 genes). We then filtered samples to retain only primary tumor, excluding all normal samples. Expression values were log2-transformed using the formula log2(x + 0.001). Subsequently, we computed the Pearson correlation coefficients between ADO expression and the expression of RNA modification genes.

Immune Microenvironment Analyses

Immune infiltration scores were estimated using the ESTIMATE algorithm, generating ImmuneScore, StromalScore, and ESTIMATEScore. Tumor-infiltrating immune cell (TIIC) fractions were assessed using the single-sample gene set enrichment analysis (ssGSEA) algorithm implemented in the GSVA R package (v1.46.0).¹⁹ Briefly, scores were computed with the gsva() function using method = “ssgsea” with the default parameters of the package: tau = 0.25, ssgsea.norm = TRUE, mx.diff = TRUE, abs.ranking = FALSE, min.sz = 1, and max.sz = Inf. Marker genes for 24 immune cell subsets were adopted from Bindea et al.²⁰ The correlation between ADO expression and immune cell infiltration scores was assessed using Spearman’s rank correlation test. Results were visualized as a heatmap generated in R (v4.2.1) using ggplot2 (v3.4.4). Associations between ADO expression and immune checkpoint genes (eg, CD276, ENTPD1, HMGB1) were also evaluated.

Immunotherapy Response Analyses

The Kaplan-Meier Plotter online database²¹ was used to investigate the prognostic relevance of ADO expression in patients across nine cancer types treated with immune checkpoint inhibitors, including anti-CTLA4 (n = 112), anti-PD-1 (n = 411), and anti-PD-L1 (n = 459). Immunotherapy response prediction in the HCC was analyzed using TIDE (Tumor Immune Dysfunction and Exclusion) algorithm, implemented with a publicly available R script.²² TIDE scores, calculated as the sum of the dysfunction and exclusion scores, were obtained with the default settings of the algorithm. Each HCC sample was classified as a predicted responder or non-responder according to the default classification cutoff of the algorithm; higher TIDE scores indicate stronger immune evasion and a lower predicted benefit from immune checkpoint blockade. TIDE scores were additionally compared between the two groups as continuous variables using the Wilcoxon rank-sum test.

Pathway and Functional Analyses

Nineteen hallmark-related gene sets were obtained from Wei et al.²³ Gene sets were converted into GMT format using GSEABase, and single-sample gene set enrichment analysis (ssGSEA) was performed using GSVA. Correlations between ADO expression and hallmark pathway scores were evaluated across cancers. For hepatocellular carcinoma (HCC), differential expression analyses were conducted between ADO-high and ADO-low subgroups, stratified by median value. Differentially expressed genes (DEGs) were identified using the DESeq R package (www.huber.embl.de/users/anders/DESeq/), with thresholds of |log2FC| > 1 and adjusted P < 0.05. KEGG pathway enrichment of upregulated or downregulated DEGs was carried out using clusterProfiler.

Drug Sensitivity Prediction

Drug sensitivity predictions were made by integrating ADO-related expression profiles with drug response data from the Genomics of Drug Sensitivity in Cancer (GDSC) database using the pRRophetic R package.²⁴ The half-maximal inhibitory concentration (IC50) of each sample was estimated by ridge regression, with all parameters set to default values. To minimize technical bias, batch effects were corrected using the ComBat algorithm, and tissue type information for all samples was incorporated into the model. In cases of duplicate gene expression values, the mean expression level was used for downstream analysis.

Cell Culture and Stable Transfection

Human hepatocellular carcinoma cells (HepG2) and murine HCC cells (Hepa1-6, its C57BL/6 background enabled syngeneic subcutaneous implantation into immunocompetent C57BL/6 mice, allowing in vivo tumor growth to be evaluated in the presence of an intact immune system) were obtained from SUNNCELL (CatLog: SNL-083 and SNL-118). Cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and maintained at 37 °C with 5% CO₂. All cell lines were routinely tested for mycoplasma contamination using the GMyc-PCR Mycoplasma Test assay (Yeasen Biotech, Shanghai). Cell line identity was authenticated by short tandem repeat (STR) profiling. ADO overexpression constructs (HepG2-OE) and shRNA-mediated knockdown vectors (Hepa1-6-SH1) were established using lentiviral transduction, followed by puromycin selection. The pLVX-IRES-Puro plasmid was used for the stable overexpression of human ADO. The pLKO.1-Puro-ADO-sh1 (5’GGTCACCTACATGCACATCTA3’), pLKO.1-Puro-ADO-sh2 (5’CGGTATGCTCAAGGTGCTGTA3’), pLKO.1-Puro-ADO-sh3 (5’ACAACCTGCACCAGATTGATG3’) and negative control (5’CCTAAGGTTAAGTCGCCCTCG3’) constructs were used for the lentivirus packaging for murine ADO knockdown selection. The pCDH-CMV-mADO(si mut)-EF1-Hygro construct was used for re-expressing ADO in knockdown cells (Hepa1-6-SH1-OE). All plasmids were purchased from Geneppl technology. Successful overexpression or knockdown was confirmed by Western blotting.

Cell Proliferation and Colony Formation Assays

Proliferation rates were assessed using a cell counting assay. Cells were seeded at 3×10³ cells per well in 6-well plates and cell numbers were recorded daily for 7 consecutive days. The average cell count of each group was calculated, and cell proliferation curves were plotted. Colony formation assays were performed by seeding 500 cells per well in 6-well plates and cultured for 14 days, followed by fixation and crystal violet staining. Colonies were quantified using ImageJ software.

Western Blotting

Cells were lysed with RIPA buffer and mixed with an appropriate amount of 5× protein loading buffer. After denaturation by heating, equal amounts of protein were separated by SDS-PAGE and transferred to PVDF membranes. Membranes were incubated overnight with primary antibodies against ADO (Proteintech), CD276 (ABclonal), HMGB1 (ABclonal), p-ERK (CST), ERK(CST), p-AKT (CST), and AKT (CST), followed by HRP-conjugated secondary antibodies. Bands were visualized with ECL detection system. For hypoxia treatment, the cells were incubated in a hypoxia (1% O2) chamber for 6 hours, then cells were lysed and proteins were extracted.

Xenograft Models

All animal experiments were conducted in Experimental Animal Center, Zhejiang Academy of Medical Sciences, in compliance with institutional ethical guidelines and approved by the Institutional Animal Care and Use Committee (IACUC) of Experimental Animal Center, Zhejiang Academy of Medical Sciences. Male C57BL/6 mice (6 weeks old, 18–22 g), obtained from Experimental Animal Center, Zhejiang Academy of Medical Sciences, were housed under specific pathogen-free (SPF) conditions with controlled temperature (22 ± 2 ℃), humidity (50–60%), and a 12-h light/dark cycle, with free access to food and water. For xenograft establishment, six-week-old male C57BL/6 mice were subcutaneously injected with Hepa1-6-NC, Hepa1-6-SH1 (5 × 10⁶ cells per mouse, 10 mice per group, 4 mice per cage). For evaluating re-expression of ADO in vivo, Hepa1-6-SH1-CTRL or Hepa1-6-SH1-OE cells were subcutaneously injected into six-week-old male C57BL/6 mice (5 × 10⁶ cells per mouse, 6 mice per group, 3 mice per cage). Mice were randomly allocated into each group. Tumor size was measured twice weekly, and tumor volume was calculated as: Volume=1/2×(length×width2). Humane endpoints were predefined as tumor volume >1500 mm³, ulceration, or impaired mobility. At the study endpoint, mice were euthanized by CO₂ inhalation followed by cervical dislocation, and tumors were excised and weighed.

Statistical Analysis

All statistical analyses were performed using R software (v4.1.0) and GraphPad Prism (v9.0). Pearson correlation analysis was applied to evaluate associations between ADO and molecular features. Comparisons between two groups were performed using the Wilcoxon rank-sum test or Student’s t-test, while comparisons among multiple groups were assessed using one-way ANOVA with post hoc corrections. Survival outcomes were evaluated by the Kaplan–Meier method, with significance determined using the log-rank test. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated by Cox regression analysis. Drug sensitivity (IC50) comparisons were performed using the Wilcoxon test. Unless otherwise specified, all tests were two-sided, and P < 0.05 was considered statistically significant. All experiments were performed with at least three independent biological replicates.

Results

The Expression and Prognostic Indications of ADO in Cancers

First, we analyzed ADO expression profiles across 31 normal tissue types using the GTEx dataset. ADO exhibited markedly elevated expression in nerve and testis, whereas comparatively low expression levels were observed in blood, liver, and pancreas (Figure 1A). Furthermore, differential expression analyses between tumor and adjacent normal tissues among cancer types based on the TCGA dataset revealed that ADO was commonly upregulated in most cancer types, including ACC, BLCA, BRCA, CESC, CHOL, COAD, ESCA, GBM, HNSC, LAML, LGG, LIHC, LUAD, LUSC, OV, PAAD, PRAD, READ, SKCM, STAD, THCA, UCS but downregulated in KIRC and UCEC (Figure 1B). We next investigated the prognostic significance of ADO expression in various cancers using univariate survival analyses. For progression-free survival (PFS), elevated ADO expression was associated with unfavorable prognosis in ACC, LIHC, and UVM, whereas it indicated a favorable prognosis in GBMLGG, LGG, and KIRC (Figure 1C). Similarly, analyses of overall survival (OS) revealed that higher ADO expression correlated with worse prognosis in ACC and LIHC, but with better prognosis in GBMLGG, LGG, and KIRC (Figure 1D).

Figure 1.

Expression and prognostic indications of ADO in cancers. (A) Expression levels of ADO across 31 normal tissue types from the GTEx dataset. (B) Differential expression of ADO between tumor and adjacent normal tissues across TCGA cancer types. (C) Univariate survival analyses of ADO expression with progression-free survival (PFS) across cancers. (D) Univariate survival analyses of ADO expression with overall survival (OS) across cancers. Statistical significance: **P < 0.01; ***P < 0.001; -P > 0.05.

Correlations Among ADO and Genetic, Epigenetic Modifications and Cancer Hallmarks

Across cancer types, ADO expression demonstrated variable associations with tumor mutation burden (TMB). Positive correlations were observed in STES and STAD, whereas a significant negative correlation was detected in CHOL (Figure 2A). Microsatellite instability (MSI) analysis revealed that ADO expression was positively correlated with GBMLGG, CESC, STES, STAD, and READ, but negatively correlated with COAD, KIPAN, PRAD, THCA, and DLBC (Figure 2B). In the context of chromatin regulation, the correlations of ADO expressions with the methylation levels of chromatin regulators were compared, including DNA methylation modifiers, histone modification regulators, and chromatin remodelers, and the proportions of positive or negative correlations were counted. LGG exhibited the highest proportion of positive correlations, while TGCT showed the highest proportion of negative correlations (Figure 2C). Regarding RNA modification, ADO expression showed almost consistent positive associations with m1A-, m5C-, and m6A-related genes (Figure 2D). These results collectively suggest that ADO is closely linked to both genetic and epigenetic regulatory processes. Functional pathway analysis using ssGSEA among cancer types revealed that ADO expression was positively correlated with G2M_checkpoint, MYC_targets, and TGFB pathways, whereas Genes_up-regulated_by_reactive_oxygen_species (ROS) was negatively correlated in most cancers. Surprisingly, ADO expression was positively correlated with nearly all cancer hallmarks in LIHC except for Genes_up-regulated_by_reactive_oxygen_species (ROS) (Figure 2E). These findings indicate that ADO may play a central role in regulating multiple functional pathways associated with tumor progression and cancer hallmark activities, especially in LIHC.

Figure 2.

Correlations among ADO and genetic, epigenetic modifications and cancer hallmarks. (A) Correlations between ADO expression and tumor mutational burden (TMB). (B) Correlations between ado expression and microsatellite instability (MSI). (C) Associations of ADO expression with methylation levels of chromatin regulators, including DNA methylation modifiers, histone modification regulators, and chromatin remodelers. (D) Correlations between ADO expression and RNA modification marker genes (m¹A, m⁵C, m⁶A). (E) Correlations between ADO expression and hallmark pathways using ssGSEA. Statistical significance: *P < 0.05.

Potential Roles of ADO in Tumor Immune Microenvironment and Immunotherapy Responses

Across cancer types, ADO expression exhibited significant associations with multiple dimensions of the tumor immune microenvironment. Analysis of ESTIMATE-derived immune and stromal scores revealed that higher ADO expression was positively correlated with elevated ImmuneScore and ESTIMATEScore in LIHC and PAAD, suggesting a more immune-enriched tumor milieu, while negatively correlated with ImmuneScore and ESTIMATEScore in several cancer types, especially in GBM and LGG (Figure 3A). In detail, we found ADO expression broadly correlated with enriched infiltration of T helper cells and Tcm cells in most cancers and positively correlated with various immune subsets in PAAD. Regarding LIHC, we also found a positive correlation of ADO expression and ImmuneScore and ESTIMATEScore, especially the infiltration levels of macrophages, T helper cells, Tem, and Th2 cells, highlighting its potential role in shaping both innate and adaptive immune responses (Figure 3B). Correlation analysis with immune-related genes, including immune checkpoints and effector molecules, further demonstrated that ADO expression was broadly positively associated with CD276, ENTPD1, and HMGB1 across diverse tumor types. Notably, ADO was commonly correlated with most immunoregulatory factors in LIHC, especially CD276, ENTPD1, and HMGB1 (Figure 3C). Survival analyses stratified by ADO expression revealed that high ADO expression was significantly associated with worse PFS after treatment with anti-PD-L1 (Figure 3D), anti-PD-1 (Figure 3E), and anti-CTLA4 (Figure 3F) treatments in pan cancers. Additionally, LIHC patients with high ADO expression had increased TIDE scores, indicating reduced potential response from immune checkpoint blockade (Supplementary Figure 1). These findings collectively suggest that ADO not only correlates with an immunologically suppressive microenvironment but also portends unfavorable prognosis, implying its potential involvement in immunotherapy responsiveness.

Figure 3.

Potential roles of ADO in tumor immune microenvironment and immunotherapy responses. (A) Correlations of ADO expression with ImmuneScore and ESTIMATEScore across cancers. (B) Associations of ADO expression with infiltration of tumor-infiltrating immune cell subsets across cancers. (C) Correlations of ADO expression with immune checkpoint molecules and effector genes across cancers. Kaplan–Meier survival analyses of ADO expression with PFS in patients treated with immune checkpoint blockade, stratified by the auto-selected best cutoff: anti-PD-L1 (D), anti-PD-1 (E), and anti-CTLA4 (F). Statistical significance: *P < 0.05.

Expression of ADO and Its Correlations with Clinicopathological Features in HCC

Pan-cancer analyses revealed widespread dysregulation of ADO and its prognostic relevance, with particularly strong associations observed in hepatocellular carcinoma (HCC). We therefore performed an in-depth investigation of ADO in HCC to assess and validate its expression, clinicopathological features, and prognostic value. In hepatocellular carcinoma (HCC), ADO expression was confirmed to be significantly upregulated in tumor tissues compared with adjacent non-tumor liver tissues across multiple independent cohorts, including ICGC, and 14 GEO datasets (Figure 4A). IHC staining in the human protein atlas (HPA) indicated a higher expression of ADO in the HCC than the normal tissue (Figure 4B), which is further confirmed by proteomic data from the CNHPP cohort (Figure 4C). Clinicopathological correlation analyses demonstrated that ADO expression was significantly associated with female gender, higher T stage, advanced pathological stage, with residue tumor status, and poorer histologic grade, while no significant differences were observed across age or race groups (Figure 4D–J). Elevated ADO predicted worse OS in three independent LIHC datasets, including TCGA (HR = 1.92, 95%CI: 1.34–2.76, P = 0.0003), GSE14520 (HR = 1.74, 95%CI: 1.10–2.74, P = 0.016), and the CNHPP cohorts (HR = 2.86, 95%CI: 0.99–8.30, P = 0.036) (Figure 4K–M). Collectively, these findings demonstrate that ADO is significantly upregulated in HCC at both transcriptomic and proteomic levels, correlates with aggressive clinicopathological features, and predicts poor survival outcomes. These results suggest that ADO may function as a potential oncogenic driver and prognostic biomarker in HCC.

Figure 4.

Expression of ADO and its correlations with clinicopathological features in HCC. (A) ADO expression in HCC vs adjacent non-tumor liver tissues across TCGA, ICGC, and GEO cohorts. (B) Immunohistochemistry (IHC) staining from the human protein Atlas (HPA) showing ADO expression in HCC and normal liver tissues. (C) ADO protein expression in HCC tumor and peri-tumor tissues from the CNHPP cohort. Correlations between ADO expression and clinicopathological features in the TCGA-LIHC cohort, including age (D), race (E), gender (F), T stage (G), pathological stage (H), residual tumor status (I), and histological grade (J). Kaplan–Meier survival analyses of ADO expression and overall survival (OS) in TCGA (K), GSE14520 (L), and CNHPP (M) cohorts. Statistical significance: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, ns: not significant.

Functional Analysis of ADO in HCC

To explore the functional role of ADO in hepatocellular carcinoma (HCC), we began by analyzing the differential expression of genes between ADO-high and ADO-low HCC tissues. The volcano plot revealed significant upregulation of genes such as CEACAM7, CA9, ESRP1, WNK2, A4GNT, SLC34A2, and EPCAM in ADO-high HCC tissues, while genes like CYP2A6, GLYAT, SLC10A1, CYP8B1, TAT, CYP3A4, and CYP2A7 were notably downregulated (Figure 5A). The heatmap analysis further confirmed these expression patterns, with distinct clustering of upregulated and downregulated genes across HCC samples (Figure 5B). Genes with elevated expression were primarily involved in KEGG pathways such as PI3K-Akt signaling, Wnt signaling, protein digestion and absorption, proteoglycans in cancer, and ECM-receptor interaction, all of which are crucial for tumor growth, metastasis, and survival (Figure 5C). Conversely, the downregulated genes were significantly enriched in pathways related to retinol metabolism, bile secretion, drug metabolism - cytochrome P450, metabolism of xenobiotics by cytochrome P450, and chemical carcinogenesis - DNA adducts, pointing to metabolic dysfunctions commonly observed in liver cancer (Figure 5D). In terms of treatment response, the drug sensitivity prediction revealed that higher expression of ADO was associated with increased sensitivity to fluorouracil and sorafenib, two common therapies for HCC (Figure 5E and F). These analyses collectively highlight the functional relevance of ADO in HCC, with its expression linked to key pathways involved in tumor progression, metabolic disruption, and sensitivity to standard therapies. ADO may thus serve as an important oncogenic driver and prognostic biomarker, providing insights for future therapeutic strategies in HCC.

Figure 5.

Functional analysis of ADO in HCC. (A) Volcano plot of differentially expressed genes between ADO-high and ADO-low HCC tissues. (B) Heatmap of top differentially expressed genes in ADO-high vs ADO-low groups. (C) KEGG pathway enrichment analysis of upregulated genes in ADO-high HCC tissues. (D) KEGG pathway enrichment analysis of downregulated genes in ADO-high HCC tissues. (E and F) Predicted sensitivity of patients to fluorouracil and sorafenib between ADO Q1 (top 25%) and ADO Q4 (bottom 25%) group. Statistical significance: ****P < 0.0001.

ADO Facilitates Cell Proliferation and Tumor Growth of HCC in vitro and in vivo

To evaluate the role of ADO in hepatocellular carcinoma (HCC) cell proliferation and tumor growth, we conducted both in vitro and in vivo experiments. In in vitro assays, HepG2 cells overexpressing ADO (HepG2-OE) exhibited significantly enhanced cell proliferation compared to control cells (HepG2-CTRL). The increase in cell number was evident over the course of 7 days, with a marked difference observed from day 3 onward (Figure 6A). Similarly, HepG2-OE cells showed increased clonogenicity, as evidenced by colony formation assays (Figure 6B and C). Western blotting analysis revealed that ADO overexpression elevated the expression of immune checkpoint protein CD276 and HMGB1, and selectively activated the ERK signaling pathway under hypoxic stress, without affecting the AKT pathway (Figure 6D). ADO expression was remarkably correlated with ERK pathway proteins including KRAS, BRAF, RAF1, MAPK1, MAPK3, and ELK1 in the TCGA-LIHC dataset (Supplementary Figure 2A–F), indicating the ERK pathway as a vital signaling in ADO-induced HCC proliferation. To further validate the role of ADO in tumor growth, the murine Hepa1-6 cell line was used to generate xenograft models. Hepa1-6 cells with ADO knockdown (Hepa1-6-SH1) (Supplementary Figure 3A) exhibited significantly attenuated cell proliferation compared to controls (Hepa1-6-NC) in vitro (Figure 6E). Upon re-expression of ADO (Hepa1-6-SH1-OE) (Supplementary Figure 3B), cell proliferation could be restored (Figure 6F). After cells were inoculated subcutaneously into mice, the tumors in the Hepa1-6-SH1 group grew more slowly than the Hepa1-6-NC group (Figure 6G and H), while Hepa1-6-SH1-OE group grew significantly faster and exhibited restored proliferation than the Hepa1-6-SH1-CTRL group in vivo (Figure 6I and J), confirming the critical role of ADO in promoting tumor growth. Collectively, these results demonstrate that ADO overexpression enhances both HCC cell proliferation and tumor growth, highlighting its potential as a therapeutic target in HCC.

Figure 6.

ADO facilitates cell proliferation and tumor growth of HCC in vitro and in vivo. (A) Comparison of cell numbers counted daily for 7 days between HepG2 cells with ADO overexpression (HepG2-OE) and controls (HepG2-CTRL). (B–C) Colony formation assays comparing HepG2-OE and HepG2-CTRL cells. (D) Western blotting showing effects of ADO overexpression on CD276, HMGB1, erk, and AKT pathways under hypoxia stress (1% O2 treatment for 6 hours). (E–F) Comparison of cell numbers counted daily for 7 days in Hepa1-6 cells with ADO knockdown (Hepa1-6-SH1 vs Hepa1-6-NC) and rescue by re-expression (Hepa1-6-SH1-OE vs Hepa1-6-CTRL). (G–H) Xenograft tumor growth assays in mice injected with Hepa1-6-NC or Hepa1-6-SH1 cells, showing tumor pictures and tumor volume curves. (I–J) Xenograft tumor growth assays in mice injected with Hepa1-6-SH1-CTRL or Hepa1-6-SH1-OE cells, showing tumor pictures and tumor volume curves. Statistical significance: **P < 0.01; ***P < 0.001; ****P < 0.0001.

Discussion

Our analyses demonstrate that ADO is aberrantly expressed across multiple malignancies and shows cancer type - specific prognostic relevance. In HCC, ADO is consistently upregulated at both transcriptomic and proteomic levels, predicts adverse outcomes, and functionally drives tumor proliferation and immune modulation. These findings highlight that ADO may serve as both a preliminary biomarker and a potential therapeutic target in liver cancer.

Prior work on hypoxia has centered almost exclusively on the role of hypoxia-inducible factors (HIFs) and their downstream transcriptional programs.^25,26 ADO, a plant oxygen sensor, was surprisingly found as a global oxygen sensor in both plants and animals by coupling oxygen availability directly to proteasomal degradation of target proteins via the N-degron pathway.^13,27 This distinct mechanism explains how ADO might influence oncogenic signaling and immune regulation independently of canonical HIF biology. Of note, till now, ADO has not been listed into hypoxia gene expression signatures when defining hypoxia in cancer.²⁸

One of the most consistent findings of our study was the prognostic impact of ADO in HCC. Across TCGA, ICGC, GEO, and CNHPP cohorts, ADO was not only upregulated at the mRNA and protein levels but also associated with advanced stage, poor differentiation, residual disease, and worse survival. This reproducibility across multiple platforms suggests that ADO is a robust marker of aggressive tumor biology in HCC. However, current prognostic models and hypoxia-directed biomarker panels for HCC have not incorporated ADO, which may limit their predictive accuracy.^29,30 Our findings suggest that inclusion of ADO could enhance risk stratification, particularly in patients with hypoxia-driven disease biology. Notably, ADO exhibited divergent prognostic associations across other malignancies, being favorable in gliomas and renal cancers but adverse in adrenocortical carcinoma and HCC. The biological basis of these divergent associations remains incompletely defined but may involve tissue-specific metabolic and microenvironmental contexts. ADO couples oxygen availability to the oxidation of cysteamine and to N-terminal cysteine (N-degron) signaling, and the consequences of its activity can be rewired by tumor-specific metabolic dependencies. For example, ADO-mediated taurine synthesis has been linked to tumor progression in pancreatic cancer,³¹ whereas studies in other settings indicate that ADO supports mitochondrial redox homeostasis through proline metabolism.³² Notably, proline metabolism itself exerts context- and tumor-type-dependent (so-called Janus-like) effects in cancer,³³ and the metabolic programs of different tumors may therefore determine whether ADO activity is growth-promoting or dispensable. These observations parallel the emerging concept that oxygen-sensing pathways can exert dual, context-dependent functions in malignancy: HIF prolyl hydroxylases (EGLN/PHD) and HIF-1α have been reported to act as tumor promoters in some cancer types and as tumor suppressors in others.^34–36 By analogy, whether high ADO expression translates into an adverse or a favorable outcome may depend on the interaction between ADO-driven metabolic and N-degron signaling and the local metabolic and immune environment of each tumor. The cancer-type-specific prognostic value of ADO should therefore be interpreted cautiously.

Beyond prognostication, ADO was also associated with genomic instability and epigenetic reprogramming. Correlations with tumor mutation burden (TMB) and microsatellite instability (MSI) were cancer type–specific, suggesting differential interactions with DNA repair mechanisms. More strikingly, ADO correlated with methylation regulators and RNA modification genes (m6A, m5C, m1A), reinforcing its role in post-transcriptional and chromatin-level regulation. RNA modifications are increasingly recognized as critical in cancers including HCC,^37,38 and ADO’s alignment with these pathways suggests oxygen sensing may directly modulate epigenetic plasticity.

Pathway enrichment revealed strong associations with G2M checkpoint, MYC targets, and TGF-β signaling - pathways central to carcinogenesis.^39–41 Our pan-cancer analyses revealed that ADO expression was negatively correlated with ROS-responsive genes in most cancers, suggesting that high ADO levels buffer tumor cells against oxidative stress. This observation aligns with recent mechanistic work showing that ADO depletion leads to toxic polyamine accumulation driven by its substrate cysteamine, causing mitochondrial hyperactivity and ROS overproduction, culminating in cell toxicity.³² Together, these findings indicate that ADO is essential for maintaining redox balance in cancer cells by restraining polyamine levels and proline catabolism, thereby limiting oxidative stress.

In parallel, ADO was linked to immune checkpoint expression, including CD276, ENTPD1, and HMGB1. These molecules are established mediators of immune escape and therapeutic resistance.^42–44 By upregulating these checkpoints, ADO appears to actively shape an immunosuppressive microenvironment, complementing its proliferative effects. This dual influence on tumor growth and immune evasion highlights ADO as a multifunctional oncogenic driver in HCC. Clinically, ADO-high tumors exhibited poor survival following immune checkpoint blockade. Elevated TIDE scores in ADO-high HCC patients further predicted reduced immunotherapy benefit. These results extend prior observations that hypoxia undermines checkpoint efficacy^45,46 and identify ADO as a non-HIF hypoxia regulator that contributes to immune resistance. This highlights the potential value of incorporating ADO expression into immunotherapy decision-making and suggests that therapeutic targeting of ADO may sensitize tumors to checkpoint inhibition.

This study has several limitations. Analyses relied on retrospective datasets, which, although validated across multiple platforms, are subject to heterogeneity. Functional validation was limited to HepG2 and Hepa1-6 cell lines and subcutaneous xenograft models, which may not fully capture the complexity of human HCC. Mechanistically, while we indicated ERK activation and checkpoint regulation downstream of ADO, the complete substrate repertoire remains undefined. The lack of ADO-targeted compounds currently limits translational progress.

Conclusion

In conclusion, this study links ADO, a hypoxia-linked oxygen sensor, with oncogenic and immunomodulatory roles in cancers especially in HCC. By acting outside the canonical HIF pathway, ADO expands the current understanding of hypoxia adaptation in liver cancer. Our findings indicate that ADO may serve as a preliminary biomarker and a preclinical candidate warranting further investigation, offering new opportunities to improve outcomes in hypoxia-driven malignancies.

Availability of Data and Materials

All datasets analyzed in this study are publicly available. TCGA and GTEx data can be accessed via the USC Xena browser (https://xenabrowser.net). ICGC data are available from the ICGC Data Portal (https://dcc.icgc.org). GEO datasets used in this study (GSE124535, GSE135631, GSE144269, GSE14520, GSE169289, GSE184733, GSE214846, GSE29721, GSE39791, GSE45267, GSE45436, GSE57957, GSE67764, GSE95698) can be obtained from NCBI GEO (https://www.ncbi.nlm.nih.gov/geo/). Proteomic data for HCC were retrieved from the CNHPP liver data portal (http://liver.cnhpp.ncpsb.org/). All other data supporting the findings of this study are available from the corresponding author upon reasonable request.

Funding Statement

This study was supported by the Zhejiang Provincial Medical and Health Science and Technology Plan (2023RC239) and The Construction Fund of Key Medical Disciplines of Hangzhou (2025HZGF06).

Ethics Approval and Consent to Participate

This study was exempt from institutional ethics review under items 1 and 2 of Article 32 of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects (effective February 18, 2023, China), because the research was based exclusively on publicly available databases including TCGA and GEO databases. The animal experimental procedures involving mice were approved by the Institutional Animal Care and Use Committee (IACUC) of Zhejiang Center of Laboratory Animals (ZJCLA) (Approval No. ZJCLA-IACUC-20040184). All procedures were performed in accordance with the National Standard of the People’s Republic of China GB/T 35,892-2018, Laboratory animal - Guideline for ethical review of animal welfare. The study is reported in accordance with ARRIVE guidelines.

Disclosure

The authors report no conflicts of interest in this work.