Radiation-Enhanced CD24 Membrane Trafficking via GPI Anchoring Mediates Antitumor Immune Evasion

Radiation enhances CD24 membrane trafficking by regulating ANAPC5/GPAA1-mediated GPI anchoring to drive cancer immune evasion, which can be circumvented by targeting CD24 to potentiate the local and abscopal antitumor effects of radiotherapy.

Abstract

Radiotherapy plays a central role in cancer treatment, and the immunostimulatory effects of radiotherapy have been increasingly recognized. A better understanding of the mechanisms underlying postradiation immune escape is needed to help overcome radioresistance. In this study, we identified that irradiated tumor cells exploit the ANAPC5/GPAA1 axis to elevate surface expression of the “do not eat me” signal CD24, inducing phagocytosis resistance and immune evasion. Mechanistically, radiation inhibited the APC/C complex, reducing ANAPC5-mediated ubiquitination of GPAA1, a catalytic subunit of glycosylphosphatidylinositol (GPI) transamidase. The subsequent accumulation of GPAA1 facilitated GPI anchoring, thereby enhancing CD24 membrane localization. Accordingly, ablation of GPAA1 or CD24 significantly potentiated the local antitumor effects of radiotherapy across multiple preclinical models, dependent on T cells and macrophages. Notably, CD24 deficiency also stimulated abscopal effects, suppressing the growth of nonirradiated tumors. Overall, this study elucidates a mechanism of radiotherapy-mediated upregulation of the innate immune checkpoint CD24, offering perspectives on radiation-induced immune escape and presenting a strategy to improve radiotherapy efficacy.

Significance:

Radiation enhances CD24 membrane trafficking by regulating ANAPC5/GPAA1-mediated GPI anchoring to drive cancer immune evasion, which can be circumvented by targeting CD24 to potentiate the local and abscopal antitumor effects of radiotherapy.

Introduction

Radiotherapy has a long-established role in cancer treatment, with approximately 50% of patients with cancer receiving radiation during their treatment course (1). Although its direct cytotoxicity is well characterized, emerging evidence highlights radiotherapy’s role as “in situ tumor vaccine” (2). This immunostimulatory capacity arises from radiation-induced immunogenic cell death (ICD), which releases damage-associated molecular patterns (DAMP) like calreticulin (CRT; an “eat me” signal) and extracellular ATP (a “find me” signal; ref. 3). Engulfment of irradiated tumor cells by antigen-presenting cells (APC) then promotes APC maturation and tumor-associated antigen (TAA) cross-presentation, priming adaptive T-cell responses (4). This cascade has the potential to induce systemic antitumor responses against nonirradiated tumor deposits, known as the abscopal effect (5). However, clinically significant abscopal responses remain exceedingly rare (6). Radiotherapy exerts direct cytotoxicity on immune cells and also recruits immunosuppressive subsets such as regulatory T cells (Treg), suggesting that radiotherapy exhibits duality in its immunomodulatory role (4). To enhance systemic antitumor responses, studies have explored combining radiation with immune checkpoint blockade targeting PD-1/PD-L1 or CTLA4 (7). Yet these attempts have yielded limited clinical benefit, suggesting unresolved immune evasion mechanisms beyond T-cell inhibitory pathways (8). A plausible explanation lies in insufficient antigen presentation, leading to inadequate T-cell activation (9).

APCs serve as critical bridge between innate and adaptive immunity (10). Among them, macrophages—the predominant immune population within solid tumors—execute dual antitumor functions: direct phagocytic elimination and antigen presentation (11). Hence, although macrophage responsiveness to radiation-upregulated “eat me” and “find me” signals can trigger antitumor immune cascade, their dysfunction abrogates the immunostimulatory effects and drives postirradiation immune evasion (12). However, tumor cells employ multiple mechanisms to evade macrophage surveillance, most notably through “do not eat me” signals such as CD47, PD-L1, β2M, and CD24 (13–16). Although CD47 emerged as the prototypical innate immune checkpoint, its ubiquitous expression on normal cells and limited efficacy in solid tumors constrain clinical utility (17, 18). CD24, a glycosylphosphatidylinositol (GPI)-anchored protein (GPI-AP), has recently been characterized as a tumor-restricted “do not eat me” signal that interacts with macrophage Siglec-10 receptor to inhibit phagocytosis (19). The immune evasion following radiotherapy suggests that CD24 represents a rational target for radioimmunotherapy combinations and prompts critical inquiry into the potential regulation of CD24 by radiation. Elucidating postirradiation dynamics of this innate immune checkpoint may provide theoretical foundation for combination therapies and holds promise to unveil novel therapeutic vulnerabilities.

Here, we interrogate radiation-induced immune evasion through the lens of innate immunity, redefining CD24 as a radiation-responsive “do not eat me” signal. Radiation enhances CD24 membrane trafficking through GPI anchoring governed by the ANAPC5/GPAA1 axis, thereby mediating phagocytosis evasion. GPAA1 or CD24 inhibition potentiated radiotherapy-induced local antitumor efficacy across multiple murine models. CD24 inhibition–radiotherapy (CD24i-RT) combination further elicited abscopal effects, achieving control of untreated distant tumors. Our findings illuminate the molecular basis of postirradiation immune escape and provide an alternative strategy to overcome radioresistance.

Materials and Methods

Cell lines and cell culture

The human pancreatic ductal adenocarcinoma (PDAC) cell lines PANC-1 (RRID: CVCL_0480) and BxPC-3 (RRID: CVCL_0186), human non–small cell lung cancer cell lines NCI-H460 (RRID: CVCL_0459) and NCI-H1299 (RRID: CVCL_0060), human monocytic cell line THP-1 (RRID: CVCL_0006), murine hepatocellular carcinoma cell line Hepa1-6 (RRID: CVCL_0327), and murine lung carcinoma cell line Lewis (RRID: CVCL_4358) were obtained from the ATCC. The murine PDAC cell line derived from late-stage primary tumors from an autochthonous mouse model of Kras^G12D- and Trp53^R172H-mutated PDAC (KPC) was generously provided by Dipanjan Chowdhury’s laboratory. NCI-H460 (RRID: CVCL_0459), NCI-H1299 (RRID: CVCL_0060), and human monocytic cell line THP-1 (RRID: CVCL_0006) were cultured in RPMI-1640 supplemented with 10% FBS and 1% penicillin/streptomycin. PANC-1 (RRID: CVCL_0480), BxPC-3 (RRID: CVCL_0186), Hepa1-6 (RRID: CVCL_0327), Lewis lung cancer (LLC; RRID: CVCL_4358), and KPC were cultured in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin. All cell lines were authenticated using short tandem repeat profiling and routinely tested for Mycoplasma contamination using the PCR-b MycoAlert Mycoplasma Detection Kit. The latest Mycoplasma testing was performed in May 2025, and all cell lines tested negative. Cells were used within 10 passages after thawing and were maintained for no longer than 4 weeks in culture before being discarded.

Radiation

For in vitro irradiation, cells were exposed to the indicated dose using a linear accelerator (Varian UNIQUE SN2236; 6 MV beam; 6 Gy/minutes). For in vivo radiotherapy, radiation was applied locally at a dose of 8 Gy to the tumor on the leg of mice using a linear accelerator (Varian UNIQUE SN2236; 6 MV beam; 6 Gy/minutes). A 1-cm-thick medical polymethyl methacrylate bolus was used to shift the peak of the 6 MV photon beam’s buildup region to the tumor center, ensuring uniform prescribed dose. Calibration and dosimetry evaluation were performed daily using an ionization chamber connected to an electrometer system. Micro thermoluminescent dosimeters were placed at the center and periphery of the irradiation field to validate the radiation dosimetry.

Animal models and evaluation of therapeutic effects

All animal studies were conducted in accordance with protocols approved by the Hubei Provincial Animal Care and Use Committee and in line with the experimental guidelines of the Animal Experimentation Ethics Committee of Huazhong University of Science and Technology. All animals were maintained in individually ventilated cages in the cancer center of Wuhan union hospital. Six- to eight-week-old female C57BL/6J mice (RRID: IMSR_JAX:000664) were purchased from Wuhan Moubaili Biotechnology Co. Mice were anesthetized with 1% pentobarbital sodium before all operations. Tumor cells (1 × 10⁶/100 μL) were inoculated subcutaneously to establish tumor models. Mice were randomly divided into different groups. Mice in the radiation treatment group were subjected to 8 Gy × 3 of radiation when their tumors reached an average volume of 50 to 100 mm³. The length (L) and width (W) of tumors were measured at intervals of 2 days, and tumor volume (V) was calculated according to the formula V = (L × W²)/2. All measurements were conducted by an experimenter blinded to both the injection conditions and the experimental cohorts. Mice were euthanized at endpoint (tumor volume ≥1,000 mm³).

Single-cell RNA sequencing analysis

The single-cell RNA sequencing (scRNA-seq) data of three primary pancreatic tumor samples were obtained from GSE281288. The scRNA-seq data of three primary non–small cell lung cancer samples were obtained from GSE274934. To control data quality, cells with more than 20% mitochondrial genes or unique molecular identifier (UMI) counts outside 100 to 150,000 or gene counts beyond 200 to 10,000 were excluded. The UMI counts of each cell were then log-normalized to transcripts per million–like values for data normalization. During feature selection, genes detected in less than 0.1% of cells per batch were filtered. For dimensional reduction and clustering, genes with quantile-normalized variance more than 0.5 and mean expression between 0.0125 and 3 were selected for principal component (PC) analysis. Significant PCs were identified using Seurat (v2.3.4) for cell clustering. Cell clusters were visualized via Uniform Manifold Approximation and Projection (UMAP) and annotated based on canonical marker expression. Expression patterns of predefined genes of interest across annotated cell subpopulations were visualized on UMAP plots. To quantify expression differences for each target gene between cell clusters, the Wilcoxon rank-sum test was applied. P values were adjusted for multiple testing using the Benjamini–Hochberg method, with adjusted P less than 0.05 defined as statistically significant. All analyses were conducted in R (v4.5).

Immunofluorescence staining

Cells were seeded in 24-well plates loaded with cell glass slides. After indicated treatment, slides were fixed with 4% paraformaldehyde at 4°C for 30 minutes, washed three times with 1× PBS, and then blocked with 10% BSA for 1 hour at room temperature. Slides were subsequently incubated with primary antibodies at 4°C overnight, followed by fluorescence-conjugated secondary antibody incubation for 1 hour at 4°C. Stained cells were visualized using a laser scanning confocal microscope.

In vitro phagocytosis assays

For flow cytometry–based phagocytosis assays, carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled tumor cells (target cells) and macrophages (effector cells) were cocultured at a 1:1 ratio in humidified incubators at 37°C with 5% CO₂. After 4 hours of coculture, cells were harvested and stained with F4/80 (BioLegend, cat. #123115, RRID: AB_893493) or CD11b (BioLegend, cat. #101211, RRID: AB_312794). Phagocytosis was defined as the percentage of F4/80⁺ CFSE⁺ or CD11b⁺ CFSE⁺ cells among F4/80⁺ or CD11b⁺ macrophages.

For immunofluorescence-based phagocytosis assays, GFP-expressing tumor cells were employed as target cells. Phagocytosis was defined as colocalization of F4/80⁺ or CD11b⁺ macrophages (red) and GFP-expressing tumor (green).

In vivo phagocytosis assays

Subcutaneous tumor models were established using GFP-expressing tumor cells. After indicated treatments, tumors were harvested, digested with collagenase and hyaluronidase for 1 hour at 37°C, and then subjected to red blood cell lysis to obtain single-cell suspensions. Cells were stained with the Zombie Violet Fixable Viability Kit, followed by antibodies against CD45 (BioLegend, cat. #103131, RRID: AB_893344), CD11b (BioLegend, cat. #101211, RRID: AB_312794), and F4/80 (BioLegend, cat. #111603, RRID: AB_3082990) according to the manufacturer’s concentrations for 30 minutes in 4°C. Phagocytosis was assessed via flow cytometry and defined as the percentage of CD11b⁺ F4/80⁺ GFP⁺ cells among CD11b⁺ F4/80⁺ macrophages.

qRT-PCR

Total RNA was extracted using TRIzol reagent then quantified using Thermo Fisher Scientific’s NanoDrop ND-1000. According to the manufacturer’s protocols, 1 μg of total RNA was reverse-transcribed into cDNA with HiScript III RT SuperMix. qRT-PCR was performed on a StepOnePlus system using ChamQ SYBR qPCR Master Mix. β-Actin was used to normalize gene expression.

Western blotting and coimmunoprecipitation

Total protein was extracted using RIPA buffer containing 1% protease and phosphatase inhibitors. Membrane protein was extracted using a Mem-PER Plus kit. Protein was then quantified using a Bicinchoninic Acid Assay Kit.

For Western blot analysis, proteins with 5 × loading buffer were boiled at 100°C for 10 minutes, separated by SDS-PAGE, and transferred to polyvinylidene difluoride membranes. Membranes were blocked with 5% nonfat milk for 1 hour at room temperature, incubated with antibodies against CD24 (Abways, cat. #DY1303, RRID: AB_3073211), GPAA1 (Proteintech, cat. #10104-1-AP, RRID: AB_2263708), and ANAPC5 (Proteintech, cat. #67348-1-Ig, RRID: AB_2882606) primary antibodies at 4°C overnight, washed with TBST, and then incubated with horseradish peroxidase–conjugated secondary antibodies for 1 hour at room temperature. Signals were visualized using enhanced chemiluminescence reagent.

For coimmunoprecipitation assay, cells were lysed with immunoprecipitation (IP) lysis solution containing 1% protease and phosphatase inhibitors. Lysates were incubated with Protein A/G agarose beads and indicated primary antibodies or IgG control at 4°C overnight. Beads were washed five times with NETN buffer, then eluted in 1× loading buffer, and boiled at 100°C for 10 minutes. Samples were used for subsequent Western blot analysis.

LC/MS-MS analysis

Cells were lysed with IP lysis solution. Lysates were incubated with Protein A/G agarose beads and anti-GPAA1 antibodies (Proteintech, cat. #10104-1-AP, RRID: AB_2263708) or IgG control at 4°C overnight. LC/MS-MS analysis was conducted using a Thermo Fisher Scientific Ultimate 3000 RSLC system paired with a Q Exactive Plus high-resolution mass spectrometer, provided by SpecAlly Life Technology Co., Ltd. Data retrieval was performed using MaxQuant software (RRID: SCR_014485) and the Andromeda algorithm. The UniProt human proteome database (RRID: SCR_002380) served as the reference, and proteins and peptides were filtered using an FDR of 1%.

AlphaFold3-based computational structure analysis

The anaphase-promoting complex/cyclosome (APC/C) structure is publicly available in the Protein Data Bank (PDB) under the identifier PDB ID: 4UI9. GPAA1 has been structurally characterized and is documented in PDB ID: 7W72.

For pairwise protein–protein docking via ClusPro, ten conformations (GPAA1 as the receptor protein and APC/C as the ligand protein) were generated and then screened based on interfacial features, with the most promising conformation selected for further analysis. All selected docking conformations were evaluated for geometric plausibility and spatial accessibility. CDC16, CDC27, ANAPC7, ANAPC1, and ANAPC5 were identified as promising binding subunits.

For electrostatic surface analysis, electrostatic properties of the interaction interfaces were analyzed using PyMOL (RRID: SCR_000305) with the APBS plugin. Interaction interfaces between GPAA1 and CDC16, ANAPC7, and ANAPC5 demonstrated good electrostatic complementarity.

For structure prediction, AlphaFold3 was utilized to further predict the complex structures of GPAA1 and CDC16, ANAPC7, and ANAPC5 subunits. Each prediction task was performed using 10 independent runs initialized with different random seeds to enhance conformational diversity. All predicted conformations were evaluated for geometric plausibility and spatial accessibility, and the best conformations were selected for further comparison. The interchain predicted TM-score metric parameter from AlphaFold3 was employed to estimate the global structural similarity between the predicted interface of two protein chains and the true interface. ANAPC5 has the highest interface confidence score and the most accurate single-chain prediction.

Flow cytometry analysis of tumor immune environment

Tumors were harvested, digested with collagenase and hyaluronidase for 1 hour at 37°C, and then subjected to red blood cell lysis to obtain single-cell suspensions. To exclude dead cells, single-cell suspensions were stained with the Zombie Violet Fixable Viability Kit. To analyze macrophages, cells were stained with antibodies against CD45 (BioLegend, cat. #157213, RRID: AB_2894427), CD11b (BioLegend, cat. #101227, RRID: AB_893233), F4/80 (BioLegend, cat. #123135, RRID: AB_2562622), CD80 (BioLegend, cat. #104713, RRID: AB_313134), CD86 (BioLegend, cat. #105123, RRID: AB_2892270), and PD-L1 (BioLegend, cat. #124307, RRID: AB_2073557) according to the manufacturer’s concentrations for 30 minutes at 4°C. To analyze T cells, a portion of cell suspensions was first stimulated with phorbol 12-myristate 13-acetate (100 ng/mL), ionomycin (100 ng/mL), and monensin (1 μg/mL) for 5 hours at 37°C with 5% CO₂ and then stained with antibodies against CD45 (BioLegend, cat. #157213, RRID: AB_2894427), CD3 (BioLegend, cat. #100245, RRID: AB_2565882), CD4 (BioLegend, cat. #100451, RRID: AB_2564591), and CD8 (BioLegend, cat. #100751, RRID: AB_2561389). Subsequently, cells were fixed and permeabilized and then stained with antibodies against IFNγ (BioLegend, cat. #505829, RRID: AB_10897937), Grzmb (BioLegend, cat. #372207, RRID: AB_2687031), and FoxP3 (BD Pharmingen, cat. #567462, RRID: AB_2916606) according to the manufacturer’s concentrations for 30 minutes at 4°C.

Macrophage and T-cell depletion

For the macrophage depletion study, clodronate liposomes were injected intraperitoneally with the dose at 150 mL per mouse every 4 days. For the CD8⁺ T-cell depletion study, mouse anti-CD8 antibodies were diluted in PBS and intraperitoneally injected with the dose at 100 mg per mouse every 4 days. The depletion efficacy was validated by flow cytometry.

Statistical analysis

Statistical analyses were carried out using GraphPad Prism v 10.0 (RRID: SCR_002798). All quantitative data are presented as mean ± SEM. For the comparison of two groups, a two-tailed Student t test was performed. For comparisons of three or more groups, one-way ANOVA with a Tukey multiple comparisons test was performed. Two-way ANOVA with a Tukey multiple comparisons test was conducted for tumor growth curves. The log-rank test was conducted for Kaplan–Meier survival analysis. The Pearson test was conducted to evaluate correlations. P < 0.05 was considered statistically significant. NS, not significant.

Results

Radiation fails to enhance macrophage phagocytosis despite inducing immunogenic signals

To systematically investigate postirradiation immune escape mechanisms, we analyzed the scRNA-seq data of murine LLC subcutaneous tumor models subjected to radiation (IR, 8 Gy × 3) and control (Ctrl) treatment (Fig. 1A). Tumor cell clusters exhibited significant enrichment of transcripts associated with “eat me” signals (e.g., Calr and Fas), “find me” signals (e.g., Cx3cl1), ICD, and antigen-processing pathways (Fig. 1B). These signals are critical for recruiting immune effectors such as macrophages to recognize and clear damaged cells. Cross-tumor validation confirmed upregulation of CRT, a canonical prophagocytic signal, in both human and murine lung and pancreatic cancer cells following radiation exposure (Fig. 1C–E; Supplementary Fig. S1A and S1B). Flow cytometric analysis of CRT in murine models further corroborated these findings in vivo (Fig. 1F; Supplementary Fig. S1C). Concurrently, radiation induced increased release of HMGB1 and ATP, functioning as proinflammatory DAMPs and “find me” signal, across multiple tumor cell lines (Fig. 1G–J; Supplementary Fig. S1D–S1G). Collectively, these results indicate that radiotherapy can effectively promote the release of immunogenic signals.

Figure 1.

Radiation fails to enhance macrophage phagocytosis despite inducing immunogenic signals. A, Schematic illustration of scRNA-seq in animal models. t-SNE, t-distributed stochastic neighbor embedding. B, scRNA-seq analysis of immunogenic markers on tumor cells from control (Ctrl) or irradiated LLC tumors (n = 3). C and D, Flow cytometry histograms and mean fluorescence intensity (MFI) of CRT on control or irradiated H460 (C) and PANC-1 (D) cells (n = 3). Normalized to the control group. E, Immunofluorescence images of CRT (red) on control or irradiated H460 cells. F, Flow cytometry histograms and mean fluorescence intensity of tumor cells CRT from control or irradiated KPC tumor models (n = 6). Normalized to the control group. G and H, Culture medium HMGB1 levels from control or irradiated H460 (G) and PANC-1 (H) cells measured by ELISA (n = 3). I and J, Extracellular ATP release from control or irradiated H460 (I) and PANC-1 (J) cells measured by ELISA (n = 3). K, scRNA-seq–derived M1 signature scores of tumor-associated macrophages from irradiated or control LLC tumors (n = 3). L, Flow cytometry histograms and quantitation of control or irradiated H460 cell phagocytosis by THP-1 macrophages (n = 3). M, Immunofluorescence images and quantitation of control or irradiated H460 cell (green) phagocytosis by THP-1 macrophages (red), with arrows indicating phagocytic events (n = 6). Scale bars, 50 μm. N, Flow cytometry dotplot and quantitation of tumor cell phagocytosis by tumor-associated macrophages from control or irradiated LLC tumor (n = 6). IR, 8 Gy irradiation. All data are presented as mean ± SEM and compared using a two-tailed Student t test. NS, not significant. A, Created in BioRender. Kong, L. (2026) https://BioRender.com/rwdzfwn.

We next interrogated whether immune cells could effectively respond to these radiation-induced immunogenic signals. Macrophage-mediated tumor cell phagocytosis and antigen presentation are central to activate antitumor immunity. However, gene set variation analysis of macrophage clusters from the scRNA-seq data revealed no significant difference in the antitumor M1 phenotype signature score after irradiation (Fig. 1K). To evaluate the functional state of macrophages, we cocultured human THP-1 monocyte-derived macrophages with CFSE-labeled H460 cells. Flow cytometry– and immunofluorescence-based quantitation demonstrated comparable phagocytic clearance of irradiated versus control tumor cells (Fig. 1L and M). Consistent results were obtained in phagocytosis assay using murine bone marrow–derived macrophages (BMDM; Supplementary Fig. S1H and S1I). In line with in vitro observations, in vivo tumor cell uptake remained equivalent between control and radiated tumor cells (Fig. 1N; Supplementary Fig. S1J). These collective findings demonstrate that radiation fails to enhance macrophage phagocytic capacity despite inducing tumor-intrinsic immunogenic signals. This paradox implies the activation of counteracting mechanisms following radiation, thereby impeding the overall efficacy.

Radiation enhances “do not eat me” signal CD24 on tumor cells

Although intrinsic “eat me” signals facilitate immune surveillance, cancer cells often overexpress “do not eat me” signals to evade macrophage-mediated clearance (20). To identify clinically relevant innate immune checkpoints, we interrogated the cellular expression profiles of “do not eat me” signals using human pancreatic cancer and non–small cell lung cancer scRNA-seq datasets. This systematic comparison revealed CD24 as a prime therapeutic target as it exhibited greater tumor specificity than CD47, PD-L1, or B2M (Fig. 2A; Supplementary Fig. S2A–S2H). We next evaluated CD24 surface level across human and murine tumor cell lines. Immunofluorescence and flow cytometry analyses revealed significant upregulation of surface CD24 24 hours after irradiation (Fig. 2B–D; Supplementary Fig. S3A–S3D). Consistent with in vitro observations, in vivo flow cytometric profiling of irradiated KPC subcutaneous tumors demonstrated elevated surface CD24 level compared with nonirradiated controls (Fig. 2E; Supplementary Fig. S3E). Time-course and dose–response experiments further confirmed that CD24 upregulation was dependent on radiation dose and persisted for at least 4 days after irradiation (Fig. 2F; Supplementary Fig. S3F and S3G). These results redefine CD24 as a radiation-responsive innate immune checkpoint.

Figure 2.

Radiation enhances tumor cell “do not eat me” signal CD24. A, UMAP plots of pancreatic adenocarcinoma with 10 clusters, NCBI Sequence Read Archive: GSE281288. CD24 expression overlaid onto UMAP. B, Immunofluorescence images of CD24 (red) on control or irradiated H460 cells. Scale bars, 25 and 5 µm. C and D, Flow cytometry histograms and mean fluorescence intensity (MFI) of CD24 surface expression on control or irradiated H460 (C) and PANC-1 (D) cells (n = 3). E, Tumor cell CD24 mean fluorescence intensity from control or irradiated KPC tumor models (n = 6). F, Flow cytometry histograms and mean fluorescence intensity of CD24 surface expression on H460 cells 48 hours after 2–18 Gy of radiation (n = 3). IR, 8 Gy irradiation. All data are presented as mean ± SEM and normalized to the control group. A two-tailed Student t test was performed for C–E. One-way ANOVA with a Tukey multiple comparisons test was performed for F. NS, not significant.

Targeting CD24 combined with radiation facilitates macrophage phagocytosis and activation

To investigate the role of enhanced CD24 signal following radiation, we first employed siRNA-mediated CD24 knockdown in H460 and PANC-1 cells, achieving more than 80% silencing efficiency, which was confirmed by flow cytometry (Fig. 3A and B). Flow cytometry– and immunofluorescence-based phagocytosis assays revealed that CD24 inhibition increased macrophage-mediated tumor cell uptake, with combined CD24 silencing and radiation further enhancing phagocytic clearance compared with either treatment alone (Fig. 3C–E). To confirm these findings, we generated stable CD24-knockout (CD24 KO) cell lines in murine LLC, PDAC (KPC), and hepatocellular carcinoma (Hepa1-6) cell lines via CRISPR/Cas9 (Supplementary Fig. S4A–S4C). CD24 ablation in combination with radiation significantly enhanced BMDM-mediated phagocytosis (Supplementary Fig. S4D–S4G).

Figure 3.

CD24 inhibition combined with radiation promotes macrophages phagocytosis and activation. A and B, Flow cytometry histograms of CD24 surface expression on H460 (A) and PANC-1 (B) cells transfected with CD24-targeting siRNA (siCD24), negative-control siRNA (siNC), or isotype control. C and D, Flow cytometry histograms and quantitation of H460 (C) and PANC-1 (D) cell phagocytosis by THP-1 macrophages, with the indicated treatments (n = 3). E, Immunofluorescence images and quantitation of H460 cell phagocytosis by THP-1 macrophages (red), with the indicated treatments (green). Arrows, phagocytic events (n = 6). Scale bars, 50 μm. F, qRT-PCR analysis of Siglec-10 in THP-1 macrophages transfected with siSiglec-10 or siNC (n = 3). G, Flow cytometry histograms and quantitation of irradiated H460 cell phagocytosis by THP-1 macrophages, with the indicated treatments (n = 3). H–J, Flow cytometry histograms and mean fluorescence intensity (MFI) of CD80 (H), CD86 (I), and PD-L1 (J) on BMDMs cocultured with tumor cells, with the indicated treatments (n = 3). IR, 8 Gy irradiation. All data are presented as mean ± SEM. One-way ANOVA with a Tukey multiple comparisons test was performed for C–E and G–J. A two-tailed Student t test was performed for F. NS, not significant.

CD24 protects cancer cells from phagocytosis through its interaction with the inhibitory receptor Siglec-10 on macrophages. Silencing Siglec-10enhanced the phagocytic uptake of irradiated tumor cells, while silencing Siglec-9, another member of sialic acid–binding Ig-like lectins family, failed to have such an effect. Furthermore, when combined with siCD24, this enhancement could not be further augmented, indicating that the radiation-upregulated CD24 signal also operates through the classical CD24–Siglec-10 axis (Fig. 3F and G; Supplementary Fig. S4H and S4I).

Beyond phagocytosis, macrophages play a critical role in antigen presentation and T-cell activation. We next investigated the expression of costimulatory molecules on BMDMs from phagocytosis assay. BMDMs cocultured with irradiated CD24-KO tumor cells revealed significantly elevated expression of CD80, CD86, and PD-L1, indicative of enhanced maturation and activation (Fig. 3H–J). Collectively, these findings suggest that the strategy of targeting CD24 in combination with radiotherapy can reverse phagocytic resistance through canonical CD24–Siglec-10 interactions and promote macrophage functional maturation and activation.

Radiation enhances CD24 membrane trafficking via GPI anchoring

To characterize CD24 dynamics following radiation, we initially assessed total CD24 protein and transcript levels after irradiation. Intriguingly, Western blot and qRT-PCR analyses failed to detect alterations in CD24 expression (Fig. 4A and B; Supplementary Fig. S5A–S5C), suggesting that radiation enhances CD24 membrane trafficking rather than biosynthesis. CD24 is a highly glycosylated protein that requires posttranslational glycosylation modification for proper subcellular localization and functional activity. The preproproteins of GPI-APs have a C-terminal signal peptide that can be recognized, cleaved, and replaced with a preassembled GPI by GPI transamidase, leading to the generation of nascent GPI-APs (Fig. 4C). These newly formed nascent GPI-APs then undergo a series of remodeling processes before being transported to and anchored on lipid rafts within the cell membrane (21). To explore GPI anchoring critical for CD24 membrane trafficking and elucidate mechanism behind radiation-induced CD24 regulation, we established H460 cell lines stably overexpressing C-terminal Flag-tagged CD24 (H460 OE; Supplementary Fig. S5D and S5E). Western blot analysis of isolated membrane and cytoplasmic fractions revealed that radiation increased the abundance of membrane GPI-anchored CD24, as detected by anti-CD24 antibody, while decreasing the levels of cytoplasmic uncleaved CD24 preproprotein, as shown by anti-Flag antibody (Fig. 4D). These findings indicate enhanced GPI-dependent membrane localization of CD24.

Figure 4.

Radiation enhances CD24 membrane trafficking via GPI anchoring. A and B, Western blot analysis of CD24 in control or irradiated H460 (A) and PANC-1 (B) whole-cell lysate. C, Schematic representation of GPI anchoring. D, Western blot analysis of CD24 in control or irradiated H460 cell membrane and cytosolic lysate. E, Proteomic analysis heatmap of differential proteins in control or irradiated H1299. F and G, Western blot analysis of GPAA1 in control or irradiated H460 (F) and PANC-1 (G) whole-cell lysate. H, Western blot analysis of GPAA1 and CD24 in H460 cells transfected with siGPAA1 or siNC. I, Flow cytometry histograms of CD24 surface expression on H460 cells, with the indicated treatment or isotype control (n = 3). MFI, mean fluorescence intensity. J, Flow cytometry histograms and quantitation of H460 cell phagocytosis by THP-1 macrophages, with the indicated treatments (n = 3). IR, 8 Gy irradiation. All data are presented as mean ± SEM and compared using one-way ANOVA with a Tukey multiple comparisons test. NS, not significant.

We further screened for candidates including GPI transamidase subunits and other reported CD24 interactors in the proteomic profiles of irradiated and control H1299 lung cancer cells to identify regulators governing radiation-induced CD24 surface accumulation. GPAA1, a key subunit of GPI transamidase responsible for catalyzing the amide bond formation between preproteins and GPI-ethanolamine (22), exhibited significant upregulation after irradiation (Fig. 4E). Western blot confirmed GPAA1 protein elevation across tumor types (Fig. 4F and G; Supplementary Fig. S5F and S5G). siRNA-mediated GPAA1 knockdown reduced the surface abundance of GPI-APs, including CD24 and CD73, without altering total protein expression (Fig. 4H and I; Supplementary Fig. S5H). Radiation restored surface levels of GPI-APs, further demonstrating GPAA1’s essential role in radiation-enhanced GPI-anchoring. Functionally, genetic inhibition of GPAA1 phenocopied CD24 ablation, enhancing phagocytic uptake when combined with radiotherapy (Supplementary Fig. S5I and S5J). Notably, this enhancement was not further augmented upon combination with siCD24, indicating that radiotherapy potentiates CD24 signaling through GPAA1, thereby enabling tumor cells to evade phagocytosis (Fig. 4J). These results establish GPAA1-mediated GPI anchoring as the molecular bridge connecting radiation exposure to CD24-dependent phagocytic resistance.

Radiation impairs APC/C-mediated ubiquitination of GPAA1 at lysine 111

Building on our identification of GPAA1 as the critical regulator of GPI-anchored CD24, we next sought to dissect the molecular mechanisms underlying GPAA1 accumulation following radiation. To this end, we employed IP coupled with mass spectrometry (IP-MS) to profile GPAA1-interacting proteins (Fig. 5A). Gene Ontology enrichment analysis of the candidate interactors revealed significant enrichment in protein K11-linked ubiquitination pathways, implicating posttranslational regulation of GPAA1 via the ubiquitin–proteasome system (Fig. 5B). Cycloheximide chase assays demonstrated that GPAA1 half-life extended upon proteasome inhibition (MG132 treatment), confirming ubiquitin-dependent degradation of GPAA1 (Fig. 5C). Radiation did not alter the transcription of GPAA1 (Supplementary Fig. S6A). Instead, it abrogated the protein degradation emerging after transcription halt (Fig. 5D) by reducing GPAA1 ubiquitination, thereby stabilizing the protein (Fig. 5E). In-depth analysis of the IP-MS dataset revealed 11 of 41 candidate proteins as components of APC/C—a multi-subunit E3 ubiquitin ligase complex—thereby implicating APC/C as the primary E3 ligase governing GPAA1 ubiquitination (Supplementary Fig. S6B; ref. 23). APC/C E3 ligase activity inhibition (apcin treatment) recapitulated the radiation-induced effects, resulting in decreased GPAA1 ubiquitination level (Supplementary Fig. S6C) and increased GPAA1 protein level (Fig. 5F), and promoted membrane localization of GPI-anchored proteins (Fig. 5G; Supplementary Fig. S6D).

Figure 5.

Radiation disrupts APC/C-mediated ubiquitination of GPAA1 at K111. A, Schematic illustration of IP-MS approach. B, GO enrichment analysis (Biological Process) of GPAA1-interacting proteins predicted in IP-MS. C and D, Western blot analysis of cycloheximide (CHX) chase assay in H460 cells treated with MG132 (C) or radiation (D). E, Western blot analysis followed by IP evaluated the ubiquitination of GPAA1 in indicated treatment groups. F, Western blot analysis of GPAA1 in H460 cells treated with apcin. G, Flow cytometry histograms of CD24 surface expression on H460 cells treated with apcin or isotype control (n = 3). MFI, mean fluorescence intensity. H, Schematic illustration of screening GPAA1-binding APC/C subunits. I, IP analysis of the GPAA1–ANAPC5 interaction. J, Western blot analysis followed by IP evaluated GPAA1 ubiquitination in H460 cells transfected with siANAPC5 or siNC. K, Western blot analysis of ANAPC5 and GPAA1 in H460 cells transfected with siANAPC5 or siNC. L, Flow cytometry histograms of CD24 surface expression on H460 cells transfected with siANAPC5, siNC, or isotype control (n = 3). M, Western blot analysis followed by IP evaluated GPAA1 ubiquitination in 293T cells transfected with different mutants. IR, 8 Gy irradiation. All data are presented as mean ± SEM. One-way ANOVA with a Tukey multiple comparisons test was performed for G. A two-tailed Student t test was performed for L. WT, wild type.

To pinpoint the direct interactor of GPAA1, we integrated IP-MS data with computational analyses, including protein–protein docking, electrostatic surface analysis, and AlphaFold3 Multimer structure predictions. Through this comprehensive approach, we systematically refined the initial pool of 11 candidate subunits and identified ANAPC5 as the primary binding partner of GPAA1 (Fig. 5H; Supplementary Fig. S7A–S7C). Endogenous coimmunoprecipitation experiments supported the ANAPC5–GPAA1 interaction (Fig. 5I). siRNA-mediated knockdown of ANAPC5 reduced GPAA1 ubiquitination (Fig. 5J), stabilizing GPAA1 (Fig. 5K), thus enhancing GPI-APs membrane trafficking (Fig. 5L; Supplementary Fig. S8A). Notably, ANAPC5 protein levels declined rapidly following radiation (Supplementary Fig. S8B), suggesting that radiation mediates GPAA1 accumulation via APC/C disruption. AlphaFold3 structural predictions mapped three lysine residues (K82, K83, and K111) as potential ubiquitination sites on GPAA1. To validate their functional relevance, we conducted exogenous IP with site-specific mutants of GPAA1 (K82R, K83R, and K111R) and identified K111 as the critical ubiquitination site (Fig. 5M). Collectively, these findings delineate a radiation-responsive ubiquitination cascade in which radiation impeded ANAPC5-mediated ubiquitination of GPAA1 at K111, leading to enhanced CD24 membrane localization.

Targeting CD24 enhances local and abscopal tumor control of radiotherapy

Having elucidated CD24 dynamics after irradiation, we focused on exploiting this radiation-responsive immune checkpoint to tackle immune evasion following radiotherapy. Flow cytometric profiling demonstrated that CD24 was overexpressed in Hepa1-6 and KPC cell lines (Supplementary Fig. S9A). We therefore established murine hepatocellular carcinoma and pancreatic cancer tumor models by subcutaneously engrafting CD24-KO or control tumor cells into C57BL/6J mice, followed by radiotherapy (8 Gy × 3) once tumors reached a volume of 50 to 100 mm³ (Fig. 6A). In CD24^high Hepa1-6 tumor models, CD24 ablation alone delayed tumor progression, and its combination with radiotherapy achieved pronounced tumor growth inhibition and prolonged survival (Fig. 6B and C). Consistent therapeutic synergy was observed in immunologically “cold” pancreatic KPC tumor models, with CD24 deficiency enhancing radiation-induced tumor regression and reducing terminal tumor burden (Fig. 6D and E). CD24^low LLC models exhibited minimal response to CD24 deletion alone and limited but significant combinatorial benefit (Supplementary Fig. S9B and S9C).

Figure 6.

Targeting CD24 sensitizes tumor to radiotherapy and elicits abscopal effect in vivo. A, Schematic illustration of radiation treatment plan in animal models. B and C, Tumor growth curves (B) and Kaplan–Meier survival plot (C) of Hepa1-6 subcutaneous tumor models, with the indicated treatment (n = 8). D and E, Tumor growth curves (D) and tumor weight (E) of KPC subcutaneous tumor models, with the indicated treatment (n = 6). F and G, Tumor growth curves (F) and Kaplan–Meier survival plot (G) of KPC subcutaneous tumor models, with the indicated treatment (n = 8). H, Schematic illustration of abscopal effect evaluation in animal models. I–K, Tumor growth curves of in situ (I) and abscopal tumors (J) and Kaplan–Meier survival plot (K) of KPC subcutaneous tumor models depicted in H (n = 6). L, Schematic illustration of antibody treatment plan in animal models. M and N, Tumor growth curves (M) and Kaplan–Meier survival plot (N) of KPC subcutaneous tumor models depicted in L (n = 7). IR, 8 Gy × 3 irradiation. All data are presented as mean ± SEM. Two-way ANOVA with a Tukey multiple comparisons test was performed for B, D, F, I, J, and M. One-way ANOVA with a Tukey multiple comparisons test was performed for E. A log-rank test was performed for C, G, K, and N. NS, not significant. A, H, and L, Created in BioRender. Kong, L. (2026) https://BioRender.com/rwdzfwn.

Given GPAA1’s role in regulating postirradiation CD24 membrane trafficking, we examined the levels of GPAA1 and ANAPC5 across cell lines and generated GPAA1-knockdown (shGPAA1) cells in the GPAA1^high KPC cell line for subsequent functional assays (Supplementary Fig. S10A–S10D). Inhibition of GPAA1 reduced CD24 levels and overcame phagocytosis resistance after radiation in vitro, effects that were reversed by CD24 overexpression (shGPAA1 + oeCD24; Supplementary Fig. S10E and S10F). We went on to examine the therapeutic efficacy in vivo. Although GPAA1 knockdown alone did not impair tumor growth, its combination with radiotherapy suppressed tumor progression and extended survival, effects abrogated by CD24 overexpression (Fig. 6F and G). These results validate the critical role of GPAA1 in CD24-mediated postirradiation phagocytosis resistance both in vitro and in vivo, highlighting therapeutic potential of targeting CD24 via GPAA1.

Clinical reports have documented “abscopal effect” of radiotherapy, in which irradiation of one tumor mass results in antitumor responses on another nonirradiated tumor lesion (24). We next aimed to explore the efficacy of CD24i-RT combination against distal tumors. One week after primary tumor inoculation on the left flanks of mice (irradiated, sgCD24 or sgVector), secondary tumors were engrafted into the right flanks (nonirradiated, sgVector; Fig. 6H). Only mice bearing CD24-deficient primary tumors exhibited delayed growth of nonirradiated secondary lesions, indicating that the combination treatment led to abscopal tumor regression (Fig. 6I–K). This systemic response prompted us to further investigate whether targeting CD24 synergizes with immunotherapy in antitumor efficacy. In KPC models, we administered anti-CD24 antibody, anti–PD-L1 antibody, or their combination, either alone or in conjunction with radiotherapy (Fig. 6L). Targeting CD24 alone failed to effectively control tumor growth but significantly sensitized tumors to radiotherapy, providing pharmacologic validation for our conclusions. Additionally, compared with anti-CD24 or anti–PD-L1 monotherapy, the triple combination regimen of anti-CD24, anti–PD-L1, and radiotherapy exerted a significant synergistic antitumor effect, highlighting a translational direction for CD24-targeted therapies (Fig. 6M and N).

Enhanced macrophage and T-cell responses orchestrate CD24i-RT synergy

In light of our previous findings that targeting CD24 potentiates radiotherapy efficacy, we delved into its underlying mechanisms. Notably, in vitro cell viability assays revealed no intrinsic radiosensitivity enhancement in CD24-KO tumor cells (Supplementary Fig. S11A–S11C), suggesting that therapeutic efficacy required the tumor immune microenvironment (TIME). Given this, we performed flow cytometry–based immune profiling of KPC tumor tissues from previous treatment models (Supplementary Fig. S12A). CD24 deficiency significantly augmented macrophage infiltration (Fig. 7A). Moreover, macrophages in the combination treatment group exhibited an antitumor activated state characterized by high expression of CD80, CD86, and PD-L1 (Fig. 7B and C; Supplementary Fig. S12B). These findings align with our in vitro phagocytosis assays, collectively demonstrating that CD24i-RT combination improved macrophage functional activation.

Figure 7.

Macrophages and T cells are required for CD24i-RT combination efficacy. A–F, Flow cytometry analysis of macrophages and T-cell subsets in KPC tumor tissues with the indicated treatment (n = 6). MFI, mean fluorescence intensity. G–J, immunofluorescence images and quantitation of activated tumor-associated macrophages (G and H) and CD8⁺ T cells (I and J) in KPC tumor tissues with the indicated treatment (n = 6). Scale bars, 100 μm. K, Schematic illustration of immune cell depletion plan in animal models. L–N, Tumor growth curves (L), tumor image (M), and tumor weight (N) of KPC subcutaneous tumor models with the indicated treatment (n = 6). IR, 8 Gy × 3 irradiation. Clo, clodronate. All data are presented as mean ± SEM. One-way ANOVA with a Tukey multiple comparisons test was performed for A–F, H, J, and N. Two-way ANOVA with a Tukey multiple comparisons test was performed for L. NS, not significant. K, Created in BioRender. Kong, L. (2026) https://BioRender.com/rwdzfwn.

Abscopal effects are reported to rely on the adaptive immune system, particularly TAA cross-priming and cytotoxic CD8⁺ T cell activation (25). Guided by this understanding, we proceeded to examine the T-cell subsets within the TIME. Combination therapy substantially enhanced intratumoral T-cell infiltration (Fig. 7D), particularly elevating cytotoxic T lymphocytes (CD8⁺ IFNγ⁺ or GrzmB⁺), indicative of amplified CD8⁺ T-cell effector capacity (Fig. 7E and F). In addition, we observed reduction of immunosuppressive Tregs (CD4⁺ FoxP3⁺) in the combination treatment group (Supplementary Fig. S12C), whereas Th1 cell (CD4⁺ IFNγ⁺) proportion and CD4⁺/CD8⁺ ratio remained unchanged (Supplementary Fig. S12D and S12E). Multispectral immunofluorescence analysis validated these observations, revealing superior macrophage and CD8⁺ T cell infiltration, along with enhanced effector activity in combination-treated tumors (Fig. 7G–J; Supplementary Fig. S12F and S12G), suggesting further activation of the adaptive immunity.

To explore whether the efficacy of CD24i-RT combination relies on therapy-responsive macrophages and CD8⁺ T cells, we depleted these cell populations in irradiated tumor models using clodronate liposomes or anti-CD8 antibodies (Fig. 7K; Supplementary Fig. S13A and S13B). Depletion of either subset substantially compromised the therapeutic efficacy of radiotherapy in CD24-deficient models, indicating that both immune cell subsets contribute to the CD24i-RT approach. Notably, depletion of CD8⁺ T cells alone was insufficient to abrogate the benefits of the combination treatment, whereas combined depletion of both subsets or depletion of macrophages alone fully reversed this antitumor effect, highlighting that macrophages are essential for the CD24i-RT regimen (Fig. 7L–N).

Low CD24 expression predicts favorable TIME and improved clinical outcomes

Building upon preclinical evidence linking CD24 to immunosuppressive microenvironments, we sought to validate its clinical relevance through multi-cohort analyses. CD24 can act as a dominant innate immune checkpoint in ovarian cancer (26). Therefore, we first interrogated the relationship between CD24 expression and TIME using multiplex immunofluorescence staining on a tissue microarray comprising 44 high-grade serous ovarian cancer tissue samples (Fig. 8A). Quantitative analysis demonstrated inverse correlations between CD24 levels and activated TIME features, with reduced CD24 levels corresponding to elevated CD8⁺ T-cell density, increased macrophage infiltration (CD68), and heightened expression of costimulatory (CD86) and immune checkpoint (PD-L1) molecules (Fig. 8B–E). Analysis of data from patients with lung squamous cell carcinoma (LUSC) from The Cancer Genome Atlas (TCGA) database confirmed pan-cancer relevance, demonstrating conserved negative associations between CD24 expression and M1 macrophage, CD8⁺ T-cell, and CD8⁺ effector memory T-cell infiltration signatures (Fig. 8F–H).

Figure 8.

Low CD24 expression correlates with increased immune cell infiltration and prolonged survival across cancer types. A–E, Immunofluorescence images (A) and correlation analysis of CD24 average fluorescence intensity (AFI) with CD8 (B), CD68 (C), CD86 (D), and PD-L1 (E) expression from an independent high-grade serous ovarian cancer (HGSOC) array (n = 44). Scale bars, 200 and 50 μm. F–H, Correlation analysis of CD24 expression levels with M1 macrophage (F), CD8⁺ T-cell (G), and CD8⁺ effector memory T-cell (H) infiltration levels in LUSC. Plotted by TIMER 2.0. TPM, transcripts per million. I, Overall survival period of patients with prostate adenocarcinoma (PRAD) receiving radiotherapy (n = 77) stratified by CD24 expression levels based on the TCGA database. J and K, Overall survival of patients with hepatocellular carcinoma (LIHC; n = 370) stratified by CD24 (J) or GPAA1 (K) expression levels based on pan-cancer database of Kaplan–Meier plotter. L and M, Overall survival of patients with non–small cell lung cancer (NSCLC; n = 1,044) stratified by CD24 (L) or GPAA1 (M) expression levels based on SurveExpress lung meta-base. A Pearson correlation test was performed for B–E. R-squared, coefficient of determination. A log-rank test was performed for I–M. Hazard ratio (HR) and its 95% confidence interval were computed via a Cox proportional hazards regression model.

Considering CD24’s role as a radiation-responsive innate immune checkpoint, we evaluated clinical outcomes in patients with prostate cancer treated with radiotherapy and found that low CD24 expression predicted improved overall survival (Fig. 8I). Our results align with prior conclusions from studies on patients with cervical squamous cell carcinoma receiving adjuvant radiotherapy (27). GPAA1 is the key molecule mediating the postirradiation CD24 membrane localization. Parallel analysis of patients with hepatocellular carcinoma (LIHC) and non–small-cell lung cancer (NSCLC) confirmed both CD24 and GPAA1 as independent prognostic factors (Fig. 8J–M). Collectively, these analyses establish reduced CD24 expression as a biomarker of immunologically active TIME and favorable survival outcomes in diverse malignancies.

Discussion

Although the immunostimulatory effects of radiotherapy have long been recognized, clinically significant systemic responses beyond the irradiated field remain exceedingly rare (28). This discrepancy suggests that radiotherapy-induced immune activation is constrained by concurrent immunosuppressive signals. Building on this insight, extensive efforts to combine radiotherapy with PD-1/PD-L1 inhibitors aim at restoring postradiation T-cell function and achieve radiosensitization (29). However, such combination therapies do not consistently yield synergistic antitumor effects, hinting at immune escape mechanisms beyond adaptive immunity that hinder radiation-induced systemic tumor control (30).

Our study established macrophage phagocytosis and activation as key prerequisites for translating radiotherapeutic cytotoxicity into systemic immunity and identified CD24 as a radiation-responsive innate immune checkpoint that blocks this essential pathway. CD24 is highly expressed and confers prognostic significance across multiple cancer types (31–33). Recent studies demonstrated that CD24 functions as “do not eat me” signal, mediating antiphagocytic effects to drive antitumor immune evasion (34). We confirmed that CD24 exhibited superior tumor specificity compared with established innate immune checkpoints CD47, B2M, and PD-L1. Targeting CD24 may confer the advantage of reduced off-target effects, circumventing “antigen sink” associated with CD47 blockade and potentially reducing toxicity risks (35). CD24Fc has demonstrated favorable safety profiles in phase II/III clinical trials for GVHD prevention and severe acute COVID treatment (36, 37). Targeting CD24 with monoclonal antibodies, bispecific antibodies, nanoparticles, and chimeric antigen receptor T cells has demonstrated promising antitumor efficacy in preclinical studies (16, 38–40). Building on this preclinical momentum, multiple anti-CD24 therapeutic modalities are now entering clinical trials for patients with advanced malignant solid neoplasms. FDA has granted Fast Track Designation to anti-CD24 monoclonal antibody PHST001 for the treatment of patients with advanced platinum-resistant ovarian cancer or in combination with chemotherapy in platinum-sensitive ovarian cancer, which reinforces the promise of CD24 as a next-generation immuno-oncology target (41). We uncovered that CD24 governed both local and systemic responses to radiotherapy in preclinical models, expanding the therapeutic scope of this target. Our animal model studies and patient cohort analyses revealed that CD24 is inversely correlated with immunologically active TIME, suggesting its potential to inform immunotherapy eligibility in clinical practice. Collectively, our study provides comprehensive clinical guidance for targeting CD24 as a therapeutic agent and utilizing CD24 as a predictive biomarker in cancer management.

In the course of characterizing CD24 postirradiation dynamics, our study unveiled a novel mechanism governing GPI-anchoring process. Radiation promoted accumulation of GPI transamidase subunit GPAA1, thereby enhancing membrane localization of GPI-APs. GPAA1 inhibition effectively restricts CD24 membrane trafficking after irradiation, allowing radiation-enhanced “eat me” and “find me” signals to promote macrophage uptake and thereby enhancing radiotherapy efficacy. Our preclinical data validate the clinical therapeutic potential of targeting GPAA1 in cancer immunotherapy. Although several CD24 antibodies blocking the CD24–Siglec‐10 axis have advanced to clinical trials, small-molecule therapeutics offer advantages in cost, administration, and circumventing Fc-mediated immunogenicity, representing a high-priority translational goal (42). Notably, radiation-enhanced GPI anchoring is not limited to CD24. In this study, we employed CD73, another GPI-AP that is involved in adenosine-mediated immunosuppression, for dual validation of this mechanism. Other GPI-APs, such as mesothelin, carcinoembryonic antigen (CD66e), and complement regulatory proteins CD55/CD59, are also overexpressed in tumors and associated with tumorigenesis, progression, or immunosuppressive TIME, making GPI anchoring a targetable liability for combinatorial radiation treatment (43, 44). However, the indispensable roles of GPI-APs in normal physiology pose significant challenges. Deficiencies in GPI biosynthesis could potentially result in neurologic deficits, T-cell immune dysfunction, and hemolytic disorders, necessitating strict targeted delivery to ensure biosafety (45, 46). Targeting the GPI pathway in cancer remains largely underexplored and requires extensive preclinical and clinical investigation.

Although we have identified APC/C subunit ANAPC5 as a critical regulator of GPAA1 protein stability, the upstream mechanisms driving APC/C dysfunction after irradiation remain unclear. Tumor metabolic reprogramming may contribute to this maladaptation as radiotherapy is known to induce metabolic shifts in cancer cells, including glutathione depletion and lactate accumulation (47, 48). Notably, glutamine and lactate have been reported to modulate APC/C-dependent cell cycle regulation, suggesting that these metabolites may link radiation exposure to APC/C dysregulation (49, 50). Further investigations into the interplay between radiation and APC/C function are needed for potential therapeutic applications in cancer management.

In conclusion, this study establishes CD24 as a radiation-responsive innate immune checkpoint governed by the ANAPC5/GPAA1 axis. We identify that “do not eat me” signal CD24 mediated the immune escape after irradiation and propose combinatorial targeting of CD24 with radiotherapy as a novel therapeutic strategy validated across preclinical models. In addition, our findings reveal GPI anchoring as a previously unrecognized vulnerability in irradiated tumors. These findings provide a translational framework for addressing radioresistance through innate immune checkpoint modulation.

Supplementary Material

Supplementary Fig. S1

Radiation alone fails to enhance phagocytosis.

Supplementary Fig. S2

CD24 is a tumor specific “don’t eat me” signal.

Supplementary Fig. S3

Radiation enhances ‘don’t eat me’ signal CD24 on tumor cells.

Supplementary Fig. S4

CD24 ablation combined with radiation promotes macrophages uptake.

Supplementary Fig. S5

Radiation enhances GPAA1 to promote CD24 membrane trafficking.

Supplementary Fig. S6

Radiation suppresses APC/C-mediated ubiquitination of GPAA1.

Supplementary Fig. S7

GPAA1 has a greater propensity to bind to ANAPC5 subunit of APC/C.

Supplementary Fig. S8

Radiation mediates GPAA1 accumulation via APC/C disruption.

Supplementary Fig. S9

Radiation mediates GPAA1 accumulation via APC/C disruption.

Supplementary Fig. S10

GPAA1 inhibition combined with radiation promotes macrophage uptake and abolished by CD24 overexpression.

Supplementary Fig. S11

Targeting CD24 does not sensitize radiotherapy in vitro.

Supplementary Fig. S12

CD24-radiation combination therapy activates macrophages and T cells.

Supplementary Fig. S13

Validation of the immune cell depletion efficiency, related to Fig. 7.

Data Availability

The scRNA-seq datasets of patients with pancreatic and non–small cell lung cancer were obtained from the Gene Expression Omnibus database at GSE281288 and GSE274934. The RNA-seq data of patients with LUSC patient were obtained from the TCGA database. All other raw data generated in this study are available upon request from the corresponding author.

Authors’ Disclosures

L. Kong reports a patent for Targeting GPAA1 or CD24: A Strategy to Enhance Radiosensitivity and Activate Radiotherapy-Induced Abscopal Effect pending. K. Yang reports a patent for Targeting GPAA1 or CD24: A Strategy to Enhance Radiosensitivity and Activate Radiotherapy-Induced Abscopal Effect pending. Y. Sun reports a patent for Targeting GPAA1 or CD24: A Strategy to Enhance Radiosensitivity and Activate Radiotherapy-Induced Abscopal Effect pending. No disclosures were reported by the other authors.

Authors’ Contributions

L. Kong: Conceptualization, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Zhou: Formal analysis, investigation, methodology, writing–original draft. W. Yuan: Formal analysis, investigation, methodology, writing–original draft. Y. Wang: Investigation. X. Liu: Investigation. J. Wang: Investigation. W. Zhong: Investigation. Q. Chen: Investigation. P. Li: Investigation. T. Pu: Investigation. Z. Feng: Visualization. Z. Zhou: Visualization. Y. Deng: Validation. W. Wei: Validation. X. Yang: Validation. J. Meng: Funding acquisition, validation. Y. Sheng: Funding acquisition, validation. C. Wan: Supervision. F. Huang: Supervision. K. Yang: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing. Y. Sun: Conceptualization, supervision, funding acquisition, methodology, project administration, writing–review and editing.