From association to intervention: Rethinking CD24’s causal role in hepatocellular carcinogenesis

Abstract

Recent longitudinal studies have demonstrated progressive upregulation of cluster of differentiation 24 (CD24) during hepatocarcinogenesis, correlating with programmed death-ligand 1 (PD-L1) expression and enhancing detection of alpha-fetoprotein-negative hepatocellular carcinoma by approximately 20%. Circulating CD24⁺ T cells show expansion patterns that could track minimal residual disease. Yet this evidence is largely observational, with gaps in functional validation, subset-specific analysis, and mechanistic links to PD-L1 transcription. Spatial mapping, survival data, and isoform distinctions remain unexplored. To establish CD24 as a bona fide therapeutic target rather than merely a correlative marker, future investigations must employ clustered regularly interspaced short palindromic repeats-based functional genomics, single-cell multi-omics integration, spatial transcriptomics, and isoform-discriminating soluble CD24 assays to definitively demonstrate causality and clinical utility.

Key Words: Cluster of differentiation 24; Hepatocellular carcinoma; Programmed death-ligand 1; Immune evasion; Biomarker; Clustered regularly interspaced short palindromic repeats; Single-cell RNA sequencing; Spatial transcriptomics

Core Tip: While cluster of differentiation 24 (CD24) is dynamically upregulated during hepatocarcinogenesis and improves the detection of alpha-fetoprotein-negative hepatocellular carcinoma, it currently lacks functional validation, resolution of specific subsets, and a clear mechanistic linkage to programmed death-ligand 1. Functional genetics, single-cell, multi-omics, and spatial analyses are needed to confirm CD24’s causality and therapeutic potential.

INTRODUCTION

Cai et al[1] provide the first longitudinal evidence that cluster of differentiation 24 (CD24) is not merely a cross-sectional marker of hepatocellular carcinoma (HCC) but is dynamically upregulated at both the mRNA and protein levels during step-wise hepatocarcinogenesis. The integration of a rat 2-FAA model with a well-characterized human hepatitis B virus-infected cohort allows the authors to link stem cell-associated CD24 signaling to immune evasion via programmed death-ligand 1 (PD-L1), thereby extending the traditional alpha-fetoprotein (AFP)-based paradigm[2,3]. The finding that serum CD24 adds 20% diagnostic sensitivity to AFP-negative HCC is of clinical significance, and the demonstration that circulating CD24⁺ T cells are expanded in patients with HCC offers a new immune-phenotypic window for monitoring minimal residual disease[4,5].

CURRENT ADVANCES AND CONTROVERSIES IN CD24 RESEARCH

Emerging paradigm of CD24 as a dual-function immune checkpoint

Recent advances have positioned CD24 as a critical node integrating innate and adaptive immune evasion mechanisms in HCC (Figure 1)[6-8]. Beyond its established role as a “don’t eat me” signal mediated through sialic acid-binding Ig-like lectin 10 (Siglec-10) on tumor-associated macrophages[9,10], emerging evidence suggests CD24 operates as a multifunctional checkpoint coordinating both macrophage phagocytosis inhibition and T-cell exhaustion. The landmark study by Barkal et al[9] demonstrated that CD24-Siglec-10 signaling recruits Src homology domain 2-containing protein tyrosine phosphatase/Src homology domain 2-containing protein tyrosine phosphatase phosphatases, suppressing cytoskeletal rearrangement required for phagosome formation. Importantly, CD24 exhibits significantly lower basal expression in normal human tissues compared to CD47, presenting a favorable therapeutic index for targeted intervention[11,12].

Figure 1 Cluster of differentiation 24-centered immune evasion network in hepatocellular carcinoma. Schematic illustration depicting cluster of differentiation 24 (CD24)-mediated “don’t eat me” signaling through sialic acid-binding Ig-like lectin 10 (Siglec-10) on tumor-associated macrophages (TAMs), concurrent programmed death-ligand 1 (PD-L1)/programmed death 1 (PD-1) axis suppression of T-cell cytotoxicity, and hypoxia-driven hypoxia-inducible factor-1 alpha (HIF-1α)-mediated transcriptional co-regulation. The diagram illustrates potential therapeutic intervention points including anti-CD24 monoclonal antibodies, CD24-targeting CAR-T cells, and dual checkpoint blockade strategies.

Controversies persist regarding the relative contribution of CD24 vs other innate immune checkpoints in HCC immune evasion. While CD47 blockade has demonstrated preclinical efficacy, its ubiquitous expression on erythrocytes limits dosing flexibility and poses on-target toxicity risks[13,14]. By contrast, CD24 expression is restricted to hepatocytes, activated immune cells, and cancer stem cells, potentially offering a wider therapeutic window[15,16]. However, recent single-cell transcriptomic analyses reveal substantial heterogeneity in CD24 expression across HCC subpopulations, with distinct cancer stem cell subsets exhibiting variable CD24 levels that correlate with differential sensitivity to anti-CD24 monotherapy[17-19].

CD24-PD-L1 crosstalk: Association vs causation

The mechanistic linkage between CD24 upregulation and PD-L1 expression represents a central unresolved question. While Cai et al[1] demonstrated longitudinal correlation, the directional causality remains ambiguous. Hypoxia-inducible factor 1 alpha (HIF-1α) emerges as a plausible integrator, simultaneously transactivating both CD24 and CD274 (PD-L1) promoters in hypoxic tumor niches[20,21]. Chromatin immunoprecipitation sequencing analyses in HCC cell lines reveal HIF-1a-binding motifs within the CD24 promoter region, suggesting convergent evolutionary pressure driving dual checkpoint expression under metabolic stress[22,23].

Alternative hypotheses propose CD24-mediated nuclear factor kappa B (NF-κB) activation as the primary driver of PD-L1 induction. In breast cancer models, CD24 ligation activates NF-κB and signal transducers and activators of transcription 3 signaling, both capable of binding the CD274 promoter[24,25]. However, HCC-specific validation remains lacking. The absence of genetic epistasis experiments - specifically whether CD24 knockout abrogates PD-L1 induction under inflammatory stimuli - represents a critical experimental gap.

Spatial heterogeneity and the ecosystem perspective

The tumor microenvironment in HCC exhibits profound spatial heterogeneity that bulk transcriptomic analyses fail to capture (Figure 2)[26-28]. Recent integration of single-cell RNA sequencing with spatial transcriptomics (ST) reveals zonal segregation of CD24-expressing malignant hepatocytes adjacent to M2-polarized tumor-associated macrophages and exhausted CD8⁺ T cells[29,30]. Notably, CD24-high tumor regions demonstrate reduced cytotoxic T lymphocyte infiltration and enhanced collagen deposition, suggesting mechanosensitive regulation of CD24 expression via integrin-mediated signaling[31,32].

Figure 2 Spatial heterogeneity of cluster of differentiation 24 expression in the hepatocellular carcinoma tumor microenvironment. A: Spatial transcriptomics visualization showing zonal distribution of cluster of differentiation 24 (CD24)-high malignant hepatocytes adjacent to M2-polarized tumor-associated macrophages (TAMs); B: Multiplex immunofluorescence demonstrating co-localization of CD24, programmed death-ligand 1 (PD-L1), and CD68 in peritumoral hypoxic niches; C: Single-cell trajectory analysis revealing developmental progression from CD24-low to CD24-high tumor cell states with concurrent acquisition of stemness markers (CD44, epithelial cell adhesion molecule [EpCAM]).

Imaging mass cytometry studies demonstrate that CD24 expression correlates with distinct immune ecosystem architectures[33,34]. CD24/Siglec-10 blocking peptide which blocked programmed death 1 (PD-1)/PD-L1 interaction as well, functioned by enhancing the phagocytosis of tumor cells by macrophages and monocytic myeloid-derived suppressor cells, and elevating the activity of CD8+ T cells for cancer immunotherapy[35,36]. These findings challenge the conventional view of CD24 as a uniform tumor marker and support a dynamic, context-dependent model where spatial distribution determines functional impact.

AUTHOR PERSPECTIVES: RETHINKING CD24’S THERAPEUTIC POTENTIAL

Beyond monotherapy: Rationale for combinatorial strategies

We posit that CD24-targeted monotherapy will prove insufficient for HCC treatment given the compensatory upregulation of alternative immune checkpoints[37,38]. Preclinical evidence supports dual blockade strategies: Simultaneous targeting of CD24 and CD47 using bispecific antibody fusion proteins (e.g., PPAB001) demonstrates synergistic enhancement of macrophage phagocytosis in human epidermal growth factor receptor 2 (HER2)-positive cancer models[39,40]. The mechanistic basis involves convergent signaling through distinct inhibitory receptors - Siglec-10 for CD24 and signal regulatory protein alpha (SIRPα) for CD47 - creating redundant “don’t eat me” barriers that require simultaneous disruption.

Moreover, the integration of CD24 blockade with adaptive immune checkpoint inhibitors presents compelling theoretical advantages. CD24-Siglec-10 inhibition enhances macrophage-mediated antigen presentation, potentially priming CD8⁺ T-cell responses that are subsequently sustained by PD-1/PD-L1 blockade[41,42]. A novel dual-targeting D-peptide capable of blocking both CD24/Siglec-10 and PD-1/PD-L1 interactions has demonstrated superior anti-tumor efficacy compared to single-agent therapy in syngeneic models[43,44].

CD24 as a liquid biopsy biomarker: Technical and biological considerations

The clinical translation of serum CD24 measurements requires rigorous attention to isoform specificity (Table 1)[45,46]. CD24 exists as both glycosylphosphatidylinositol-anchored membrane protein and soluble form generated by A disintegrin and metalloproteinase 10/A disintegrin and metalloproteinase 17-mediated shedding[47,48]. Soluble CD24 retains Siglec-10 binding capacity and may function as a decoy molecule, competitively inhibiting therapeutic antibody efficacy while maintaining immune suppressive activity[49,50]. We advocate for the development of isoform-specific enzyme-linked immunosorbent assays capable of distinguishing membrane-bound vs shed CD24, analogous to established methodologies for soluble PD-L1 quantification.

Table 1 Multi-omics approaches to validate cluster of differentiation 24 causality in hepatocellular carcinoma.

Experimental strategy	Technical platform	Expected outcome	Clinical translation
Genetic knockout	CRISPR-Cas9 hydrodynamic transfection (Fah-/- mice)	Reduced tumor incidence, restored anti-PD-1 sensitivity	Patient selection for combination therapy
Single-cell profiling	scRNA-seq + CITE-seq + scATAC-seq	CD24+ subset resolution, regulatory network mapping	Biomarker discovery for liquid biopsy
Spatial mapping	10 × Visium + IMC (43-marker panel)	Zonal CD24 distribution, immune ecosystem architecture	Image-guided therapeutic targeting
Functional validation	In vivo CRISPR screening (AAV-sgRNA library)	Genetic interaction map, synthetic lethal targets	Rational combination design
Isoform quantification	Isoform-specific ELISA, mass spectrometry	sCD24 vs mCD24 dynamics, ADAM10/17 activity	Monitoring therapeutic response

AAV: Adeno-associated virus; ADAM10: A disintegrin and metalloproteinase domain-containing protein 10; ADAM17: A disintegrin and metalloproteinase domain-containing protein 17; CITE-seq: Cellular indexing of transcriptomes and epitopes by sequencing; CRISPR: Clustered regularly interspaced short palindromic repeats; CRISPR-Cas9: Clustered regularly interspaced short palindromic repeats-associated protein 9; ELISA: Enzyme-linked immunosorbent assay; IMC: Imaging mass cytometry; mCD24: Membrane-bound CD24 (or membrane CD24); PD-1: Programmed cell death protein 1; scATAC-seq: Single-cell assay for transposase-accessible chromatin using sequencing; sCD24: Soluble CD24; scRNA-seq: Single-cell RNA sequencing; sgRNA: Single-guide RNA.

Circulating tumor cells (CTCs) expressing CD24 represent an underexplored liquid biopsy dimension. Single-cell transcriptomic profiling of HCC CTCs reveals enrichment of stemness-associated gene signatures including CD24, insulin-like growth factor 2, and AFP[51,52]. The integration of CD24⁺ CTC enumeration with cell-free DNA methylation markers (e.g., septin 9, homeobox A1) may enhance early HCC detection sensitivity beyond current AFP-based surveillance protocols[53,54].

FUTURE RESEARCH DIRECTIONS

Functional genomics and in vivo clustered regularly interspaced short palindromic repeats screening

The transition from association to causation necessitates systematic genetic perturbation studies (Table 1). Hydrodynamic transfection of clustered regularly interspaced short palindromic repeats (CRISPR)-associated protein 9 single-guide RNA libraries targeting Cd24a in Fah^-/^- mice enables autochthonous hepatocarcinogenesis with concurrent gene knockout[55-58]. Pooled mutagenesis screens comparing tumor incidence, progression kinetics, and immune infiltration patterns between Cd24a-wildtype and knockout backgrounds will definitively establish CD24’s role in malignant conversion.

In vivo CRISPR screening platforms have identified novel immunotherapy resistance mechanisms in liver cancer models. Wang et al[55] demonstrated that knockout of chromatin remodeling genes (AT-rich interaction domain 1A, lysine-specific methyltransferase 2D) sensitizes tumors to anti-PD-1 therapy, whereas loss of antigen presentation machinery (B2m) confers resistance[59,60]. Extending these approaches to evaluate CD24 genetic interaction with established immune evasion pathways will prioritize combinatorial therapeutic targets.

Single-cell multi-omics and spatial systems biology

We advocate for comprehensive single-cell multi-omic profiling of HCC specimens, integrating transcriptomics, chromatin accessibility (single-cell assay for transposase-accessible chromatin using sequencing), and surface proteomics (cellular indexing of transcriptomes and epitopes by sequencing) to resolve CD24 regulatory networks[61,62]. Particular attention should focus on CD24⁺ T-cell subsets, which may represent distinct activation states rather than uniform tumor-permissive populations. Multi-color flow cytometry panels discriminating CD24 expression across CD4⁺ helper, CD8⁺ cytotoxic, regulatory T, and exhausted T-cell subpopulations will clarify whether CD24 marks functional anergy or antigen-experienced activation[63,64].

ST platforms (10 × Visium, Xenium) enable mapping of CD24 expression within architectural tissue contexts. Correlation with multiplex immunofluorescence (CD24, PD-L1, CD68, CD8, alpha-smooth muscle actin) will reveal whether CD24-high regions correspond to fibrotic barriers, hypoxic zones, or invasive fronts - each implying distinct therapeutic vulnerabilities[65,66].

Therapeutic development and clinical translation

Chimeric antigen receptor (CAR) T-cell therapy targeting CD24 represents an innovative approach for HCC treatment (Table 2)[67,68]. Preclinical studies in triple-negative breast cancer demonstrate that CD24-specific CAR-T cells (24BBz) exhibit antigen-specific cytotoxicity and tumor regression in xenograft models[69,70]. However, T-cell exhaustion limits sustained efficacy, suggesting that armored CAR constructs secreting PD-1 antibodies or CRISPR-mediated PD-1 knockout may enhance durability[71,72].

Table 2 Comparative analysis of “don’t eat me” signals in hepatocellular carcinoma immunotherapy.

Checkpoint	Ligand/receptor	Expression pattern	Clinical development	Advantages	Limitations
CD24	Siglec-10	CSCs, activated immune cells, hepatocytes	Preclinical (SWA11 mAb, CAR-T)	Low normal tissue expression, dual innate/adaptive function	Heterogeneous expression, shedding-mediated decoy effects
CD47	SIRPα	Ubiquitous (high on RBCs)	Phase I/II (Hu5F9-G4, magrolimab)	Well-established biology, broad tumor coverage	On-target anemia, dosing limitations
PD-L1	PD-1	Inducible on tumor and immune cells	Approved (atezolizumab, durvalumab)	Proven efficacy, established biomarkers	Primary resistance, immune-related adverse events
MHC-I	LILRB1/2	Variable across HCC subtypes	Early preclinical	Alternative escape mechanism	Complex regulation, technical challenges

CAR-T: Chimeric antigen receptor T cell; CSCs: Cancer stem cells; HCC: Hepatocellular carcinoma; Hu5F9-G4: Humanized anti-CD47 monoclonal antibody (5F9-G4; magrolimab); LILRB1/2: Leukocyte immunoglobulin-like receptor subfamily B member 1/2; MHC-I: Major histocompatibility complex class I; PD-1: Programmed cell death protein 1; PD-L1: Programmed death-ligand 1; RBCs: Red blood cells; Siglec-10: Sialic acid-binding Ig-like lectin 10; SIRPα: Signal regulatory protein alpha; SWA11: Anti-CD24 monoclonal antibody (clone SWA11).

Antibody-drug conjugates combining anti-CD24 specificity with cytotoxic payloads offer another therapeutic avenue. The recent development of glypican-3-CD24 dual-targeting antibody-drug conjugates demonstrates anti-cancer activity through antibody-dependent cellular cytotoxicity mechanisms[73], supporting expanded evaluation in HCC models.

MAJOR SCIENTIFIC LIMITATIONS AND HOW TO ADDRESS THEM

Lack of functional perturbation

The study infers that CD24 “promotes” progression solely from correlation and expression kinetics[1]. No genetic or pharmacologic manipulation (e.g., clustered regularly interspaced short palindromic repeats deletion of Cd24a in 2-FAA rats or therapeutic CD24 blockade) is performed. Consequently, the directional causality between CD24 upregulation and malignant conversion remains unproven and requires further investigation. Future work should employ hydrodynamic transfection of Cd24a-single-guide RNA plasmids in Fah^-/^- rats or inject anti-CD24 monoclonal antibody (clone SWA11) to determine whether loss or inhibition of CD24 reduces tumor incidence or restores PD-L1 inhibitor responsiveness[9,39].

Over-interpretation of CD24⁺ T cells

The authors equate the abundance of CD24⁺ lymphocytes with a tumor-permissive state without discriminating between CD24-high B cells, exhausted CD8⁺ T cells, or regulatory T-cell subsets[17,63]. CD24 is a well-established co-stimulatory molecule on B cells, and its presence on T cells may simply reflect activation rather than a tumor-driving stemness program. Multi-color flow panels (CD3, CD19, CD4, CD8, PD-L1, T-cell immunoglobulin and mucin-domain containing protein 3) and single-cell RNA sequencing are required to clarify which compartment contributes to the observed CD24 signal and whether these cells are functionally anergic or cytotoxic[64].

PD-L1 correlation without mechanistic depth

Although PD-L1 expression parallels CD24, no data demonstrate that CD24 directly transactivates PD-L1 transcription[1,24]. Previous work in breast cancer has shown that CD24 ligation activates NF-κB and signal transducers and activators of transcription 3, both of which can bind the CD274 promoter[25]. Chromatin immunoprecipitation (NF-κB p65, signal transducers and activators of transcription 3) and luciferase reporter assays using the 1.2-kb CD274 promoter region should be applied to HCC cell lines (Huh7-CD24-high vs clustered regularly interspaced short palindromic repeats-CD24-knockout) to explore and establish this regulatory axis[20,22].

Absence of spatial context

Serum and whole-liver homogenate measurements obscure the zonal distribution of CD24[26,28]. ST (10 × Visium) or multiplex immunofluorescence (CD24, PD-L1, AFP, CD8) could reveal whether CD24 is enriched in peritumoral hypoxic niches where PD-L1 is concurrently upregulated, as recently shown in colorectal liver metastases[30].

Statistical overfitting of survival data

The multivariate Cox model contains nine variables for 129 patients with HCC, approximating 14 events per variable - below the recommended 20:1 ratio[46]. Bootstrap validation (≥ 1000 iterations) or least absolute shrinkage and selection operator penalized regression should be used to verify that CD24 remains a reliable independent predictor[45].

Failure to explore circulating CD24 isoforms

CD24 exists in both glycosyl-phosphatidylinositol-anchored and soluble forms generated by A disintegrin and metalloproteinase/17 shedding[47,48]. Soluble CD24 can directly engage Siglec-10 on tumor-associated macrophages, thereby suppressing anti-tumor immunity[49]. Western blots under non-reducing conditions or enzyme-linked immunosorbent assay with isoform-specific capture antibodies would clarify which molecular species accumulates in sera and whether A disintegrin and metalloproteinase inhibition (GI254023X) attenuates the immune-suppressive effect[50].

CONCLUSION

Cai et al[1] provide valuable longitudinal data on CD24 upregulation during cancer progression; however, showing that CD24 increases over time does not prove it drives the disease. We emphasize that functional genetic validation through CRISPR-mediated perturbation, single-cell multi-omic deconvolution, and promoter-level mechanistic studies are indispensable for elevating CD24 from a passive correlate to an active driver of HCC progression.

Combining ST with imaging mass cytometry could reveal how CD24 distribution varies across tumor regions underlying CD24-mediated immune evasion, particularly regarding its zonal distribution within hypoxic peritumoral niches. CD24 blockade alone will likely fail because tumors can compensate by upregulating other checkpoint pathways. We advocate for rational combination strategies integrating CD24 blockade with adaptive immune checkpoint inhibitors or dual innate checkpoint blockade using bispecific antibodies. Furthermore, the development of CD24-directed CAR-T cells and antibody-drug conjugates warrants expedited preclinical evaluation in autochthonous HCC models.

Ultimately, successful clinical translation necessitates prospective validation of CD24 as a liquid biopsy biomarker for minimal residual disease monitoring and therapeutic response assessment. Multi-center consortium efforts employing standardized isoform-specific assays will be critical to harmonize biomarker thresholds and guide patient stratification for CD24-targeted interventions, bringing us closer to truly personalized treatment for liver cancer.