Novel insights into the function of CD24: A driving force in cancer

CD24 is a highly glycosylated protein with a small protein core that is linked to the plasma membrane via a glycosyl‐phosphatidylinositol anchor. CD24 is primarily expressed by immune cells but is often overexpressed in human tumors. In cancer, CD24 is a regulator of cell migration, invasion and proliferation. Its expression is associated with poor prognosis and it is used as cancer stemness marker. Recently, CD24 on tumor cells was identified as a phagocytic inhibitor (“do not eat me” signal) having a suppressive role in tumor immunity via binding to Siglec‐10 on macrophages. This finding is reminiscent of the demonstration that soluble CD24‐Fc can dampen the immune system in autoimmune disease. In the present review, we summarize recent progress on the role of the CD24‐Siglec‐10 binding axis at the interface between tumor cells and the immune system, and the role of CD24 genetic polymorphisms in cancer. We describe the specific function of cytoplasmic CD24 and discuss the presence of CD24 on tumor‐released extracellular vesicles. Finally, we evaluate the potential of CD24‐based immunotherapy.

thus allowing tumor cells mimic the correct player? In the immune system, CD24 is important in the regulation of cell proliferation and clonal expansion. 4 It was recently reported that CD24 on tumor cells can act as a brake on the immune system. Bakal et al pointed out a novel role for CD24 at the interphase of the immune system and tumor cells. In search for new phagocytosis inhibitors, they observed that CD24 on ovarian cancer (OvCa) or triple-negative breast cancers (TNBC) acted as an antiphagocytic surface protein, which was termed a "don't eat me" signal. 8 CD24 on tumor cells interacted with Siglec-10 on macrophages. 8 The CD24-Siglec-10 axis is well known in the immune system and is exploited as a novel target for dampening of overshooting immune reactions.
Rapid progress on the functional role of CD24 has been made over the last decade. Here, we summarize the present status of the research. We outline some basic features of CD24 and then focus on the research progress with special emphasis on its role in cancer.
2 | CD24 BASIC FINDINGS 2.1 | What does the CD24 protein look like?
The murine precursor molecule and homolog for human CD24 is the mouse heat stable antigen (HSA). It was discovered and defined in the early 1980s using a panel of rat monoclonal antibodies (mAb) from different labs, including M1.69, J11D, B2A2, 79 and others. These mAbs recognized a cell surface molecule on mouse hematopoietic cells. The designation "heat stable antigen" reflected the antigen's resistance to heat. 9 In many studies during the late 1980s mAbs to HSA were used to investigate the maturation of hematopoietic cells in the mouse. [10][11][12][13] Biochemical analysis showed that the antigen migrated at a molecular mass of 40 to 70 kDa in SDS-PAGE. 14 In the beginning of the 1990s, HSA was cloned by two groups. 15,16 Surprisingly, the mRNA encoded a small peptide containing only 30 amino acids after removal of the signal sequence and displacement of the C-terminal region by the GPI-anchor.
Thus, almost all of the mass of HSA was made up by extensive N-and O-linked glycosylation at multiple sites of the short peptide. 15,16 In 1991, Kay and Humphries cloned the human CD24 antigen that turned out to be the human ortholog to mouse HSA. 17  of the protein and cleaves it at the ω-site with a preference for small nonpolar amino acids in this position. The preassembled anchor is then transferred to the protein. 18,19 In polarized epithelial cells, protein homooligomerization is a key step determining the apical sorting of GPI-anchored proteins. 20 The presence of both lipid anchor and protein portions confers unique trafficking features to these proteins. This allows them to partition into membrane microdomains enriched in cholesterol and sphingolipids (lipid rafts). 20,21 A number of anti-human CD24 mAbs were characterized in the V International Workshop on Human Leucocyte Differentiation Antigens. 1 These include those recognizing the central Leucine-Alanine-Proline (LAP) core (SWA11, ML5, OKB2) or those requiring CD24-associated sialic acid for binding (BA-1, HB-9, VIB-E3, SN3). 22 Interestingly, a recent biochemical analysis suggested that the epitope for mAb SN3 is likely present in the O-glycans or in the glycan core of the GPI-anchor, and therefore is not entirely CD24 specific. 23 The CD24 core protein can be modified with carbohydrate-defined CD-antigens, such as α2-6 sialylated polylactosamine structures. 24 The highly variable glycosylation makes it essential to validate mAb specificity before drawing conclusions about CD24.
CD24 expression has been reported in many cell types in the mouse. These include hematopoietic cells (transiently in T cells, constitutively in B lymphocytes, thymocytes, erythrocytes, neutrophils, dendritic cells, macrophages and others; for a review, see Reference 4) and nonhematopoietic cells (ie, neural cells; for a review, see Reference 25).
In humans, CD24 is expressed on B lymphocytes, monocytes, granulocytes and on lymphoid tumor cell lines. In contrast to the mouse, human erythrocytes do not express CD24. 17 Interestingly, a genomic study revealed the conservation of CD24 across many mammalian species and its appearance prior to the reptilian-avian divergence. 26 2.2 | What do we know about CD24 associated glycans?
Given the high degree of glycosylation (CD24: 14 O-and two N-glycosylation sites, HSA: seven O-and four N-glycosylation sites), it is clear that an understanding of CD24 biology requires knowledge about associated glycans. Like other membrane proteins, the glycosylation of CD24 is cell-type specific depending on the cellular repertoire of glycosyltransferases. 14 Aberrant glycosylation can occur in cancer cells due to cancer-associated upregulation of certain glycosyltransferases. [27][28][29] Unfortunately, there is not a complete structure of CD24 associated glycans. Some studies were carried out to gain insights. N-and O-linked glycans of mouse brain CD24 were studied. 30,31 For Olinked glycans, the analysis revealed a diverse mixture of mucin-type and O-mannosyl glycans carrying, in part, functionally relevant epitopes, such as 3-linked sialic acid, disialyl motifs, Le(X), sialyl-Le(X) or HNK-1 units. For N-linked glycans, a highly heterogeneous pattern mainly includes complex type glycans expressing distinct carbohydrate epitopes, like 3-linked sialic acid, Le(X) or blood group H antigens, bisecting N-acetylglucosamine residues, and N-acetyllactosamine repeats as well as high-mannose and hybrid type species.
In another study, N-glycans released from mouse and human CD24 from lymphoblastoma, neuroblastoma and astrocytoma cell lines or from mouse brain homogenate were analyzed and compared. 32 The authors described the presence of fucosylated and sialylated complex and hybrid N-glycans. Finally, Motari et al investigated the structure of CD24-Fc (see below) by matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry. 33 The majority of the oligosaccharides were Neu5Acα-2,3/6Galβ-1,3GalNAc (ie, sialyl-T). Although the study provided valuable data, it is clear that CD24-Fc is released via the secretory pathway. The glycosylation pattern may differ from that of membrane-associated CD24.
In summary, the information is incomplete, most likely due to inherent technical difficulties in resolving such complex structures.

| Which are the CD24 binding partners?
Several binding partners for CD24 have been identified ( Table 1). The study of Kadmon et al suggested that mouse CD24 (ie, nectadrin) is a cell adhesion glycoprotein exhibiting homophilic binding. 34 It is possible that this type of interaction is carbohydrate-mediated and reflects the ability of GPI-anchored proteins to cluster in lipid rafts. 20 In mouse brain, CD24 can interact in a carbohydrate-dependent manner with the neural adhesion molecule L1CAM (CD171), leading to the inhibition of neurite growth and altered L1CAM signaling. [35][36][37] CD24 was identified as a ligand of P-selectin in mouse myeloid and endothelial cells. 41,42 In human cancer cells, the CD24-P-Selectin interaction supported the rolling of breast carcinoma cells on endothelial cells and adhesion to platelets 38,39,46 and facilitated adhesion of ovarian cancer cells to mesothelium. 47 One study also reported the rolling of CD24 positive MCF-7 cells on immobilized E-Selectin. 43 To bind to P-selectin, CD24 required appropriate modifications with carbohydrates, such as sialyl-Lex. 40 Tumor cells forced to overexpress CD24 and FucT7 (synthesizes sialyl-Lex) bound to endothelial cells and platelets in vitro and in vivo, as mediated by P-selectin. 40 The CD24-P-Selectin pathway may enhance the metastatic spread of tumor cells. 40 In addition to P-selectin, CD24 can interact with Siglecs, a class of sialic acid binding receptors on immune cells. 23,44 Siglec-5 and Siglec-10 are important immune-inhibitory receptors on monocytes, granulocytes and lymphocytes. 48 Initial experiments in mice identified CD24 in a trimolecular complex together with Siglec-10 and the danger-associated molecular pattern (DAMP) molecule HMGB1. 44 The CD24 and Siglec-10 axis selectively apparently repressed tissue damage-induced immune responses. 44 The sialic-acid-based pattern recognition may discriminate infections from tissue injuries. 49 Importantly, the diverse and heterogeneous glycosylation of CD24 might bind a diverse array of DAMPs. Mass spectrometry of CD24-associated proteins revealed several prominent DAMPs including heat shock protein (HSP)70, HSP90 and Nucleolin. 44 Interestingly, human CD52, a small GPI-anchored molecule that is structurally similar to CD24, can also bind via its glycans to the proinflammatory B box of HMGB1 to engage the Siglec-10 receptor and suppress human T cell function. 50 Specific glycoforms were also required for the immune suppressive activity of soluble CD52. 51 In humans, purified CD24 derived from tumor cell lines can bind to Siglec-5 23 whereas CD24 isolated from human placenta clearly bound to Siglec-10. 45 Finally, as discussed below, CD24 on OvCa and TNBC cells bound to Siglec-10 on human monocytes. 8 These findings suggest that Siglecs can exquisitely discriminate between different glycoforms of CD24.

| How does CD24 act as a signaling molecule?
Antibody-mediated crosslinking experiments showed that CD24 itself has signaling capacity. 52 An increase in free cytoplasmic Ca 2+ was observed in B lymphocytes and monocytes, and augmented hydrogen production was induced by CD24 crosslinking in granulocytes. 52 As discussed above, CD24 is localized in detergent-resistant membrane domains (DRMs; also termed lipid rafts). These membrane T A B L E 1 Ligands to CD24 such as the CXCR4 receptor and β1-integrin. 58,59 The presence of CD24 inactivated CXCR4 signaling in cancer cells by excluding the latter from rafts. 58 In another example, CD24 reportedly recruited β1-integrin into rafts and augmented cell adhesion and migration. 59 Interestingly, a role of CD24 in the regulation of β1-integrins was noted in B lymphocytes from CD24-knockout mice or in tumor cells derived from knockout mice retransfected with CD24. 60 CD24 crosslinking with mAbs can induce a variety of effects in mouse and human cells. In mouse myeloid cells 54 or lymphocytes, 61 integrin-mediated cell binding can be induced. Of note, cell binding could also be induced by membrane raft-crosslinking with cholera toxin, which binds to GM1 lipids. 61 Triggering of apoptosis by antibody-mediated crosslinking has been demonstrated. 62 Another study demonstrated a link between intracellular CD24 and p53 inactivation. 83 ARF is a tumor suppressor gene that supports p53 function by inactivating MDM2, the E3 ligase that causes p53 ubiquitinylation and degradation. 83 The nucleolar protein NPM interacts with ARF and protects it from degradation, which protects p53 from MDM2-promoted degradation. The use of prostate cancer cell lines as a model system indicated that CD24 (most likely unprocessed) promoted cell proliferation and cell cycle progression and disrupted the ARF-NPM interaction, which destabilized p53. Additional animal experiments and clinical data supported the conclusion that expression of CD24 enables mutational and viral oncogene-mediated p53 inactivation. 83 The authors suggested that the functional elimination of CD24 might restore p53 wild-type function in cancer.
A recent study reported that the CD24-p53 link can maintain intrahepatic macrophages in a murine model of diethylnitrosamineinduced hepatocellular carcinoma (HCC). In contrast, in CD24−/− mice the level of p53 was increased in macrophages, resulting in their depletion following diethylnitrosamine treatment. The results established a critical role for macrophages in suppressing HCC development and suggested a functional role of CD24 and p53 in this process. 84 In summary, the synthesis and processing of CD24 must be more intensively investigated in various cancers. Clarification of the biological role of the GPI-anchor replaced C-terminal portion of CD24 is also necessary.

| What is known about CD24 genes and genetic polymorphisms?
In humans, the CD24 gene is located on chromosome 6q21. Three intronless pseudogenes are located on chromosome 15q21-q22 and Yq11 was discovered by in situ hybridization. 85 Two homologs exist.
One is located on chromosome 1p36 and one has been tentatively mapped to chromosome 20q11. 85 In the mouse, CD24 was mapped to chromosome 10. 86 In addition, two other genes appeared to be intronless retroposons. Despite numerous sequence changes, an open reading frame was maintained.
However, no expression of these pseudogenes was detected. 16,86 Ayre et al recently demonstrated that the diversity of the CD24 genomic structure between and within species, with varying numbers of exons, introns and the presence of untranslated regions (UTRs). Of note, the authors found no obvious criteria distinguishing CD24 genes from those annotated as CD24-like. 26 Studies of autoimmune diseases have identified two polymorphisms within the CD24 gene as modifiers of disease risk and progression. 87 Several studies have investigated the role of CD24 SNPs in cancer incidence and progression. A significant association of CD24 genetic variants with onset and progression was noted in prostate cancer. 96 In esophageal cancer, the incidence of metastasis in regional lymph nodes was markedly higher in patients carrying the V/V variant compared to those not carrying it. 97 The influence of CD24 polymorphisms on breast cancer prognosis and risk was investigated in a cohort of 2514 patients and 4858 controls. 98 The CD24 V/V genotype affected the outcome, but not risk, of primary breast cancer. 98 The clinical relevance of CD24 polymorphisms and their potential to predict a pathologic complete response (pCR) to neoadjuvant chemotherapy was studied in primary breast cancer. 99 The CD24 V/V genotype was the only significant predictor of pCR following doxorubicin treatment. Interestingly, a significant correlation was evident between CD24 V/V and intratumoral lymphocyte aggregates. 99 However, these findings were not confirmed in another study in breast cancer where pCR to anthracycline-and taxane-based neoadjuvant chemotherapy was investigated. 100 A metaanalysis of the role of CD24 in cancer showed that two SNPs (rs52812045 and rs3838646) did not modify the risk of cancer. 101 In summary, CD24 SNPs have not been proven to be strong and independent predictors of cancer incidence, progression and responsiveness to therapy. These findings differ from the observations in autoimmune disease and suggest that the CD24 polymorphisms affect immune cell functions more profoundly than tumor cell characteristics. Finally, a prostate cancer study described that tumors with P170 (T) or P-534(C) alleles had a 2-fold increased protein expression of CD24 compared to tumors with P170(C) or P-534(A) alleles. Likewise, tumors with a combination of P170(T/T) and P-534(C/C) genotypes were associated with a high mRNA level of CD24. 96 Thus, it appears that small changes in expression levels of CD24 can result in measurable physiological effects.

| HOW IS CD2EXPRESSION REGULATED?
In mice, CD24 displays tissue-specific as well as developmental regulation. During the maturation of several hematopoietic lineages, HSA expression is generally high in immature precursor cells and low or absent in terminally differentiated cells. The sequence and methylation analysis of the murine CD24 promoter revealed characteristics of both "housekeeping" and tissue-specific promoters, including multiple putative SP-1 and AP-2 consensus binding sites. 86 As mentioned above, SP-1 binding sites were also reported in the human CD24 promoter. 95 In human urothelial carcinoma, CD24 expression is under the control of androgen receptor. Androgen treatment led to increased CD24 promoter activity. Androgen receptor and androgen-response elements upstream of the CD24 start codon were considered responsible. 102 In contrast, in primary breast cancer, CD24 was downregulated and involved estrogen receptor-α and two estrogen responsive elements in the CD24 promoter, one of which was able to bind estrogen receptor-α. 103 In another study on breast cancer cell lines MCF-7, MCF-10 and MD-MB-231, the overexpression of Twist suppressed the expression of CD24. 104 Twist promotes the generation of a breast CSC phenotype characterized by the high expression of CD44, little or no expression of CD24, and increased aldehyde dehydrogenase 1 activity. 104 In contrast, in breast cancer cells the truncated glioma-associated oncogene homolog 1, known as the terminal effector of the Hedgehog pathway, can upregulate CD24 gene expression, thereby contributing to enhanced migration and invasiveness. 105 In human bladder cancer, CD24 is reportedly an effector of hypoxia-inducible factor 1 driven tumor growth and metastasis mediated by a functional hypoxia response element in the CD24 promoter. 106 In colorectal cancer, CD24 expression is apparently under the control of the β-catenin/TCF-dependent transcription machinery via the activation of cyclooxygenase-2. 107 In contrast, CD24 expression in melanoma cell lines is low but can be induced by BRAF inhibitors, such as Vemurafenib. 108 The upregulation of CD24 is reportedly mediated by SOX2 that can bind to the CD24 promoter. 108 The forced overexpression of either SOX2 or CD24 can significantly increase the resistance to BRAF inhibitors, while SOX2 or CD24 knockdown rendered cells more treatment sensitive. 108 A recent study involving HCC examined in more detail transcripts from the CD24 gene located on chromosome 6q21. 109 Two transcripts, termed CD24A and CD24B, were identified. They differed in length and only shared approximately 47% amino acid identity. 109 The variant CD24A, but not CD24B, which was barely detected by qPCR and western blotting, was significantly upregulated in HCC tissue. 109 When overexpressed in recipient cells, both CD24A and CD24B con- OvCa, lung cancer and prostate cancer. CD24 gene amplification was observed in carcinoma of breast, ovarian and lung, but not in the prostate. Importantly, the copy number amplification was strongly correlated with CD24 mRNA overexpression and gene amplification seemed to be the most influential genetic alteration for the prognosis of BRCA. 110

| Is CD24 a good marker for EVs?
Cells in the body (but also tissue cultured cells) release vesicles into the surroundings. These are termed EVs. [111][112][113][114][115] This term refers to all vesicles released by various cell types including erythrocytes, platelets, leukocytes and cancer cells. 116,117 The process of EV secretion is particularly active in proliferating cells, such as cancer cells. 116,117 Small vesicles (50-150 nm)

| Is soluble CD24-Fc a novel immunotherapeutic?
In line with its expression as a hematopoietic cell surface antigen, the function of CD24 has been intensively studied in the immune system.
The important findings have been already summarized in a previous review 4 and will not be mentioned again.
It was initially shown that mouse CD24 on antigen presenting cells can act as a CD28-independent costimulatory molecule for both CD4 and CD8-T cell responses. [129][130][131][132] Furthermore, CD24 on T cells was found to be essential for T cell homeostatic proliferation. 133 Considering the various possible binding partners (see Table 1 Phagocytosis is an important mechanism that allows the innate immune system to clear the body of dangerous or toxic substances.
Phagocytic cells include neutrophils, monocytes, macrophages, dendritic cells, osteoclasts and eosinophils. 140 In the cancer setting, monocytes and macrophages participate in radio-and chemotherapy by removing apoptotic and necrotic tumor cells and debris with or without inflammation. 141 Moreover, tumor antigen-induced cellular immunity is dependent on phagocytosis by antigen presenting cells. 140 A recent paper has shed light on the adverse function of the CD24-Siglec-10 axis in human cancer. Barkal

et al demonstrated that in
OvCa or TNBC CD24 can act as an antiphagocytic surface protein (ie, "don't eat me" signal). CD24 on tumor cells reportedly interact with Siglec-10 on tumor-associated macrophages. 8 The authors described that genetic ablation or therapeutic blockade of CD24 resulted in macrophage-dependent reduction of tumor growth in vivo and increased survival. 8 The discovery of immune checkpoints as targets for successful therapeutic interventions has transformed oncological practice in the last decade. In addition to Programmed cell death protein 1 and Cytotoxic T-lymphocyte-associated protein 4, "don't eat me" signals were discovered as valuable targets. Such surface molecules include CD47, programmed cell death ligand and β2-microglobulin associated with class I major histocompatibility complex molecules. 8 mAbs to these molecules that antagonize the interaction of the "do not eat me" signals with their macrophage-expressed receptor (termed phagocytic checkpoints) have demonstrated therapeutic potential in several cancers. 142 Barkal The functional role of CD24 in oncoimmunology. A, CD24 promotes tumor progression by activating signaling molecules involved in proliferation and survival of cancer cells. CD24-mediated raft clustering leads to activation of Src kinases that are of central importance for downstream signaling. Activated Src can phosphorylate STAT3, FAK but can also trigger other signaling pathways. B, By interaction with Siglec-10 on the surface of phagocytic cells, CD24 protects cancer cells from phagocytosis. C, CD24 is expressed on tumor derived EVs and could protect EVs from phagocytosis. D, Antibodies against CD24 could be used to counteract the function of CD24 on the surface of tumor cells or EVs and restore phagocytosis. CD24-Fc used as immune-therapeutic could trigger immune suppression in Siglec-10 expressing immune cells (T-, B-and phagocytic cells) and thereby protect from autoimmunity. In the cancer setting this could be detrimental. However, CD24-Fc might also act as soluble inhibitor for CD24-Siglec-10 interaction In summary, these collective data indicate that although CD24 is widely used as marker, we are only starting to understand its relevance and function in CSCs.

| Can CD24 be a target for cancer immunotherapy?
Due to the abundant expression in human cancer, and its putative role as a CSC marker, CD24 might be a good target for cancer therapy.
However, its expression on immune cells might cause deleterious side effects. Despite this, several attempts have sought to analyze the efficacy of anti-CD24 based tumor therapy in preclinical models. These models are summarized in Table 2.
Different formats were used for therapy. These included unconjugated mAbs, where the effector mechanism mainly relies on antibody-dependent cell-mediated cytotoxicity of SWA11, 153,159 ALB9, 160 antibody-drug conjugates, [161][162][163] bispecific antibodies, 164 chimeric antigen receptor T cells 166 or chimeric antigen receptornatural killer cells. 165 In all these experiments more or less efficient killing of tumor target cells was observed in vitro or in vivo in human tumors grafted into immunodeficient mice. It remains to be seen if any of these approaches can also efficiently eliminate TICs in vivo.
Blocking of CD24 to prevent phagocytic inhibition via Siglec-10 might be feasible with naked CD24 mAbs.

| CONCLUDING REMARKS
In the present review, we have summarized recent advances of CD24 in human tumors. A graphical summary is presented in Figure 1. We have tried to be as careful as possible not to overlook essential contributions. If we have, we apologize.
As in any progress report, a number of important questions and problems remain. Although the original characterization of CD24−/− mice did not reveal a dramatic phenotype, 167 it is surprising to see how many cellular defects were observed both in immunological 133,135,168,169 and nonimmunological studies. 107,170,171 It is unclear to what extend these observations are due to the lack of CD24-mediated raft signaling or to a failure of interaction with cognate ligands.
The processing of CD24 in cancer has not been carefully studied.
In particular, the function and fate of the GPI-anchor displaced tail and the presence of cytoplasmic CD24 is of interest. If indeed cytoplasmic CD24 interferes with p53 stability, it is important to know which part of CD24 is involved. Due to its intensive and variable glycosylation, CD24 may have additional ligands, as suggested. 26 It will be interesting to study whether other Siglecs or Galectins that are endowed with carbohydrate binding ability are able to interact with CD24 glycoforms.
Finally, the use CD24 as a target for cancer therapy has just begun. Time will tell if this approach is feasible.

ACKNOWLEDGMENT
The authors thank Dr Kai Doberstein and Dr Jacky Trotter for helpful comments on the article. Open access funding enabled and organized by Projekt DEAL.