In the present study, we have analyzed immunohistochemical staining characteristics and the prognostic value of CD24 expression in NSCLC with a special emphasis on the comparison of the CD24 antibodies SWA11 and SN3b. The most important result of our study is that the prognostic relevance of CD24 is critically dependent on the careful consideration of sub-cellular compartments and the epitope specificity of the antibody used.
Overall, about one third of the NSCLC cohort revealed a significant CD24 expression (either cytoplasmic or membranous). These results are in line with the findings of other studies. In another NSCLC cohort, CD24 (SN3b) expression was found in 33 % of the samples (87 of 267 cases) . Consistent with those results, we have found similar rates of high CD24 expression levels (35 % of the cases) for SWA11. Originally, we would have expected lower rates than those found by Lee et al, as they used the antibody SN3b, that also recognizes yet unidentified other glycoproteins next to CD24. Furthermore, they used whole mount sections instead of tissue microarrays. A possible explanation for rather equal detection rates would be the fact that it has been demonstrated that the epitope recognized by SN3b is indeed present in CD24, but is not found in all glycoforms of CD24 . In contrast to the commonly used mAb SN3b, mAb SWA11 binds to the protein core of CD24 and does not depict other glycan moieties next to CD24. The protein core of CD24 is linear, consisting of the amino acid sequence leucine-proline-alanine (LAP) next to a glycosyl-phosphatidylinositol anchor .
CD24 expression has been associated with disease progression and cancer-related death in the majority of malignant tumors [2, 3, 16, 17], although a caveat to these data is that most of these studies are based on the supposedly less specific CD24 clone SN3b. Lee et al demonstrated a significant association between CD24-high expression (SN3b) and shorter patient survival times. Furthermore, Lee and colleagues and ourselves in former studies referred the results to cytoplasmic CD24 expression [2, 3].
In invasive ovarian carcinoma, patients carrying tumors with cytoplasmic CD24 expression showed a significantly shorter mean survival time of 37 months versus patients with tumors without cytoplasmic expression of CD24 (98 months) . Also in NSCLC, expression of CD24 has been claimed to be an independent prognostic marker of shorter patient survival times, especially in AC .
Recently, CD24 expression has been addressed as a putative stem cell marker in NSCLC . In that context particular attention has to be paid to the co-expression of CD44. Sterlacci et al. demonstrated that the phenotype CD24−/CD44+ did not show a significant difference in overall survival for the entire NSCLC cohort when compared with the CD24+/CD44 −population. However, when stratified according to histology, AC displaying the putative cancer stem cell (CSC) signature CD24−/CD44+ had a significantly shorter overall survival than CD24+/CD44− AC. However, these findings could not be ascertained as an independent factor, when calculated by multivariable analysis . Since the overwhelming evidence of the pro-tumorigenic properties of CD24 is independent of the CD24low/CD44-high stem cell definition, we focused on CD24 expression alone and did not include CD44 expression data in the present study.
Our results are only partly consistent with the published data. We were able to demonstrate that membranous staining pattern (using the mAb SWA11) was associated with a poorer overall survival and we revealed an increased risk of lymphonodular spread in the subgroup of CD24-high tumors, being in accordance with the published data. Nonetheless, our results are also partially conflicting with previous immunohistochemical data, as we could not confirm a significant correlation with patient survival for cytoplasmic expression of CD24, neither for SWA11 nor for SN3b.
The underlying biological mechanisms of CD24 promoted tumor progression are still incompletely characterized, although a growing number of studies have contributed to our comprehension [5–7, 17–20]. Expression of CD24 may provide an enhanced capability of tumor cells to adhere to activated endothelial cells mediated by its P-selectin binding site  or alter cellular signaling . Aigner and colleagues showed that CD24 functions as a ligand for P-Selectin under physiological flow conditions, using a plate flow chamber assay. In their study, CD24 proved to be necessary for mediation of rolling on P-Selectin, as low expression levels or cleavage of CD24 resulted in inhibition of attachment and rolling, in a breast carcinoma cell line . The precise mechanisms of ligand binding have still to be elucidated. In particular, CD24 does not contain the sulfated tyrosine residue of the P-Selectin glycoprotein ligand 1 (PSGL-1), i.e. another P-Selectin ligand . Another mechanism of CD24 binding focuses on the observed association of CD24 with the sulfate-containing epitope HNK-1 which is also recognized by P-Selectin. This observation may lead to the assumption that HNK-1 mediates CD24 binding . Enhanced disease progression as a result of metastatic spread with poorer survival rates may therefore be reasonable [6, 7]. As known, cells of hematogenously metastasizing tumours attach to platelets in the bloodstream [23–25]. Activated platelets express P-Selectin. Therefore, CD24-positive cells probably attach to activated platelets, containing P-Selectin on their surface, at the point, when the primary tumour invades into the vascular system . Moreover, CD24-mediated tumor propagation has also been associated with an increase of local invasiveness: CD24 mediated invasion of cancer cells has been hypothesized as a result of increased contractile forces as indicated by the findings of A125 human lung cancer cells with different CD24 expression levels using CD24-high and CD24-low transfectants in three-dimensional extracellular matrix (ECM) invasion assays . The percentage of invasive cells and their invasion depth was increased in CD24-high cells compared with CD24-low cells. Conversely, knockdown of CD24 and of the ß1-integrin subunit in CD24-high cells decreased their invasiveness, indicating that the increased invasiveness is CD24- and ß1-integrin subunit-dependent . Interestingly, besides acting as a ligand for P-Selectin, recently it has been proven that CD24 expression also indirectly stimulates cell adhesion to fibronectin, collagens 1 and 4, and laminin through activation of α3β1 and α4β1 integrin activity. Sleeman and colleagues have reported that β1 integrins colocalize focally with CD24. This suggests a direct interaction between CD24 and β1-containing integrins. In their study they show that CD24 interacts with c-scr, leading to stabilization of the kinase-active form of c-scr, which is necessary for sufficient activation of integrin adhesion to extracellular matrix components such as fibronectin. Thus, CD24 mediates cell adhesion in a P-Selectin dependent and a P-Selectin independent manner . Next to its influence on cell adhesion, metastasis and on invasion, CD24 also serves as a mediator of proliferation. Apparently, a depletion of CD24 by siRNA leads to a significant decrease of cell numbers in several cell lines as well as to a reduction of their clonogenicity . These experimental results provide a functionally well compatible explanation for the observed clinicopathological correlation of CD24 expression and poorer overall survival resp. increased occurrence of lymphonodular spread in our study.
Notably, CD24 may also provide a promising target for individualized therapy strategies beyond the scope of prediction. For example, CD24 specific antibodies have been applied in the treatment of the transplantation associated B-cell proliferative syndrome . Moreover, mAb SWA11 has recently been shown to have a beneficial effect on anti-cancer treatment when used as addition to gemcitabine treatment in an A549 lung cancer model . Pretreatment with mAb SWA11 led to a significant retardation of carcinoma growth compared to monotherapy with gemcitabine, which was attributed to faster internalisation of tumour antigen-bound therapeutic antibodies and alterations in the intratumoural cytokine milieu. Increased levels of intratumoural chemoattractants such as CXCL9/MIG and CCL2/MCP-1 were observed, in accordance to a heightened infiltration of xenografts by macrophages, possibly gained through the involvement of the antibody-dependent cell-mediated cytotoxicity. 
Some parts of our results, however, are conflicting with previous data, as we could not reproduce a significant correlation with patient survival for cytoplasmic expression of CD24 for SWA11. Vice versa, we revealed that only the membranous staining pattern was indicative for poorer overall survival (p = 0.007). As a minimal discordance to findings of former NSCLC studies [2, 3], our SWA11 based study could not confirm an especially relevant prognostic value of CD24 for pulmonary adenocarcinomas. Still, this study demonstrates a small trend towards a subtype-dependent prognostic relevance of membranous CD24 expression. Larger cohorts will be necessary for a more substantial statistical power concerning its prognostic relevance for AC.