CD24 affects CXCR4 function in pre-B lymphocytes and breast carcinoma cells

CD24, also known as heat-stable antigen (HSA) in mice, is a small heavily glycosylated cell-surface protein that is linked to the membrane by a glycosyl-phosphatidylinositol (GPI-) anchor (Pierres et al., 1987; Kay et al., 1990; Alterman et al., 1990). Mouse CD24 has a protein core of 27 amino acids with seven potential glycosylation sites, whereas human CD24 consists of 31 amino acids with 16 potential O- and N-glycosylation sites. Owing to this extensive glycosylation, CD24 has mucin-like characteristics (reviewed by Kristiansen et al., 2004b).

CD24 is expressed in mouse hematopoietic cell subpopulations including B lymphocytes, the majority of thymocytes, erythrocytes and neutrophils. Because of its lineage-specific and developmentally regulated expression, CD24 was traditionally used as a differentiation marker for B- and T-cell ontogeny (Poncet et al., 1996; Egerton et al., 1990; Lu et al., 1998). Later studies revealed that CD24 is not exclusively expressed by hematopoietic cells but is also present in the developing brain as well as in a broad range of epithelial cells (Shirasawa et al., 1993; Poncet et al., 1996; Magnoldo and Barrandon, 1996; Maric et al., 1996). In humans, CD24 is not expressed on erythrocytes or thymocytes but is present on a wide variety of malignancies including B-cell lymphoma, renal cell carcinoma, small-cell and non-small-cell lung carcinoma, nasopharyngeal carcinoma, hepatocellular carcinoma, bladder carcinoma and glioma, epithelial ovarian and breast cancer (reviewed by Kristiansen et al., 2004b).

CD24-knockout mice are viable and display no obvious defects. However, B lymphocytes from these mice displayed a slight defect in bone marrow maturation with reduced numbers of pre-B lymphocytes, but normal numbers of mature B lymphocytes, in the periphery (Nielsen et al., 1997). This suggested a role for CD24 in stimulating proliferation and maturation of pre-B lymphocytes within the bone marrow. It has also been described that CD24 expression on T cells is required for optimal proliferation in a lymphopenic host (Li et al., 2004). On the other hand, cross-linking of CD24 is able to induce apoptosis in a human B-cell subset during the early activation stage (Suzuki et al., 2001). Thus, probably both overexpression and lack of CD24 have an influence on cellular function.

Functionally, CD24 has been demonstrated to serve as a ligand for P-selectin in mouse myeloid and in human tumour cells (Aigner et al., 1997). The role of human CD24 as a P-selectin ligand depends on the appropriate modification of glycans as only sialylLe^x-modified CD24 can promote rolling and tumour cell colonization of the lungs (Friedrichs et al., 2000). The P-selectin ligand function has been preserved in mice and humans. For cancers of the breast, ovary, colon and prostate it has been demonstrated that CD24 expression is associated with poor prognosis and shortened survival time (Kristiansen et al., 2002; Kristiansen et al., 2003; Kristiansen et al., 2004a). It has also been proposed that CD24-mediated binding to P-selectin on endothelial cells and platelets could facilitate the exit of tumour cells from the bloodstream and hence favour metastasis (Friedrichs et al., 2000; Kristiansen et al., 2004a).

A recent study of primary breast cancer cells isolated from pleural infusions suggested that CD24 might be a potential marker for breast tumour stem cells (Al-Hajj et al., 2003). It was shown that the ability to form tumours in NOD/SCID mice was greater in the CD44⁺CD24^low/–-expressing cell fraction compared with the CD44⁺CD24^high fraction. Owing to the enhanced tumour-forming ability, the CD44⁺CD24^low/– fraction was proposed to represent `breast tumour stem cells' (Al-Hajj et al., 2003). The mechanism of growth regulation by CD24 is as yet undefined.

Chemokines are a superfamily of cytokine-like small peptides that are subdivided into four distinct classes depending on the arrangement of cysteine residues (Rossi and Zlotnik, 2000). Chemokines and their respective receptors, a family of seven transmembrane-spanning heterotrimeric G-protein-coupled molecules, are known to play an important role in immune and inflammatory responses as well as hematopoiesis and HIV infection (Taub et al., 1995; Gerard and Rollins, 2001; Moser et al., 1998; Broxmeyer, 2001; Berger et al., 1999). The stromal-cell-derived factor 1α (SDF-1α, also known as CXCL12) is the only known ligand for CXCR4 and was first characterized as a pre-B-cell growth-stimulating factor that is essential for B-cell lymphopoiesis, myelopoiesis, cardiogenesis and embryogenesis (Nagasawa, 1996). The expression of CXCR4 in a wide variety of tumours including breast cancer, coupled with the expression of SDF-1α in sites of breast cancer metastasis such as the lungs, liver and bone, suggests a crucial role for CXCR4/SDF-1 in organ-specific breast cancer metastasis (Muller et al., 2001; Smith et al., 2004).

For optimal signalling, CXCR4 must be embedded in membrane lipid rafts (Manes et al., 2001; Shamri et al., 2002; Nguyen and Taub, 2002; Wysoczynski et al., 2005). Here we demonstrate that the presence of CD24 in lipid rafts alters the raft residence and response of CXCR4. Our results in mouse pre-B cell lines suggest that this occurs because of changes in membrane cholesterol, affecting CXCR4-triggered cell migration and ERK1/2 phosphorylation. Furthermore, altered lipid raft residence and chemotactic migration in response to SDF-1 was demonstrated in MDA-MB-231 breast carcinoma cells expressing high and low levels of CD24. In addition we show that low expression of CD24 augmented the growth of tumours from MDA-MB-231 breast cancer cells in NOD/SCID mice. As CXCR4 is implicated in metastasis and growth of breast and other cancers, the regulatory role of CD24 on CXCR4 signalling may be of great interest for future research.

The production and characterization of CD24-knockout mice were described in detail before (Nielsen et al., 1996). We isolated CD19⁺ B lymphocytes via depletion of non-B lymphocytes from the bone marrow and spleen of CD24^–/– or CD24^+/+ mice. The enrichment was monitored by FACS analysis with appropriate markers (Fig. 1A). We compared the SDF-1-mediated migration of CD19⁺ cells derived from the bone marrow and spleen of CD24^+/+ and CD24^–/– mice. Bone-marrow-derived cells from CD24^–/– mice showed an ∼2.8-fold enhancement in migration compared with cells derived from CD24^+/+ mice (Fig. 1B). The difference in SDF-1-induced migration was not observed in CD19⁺ cells isolated from the spleen (Fig. 1C).

Having established that CD24 can affect SDF-1-mediated cell migration of bone-marrow-derived B cells, we investigated whether this was also detectable in pre-B-cell lines. The pre-B lymphocytic cell line N232.18 (CD24^–/–) was established from CD24-knockout mice and retransfected with a CD24 expression plasmid resulting in the cell line 18H18⁺ (CD24^+/+) (Hahne et al., 1994). Additionally, a mock-transfected line 18H18^– (CD24^–/–) was established. As depicted in Fig. 2A, the cell lines differed in CD24 expression but were similar in expression of CXCR4. In chemotactic migration assays towards SDF-1, we showed that N232.18 cells and 18H18^– cells migrated in response to SDF-1 in a dose-dependent manner (Fig. 2B). By contrast, CD24^+/+ cells were only weakly responsive to SDF-1 and showed a ∼50% reduction in chemotaxis compared with CD24^–/– cells (Fig. 2B).

In addition to SDF-1 we used the chemokines IP-10 (CXCL10), MIP-3α (CCL20) and BLC (CXCL13) to investigate whether CD24 was able to influence migration. All three chemokines failed to induce migration of CD24^+/+ and CD24^–/– pre-B-cell lines (Fig. 2C). Similar results were obtained with isolated CD19+ B cells from bone marrow (data not shown).

We also examined the expression of CXCR4 by biochemical means. Fig. 2D (upper panel) shows the analysis of cell lysates of the three cell lines indicating that the CXCR4 expression was comparable. We next investigated whether there was a change in the binding affinity of SDF-1 to its receptor by carrying out ligand binding with biotinylated SDF-1, then detecting the bound ligand by FACS analysis. SDF-1 binding was indeed slightly elevated in both CD24^–/– cell lines compared with CD24^+/+ cells (Fig. 2E).

It has been reported that in lymphoid cells, the lack of GPI-anchored proteins can cause an increase in cellular cholesterol (Abrami et al., 2001). As CD24 is a prominent GPI-anchored molecule in pre-B cells, we hypothesized that the lack of CD24 might have a similar effect. To test this, we measured the cellular cholesterol content in CD24^–/– and CD24^+/+ cell lines. The amount of cellular cholesterol was significantly higher in the two CD24^–/– lines compared with CD24^+/+ cells (Fig. 3A). Cellular cholesterol can be visualized by staining with the polyene antibiotic Filipin that specifically binds to unesterified cholesterol (Muller et al., 1984). We used FACS analysis of Filipin-stained cells to assess membrane cholesterol levels in CD24^+/+ versus CD24^–/– cells. The mean fluorescence intensity of stained cells was significantly higher in CD24^–/– cells compared with CD24^+/+ cells (Fig. 3B).

The function of the chemokine receptor CXCR4 is sensitive to membrane cholesterol extraction (Nguyen and Taub, 2002; Nguyen and Taub, 2003). As CD24 appeared to affect SDF-1 responsiveness, by possibly altering cholesterol levels, we examined whether the rate of cell migration was sensitive to cholesterol manipulation. Therefore we incubated 18H18⁺ cells with soluble cholesterol to restore the diminished SDF-1 signalling. This treatment resulted in an elevation of chemotactic migration by approximately 40% compared with untreated cells (Fig. 4A, left panel). The high-cholesterol-containing cell lines N232.18 and 18H18^– were treated with the HMG-CoA-reductase inhibitor fluvastatin in the presence of lipoprotein-deficient serum. This treatment inhibits cholesterol biosynthesis as well as cholesterol uptake. Both CD24^–/– cell lines became less responsive to SDF-1 and displayed a reduction in chemotactic migration by 80% and 60%, respectively, compared with untreated cells (Fig. 4A, middle and right panels). Reduced SDF-1-triggered migration was also seen after treatment of N232.18 and 18H18^– cells with the cholesterol extracting agent methyl-β-cyclodextrin (MCD; data not shown).

CXCR4 signalling via SDF-1 activates the MAP-kinase pathway by phosphorylation of ERK (Ganju et al., 1998). SDF-1 triggered ERK phosphorylation in CD24^–/– cells but not in CD24^+/+ 18H18⁺ cells (Fig. 4B). To examine whether the ERK phosphorylation depended on cholesterol, we repeated the treatment with the HMG-CoA-reductase inhibitor fluvastatin for 72 hours with CD24-negative (18H18^–) and CD24-positive (18H18⁺) cell lines. Fluvastatin treatment abolished SDF-1-induced ERK1/2 phosphorylation compared with untreated cells (Fig. 4C). We also examined the potential of cholesterol loading to restore the SDF-1-induced phosphorylation of the CXCR4 downstream target ERK1/2 in CD24-positive 18H18⁺ cells. Incubation with soluble cholesterol leads to a partial restoration of ERK1/2 phosphorylation in response to SDF-1 (data not shown).

As it has been published that CXCR4 is an important factor in breast cancer metastasis (Muller et al., 2001) we wanted to know whether the results obtained from pre-B cells could be transferred to a breast cancer cell line model. Therefore, we overexpressed CD24 in MDA-MB-231 breast cancer cells and established cells with CD24^high and CD24^low expression levels at the cell surface. Fig. 5A (left panel) depicts the different CD24 levels in the two MDA-MB-231 sub-lines, as examined by FACS analysis. Overexpression of CD24 did not affect the level of other antigens such as the integrin α₅β₁ or L1 adhesion molecule (data not shown). Interestingly, as measured by Filipin staining, no significant difference in membrane cholesterol was detected (data not shown).

Both cell lines expressed low levels of CXCR4 at the cell surface (Fig. 5A, lower left panel). We therefore overexpressed a CXCR4-GFP fusion protein by retroviral transduction in both sub-lines. Overexpression followed by FACS selection led to a clear enrichment of CXCR4 at the cell surface of the CD24^high and CD24^low sub-lines (Fig. 5A, lower right panel). Next we analysed the chemotactic migration of all cell lines in response to SDF-1. Despite their low level of endogenous CXCR4 at the cell surface, the non-transduced MDA-MB-231 cells showed a chemotactic response to SDF-1 (Fig. 5B, left panel). The CD24^low-expressing MDA-MB-231 cells showed a ∼1.3- to 2-fold increase in migration compared with the MDA-MB-231 CD24^high sub-line. This difference in migration behaviour owing to the expression of CD24 was further confirmed in the CXCR4-GFP-overexpressing cell lines. Fig. 5B (right panel) depicts an overall increase in motility owing to overexpression of CXCR4-GFP and approximately twofold elevated migration in response to SDF-1 in MDA-MB-231 CD24^low cells compared with the corresponding CD24^high-expressing cells.

To further corroborate these results, we used siRNA-mediated depletion of CD24 in CD24^high-expressing cells. Transfection of the CD24-specific siRNA led to a significant reduction of CD24 expression on the cell surface as revealed by FACS staining (Fig. 5C, right panel). The CD24-depleted cells showed approximately threefold elevated migration compared with control-siRNA-transfected cells (Fig. 5C, left panel). Thus, CD24 expression levels can interfere with CXCR4-mediated cell migration.

The SDF-1/CXCR4 axis can regulate the proliferation of progenitor cells (Smith et al., 2004; Zeelenberg et al., 2003; Orimo et al., 2005) but also promotes the growth of primary and metastatic breast cancer cells (Smith et al., 2004.) We examined the tumour growth of our breast cancer cell lines in NOD/SCID mice. We injected subcutaneously MDA-MB-231 CD24^high and MDA-MB-231 CD24^low as well as MDA-MB-231 CD24^high CXCR4-GFP and MDA-MB-231 CD24^low CXCR4-GFP cells. The tumour volume was monitored over 40 days after which the mice were sacrificed and tumours were prepared. MDA-MB-231 CD24^low cells formed tumours of approximately fivefold larger volume compared with the CD24^high variant (Fig. 6A). MDA-MB-231 CD24^low CXCR4-GFP tumours developed a significantly (approximately threefold) increased volume compared with that of the CD24^high variant (Fig. 6B). Therefore it seems that low CD24 expression favours tumour growth in NOD/SCID mice possibly by its influence on CXCR4 signalling.

We also determined the influence of SDF-1 on the growth of MDA-MB-231 cells in vitro using proliferation assays. MDA-MB-231 CD24^high and MDA-MB-231 CD24^low breast cancer cells were incubated with different concentrations of SDF-1 and cell numbers were determined after 24 hours. In the presence of SDF-1 MDA-MB-231 CD24^low cells showed significantly higher proliferation compared with MDA-MB-231 CD24^high cells (Fig. 6C).

We finally assessed the distribution of the CXCR4 receptor at the cell surface biochemically. CXCR4 is localized within lipid rafts that are rich in cholesterol (Nguyen and Taub, 2002; Wysoczynski et al., 2005). Therefore, we examined the distribution of CXCR4 and CD24 in pre-B cell lines by sucrose density centrifugation after lysis in cold Triton X-100 buffer. As expected for a GPI-anchored protein, CD24 was detected in the lipid raft fractions of CD24^+/+ pre-B cells (Fig. 7A). CXCR4 was observed in the intermediate non-raft fractions but not within the lipid raft fractions of the gradient (Fig. 7B). By contrast, in both CD24^–/– cells the CXCR4 receptor was partially found within lipid rafts (Fig. 7B). To control for equal loading, the western blot was reprobed with an antibody against the raft-associated kinase Fyn. In all pre-B-cell lines Fyn was equally distributed within lipid raft and in intermediate non-raft fractions (Fig. 7B).

To determine whether there was also a different CXCR4 lipid raft pattern in breast cancer cells, we isolated lipid raft fractions of CD24^high and CD24^low-expressing breast cancer cell lines. CXCR4 was present within the lipid raft fractions of MDA-MB-231 CD24^low cells (Fig. 7C, lower panel). By contrast, CXCR4 was absent in rafts of CD24^high-expressing MDA-MB-231 cells (upper panel). The distribution of the raft-associated kinase Fyn was equal in both sub-lines. We conclude that CD24 affects the raft residence of the CXCR4 receptor in both pre-B cells and in breast carcinoma cells.

In the present report we demonstrate that (1) CD24 expression in pre-B cells affects the responsiveness of the CXCR4 receptor; (2) a similar observation was made in MDA-MB-231 breast carcinoma cells; (3) siRNA-mediated depletion of CD24 in MDA-MB-231 CD24^high-expressing cells enhances CXCR4-receptor-mediated cell migration; (4) in pre-B cells, but not in carcinoma cells, the absence of CD24 affects the level of cellular cholesterol; (5) in carcinoma cells CD24 expression reduced tumour growth in NOD/SCID mice; (6) CD24-expressing cells differed from non-expressing cells in the residence of CXCR4 in membrane rafts. These findings suggest that CD24 regulates CXCR4 receptor responsiveness by excluding it from membrane rafts. Our findings establish a new mechanism by which the function of CXCR4 can be modulated without affecting protein expression levels.

Our study was inspired by the work of Abrami and co-workers who show that in mouse BW5147 cells deficient in GPI-anchor synthesis, the level of cellular cholesterol was enhanced (Abrami et al., 2001). By contrast, in CHO cell clones rendered resistant to proaerolysin and displaying also a lack in the synthesis of GPI-anchored proteins, a similar influence on cholesterol levels was not observed. Instead, these clones revealed an enhanced content of caveolin-1 (but not of other raft-associated proteins such as flotillin-1). It was argued that caveolin-1 was capable of counterbalancing the loss of GPI-anchored proteins. In our study we hypothesized that the loss of a major B-cell GPI-anchored protein such as CD24 might cause a similar effect in pre-B cells. Indeed, we found a significantly enhanced level of cholesterol in CD24-deficient cells. The observed effect was surprising as murine pre-B cells express other GPI-anchored proteins such as CD48, CD59, CD157 or Ly-6A/E. We speculate that CD24 represents a major GPI-anchored protein and that the lack of expression may cause similar effects to those observed in cells in which GPI-anchor synthesis is blocked.

Cholesterol homeostasis in cells is subject of complex regulations consisting of uptake via the LDL receptor, export via HDL and lipoproteins or membrane vesicles and de novo biosynthesis (reviewed by Simons and Ikonen, 2000). It is presently unclear, by which mechanisms CD24^–/– cells acquire elevated cholesterol. Our preliminary results, using fluorescent LDL-uptake assays, suggest that both CD24^–/– and CD24^+/+ cells bind and take up LDL and that CD24^+/+ cells display a higher rate (H.S., unpublished results). This is expected given the lower level of cholesterol in these cells. As lymphoid cells also display low levels of cholesterol de novo synthesis, it is likely that the regulation is mostly via the export of cholesterol. Further experiments are needed to more closely define this mechanism.

Previous studies have shown that the function of the CXCR4 receptor is sensitive to cholesterol manipulation (Nguyen and Taub, 2002 and Nguyen and Taub, 2003). These authors claimed that normal cholesterol is crucial to the conformational integrity and function of the chemokine receptors CXCR4 and CCR5. In our study, we investigated the possibility of a link between cholesterol levels and CD24 expression and used SDF-1-triggered cell migration as a sensitive readout. Indeed, the influence of CD24 on cell motility was evident in pre-B-cell lines and also in freshly isolated CD19⁺ pre-B cells from the bone marrow but not from the spleen. Interestingly, the phenotype of CD24^–/– mice displays a slight block in the differentiation of B cells and a reduction in cell number of pre-B cells in the bone marrow compared with CD24^+/+ mice (Nielsen et al., 1997). However, numbers of mature B-cells were normal in the spleen of CD24^–/– mice. The results presented in this report can offer an explanation for these findings. It is possible that SDF-1 presented by the bone marrow stroma of CD24^–/– animals may lead to hyperactivation of the CXCR4 receptor owing to the lack of CD24. This could result in aberrant motility and guidance within the bone marrow of CD24^–/– cells. It is also possible that the effects caused by the absence of CD24 are less prominent in mature B cells in the periphery compared with the bone marrow which could be explained by an antagonistic effect of other GPI-anchored molecules.

CD24 also modulated the responsiveness of the CXCR4 receptor in MDA-MB-231 breast carcinoma cells. In this cellular system of high- or low-expressing variants, the total cellular cholesterol content was not altered by CD24. This could be due to inherent differences between lymphoid and epithelial cells or to the presence of caveolins as suggested previously for CHO cells (Abrami et al., 2001). We observed that CD24^low cells migrated better in response to SDF-1, showed increased proliferation in the presence of SDF-1 and exhibit enhanced tumour growth in NOD/SCID mice. Interestingly, we did not observe elevated growth in CXCR4-overexpressing cells in our NOD/SCID mice experiments (Fig. 6A,B). This is in contrast to earlier reports which showed that overexpression of CXCR4 in MDA-MB-231 breast cancer cells led to increased tumour volume in NOD/SCID mice (Darash-Yahana et al., 2004). The reason for this discrepancy could be that breast cancer cells used in this previous report possess no endogenous CXCR4 expression whereas our MDA-MB-231 cells show moderate levels of functional CXCR4 receptor endogenously on the cell surface (Fig. 5A,B). This may lead to saturation, because the endogenous CXCR4 receptor is already able to promote tumour growth in such a way that cannot be further augmented by additional overexpression of CXCR4.

Importantly, in the pre-B cell and in the breast cancer cell-line system we observed that the presence of CD24 caused a depletion of the CXCR4 receptor from membrane rafts. It is well established that lipid rafts represent important platforms to couple receptors to signalling machinery (Harder and Engelhardt, 2004). Owing to the presence of kinases, G-proteins and other signalling components, lipid rafts play a pivotal role in signal transduction (Brown and London, 1998; Kurzchalia and Parton, 1999). Cholesterol is an integral component of membrane rafts and GPI-anchored proteins such as CD24 are exclusively present in rafts.

The importance of the incorporation of the CXCR4 receptor into membrane lipid rafts has recently been analysed (Wysoczynski et al., 2005). Molecules such as fibrinogen and fibronectin have a priming effect on hematopoietic stem/progenitor cells. The interaction between these molecules, receptors and other adhesion molecules on the cell surface seems to be crucial for increasing the incorporation of CXCR4 into membrane lipid rafts. The described raft residence of CXCR4 is in line with other reports which show that the inability of SDF-1 to bind to non-raft-associated CXCR4 may play a regulatory role in maintaining receptor activity and the migratory potential of cells (Nguyen and Taub, 2002). In the context of α₄ integrin activation in lymphocytes, it has been described that CXCR4 has to be embedded within lipid rafts. CXCR4, as a G-protein coupled receptor, requires intact membrane rafts to convert chemokine signalling into productive α₄ integrin avidity stimulation (Shamri et al., 2002). Our data are in agreement with these earlier findings. We observed in pre-B cells that owing to the depletion of CXCR4 from rafts there was also an uncoupling of the receptor from downstream signalling components. In CD24^+/+ 18H18⁺ cells, no phosphorylation of ERK in response to SDF-1 was observed. As summarized in the model presented in Fig. 8, we propose that CD24 affects CXCR4 function by modulating the raft residence of the receptor.

Previous studies have shown that protein tyrosine phosphatases (SHIP1 and SHIP2), as well as membrane-expressed hematopoietic phosphatase CD45 are also involved in the modulation of CXCR4 signalling (Fernandis et al., 2003). SDF-1-triggered migration is impaired in SHIP1/2-knockout mice (Chernock et al., 2001) and lymphocytes negative for CD45 show reduced chemotaxis towards SDF-1 (Fernandis et al., 2003). Interestingly, following SDF-1 stimulation, CD45 was found to colocalize with CXCR4 in lipid rafts (Fernandis et al., 2003). Other molecules have been identified that may increase the sensitivity/responsiveness of CXCR4-positive cells to SDF-1 such as complement cleavage fragments, platelet-derived microvesicles, hyaluronic acid, fibronectin or soluble uPAR (Kucia et al., 2004; Wysoczynski et al., 2005). Factors such as LPS, heparin or other chemokines (MIP1α or RANTES) can desensitize the CXCR4 receptor whereas the action of proteinases such as CD26 (dipetidylpeptidase IV) under inflammatory conditions can silence CXCR4/SDF-1 signalling by cleavage of the ligand or receptor (Kucia et al., 2004; Epstein, 2004). The mechanism of CXCR4 silencing by CD24 is novel as it does not interfere with protein expression but is mediated by uncoupling of the receptor.

The SDF-1/CXCR4 axis was initially studied for its importance in initiating cell trafficking. The biological effects of this interaction were studied in the context of cell motility during inflammation (Gerard and Rollins, 2001; Moser et al., 1998), tissue regeneration (Kucia et al., 2004), HIV infection (Berger et al., 1999) and lately also concerning the ability to promote cell growth and survival of progenitor and tumour cells (Smith et al., 2004; Muller et al., 2001; Kucia et al., 2004; Epstein, 2004). Many tumour cells express functional CXCR4 receptor and several recent reports have shown that CXCR4 receptor expression predicts a poor outcome for cancers of the breast (Muller et al., 2001), skin (Scala et al., 2005), ovary (Scotton et al., 2002) and prostate (Darash-Yahana et al., 2004). It is believed that CXCR4 can direct tumour cells into SDF-1-rich tissues or organs of the body such as bone marrow and lymph nodes (Muller et al., 2001). SDF-1 is also secreted by the liver and kidneys and the central nervous system (Stumm et al., 2002) and secretion can be enhanced by tissue damage. Tumour cells, in particular so called `tumour stem cells' may lodge and grow in SDF-1-rich niches (Epstein, 2004).

In particular, when present in the cytoplasm of tumour cells, CD24 expression has also been associated with poor prognosis in breast (Kristiansen et al., 2003), ovarian (Kristiansen et al., 2002; Choi et al., 2005) and prostate cancer (Kristiansen et al., 2004a). Interestingly, a recent study on primary breast cancer cells isolated from pleural infusions of tumour patients has shown that the ability to form tumours in NOD/SCID mice is far greater in the CD44⁺CD24^low/– fraction compared with the CD44⁺CD24^high fraction (Al-Hajj et al., 2003; Abraham et al., 2005). Xenografted tumours appearing in mice injected with CD44⁺CD24^low/negative cells were again heterogeneous in expression and ranged from CD24^high to CD24^low expression (Al-Hajj et al., 2003). Owing to its enhanced tumour-forming ability, the CD44⁺CD24^low/– fraction was considered to represent `breast tumour stem cells' (Al-Hajj et al., 2003). A recent study has challenged this view by demonstrating that the percentage of CD44⁺CD24^low/– cells in breast carcinoma tissue samples is highly variable (between 0-80%) (Abraham et al., 2005). The size of this sub-fraction had no influence on overall survival time but was favourable for distant metastasis. Our study suggests that low CD24 expression levels can promote the growth ability of breast tumour cells by augmenting CXCR4 responsiveness.

In summary, our results demonstrate for the first time a novel function of CD24 as a regulator of CXCR4 responsiveness in lymphoid cells and in carcinoma cells. This effect is due to the unique localization of CD24 in rafts. The mechanism by which CD24 uncouples CXCR4 signalling and the possibility that CD24 can regulate other raft-associated cell functions warrants further investigation.

The following antibodies were used: PE-conjugated monoclonal antibody (mAb) M1/69 to mouse CD24 and the corresponding PE conjugated isotype control antibody from Becton Dickinson (Heidelberg, Germany), mAb 79 to mouse CD24 (Kadmon et al., 1994) and mAb SWA11 to human CD24 (Weber et al., 1993); biotinylated mAb to mouse CXCR4 from Becton Dickinson (clone 2B11), mAb to mouse CXCR4 from Capralogics (Hardwick, MA) (CI00116) and ProSci (Poway, CA) (1012), and to human CXCR4 from R&D Systems (Wiesbaden, Germany) (MAB 172, clone 44716) and Chemicon (Hofheim, Germany) (AB1847); mAb to phospho-specific ERK1/2 (clone 20A) and ERK 1 (clone MK12) from Becton Dickinson; polyclonal Ab to Calnexin from Stressgen (Victoria, Canada) (# SPA-860); polyclonal Ab to Fyn from Santa Cruz Biotechnologies (sc-016); PE-conjugated mAb to mouse CD45 (CD45R/B220) from Becton Dickinson (553090). SDF-1α, biotinylated SDF-1α, streptavidin-FITC, IP-10, MIP-3α and BLC were all obtained from R&D Systems.

The LZRS-CXCR4-GFP-IRES-Zeocin construct (kind gift from Dr Hordijk, University of Amsterdam, The Netherlands) was transfected into amphotrophic Phoenix packaging cells via calcium phosphate transfection to produce retroviruses.

MDA-MB-231 CD24^high or CD24^low were infected with retrovirus-containing supernatant in the presence of polybrene (Sigma, Taufkirchen, Germany) and stably transduced cells were sorted for CXCR4 expression by FACSVantage.

The production and characterization of the CD24^–/– pre-B lymphocytic cell line N232.18 and the CD24 retransfected sub-line 18H18⁺ have been previously described (Hahne et al., 1994). 18H18^– cells were obtained from 18H18⁺ cells by selecting CD24-loss variants using FACS. Pre-B cells were cultured in RPMI 1640 containing 10% fetal calf serum, glutamine and penicillin/streptomycin. CD24^–/– mice (C57BL/6) were initially obtained from Peter Nielsen (Max-Planck Institute for Immunobiology, Freiburg, Germany) and bred at the DKFZ. Bone marrow and spleen cells were collected from 6- to 8-week-old female mice and CD19+ B lymphocytes were isolated using a mouse B-cell isolation kit (Miltenyi, Bergisch-Gladbach, Germany). MDA-MB-231 CD24^high cells were established by transfection with a CD24-pcDNA3.1 expression plasmid with SuperFect from Qiagen (Hilden, Germany). MDA-MB-231 CD24^high and MDA-MB-231 CD24^low breast cancer cell lines were maintained in DMEM with 10% fetal calf serum, glutamine and penicillin/streptomycin.

Cells were washed, resuspended in cold PBS containing 5% fetal calf serum, and then incubated with mAb to CD24 or CXCR4 for 30 minutes followed by washing and incubation for 20 minutes with PE-conjugated IgG secondary antibodies (Jackson ImmunoResearch). Mouse CXCR4 was detected using biotinylated CXCR4 mAb and Streptavidin conjugated PE (Jackson ImmunoResearch). Cells were analysed with a FACScan or sorted with FACSVantage (Becton Dickinson, Heidelberg, Germany). For data analysis, Cellquest or FlowJow software was used.

Cells were fixed with 1% formaldehyde/PBS and then stained with Filipin III (Sigma F4767; Taufkirchen, Germany) at a concentration of 0.05 mg/ml in 10% FCS/PBS for 1 hour at room temperature in the dark. Filipin was measured with FACSVantage or FACSDiva instruments using a UV laser (Becton Dickinson).

For cholesterol quantification, lipids were extracted by incubation in hexane/isopropanol (3:2) for 30 minutes on ice. Following overnight evaporation of the solvent, the dried lipids were resolved in absolute ethanol and the relative cholesterol content was measured with an Amplex Red cholesterol assay from Molecular Probes (PoortGebouw, Netherlands). The fluorometric method was used according to the manufacturer's instruction.

To inhibit cholesterol uptake as well as cholesterol biosynthesis, cells were cultured in RPMI medium containing 10% lipoprotein-deficient serum (Sigma) in the presence of the HMG-CoA-reductase inhibitor fluvastatin (kind gift from Jörg Kreuzer, University of Heidelberg, Germany) at 0.5 μM for 72 hours. For cholesterol loading, 18H18+ cells were incubated with soluble cholesterol (Sigma) at 30 μg/ml for 30 minutes at 37°C.

Around 5×10⁵ pre-B cells in 100 μl RPMI/0.5% BSA were added to Transwell inserts (Costar) with a 5 μm pore size. Serially diluted recombinant SDF-1α or other chemokines were added to the lower chamber and pre-B cells were allowed to migrate up to 16 hours. Transwell inserts were removed and the migrated cells in the lower chamber were counted with a Coulter Counter Z2 (Beckman Coulter; Krefeld, Germany). MDA-MB-231 cells were serum-starved overnight before application to Transwell inserts with an 8 μm pore-size. To facilitate migration, the underside of the filters were coated with fibronectin (7.5 μg/ml). Approximately 3-5×10⁵ cells in 100 μl RPMI/0.5% BSA were suspended in the upper chamber. Recombinant SDF-1α was added to the lower chamber, then after 4-6 hours of incubation, the cells on the upper surface of the filters were removed by wiping with cotton swabs. Migrated cells on the lower surface of the filter were detached using 100 mM EDTA in PBS and counted with a Coulter Counter Z2.

Control siRNA and CD24 siRNA duplex (sense, 5′-ACA ACA ACU GGA ACU UCA A dTdT-3′) were purchased from MWG Biotech (Ebersberg, Germany). CD24 and control siRNAs were transfected into MDA-MB-231 CXCR4-GFP CD24^high cells at a final concentration of 100 nM with Oligofectamine transfection reagents (Invitrogen, Karlsruhe, Germany).

Cells were lysed in ice-cold lysis buffer in 25 mM Tris-HCl pH 8.0 containing 1% Triton X-100 (Roche), 1 mM PMSF, 10 μg/ml aprotinin, pepstatin and leupeptin, 100 mM NaF and 10 mM Na₃VO₄ for 30 minutes on ice. The lysate was mixed with an equal volume of 85% sucrose (w/v in TBS), and a step gradient was prepared by overlaying with 35% sucrose in TBS followed by a final layer of 5% sucrose. The gradient was centrifuged for 20-22 hours at 200,000 g using a Beckman SW60 rotor. Fractions of 0.5 ml were collected from the top of the gradient and precipitated with chloroform/methanol as described previously (Wessel and Flügge, 1984). For CD24 detection, fractions were precipitated with a tenfold volume of acetone and then washed with a fivefold volume of 50% acetone/H₂O. Fractions were then boiled in the presence of SDS sample buffer under reducing conditions. Equal protein loading of the cell lysates was standardized with the Biorad protein determination kit. Samples were separated on 10% SDS-PAGE gels and transferred to Immobilon membranes using semi-dry blotting. After blocking with 5% skimmed milk in TBS, membranes were probed with primary antibodies followed by horseradish peroxidase-conjugated anti-mouse or anti-rabbit secondary antibodies and ECL detection (Amersham-Pharmacia, Freiburg, Germany).

Approximately 1×10⁷ cells were injected subcutaneously into the left and right flank, respectively, of female NOD/SCID mice. Tumour growth was monitored every 3-4 days over 40 days at which point the experiment was terminated and tumours were collected for histological evaluation. At different time points the tumour was measured and the volume was calculated using the formula: V=(L×W²)π /6.

Around 5×10⁴ cells were grown in triplicate on 24-well plates in DMEM/10% FCS. After 24 hours, medium was removed and replaced with serum-free medium with or without 10, 20 or 30 nM SDF-1. Cells were cultured for additional 24 hours and then harvested and counted with a Coulter Counter Z2.

Statistical significance was assessed using a Student's t-test. The correlation coefficient was calculated using Microsoft Excel software.

We thank Verena Gschwend for help with isolating bone-marrow-derived B cells; Klaus Hexel for FACS sorting and Filipin measurement and Michael Sanderson for helpful discussions.