Chondroitin sulfate glycosaminoglycans as major P-selectin ligands on metastatic breast cancer cell lines


  • Behjatolah Monzavi-Karbassi,

    Corresponding author
    1. Arkansas Cancer Research Center and Department of Pathology, University of Arkansas for Medical Sciences, Little Rock, AR
    • Arkansas Cancer Research Center, University of Arkansas for Medical Sciences, 4301 West Markham St., Slot 824, Little Rock, AR 72205, USA
    Search for more papers by this author
    • Fax: +1501-526-5934.

  • J. Steven Stanley,

    1. Arkansas Cancer Research Center and Department of Pathology, University of Arkansas for Medical Sciences, Little Rock, AR
    Search for more papers by this author
  • Leah Hennings,

    1. Arkansas Cancer Research Center and Department of Pathology, University of Arkansas for Medical Sciences, Little Rock, AR
    Search for more papers by this author
  • Fariba Jousheghany,

    1. Arkansas Cancer Research Center and Department of Pathology, University of Arkansas for Medical Sciences, Little Rock, AR
    Search for more papers by this author
  • Cecile Artaud,

    1. Arkansas Cancer Research Center and Department of Pathology, University of Arkansas for Medical Sciences, Little Rock, AR
    Search for more papers by this author
  • Saeid Shaaf,

    1. Arkansas Cancer Research Center and Department of Pathology, University of Arkansas for Medical Sciences, Little Rock, AR
    Search for more papers by this author
  • Thomas Kieber-Emmons

    1. Arkansas Cancer Research Center and Department of Pathology, University of Arkansas for Medical Sciences, Little Rock, AR
    Search for more papers by this author


The metastatic breast cancer cell line, 4T1, abundantly expresses the oligosaccharide sialylated Lewis x (sLex). SLex oligosaccharide on tumor cells can be recognized by E- and P-selectin, contributing to tumor metastatic process. We observed that both selectins reacted with this cell line. However, contrary to the E-selectin reactivity, which was sLex dependent, P-selectin reactivity with this cell line was sLex-independent. The sLex-Neg variant of the 4T1 cell line with markedly diminished expression of sLex and lack of sLea, provided a unique opportunity to characterize P-selectin ligands and their contribution to metastasis in the absence of overlapping selectin ligands and E-selectin binding. We observed that P-selectin binding was Ca2+-independent and sulfation-dependent. We found that P-selectin reacted primarily with cell surface chondroitin sulfate (CS) proteoglycans, which were abundantly and stably expressed on the surface of the 4T1 cell line. P-selectin binding to the 4T1 cells was inhibited by heparin and CS glycosaminoglycans (GAGs). Moreover, Heparin administration significantly inhibited experimental lung metastasis. In addition, the data suggest that surface CS GAG chains were involved in P-selectin mediated adhesion of the 4T1 cells to murine platelets and human umbilical vein endothelial cells. The data suggest that CS GAGs are also the major P-selectin-reactive ligands on the surface of human MDA-MET cells. The results warrant conducting clinical studies on the involvement of cell surface CS chains in breast cancer metastasis and evaluation of various CS types and their biosynthetic pathways as target for development of treatment strategies for antimetastatic therapy of this disease. © 2006 Wiley-Liss, Inc.

Breast cancer metastasis represents a terrible milestone associated with a poor prognosis. The multistep process of metastasis includes release of malignant cells from the primary neoplasm, migration of cancer cells into the blood circulation, interaction with platelets and leukocytes in circulation, adhesion to the vascular endothelium in distant organs and growth of the disseminated cancer cells within the vessels or within the tissue following extravasation.1, 2, 3 Each step in this process requires different types of interactions between cancer cells and the host microenvironment.

A widely accepted hypothesis is that cancer cells exploit the adhesion molecules used by hematopoietic cells to migrate into distant organs.4 In particular, the selectin family of adhesion molecules whose ligands, sialyl Lewis x (sLex) and sialyl Lewis a (sLea), are expressed at elevated levels on various cancer cells of human and murine origin5, 6, 7, 8, 9 and play a significant role in cancer metastasis. The oligosaccharides sLex and sLea are the prime ligands for P- and E-selectin and many published works underscore the relative importance of interactions between sLex/a and these vascular receptors in the hematogenous spread of cancer cells.7, 10, 11 Studies investigating the survival of postsurgical patients with various carcinomas demonstrate that survival correlates inversely with tumor expression of sLex.12, 13, 14 This supports the view that expression of sLex on tumor cells positively affects metastatic potential. This effect was thought to be mediated via a heterotypic cell adhesion between tumor sLex and P- and E-selectins on platelets and endothelium.15, 16, 17 In this regards, various strategies have been taken to block metastasis by targeting these interactions.18, 19, 20, 21, 22 An accumulating body of evidence indicates that P-selectin plays a crucial role during hematogenous metastasis, and this lectin has been shown to bind to several human cancer cell lines, such as colon cancer, lung cancer, breast cancer, malignant melanoma, gastric cancer, tongue squamous cancer and neuroblastoma.8, 15, 22, 23, 24, 25, 26, 27, 28, 29

We have shown recently, in the 4T1 spontaneous metastasis model, that deficiency in sLex/a oligosaccharides was associated with an increase in metastatic potential of these breast cancer cells.3 The data suggested that this deficiency was associated with diminished homotypic adhesion and higher motility of the tumor cells.3 However, while a decrease in homotypic adhesion can help in the release of metastatic cells from the primary tumor, the ability of the circulating tumor cell to extravasate and form secondary metastatic lesions may be disrupted if their adhesion to endothelial cells and platelets is also decreased. We have shown that P-selectin reacted with both the parental 4T1 cells and the sLex-Neg variant almost equally.3 To further elucidate the relative role of P- and E-selectin ligands in metastasis through homo- or hetero-typic adhesion, we characterized P-selectin ligands on the mammary cell line 4T1, which is purported to mimic stage IV human breast carcinoma.30 We show here that while P-selectin reacts strongly with 4T1 cells, its reactivity does not depend on sLex/a expression. This explains how reduced sLex/a levels can decrease homotypic adhesion without affecting heterotypic adhesion resulting in an increase in metastatic potential. In this regard, the role of sLex oligosaccharide in tumor metastasis must be reevaluated in light of the identification of an increasing number of ligands for the selectin family of adhesion molecules and the aberrant glycosylation present in tumor cells.

Cell surface proteoglycans (PGs), another class of surface adhesion molecules composed of glycosaminoglycan (GAG) side chains covalently bound to a protein core, may play a dual role in invasion and metastasis of tumor cells.31 Some evidence suggest that tumor cell surface PGs inhibit metastasis via promoting cell–cell and cell–extracellular matrix adhesion,32 while others imply a prometastatic role for the same molecules.33 It is known that heparan sulfate (HS) and chondroitin sulfate (CS) PGs are P-selectin ligands.28, 34, 35, 36 Therefore, it is reasonable to expect that they play a role in tumor cell adhesion to platelets and endothelium, promoting extravasation of tumor cells once in circulation.4 Here, we demonstrate that CS PGs on the surface of 4T1 cells are major P-selectin ligands, which are involved in prometastatic heterotypic adhesion of tumor cells to platelets and endothelial cells.


CS, chondroitin sulfate; GAG(s), glycosaminoglycans(s); HS, heparan sulfate; HUVECs, human umbilical vein endothelial cells; PG(s), proteoglycan(s); sLea, sialyl Lewis a; sLex, sialyl Lewis x.

Material and methods

Reagents, antibodies and cell lines

The mAb KM-93 was obtained from Kamiya Biomedical, Seattle, WA. FH6 and CSLEX1 were from GlycoTech, Gaithersburg, MD. The anti-CS mAb 2-B-6 and anti-HS mAb F58-10E4 (10E4) were from Associates of Cape Cod, (Falmouth, MA). Fluorescence-conjugated anti-human IgG and anti-mouse IgG/IgM were obtained from Sigma (St. Louis, MO). Recombinant E- and P-selectin/Fc (human IgG) were from R&D systems (Minneapolis, MN). The biotinylated Maackia amurensis lectin II (MAL-II) and Sambucus nigra lectin (SNA) were obtained from Vector laboratories (Burlingame, CA). The pIRES-EGFP vector was obtained from BD-Clontech Laboratories, (Palo Alto, CA). The murine breast tumor cell line 4T1 was obtained from ATCC (Manassas, VA). The sLex-Neg variant of the 4T1 cell line was previously isolated by in vitro enrichment of a subpopulation negative for the mAb KM93 binding.3 The human MDA-MET cell line, which has been selected in vivo from MDA-MB-231 cell line for its bone colonizing phenotype,37 was kindly provided by Dr Larry Suva of our institution. Neuraminidase (Vibrio cholerae) was from Sigma. Chondroitin sulfate A (CS A, chondroitin-4-sulfate) and Chondroitin sulfate B (CS B, dermatan sulfate) were from CALBIOCHEM (EMD Biosciences, San Diego, CA). Chondroitin sulfate C (CS C, chondroitin-6-sulfate) was obtained from Sigma. Chondroitin sulfate E (CS E) was from Associates of Cape Cod. Heparinase III and chondroitinase ABC were from Sigma or Associates of Cape Cod. All protease inhibitors were obtained from sigma.

Transfection of the FTIII gene into 4T1 cells

The 1083 bp coding fragment of human fucosyl transferase III (FTIII) gene in pCDNA3 plasmid was kindly provided by Dr. Insug O-Sullivan (University of Illinois). The coding sequence was further adapted for cloning between EcoRI and Xho I restriction sites by PCR using the following primers; 5′ CGA GAA TTC TCA GGT GAA CCA AGC CGC TAT G 3′ EcoRI and 5′ CGA CTC GAG ATG GAT CCC CTG GGT GCA 3′ Xho I. The amplified fragment was digested and inserted into the MCS of pIRES-EGFP vector to make FTIII/ pIRES-EGFP construct. 4T1 cells were then transfected with this construct or pIRES-EGFP vector alone using lipofectamine™ 2000 (invitrogen co., Carlsbad, CA) transfection reagent. The pIRES-EGFP vector contains an internal ribosome entry site (IRES) between the MCS and the EGFP (enhanced green fluorescent protein) coding region. This allows the FTIII gene (cloned into the MCS) and the EGFP gene to be translated from a single bicistronic mRNA.

Flow cytometry

Cells were passed to new flasks 24 hr before measuring the bindings. The subconfluent monolayer of cells was detached with the GIBCO® enzyme-free cell dissociation buffer (Invitrogen co.) and washed with Dulbecco's phosphate buffered saline with Ca++ and Mg++ (Mediatech, Herndon, VA). Cells were transferred to FACS buffer (Dulbecco's phosphate buffered saline, 1% BSA and 0.1% sodium azide), counted and adjusted to ∼1–2 × 106/ml. Monoclonal antibodies were added to a final concentration of 10 μg/ml or as stated in figure legends. Cells were incubated on ice for 30 min, washed twice with FACS buffer, before the addition of FITC-conjugated goat anti-mouse immunoglobulin for monoclonal analysis. Recombinant E- and P-selectin/Fc (human IgG) molecules and FITC/Alexa Fluor/PE-conjugated anti-human IgG were used for binding analyses in flow cytometry assays. Human and murine recombinant selectins were used for human and murine cells, respectively. Recombinant selectins were used at 2–15 μg/ml concentrations depending on the assay as stated in figure legends. Human IgG binding to human and murine cell lines was tested several times without noticeable binding, which was similar to the secondary Ab, therefore staining with secondary Ab is shown as background in majority of flow cytometry assays. Binding of recombinant P-selectin to cells was examined after treatment with neuraminidase, heparinase and chondroitinase ABC, as explained below, or in the presence of various concentrations of heparin and CS A, B, C and E. For lectin binding assays, cells were incubated with biotinylated lectins at a final concentration of 10 μg/ml and staining was completed by addition of FITC-conjugated streptavidin (2 μg/ml). Acquisition and analysis of data was performed either using a FACScan analyzer and by CellQuest software (BD Immunocytometry systems, Mansfield, MA) or on an EPICS®XL™ flow cytometer (Beckman Coulter, Fullerton, CA).

Inhibition of sulfate biosynthesis

Cells were washed with sulfate-free DMEM medium (Hyclone, Logan, UT) supplemented with 10% dialyzed FBS and 100 mM sodium chlorate (Sigma) and cultured in the same medium for 2 hr. The medium was then refreshed and incubation was continued overnight. These treated cells were harvested with the GIBCO® cell dissociation buffer (Invitrogen co.), washed and resuspended in FACS buffer for further analyses by flow cytometry.

Neuraminidase, pronase, heparinase and chondroitinase treatment

In some instances, tumor cell surfaces were pretreated before probing by flow cytometry. Removal of glycosaminoglycans (GAGs) was performed by treatment of 2 × 105 cells with heparinase III (1 unit/ml) and/or chondroitinase ABC (1 unit/ml) in 500 μl of HBSS buffer, in the presence of a cocktail of protease inhibitors (10 μg/ml Leupeptine, 10 μg/ml pepstatin, 20 μg/ml aprotinin and 10 mM benzamidine), for 1 h at 37°C. Alternatively, cells were treated with 500 μg/ml pronase (EMD biosciences, San Diego, CA) in HBSS buffer for 45 min at 37°C. Removal of sialic acid was performed by incubating cells with 100 mU/ml neuraminidase from Vibrio cholerae (Sigma) at 37°C for 1 hr.

Histopathologic evaluation of P-selectin binding

P-selectin histochemistry was performed as described earlier.3 Briefly, primary tumors and lungs were harvested from mice inoculated with the sLex-Neg 4T1 cells at 21 days post inoculation, placed in O.C.T (optimal cutting temperature compound (Ted Pella, Redding, CA) and frozen in liquid nitrogen. Frozen sections (5 μm) were fixed for 10 min in cold acetone then washed with cold DPBS (Cellgro® Mediatech, Herndon, VA). Endogenous peroxidase was blocked by immersion in 0.3% (w/v) hydrogen peroxide in absolute methanol for 15 min followed by DPBS wash. Nonspecific binding was blocked by incubating with DPBS+1% BSA at RT for 20 min. Sections were then incubated with recombinant mouse P-Selectin/human FC chimera (R&D systems, Minneapolis, MN) for 30 min in DPBS + 0.2% BSA at room temperature (RT) and washed in DPBS. Sections were incubated with anti-human IgG (Fc specific) peroxidase conjugate (1/300 dilution) for 15 min at RT followed by DPBS wash. Sections were incubated with diaminobenzidine solution (DAB) for 5 min at RT, washed with distilled water, counterstained with methyl green, mounted and examined under a light microscope. The recombinant P-selectin was omitted in negative controls to rule out nonspecific binding of the secondary antibody.

HUVEC adhesion assay

Clonetics™ human umbilical vein endothelial cell (HUVEC) system was used (Cambrex bio sciences, Walkersville, MD). Monolayer of HUVECs was prepared according to instructions. HUVECs then were incubated with supplied medium supplemented with IL-4 (20 ng/ml) for 24 hr. Medium was replaced with similar medium supplemented with Prostaglandin E2 (PGE2) for 10 min. Calcein AM-labeled (Molecular Probes, Eugene, Oregon) 4T1 cells left untreated or were treated with chondroitinase/heparinase, then added to the activated monolayers of HUVECs. Cells were coincubated at 37°C for 30 min and then unbound 4T1 cells were removed by washing gently with prewarmed medium. Then PBS was added to all wells, fluorescence intensity was measured using FLx 800 Microplate Fluorescence Reader (Bio-Tek Instruments, Winooski, Vermont) and percentage of adhesion was calculated.

Platelet adhesion assay

Blood was collected into sodium citrate (0.38% w/v) from naïve mice, and platelets were isolated from plasma by centrifugation. Platelets were washed and labeled with 5 μM final Calcein AM (Molecular Probes, Eugene, Oregon) for 15 min at 37°C. Platelet were then washed and incubated with 1 U/ml thrombin (Haematologic Technologiesc, Essex Junction, VT) at 37°C for 10 min. Heparin (Baxter Healthcare Corp., Deerfield, IL, 100 U/ml final concentration) was then added to the mix-platelet/thrombin, and incubated with 4T1 cells in flow cytometry tubes, for 15 min at RT, then the cell mixture acquired and analyzed by flow cytometry.

In vivo tumor model and metastasis detection

BALB/c female mice (6- to 8-week-old) were obtained from Harlan (Indianapolis, IN). Mice were injected subcutaneously with 100 μl PBS or heparin (Baxter Healthcare Corp., 100 unit/mouse) in 100 μl of PBS and then, 30 min later, tumor cells (2.5 × 104) were inoculated into tail vein. Mice from each studied group were injected in alternating order with gently resuspended cells. Mice were sacrificed at day 26 post tumor transfer and lung metastases were detected by growing serially diluted minced samples in medium containing 6-thioguanine as described.3

Statistical analysis

All experiments were repeated at least 2 times. The Student's t test was used to compare differences between means. χ2test was used for comparisons made between animal groups regarding lung metastases using EXCEL® software. Differences were considered significant if p was <0.05.


Transfection with human fucosyltransferase III induced expression of FH6- and CSLEX1-reactive sLex epitope variants with no effect on P-selectin binding

We have recently shown that 4T1 cells are deficient in sLea expression making the cell line a good candidate to study the involvement of sLex-mediated adhesion properties.3 There are several mAbs defined as KM93, FH6 and CSLEX1 that recognize sLex. These mAbs have been demonstrated to recognize different forms of the sLex antigen.38, 39, 40 FH6 is specific for an extended form of sLex38, while CSLEX1 and KM93 antibodies both recognize the sLex tetrasaccharide; however, the nature of the molecules carrying the carbohydrate determinant is known to affect the reactivity of these 2 antibodies.41 Among the above antibodies only KM93 reacts with the 4T1 tumor cell surface and positive and negative enrichment for KM93-reactive sLex does not change P-selectin binding.3 We speculated that discrepancy between P-selectin binding and surface expression of KM93-reactive sLex could be due to the specificity of KM93 for adjacent structures. In other words, CSLEX-1 or FH6-reactive sLex epitopes may differentially react with P- and E-selectin because of possible variations in lipid or peptide backbones.

To test this hypothesis, a broader array of cell surface sLex epitopes was needed as suggested by Kanoh et al.42 Therefore, we transfected 4T1 cells with fucosyltransferase III (FTIII) to expand the expression of other sLex epitopes (Fig. 1a). The binding of KM93 mAb was increased on FTIII-transfected cells. Most importantly, CSLEX1 and FH6 reactive epitopes were expressed at detectable levels in the transfected cells. The antibody binding data indicate that transfection with FTIII encoding sequence increased the expression of various sLex epitopes.

Figure 1.

Analysis of anti-sLex mAbs and mouse E-/P-selectin binding to the transfected 4T1 cells by flow cytometry. (a) FTIII, 4T1(FTIII) and vector transfected, 4T1(EGFP), cells were incubated with indicated mAbs and then were washed and binding was visualized by incubating with FITC-conjugated goat anti-mouse secondary antibody. (b) For E-/P-selectin binding, cells were first incubated with IgG-chimeric proteins (15 μg/ml) and then stained with FITC-conjugated goat anti human IgG. 4T1(EGFP) and 4T1(FTIII) cells are illustrated by open and filled histograms, respectively. Each experiment was repeated 3 times and histograms of one representative are depicted.

We further examined P- and E-selectin reactivity with parental and transfected cells. Cells were incubated with recombinant mouse E- and P-selectin/Fc (human IgG) chimeras and assayed for binding by flow cytometry. An increase in E-selectin binding was observed after FTIII transfection (Fig. 1b). This was expected given the increase in KM93, CSLEX1 and FH6 binding after transfection. P-selectin, however, bound very well to the parental 4T1 cells and the binding did not increase for the transfected cells. These data confirm that P-selectin binding to 4T1 cells is not dependent on the expression of sLex on the tumor cell surface.

Sialylated ligands are not a major source in P-selectin binding to the 4T1 cells

To further investigate the nature of P-selectin-reactive carbohydrate ligands on the 4T1 cell surface, the dependence of E- and P-selectin binding to 4T1 cells on divalent cation concentration was examined. Consistent with what we have reported before,3 both lectins bound to the 4T1 cells, with P-selectin showing very strong reactivity (Fig. 2a). Contrary to the E-selectin reactivity, P-selectin reactivity was not blocked by a low concentration of EDTA. EDTA inhibited P-selectin reactivity only at high concentrations indicating that it affects the P-selectin binding via a mechanism unrelated to its divalent cation chelating properties. Others studying P-selectin interaction with HS chains, observed similar phenomenon and hypothesized that EDTA might impose its effects by virtue of its inherent polycarboxylate structure which mimics the high charge density of HS-like chains.34

Figure 2.

Characterization of P-selectin binding to 4T1 cells using EDTA and neuraminidase. (a) Cells were harvested and first incubated with IgG chimeric P- or E-selectin (15 μg/ml) in the absence or presence of different concentrations of EDTA, and then stained with FITC-conjugated goat anti-human IgG. Mean fluorescence intensity for each histogram is shown. Human IgG was used as negative control at 15 μg/ml. (b) Cells were treated with neuraminidase at 100 mU/ml for 1 hr at 37°C and binding of sialic acid-specific lectins MAL-II and SNA andP-selectin was examined as indicated. Filled histograms represent staining with secondary antibody or streptavidin-FITC only. Continuous (dark line) and dotted (light line) histograms represent the lectin reactivity without and with neuraminidase treatment, respectively. Each assay was repeated 4 times with similar results.

We further treated cells with neuraminidase to determine whether there is a relationship between sialylation and reactivity of P-selectin. Although neuraminidase treatment reduced binding of the lectins MAL-II and SNA to 4T1 cells, it did not change the P-selectin reactivity with the cells (Fig. 2b). MAL-II and SNA bind to particular carbohydrate structures that contain sialic acid. Theses results provide further evidence that E- and P-selectin react with separate ligands on the surface of 4T1 cells and P-selectin, in contrast to E-selectin, binds to unsialylated ligands on these cells in a Ca2+-independent manner.

Sulfated proteoglycans as major cell surface P-selectin ligands

To further characterize the P-selectin ligands on the surface of 4T1 cells, cells were treated with pronase or were grown in sulfate-free medium in the presence of sodium chlorate to inhibit sulfate biosynthesis. Treatment with pronase dropped the P-selectin reactivity almost to the levels of the negative control, indicating the proteinaceous nature of the ligands (Fig. 3a). Growing the cells in sulfate-free medium containing sodium chlorate led to elimination of P-selectin binding in a majority of the cells, indicating that most P-selectin ligands on the 4T1 cells are sulfated (Fig. 3b). Similar pattern was observed for the mAb 10E4 reactivity with cells (Fig. 3b). mAb 10E4 reacts with an epitope that occurs in native heparan sulfate (HS) GAG chains, which can be destroyed by N-desulfation of the GAG.43 Sodium chlorate treatment down-regulates various HS sulfation reactions, including N-sulfation.44 Culture of cells in chlorate-supplemented medium has been widely used in experimental work addressing HS binding interactions, including its binding to P-selectin.29, 45, 46 Therefore, we used sulfate-dependent 10E4 reactivity with cells as a control in sulfation inhibition assay.

Figure 3.

(a) Pronase treatment reduced P-selectin reactivity with 4T1 cells. P-selectin binding to the 4T1 cells (15 μg/ml, thin line histogram) was sharply reduced (dotted line) to the level of secondary antibody binding (thick solid line) after pronase treatment. (b) 4T1 cells were harvested and washed with sulfate-free DMEM medium for several times. Cells then were cultured in sulfate-free DMEM supplemented with 10% dialyzed FBS and 100 mM sodium chlorate. The medium was replaced with the same fresh medium after 4 hr of incubation and then incubated for another 18 hr. Then binding of 10E4 mAb (10 μg/ml) and P-selectin (15 μg/ml) was determined and compared with cells grew in normal medium as indicated. (c) Cells were left untreated or treated with a mixture of heparinase/chondroitinase at 37°C for 1 hr and P-selectin (15 μg/ml) reactivity was determined. For treatment 2 × 105 cells were resuspended in 0.5 ml of HBSS and incubated with 1 unit/ml of heparinase and 1 unit/ml of chondroitinase ABC. All the above assays were repeated at least 3 times and results of 1 representative experiment out of 3 are depicted.

Sulfated GAGs-like HS and CS are carbohydrate moieties of PGs, which serve as P-selectin ligands.34, 36 Treatment of 4T1 cells with a mixture of the glycosaminoglycan-cleaving enzymes, heparinase and chondroitinase, decreased P-selectin binding (Fig. 3c). These data clearly suggest that P-selectin ligands on the surface of 4T1 cells are sulfated PGs, most likely the GAGs HS or CS.

P-selectin ligands are stably expressed on the surface of 4T1 cells and are involved in interaction with HUVECs

Our data indicate ample expression of P-selectin ligands on the surface of the 4T1 cells in vitro. To examine the stability of expression in vivo, we stained pathological samples from primary tumor and lung lesions (Fig. 4). P-selectin ligands were observed to be significantly and stably expressed on the surface of tumor cells in the primary and secondary lung lesions. The results suggest that P-selectin ligands may play a role in hematogenous metastasis in this syngeneic breast cancer model.

Figure 4.

P-selectin histochemical binding to the primary mass and metastatic pulmonary tumors. P-selectin ligands were expressed uniformly and strongly on cells of the primary mass in sLex-Neg cells. P-selectin ligands were very strongly expressed on metastatic cells in lung sections. The ligands were detected using recombinant mouse P-Selectin/human FC chimeric (2 μg/ml) followed by anti-human IgG (Fc specific) peroxidase conjugate. Bar equals 20 μm.

We examined 4T1 cell binding to HUVECs and observed that interaction of P-selectin on HUVECs and its ligands on 4T1cells play an important role in the cell adhesion (Fig. 5). Stimulating surface expression of P-selectin on HUVECs led to an increase in adhesion to the 4T1 cells. The adhesion was significantly inhibited by treatment of the 4T1 cells with the mixture of heparinase and chondroitinase. There was background adhesion to HUVECs, which was also significantly inhibited by treating the 4T1 cells with heparinase/chondroitinase mix, implying a constitutive presence of P-selectin on the HUVECs under our experimental conditions. We further confirmed this point by examining P-selectin expression on HUVECs (data not shown). A low constitutive expression of P-selectin was detected on 10% of cells, which was elevated to a more intense staining on about 20% of cells after treatment with IL-4 and PGE2. Adhesion was clearly enhanced after P-selectin induction on HUVECs and suppressed after heparinase/chondroitinase treatment of tumor cells (Fig. 5).

Figure 5.

Involvement of P-selectin ligands in binding to HUVECs. Monolayers of HUVECs were prepared and left untreated or treated for induction of surface presence of P-selectin receptor before conducting the experiment. The 4T1 cells were labeled with Calcein AM and left untreated or treated with a mixture of heparinase/chondroitinase enzymes and added to the HUVEC monolayers and incubated at 37°C for 45 min. The unbound cells were removed by gentle washing. Percentage of adhesion was calculated based on mean fluorescence intensities and presented as average of 11 replications. Bars represent SD based on 11 replications. A representative experiment out of 2 is shown. Paired Student's t test was used to compare the means.

Heparin inhibits both P-selectin binding to the tumor cells and tumor cell-platelet interactionted by P-selectin

It has been shown that P-selectin can bind to immobilized heparin.47 An inhibitory role for heparin in P-selectin interaction with its ligands on cell surface has been demonstrated.17, 34 We tested heparin's ability to inhibit P-selectin's interaction with the cell surface in vitro. Recombinant P-selectin was incubated with heparin and then the mixture was added to cells to test the binding (Fig. 6a). Heparin efficiently inhibited P-selectin binding to cells in a dose-dependent manner. Mouse platelets can interact with tumor cells in a P-selectin dependent fashion.8, 17 To examine whether heparin can block such an interaction, we mixed 4T1 cells with calcein-AM-labeled mouse platelets in the presence of mouse thrombin with or without heparin. Mouse thrombin was added to stimulate relocation of P-selectin to platelet surface and tumor cells were then analyzed by flow cytometry for calcein AM staining, indicating platelet attachment. Thrombin-treated platelets showed binding to tumor cells, which was reduced in the presence of heparin (Fig. 6b). The data suggest that tumor cell-platelet interaction is P-selectin mediated and can be blocked by heparin.

Figure 6.

Effects of heparin on P-selectin interaction with tumor cells. (a) P-selectin (10 μg/ml) was incubated with serial dilutions of heparin at room temperature with gentle shaking for 30 min. 4T1-sLex-Neg cells were harvested and the mixture then was added to the cells and incubated on ice for 30 min. Anti-human-FITC was then added at 1:40 dilution and incubation continued for another 40 min. (b) Mouse platelets were isolated from citrated blood and calcein-AM labeled and pulsed with thrombin. Tumor cells were detached using trypsin free buffer and washed in DPBS buffer. Cells then were mixed with platelets (2 × 105 tumor cells with 6 × 106 platelets) with or without heparin and incubated at room temperature for 5 min. Tumor cells then were analyzed by flow cytometry to quantitate the attachment of labeled platelets with or without preactivation with thrombin in the presence or absence of heparin. Percentage of positive cells for each histogram is shown. Assays were repeated 2 times and results of one representative assay are depicted. Percentage of positive cells for each histogram is shown.

Heparin inhibits hematogenous metastasis of breast cancer cells to the lung

Many studies documented an inhibitory role for heparin in cancer metastasis.19, 48, 49, 50, 51, 52, 53 Borsig et al.17 have suggested that heparin treatment attenuated tumor metastasis in mice by inhibiting P-selectin-mediated interactions of platelets with carcinoma cell-surface mucin ligands. To establish a functional correlation between P-selectin ligand expression of breast 4T1 cells and their metastatic ability in vivo, we injected mice with 100 units of heparin 30 min before tumor cell inoculation. Mice were sacrificed 26 days after tumor inoculation, lungs were harvested and metastatic cells were detected by clonogenic assay. We observed a complete absence of metastases in lung of majority of mice (6 mice out of total of seven) injected with heparin (Table I). All mice that were injected with PBS as control developed lung metastasis. Thus, blocking of P-selectin interaction with its ligand in vivo significantly prevented the establishment of metastatic foci.

Table I. Number of Mice Detected Positive For established Lung Metastases
TreatmentPositive (total)
  • *

    p = 0.0023 when compared with PBS-treated group by χ2 test.

PBS6 (6)
Heparin1* (7)

Chondroitin sulfate GAGs as major P-selectin ligands on the murine and human breast cancer cell lines

It is reported that a CS PG, versican, that was derived from a human renal adenocarcinoma cell line binds to P-selectin via its CS chains.35 To further explore the nature of P-selectin ligands on the 4T1 breast tumor cell line, we used heparinase and chondroitinase ABC separately in P-selectin binding assays. The data indicates a major role for CS in P-selectin binding to the 4T1 cells (Fig. 7a). In separate assays, effects of heparinase and chondroitinase activity on cell surface HS and CS was examined using mAbs 10E4 and 2-B-6. mAb 10E4 binds to an epitope on HS and has been used widely as a research tool in studying HS functions.29, 45, 46 mAb 2-B-6 reacts with 4 sulfated chondroitin and dermatan sulfate following Chondroitinase ABC digestion of various proteoglycans (PGs) and an increase in reactivity of this Ab after Chondroitinase ABC treatment has been demonstrated.54, 55 The Ab recognizes the CS “stubs”, which remain after digestion with Chondroitinase ABC. Treating cells extensively with heparinase III completely abrogated binding of mAb 10E4 to cells, whereas no effect was observed on binding of P-selectin to the cells (Fig. 7b). Cells then were treated briefly with Chondroitinase ABC and reactivity of P-selectin and mAbs 10E4 and 2-B-6 was detected (Fig. 7c). While Chondroitinase ABC treatment did not affect 10E4 binding, it reduced P-selectin binding and increased 2-B-6 reactivity, as was expected. The results demonstrate that digestion with Chondroitinase ABC, and not Heparinase, destroyed P-selectin-specific epitope(s).

Figure 7.

Chondroitin sulfates are P-selectin ligands on 4T1 cells. (a) 4T1 cells were harvested with cell dissociation buffer and treated with chondroitinase or heparinase or a mixture of both for 1 hr™ at 37°C, then P-selectin (10 μg/ml) reactivity was detected. Percentage of positive cells for each histogram is shown. (b) Cells were incubated with heparinase III (1 U/ml) in the presence of a protease inhibitor cocktail, see Material and Methods section, and reactivity of P-selectin (10 μg/ml) and mAb 10E4 (7 μg/ml) was detected by flow cytometry. Filled histogram represents staining with secondary antibody (p-selectin assay) or mouse IgM (10E4 mAb assay). Continuous (dark line) and dotted (light line) histograms represent binding of the indicated probes to untreated and heparinase-treated cells, respectively. (c) Cells were incubated with Chondroitinase ABC (0.5 unit/ml) for 40 min and then reactivity of P-selectin and mAbs 10E4 and 2-B-6 was measured by flow cytometry. Filled histogram represents negative control as explained in 7B or mouse IgG for 2-B-6 staining. Continuous (dark line) and dotted (light line) histograms represent binding of the indicated probes to untreated and chondroitinase ABC-treated cells, respectively. These experiments were repeated 3 times with similar results.

We further used various chondroitin sulfates to block the interaction, including chondroitin sulfates A, B, C and E. Among those tested, in keeping with the data reported by others,36 only CS B (dermatan sulfate) and CS E inhibited P-selectin binding to the cells (Figs. 8a and 8b). CS A (Fig. 8a) and C (data not shown) did not show inhibitory effects even with the low amount of P-selectin used. CS B showed inhibitory effects only at higher concentrations, while CS E was a more potent inhibitor with a complete inhibition of binding at a concentration of 0.5 mg/ml in the presence of 10 μg/ml of P-selectin (Figs. 8a and 8b). The effective dose of heparin, 120 units (Fig. 6a), corresponds to 0.7 mg/ml heparin, which is close to the CS E blocking concentration of 0.5 mg/ml.

Figure 8.

Inhibitory effect of various CS types on P-selectin binding to 4T1 cells. Recombinant P-selectin was incubated with serial concentrations of CS A, CS B (a) or CS E (b) for 30 min at room temperature. Cells were harvested with trypsin-free buffer, washed and resuspended in flow cytometry buffer. P-selectin chondroitin sulfate mix was added to the cells and binding was visualized using FITC-anti-human IgG. P-selectin was used at 5 μg/ml (a) and 10 μg/ml (b). (c) CS chains are the major P-selectin ligands on human MDA-MET breast cancer cells. Cells were harvested with cell dissociation buffer and treated with Chondroitinase ABC or heparinase or a mixture of both for 1 hr at 37°C. Then P-selectin reactivity was detected using recombinant human P-selectin at 15 μg/ml (dark histograms) and 3 μg/ml (light histograms). Percentage of positive cells for each histogram is shown. Results of one representative assay out of 2 are depicted.

Major component of CS GAGs consist of hexosamine, N-acetyl-D-galactosamine (GalNAc) and hexuronic acid, D-glucuronic acid (GlcUA) or L-iduronic acid (IdoUA), units, which are arranged in alternating sequences to form repeating disaccharide chains. The repeating units contain a number of sulfate constituents at various positions, which create a considerable degree of structural and functional diversity. The repeating disaccharide unit for CS A, B, C and D are [GlcUA-GalNAc (4S)], [IdoUA-GalNAc (4S)], [GlcUA-GalNAc (6S)] and [GlcUA-GalNAc (4S, 6S)], respectively. CS B can be 2-O sulfated on the IdoUA residues.56 Apparently over sulfation and generated negative charges might have a role in inhibitory potential of CS types.

Taken together, these data indicate that CS is the major P-selectin ligand on the surface of the 4T1 cells. However, only over sulfated CS E is able to effectively inhibit the interactions. We further tested bone-colonizing human breast cancer cell variant MDA-MET37 for expression of P-selectin ligands. Using low and high concentrations of P-selectin, we observed that P-selectin bound to cells and that the binding was decreased after Chondroitinase ABC treatment (Fig. 8c). Heparinase had little effect in reducing P-selectin binding and a mixture of heparinase and chondroitinase did not perform better than chondroitinase alone. This data suggests that CS PGs are the major P-selectin ligands on the surface of human breast cancer cells.


We have recently shown, in the 4T1 spontaneous metastatic model of breast cancer, that the presence of sLex/a oligosaccharides is not required for metastasis to the lung.3 Although sLex/a oligosaccharides are common ligands for both E- and P-selectin, these two lectins did not correlate in their reactivity with the 4T1 cells. We observed that P-selectin bound to the 4T1 cells strongly, and the binding was not affected by sorting for sLex oligosaccharide by KM93 antibody or even by FTIII gene transfection. E-selectin binding can be predicted by reactivity of anti-sLex antibodies, indicating that E-selectin binding is predominately sLex dependent. However, P-selectin binding did not correlate with either E-selectin or sLex-reactive antibodies, suggesting that much of the P-selectin binding is not to sLex or other related oligosaccharides. Similar results have been reported by Kobzdej et al.57 where the forced fucosylation of intact murine neutrophils with exogenous FTVI increased the binding of the sLex reactive antibodies CSLEX1 and HECA-452, but did not increase E- or P-selectin binding. They suggested that the endogenous fucosyltransferases selectively fucosylate specific substrates that bind preferentially with E- and P-selectin. Our results are similar in that we observe an increased binding with sLex reactive antibodies and E-selectin but no increase in P-selectin binding. The P-selectin binding in 4T1 appears to be dependent upon structures other than sLex or sLea ligands with increased sLex expression having almost no effect on P-selectin binding. The data suggest that sLex epitope variants on the 4T1 cells may not be presented on appropriate carrier molecules to be recognized by P-selectin. Such selectivity of molecules serving as potential P-selectin ligands has been reported.27

The role of P-selectin ligands in heterotypic adhesion is a critical component determining the efficiency of tumor cell dissemination.8, 17 P-selectin is represented as a relatively promiscuous receptor, as it binds to a variety of heterogeneous ligands.11 Characterization of P- and E-selectin ligands is crucial for the assessment of metastatic risk and the development of possible ways of dealing with metastatic disease. Since sLex has been introduced as an overlapping ligand for selectins, it has been evaluated widely as a therapeutic target to inhibit tumor metastasis.18, 19, 20, 21, 22 We observed that a significant amount of P-selectin binding was both Ca2+-independent and sialic acid-independent, confirming that sLex is not a P-selectin ligand on 4T1 cells. Similarly, the EDTA resistant (Ca2+-independent), nonfucosylated, nonsialylated P-selectin ligands observed by Varki's group in murine colon adenocarcinoma cell line MC-38 implicated sulfated glycolipids.58 Heparan sulfate-like PGs on the surface of human colon carcinoma and melanoma cell lines have also been suggested to function as P-selectin ligands.22, 28 In this regard, others have shown that P-selectin binds a large range of heparin and HS fragments in a Ca2+-independent manner.34

Taken together, E-selectin binding to the 4T1 cell line is restricted to sLex or closely related structures while P-selectin binding can involve a varied group of compounds, including Ca2+-independent binding to non-Lewis structures. Characterization of P-selectin binding to the 4T1 cells showed that this interaction is sulfur-dependent and heparinase/chondroitinase sensitive, implying that P-selectin ligands on the surface of the 4T1 cells are mostly HS/CS PGs. Consistent with our data, others found that P-selectin binding to the breast cancer cell line ZR-75-30 was sulfation-dependent.29 Further characterization of the 4T1 surface ligands clearly indicate that particular CS galactosaminoglycan polymers can be the major P-selectin ligands expressed on this cell line. Our data are in agreement with others who have shown that a CS PG, derived from the renal adenocarcinoma ACHN, versican, acts as P-selectin ligand and that the binding can be mediated by an interaction between CS GAG chains and the lectin domain of P-selectin.35 Consistent with other reports, we observe that, among various CS types, only CS B and CS E are able to inhibit the interaction in addition to heparin.35, 36 However, our observation of calcium independent P-selectin reactivity distinguishes our findings from those of these reports.

We observed the stable expression of P-selectin ligands on 4T1 cells in vivo, which based on extensive published literature, suggests that these ligands contribute to the metastatic behavior of this cell line. We found that cell surface P-selectin ligands indeed contribute to binding of the 4T1 cells to platelets and HUVECs. Nonsialylated P-selectin ligands, with no reactivity toward E-selectin, could play the crucial role in tumor cell migration in spontaneous models. While down modulation of sLex may abrogate E-selectin reactivity and deteriorate tumor cell–cell adhesion,3 intact P-selectin reactivity with HS/CS can compensate for such abrogation and facilitate microemboli formation and adhesion to the endothelial cells, promoting tumor cell arrest in vasculature and extravasation.

Heparin is being used as anticoagulant treatment of venous thromboembolism in cancer patients, where it has been shown to improve patient survival by mechanisms not explained by anticoagulation.59 Our data clearly demonstrate that heparin inhibited P-selectin binding to the 4T1 cells, and it blocked P-selectin mediated adhesion of platelets to this tumor cell line and that the administration of heparin indeed inhibited lung metastasis in vivo. Other investigators have shown that heparin inhibits experimental lung metastasis.53 However, heparin as an anticoagulant has the possibility of causing bleeding at high doses, which may restrict the potential clinical application of this drug in preventing metastasis.

Inhibition of interaction between P-selectin with its various ligands on tumor cells may have potential value as an antimetastatic therapeutic strategy. Competition studies suggest that heparin and CS interaction may involve a region of the P-selectin molecule very close to the lectin-binding site for sLex. In this regard, others have shown that heparin is capable of blocking P-selectin binding to various tumor cells with various surface ligands, including sLex, sulfated glycolipids and HS PGs.17, 22, 28, 58 Here, we show that heparin blocks P-selectin binding to tumor cell surface CS GAG chains. Consistent with our data, it has also been shown that P-selectin binding to CS chains of versican can be inhibited by heparin.35 In addition, it has been shown that the binding of this CS PG to P-selectin was inhibited by sLex, which is in agreement with the notion that CS binding to the lectin domain of P-selectin is similar to sLex binding.35

CS has also been shown to inhibit binding of P-selectin to cell surface HS PGs28, which along with our current data suggest that over sulfated CS types may be used to block P-selectin binding to any of its ligands on tumor cells. Such broad specificity can be explained by recognition of a clustered epitope by P-selectin.34 Our data in agreement with others indicate that P-selectin ligands are expressed on human breast cancer cell lines.7, 27, 29 While we observed that CS B and CSE inhibit P-selectin binding to cells, this does not necessarily mean that CS B and CS E chains are relevant structures on the cell surface. CS PGs, located on the cell surface and in the ECM, are expressed by the majority of mammalian cells. To be used as P-selectin ligands for cancer cell metastasis, CS chains are expected to be presented with tumor-specific modified structures. Various CS types have the potential to display an enormous structural diversity on the cell surface by embedding multiple overlapping sequences constructed with distinct disaccharide units modified by different pattern of sulfation. One major conclusion that can be drawn from our data is that over sulfation is essential for the observed inhibitory effects of CS types. In this regard, the low inhibitory role of CS B may be due to sulfation levels of disaccharide constituents as previously suggested.36 Sulfated moieties contain negative charges, which play an important role in P-selectin recognition. However, both sLex oligosaccharide and HS chains also contain negative charges and are avidly expressed on the surface of 4T1 cells but do not contribute to P-selectin binding. We presume that P-selectin may recognize a minor constituent of CS chains that has specific modification by sulfation. In this regard, Kawashima et al.36 showed that CS chains containing [GlcA-GalNAc(4,6-O-disulfate)] and [IdoA-GalNAc(4,6-O-disulfate)] bind P-selectin with high affinity. In addition, the protein core and its other carbohydrate components may also play a role in P-selectin recognition.

Taken together, the data imply that P-selectin interaction with CS GAGs can be targeted for treatment of metastatic breast cancer. Notably, CS is a popular dietary supplement being used for the treatment of osteoarthritis, and it is reasonable that both heparin and CS be chosen as candidates for more detailed preclinical/clinical studies directing toward treatment of breast cancer metastases.


This study was supported by a DOD grant (DAMD17-0101-0366) and a NIH grant (CA089480) both to T.K.E. We thank Dr. O-Sullivan, University of Illinois, for the FTIII gene construct.