CD44 variant isoforms expressed by breast cancer cells are functional E-selectin ligands under flow conditions

Abstract

Adhesion of circulating tumor cells to vascular endothelium is mediated by specialized molecules that are functional under shear forces exerted by hematogenous flow. Endothelial E-selectin binding to glycoforms of CD44 mediates shear-resistant cell adhesion in numerous physiological and pathological conditions. However, this pathway is poorly understood in breast cancer and is the focus of the present investigation. All breast cancer cell lines used in this study strongly expressed CD44. In particular, BT-20 cells expressed CD44s and multiple CD44v isoforms, whereas MDA-MB-231 cells predominantly expressed CD44s but weakly expressed CD44v isoforms. CD44 expressed by BT-20, but not MDA-MB-231, cells possessed E-selectin ligand activity as detected by Western blotting and antigen capture assays. Importantly, CD44 expressed by intact BT-20 cells were functional E-selectin ligands, regulating cell rolling and adhesion under physiological flow conditions, as found by shRNA-targeted silencing of CD44. Antigen capture assays strongly suggest greater shear-resistant E-selectin ligand activity of BT-20 cell CD44v isoforms than CD44s. Surprisingly, CD44 was not recognized by the HECA-452 MAb, which detects sialofucosylated epitopes traditionally expressed by selectin ligands, suggesting that BT-20 cells express a novel glycoform of CD44v as an E-selectin ligand. The activity of this glycoform was predominantly attributed to N-linked glycans. Furthermore, expression of CD44v as an E-selectin ligand correlated with high levels of fucosyltransferase-3 and -6 and epithelial, rather than mesenchymal, cell phenotype. Together, these data demonstrate that expression of CD44 as a functional E-selectin ligand may be important in breast cancer metastasis.

during hematogenous metastasis, tumor cells dissociate from a primary tumor, migrate through the tissue space, and enter the circulatory system. The blood-borne circulating tumor cells (CTCs) travel to distant sites, where they adhere to the endothelial cells lining the vessel wall, extravasate, and form metastatic colonies if directed by niche factors (6, 20, 30). Determining the mechanisms underlying these steps can provide insights into development of novel diagnostics, prognostics, and therapeutics for cancer. Of particular interest is elucidation of the molecular mechanisms by which metastatic cells adhere to the endothelium while resisting the disruptive shear exerted by blood flow.

The shear-resistant adhesion of cancer cells is hypothesized to be similar, in part, to leukocyte recruitment at sites of inflammation, which is mediated by selectins and their ligands (6, 12, 14, 18, 23, 30, 46, 47). Several in vivo and in vitro studies have implied that endothelial E-selectin is involved in breast, colon, and prostate cancer cell adhesion to blood vascular endothelium (6, 8, 11, 12, 14, 16, 31, 46, 47). Interestingly, E-selectin is constitutively expressed on human bone marrow microvascular endothelium (44), and bone marrow is one of the most frequent sites of cancer metastasis (55). Thus there has been an increased focus on determining the molecules on cancer cells that can serve as E-selectin ligands (6, 8, 11, 12, 14, 18, 31, 46, 47). One such entity is CD44 (12, 18, 19, 24, 34, 51), which is a family of proteins encoded by a single gene comprising 20 exons. Exons 1–5 and 15–20 generate the standard form (CD44s, typically 90–100 kDa), which lacks any variable region, while exons 6–14 are alternatively spliced to produce CD44 variant (CD44v) isoforms (>100 kDa), designated CD44v1-10 (27). Sialofucosylated carbohydrate moieties terminally presented on N- and O-glycans linked to core CD44 molecules ultimately confer E-selectin ligand activity (24, 26, 41). Such CD44 glycoforms were first identified on hematopoietic cells as reactive to E- and L-selectin and, hence, termed hematopoietic cell E-/L-selectin ligand (HCELL) (17, 18, 42). Notably, the selectin ligand activity of CD44 of hematopoietic cells was primarily associated with the N-glycosylated, sialofucosylated CD44s isoform (17, 18, 42), whereas O-glycosylated, sialofucosylated CD44v isoforms are the prevalent E-selectin ligands expressed by colon carcinoma cells (24, 27, 41).

Several lines of evidence have implicated CD44 in breast cancer metastasis. Particularly, CD44 expression has been detected on numerous breast cancer cell lines and primary tumors (1, 38, 57), as well as breast CTCs (50). In addition, CD44 has been recognized as a breast cancer stem cell marker (3, 4), and it is hypothesized that a subpopulation of CD44-expressing tumor cells possess high metastatic potential (35). While these investigations link CD44 to breast cancer metastasis, others give insight into the role of breast cancer cell CD44 as E-selectin ligands. For instance, E-selectin ligand activity of CD44 under static (no-flow) conditions has been observed during postadhesion transendothelial migration of breast cancer cells (57). However, evaluation of breast cancer cell CD44 isoforms as functional E-selectin ligands operational under physiological flow conditions is lacking. We previously reported that breast cancer cells adhere to endothelium under flow conditions predominantly via E-selectin-E-selectin glycoprotein ligand(s) interactions (46, 47). Hence, we hypothesized that breast cancer cells express glycoform(s) of CD44 that serve as functional E-selectin ligand(s) that mediate(s) breast cancer cell adhesion to vascular endothelium under physiological flow conditions. Therefore, in the present study we characterized CD44 expressed by breast carcinoma cell lines and assessed their E-selectin ligand activity under flow conditions.

MATERIALS AND METHODS

Cell culture.

The BT-20 breast cancer cell line [American Type Culture Collection (ATCC), Manassas, VA] was grown in minimum essential medium (MEM; Invitrogen, Carlsbad, CA) with 10% FBS and 1× penicillin-streptomycin. MDA-MB-231, Hs-578T, and MDA-MB-468 breast cancer cell lines (all from ATCC) were cultured in Dulbecco's modified Eagle's medium (Invitrogen) with 15% FBS and 1× penicillin-streptomycin. E-selectin-transfected Chinese hamster ovary (CHO-E) cells (a generous gift from Dr. Robert Sackstein, Harvard Medical School, Boston, MA) were maintained in MEM supplemented with 10% FBS and 0.1 mM nonessential amino acids (Invitrogen). The LS-174T colon carcinoma cell line (ATCC) was cultured in MEM supplemented with 10% FBS, 0.1 mM nonessential amino acids, and 1 mM sodium pyruvate (Invitrogen).

Antibodies and chimera constructs.

Anti-human CD44 (515), CD62E (68-5H11), and FITC-conjugated CD44 (G44-26) monoclonal antibodies (MAbs) and all isotype controls were obtained from BD Biosciences (San Jose, CA). Anti-human CD44 (2C5), CD44v3 (3G5), and CD44v4/5 (3D2) MAbs and recombinant chimeric mouse E-selectin-human IgG1 Fc constructs (E-Ig chimera) were obtained from R & D systems (Minneapolis, MN). Anti-human MAbs against variant isoforms v3 (VFF 27), v4 (VFF 11), v5 (VFF 8), v6 (VFF 7), v7 (VFF 9), v7/8 (VFF 17), and v10 (VFF 14) of CD44 were obtained from AbD Serotec (Raleigh, NC). Anti-human E-cadherin and N-cadherin MAbs were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). FITC- and alkaline phosphatase (AP)-conjugated secondary antibodies were obtained from Southern Biotech (Birmingham, AL).

Cell treatments.

BT-20 cells were cultured for 48 h with 2 mM deoxymannojirimycin (DMJ; Sigma-Aldrich, St. Louis, MO) or 1 mM benzyl-N-acetyl-α-galactosaminide (Bzl-GalNAc; Sigma-Aldrich) to inhibit N- or O-linked glycosylation of glycoproteins, respectively (14).

Quantitative RT-PCR.

Extraction and purification of RNA were performed using the RNeasy Plus kit (Qiagen, Valencia, CA) following the manufacturer's protocol. The RNA purity and concentration were evaluated using a NanoVue Plus (GE Healthcare Biosciences, Piscataway, NJ). Purified RNA was treated with RNase inhibitor (Applied Biosystems, Foster City, CA) and DNase I (New England Biolabs, Ipswich, MA) to remove genomic DNA. RNA was reverse-transcribed by a high-capacity reverse transcription kit (Applied Biosystems) following the manufacturer's instructions. PCR was performed on cDNA, synthesized from 75 ng of RNA by SYBR Green FastMix (Quanta BioSciences, Gaithersburg, MD) chemistry, and monitored by an iCycler iQ5 real-time PCR instrument (Bio-Rad Laboratories, Hercules, CA). Forward and reverse primers were purchased from Integrated DNA Technologies (Coralville, IA). Gene expression data were normalized to the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (40).

Cell lysis and immunoprecipitation.

Cells were lysed in buffer containing 1% Triton X-100, 0.02% NaN₃, 150 mM NaCl, 0.5 mM Tris (pH 10.4), 1 mM EDTA, and protease inhibitor cocktail (Roche Applied Sciences, Indianapolis, IN). CD44 was immunoprecipitated from cell lysates by incubation with anti-CD44 MAb (515) and protein G-agarose beads (Invitrogen) overnight at 4°C under constant rotation. Protein G beads were subsequently washed with lysis buffer and then incubated with Laemmli reducing sample buffer and heated to 100°C for 5 min to release CD44 (47).

To purify E-selectin-reactive proteins, cell lysates were first precleared of nonspecific antigens by incubation with human IgG isotype control and protein G beads. The precleared lysate was incubated overnight with E-Ig chimera and protein G beads at 4°C with constant rotation. After sufficient washes, E-Ig chimera-reactive antigens were eluted from protein G beads using elution buffer [5 mM EDTA, 50 mM Tris (pH 7.4), and 0.1% Triton X-100]. The immunoprecipitated material was resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and then subjected to Western blot analysis (24, 47).

RNA interference.

CD44 silencing in BT-20 cells was achieved using MISSION shRNA lentiviral transduction particles (Sigma-Aldrich) prepared from pLKO.1 vector (TRC version 1). The concentration of viral particles was optimal at 10 multiplicity of infection, and cells transfected with specific sequence or with empty vector were puromycin-selected (47). Five different specific clones were screened, and a clone leading to significantly reduced cell surface expression of CD44 by flow cytometry was chosen for experiments: the oligonucleotide sequence 5′-CCGGCGCTATGTCCAGAAAGGAGAACTCGAGTTCTCCTTTCTGGACATAGCGTTTTTG-3′ (TRCN0000057567), where the underlined portion is sense and the italicized portion is antisense.

SDS-PAGE and Western blotting.

SDS-PAGE of cell lysates or immunoprecipitates was performed on 4–15% Tris·HCl precast gels (Bio-Rad Laboratories) under reducing conditions, and proteins were subsequently transferred to polyvinylidene difluoride membrane (Bio-Rad Laboratories). Membranes were blocked with FBS overnight to reduce the background noise and incubated with appropriate primary antibodies or isotype controls and then with AP-conjugated secondary antibodies (47). Signals were developed using Western Blue AP substrate (Promega, Madison, WI).

Flow cytometry.

Cells were washed with blocking buffer and incubated with unconjugated primary antibody or isotype control for 30 min at 4°C. Cells were washed and stained with secondary antibody for 30 min at 4°C (12, 46, 47). Alternatively, cells were directly labeled with FITC-conjugated specific antibody or isotype control. Cells were washed and analyzed by a FACSort or FACSAria special-order research product flow cytometer/sorter (BD Biosciences).

Flow adhesion assay and antigen capture.

The experimental setup for the flow adhesion assays consisted of a parallel-plate flow chamber (Glycotech, Rockville, MD) placed on a Nikon TE300 inverted microscope equipped with a CCD video camera and a video recorder (32, 46, 47). Untreated, vector, or CD44-silenced BT-20 cells (1 × 10⁶ cells/ml) were perfused over a monolayer of CHO-E cells for 4 min at a bone marrow microvascular wall shear rate of 100 s⁻¹ (37). Subsequently, for detachment analysis, the shear rate was sequentially doubled in time steps of 30 s until the final shear rate, 3,200 s⁻¹. To determine percent attachment, the number of adherent cells was counted at the end of the time intervals and normalized with respect to the number of adherent cells at the end of 100 s⁻¹. Cell rolling velocity was calculated by capture of still images from video over 5 s and determination of distance traveled by the cell over that time period using ImageJ software (14, 47).

The antigen capture assay is described elsewhere (22). Briefly, 20 μg/ml MAb was incubated on a petri dish at 37°C for 2 h. The MAb substrate was blocked by 1% BSA overnight at 4°C. Lysate equivalent to 2 × 10⁶ cells was applied to the spot using cloning chambers and incubated overnight at 4°C. CHO-E cells (1 × 10⁶/ml) were perfused over captured antigens for 2 min at 100 s⁻¹ using the parallel-plate flow chamber setup described above. The number of adhering cells included all cells bound to the surface at the end of perfusion (14, 45).

Fluorescence microscopy.

Cells grown on tissue culture-grade Millicell EZ slides (Millipore, Billerica, MA) were washed and fixed in 4% methanol-free paraformaldehyde in Dulbecco's PBS. The slides were incubated with primary antibody or isotype control for 1 h at 4°C, then washed and incubated with Alexa Fluor 488-conjugated secondary antibody for 1 h at 4°C (46, 47). A drop of ProLong Gold antifade reagent with DAPI (Invitrogen) was added before coverslips were mounted. Slides were imaged using a ×40 objective under wide-field fluorescence using an inverted microscope (DMI 6000, Leica Microsystems, Wetzlar, Germany) and SimplePCI software (Hamamatsu, Sewickley, PA).

Statistics.

Values are means ± SE for at least three independent experiments. Statistical significance (P < 0.05) between control and sample was tested by paired Student's t-test, except where indicated. Multiple samples were compared with one-way ANOVA coupled with post hoc Tukey's multiple-comparison test (P < 0.05).

RESULTS

Breast cancer cell lines express CD44 isoforms.

Previously, we showed that shear-resistant adhesion of breast cancer cell lines is mediated by E-selectin and breast cancer cell glycoprotein ligands (47). It has also been shown that colon cancer, prostate cancer, and acute myelogenous leukemia (AML) cells express glycoforms of CD44 as E-selectin ligands under flow conditions (8, 12, 18, 24). Therefore, BT-20, MDA-MB-468, MDA-MB-231, and Hs-578T breast cancer cell lines were initially screened for CD44 expression using an anti-CD44 MAb (515) that recognizes CD44s and CD44v (18, 24, 25). Consistent with previous reports (1, 38, 45), flow cytometric analysis showed that each of these breast cancer cell lines robustly expresses CD44 (Fig. 1A). To quantitatively compare the expression of CD44 isoforms at the mRNA level, quantitative RT-PCR (qRT-PCR) was performed using primers designed for the specific isoforms (49). Generally, higher levels of CD44v isoforms were expressed by BT-20 and MDA-MB-468 cells than by MDA-MB-231 and Hs-578T cells. Conversely, similar or slightly higher levels of CD44s were expressed by MDA-MB-231 and Hs-578T cells (Fig. 1B) than by the other two breast cancer cell lines.

Fig. 1.

CD44s and CD44v are expressed by breast cancer cells. A: breast cancer cells were labeled with anti-CD44 MAb (515, filled curves) or isotype control (open curves) and analyzed by flow cytometry. B: quantitative RT-PCR (qRT-PCR) was performed on breast cancer cell mRNA using primers designed for detection of specific CD44 isoforms. Data are presented as expression relative to the housekeeping gene GAPDH. Values are means ± SE; n = 4 independent experiments. *P < 0.05 by one-way ANOVA coupled with Tukey's multiple-comparison test.

The breast cancer cell lines were also probed by flow cytometry to find expression of CD44 variants at the protein level. In line with the qRT-PCR data (Fig. 1B), BT-20 cells showed positive expression of multiple CD44v isoforms (Fig. 2A). Particularly, CD44v3-6 were strongly expressed on the surface of these cells, while other isoforms were also present but at much lower levels. MDA-MB-231 cells (Fig. 2B) expressed limited levels of the variant isoforms, implying that the CD44 proteins (Fig. 1A) on these cells are mainly CD44s. These qRT-PCR and flow cytometric data collectively demonstrate expression of higher levels of CD44v isoforms by BT-20 and MDA-MB-468 cells than by MDA-MB-231 and Hs-578T cells.

Fig. 2.

Differential levels of CD44v isoforms are expressed on the surface of breast cancer cells. BT-20 (A) and MDA-MB-231 (B) cells were labeled with MAb against specific CD44v isoforms (filled curves) or isotype control (open curves) and analyzed by flow cytometry. Data are representative of 3 independent experiments.

CD44 expressed by BT-20, but not MDA-MB-231, cells is an E-selectin ligand.

We chose to conduct further experiments with BT-20 and MDA-MB-231 cells, which represent cell lines expressing two different levels of CD44v isoforms. Cell lysates and CD44 immunoprecipitates (MAb 515) were subjected to Western blot analysis using another general, multiple-isoform anti-CD44 MAb (2C5). BT-20 lysates revealed three distinct CD44 bands at ∼95, ∼110, and ∼150 kDa, whereas MDA-MB-231 lysates displayed a single prominent band at ∼95 kDa (Fig. 3A). The bands at ∼95 kDa are presumed to be CD44s, and bands at >100 kDa are most likely CD44v (27). Additionally, Western blotting of immunoprecipitates from both cell lines showed molecular weight bands of CD44 similar to those observed with cell lysates (Fig. 3A). The intensity of the ∼110-kDa isoform of BT-20 lysate was weak, which is likely due to its low expression. The Western blot results thus confirm that BT-20 cells express multiple CD44 isoforms, including high-molecular-weight CD44v, and that MDA-MB-231 cells mainly express CD44s.

Fig. 3.

CD44 expressed by BT-20 cells possesses functional E-selectin ligand activity. A: cell lysate from 3 × 10⁶ cells or CD44 MAb (515) immunoprecipitates (ippt) from 1 × 10⁷ lysed cells were subjected to Western blotting with anti-CD44 MAb (2C5). β-Actin loading controls are shown for cell lysates, but not for immunoprecipitates (indicated by NA). Data are representative of 3 independent experiments. B: E-Ig chimera immunoprecipitates from 1 × 10⁷ lysed cells were subjected to Western blotting with anti-CD44 MAb (2C5) or isotype control. Data are representative of 3 independent experiments. C: E-selectin-transfected Chinese hamster ovary (CHO-E) cells were labeled with anti-CD62E MAb (filled curve) or isotype control (open curve) and tested by flow cytometry. D: CHO-E cells were perfused over antigens captured by CD44 MAb (515) or isotype control (mIgG1) at a bone marrow microvascular wall shear rate of 100 s⁻¹. Values are means ± SE; n = 5. *P < 0.05 vs. mIgG1. $P < 0.05 vs. BT-20.

To initially screen for E-selectin ligand activity of CD44, Western blot analysis of E-Ig chimera immunoprecipitates was carried out using anti-CD44 MAb (2C5) or an isotype control. As shown in Fig. 3B, bands for BT-20 at ∼95, ∼110, and ∼150 kDa were detected by the anti-CD44 MAb, but not by the isotype control. Note that these bands (Fig. 3B) correspond to molecular weights for CD44 similar to those shown in Fig. 3A. In contrast to the results observed with BT-20 cells, probe of the E-Ig chimera immunoprecipitates from MDA-MB-231 cells did not reveal any detectable CD44 bands (Fig. 3B). These results suggest that the CD44 isoforms expressed by BT-20 cells are E-selectin ligands but the CD44s expressed by MDA-MB-231 cells has limited, if any, E-selectin ligand activity.

We next sought to determine if CD44 expressed by breast cancer cells are functional E-selectin ligands under physiological flow conditions. Antigens captured from BT-20 cell lysates using anti-CD44 (515) MAb were adsorbed on a petri dish. CHO-E cells, a cell line stably expressing E-selectin (Fig. 3C), were subsequently perfused over the adsorbed antigens at a bone marrow vasculature wall shear rate of 100 s⁻¹. CHO-E cells robustly adhered to CD44 immunoprecipitated from BT-20 cells and LS-174T colon cancer cells (positive control) (12) (Fig. 3D). In contrast, CHO-E cells exhibited negligible adhesion to antigens captured with an isotype control (Fig. 3D). In addition, treatment of CHO-E cells with an E-selectin function-blocking MAb (anti-CD62E MAb) prior to perfusion through the flow chamber significantly diminished (>7-fold reduction) CHO-E cell adhesion to CD44 from BT-20 cells (data not shown). In combination, these results show that adhesion of CHO-E cells to CD44 from BT-20 cells is specifically mediated by E-selectin. In contrast, CD44 isolated from MDA-MB-231 cells supported relatively low levels of CHO-E cell adhesion. The level of adhesion was only 12% of that observed for CD44 isolated from BT-20 cells. These data collectively demonstrate that CD44 isolated from BT-20 cells is a highly avid E-selectin ligand under physiological flow conditions but CD44 from MDA-MB-231 cells does not express ligand activity at an appreciable level.

CD44 on intact BT-20 cells is a functional E-selectin ligand under physiological flow conditions.

To investigate the E-selectin ligand activity of CD44 on intact BT-20 breast cancer cells, BT-20 cells transduced with CD44 shRNA or empty vector were perfused over a monolayer of CHO-E cells. Initially, the efficacy of CD44 silencing was verified by flow cytometry, which showed that CD44-silenced BT-20 cells expressed a significantly lower level (∼50% reduction) of CD44 than vector control cells (Fig. 4A). Furthermore, the adhesion of wild-type, vector control, and CD44-silenced BT-20 cells to CHO-E cells was specifically mediated by E-selectin, as the cells adhered to CHO-E cells, but not to CHO-E cells treated with anti-CD62E MAb (data not shown). Notably, the rolling velocities of silenced cells at 400, 800, and 1,600 s⁻¹ were significantly higher than vector cells (Fig. 4B), revealing involvement of E-selectin-CD44 ligation in control of the cell rolling velocity. Furthermore, CD44-silenced BT-20 cells were less stably adherent than vector cells; this was especially apparent at higher shear rates. For example, ∼60% of vector cells were adherent, whereas only ∼30% of the silenced cells remained attached at 800 s⁻¹ (Fig. 4C). These data demonstrate that CD44 expressed by intact BT-20 cells possess functional E-selectin ligand activity under physiological flow conditions.

Fig. 4.

CD44 silencing in BT-20 cells reduces adhesion to E-selectin-expressing cells. A: BT-20 cells transfected with empty vector or shRNA against CD44 were labeled with anti-CD44 MAb (G44-26) or isotype control and analyzed by flow cytometry. *P < 0.05 vs. isotype. $P < 0.05 vs. vector. B: rolling velocities of vector control or CD44-silenced BT-20 cells over CHO-E cells. Values are means ± SE; n = 15 cells. *P < 0.05 vs. vector. C: vector control or CD44-silenced cells were perfused at a wall shear rate of 100 s⁻¹, and shear rate was increased stepwise every 30 s. Data are presented as percentage of attached cells at the end of each step. Values are means ± SE; n = 5 independent experiments. *P < 0.05 vs. vector.

BT-20 cell CD44v isoforms are sufficient for shear-resistant adhesion of CHO-E cells.

To investigate whether specific CD44v isoforms are sufficient for functional E-selectin ligand activity, antigens immunopurified using MAbs against specific variants were adsorbed onto tissue culture dishes, and CHO-E cells were perfused over the captured antigens at 100 s⁻¹. Since BT-20 cells mainly expressed CD44v3-6 isoforms on the cell surface (Fig. 2), only these isoforms were tested for E-selectin ligand activity. Notably, CHO-E cells strongly adhered to CD44v3 and CD44v4/5 but barely adhered to antigens isolated with CD44v6 or the isotype control (Fig. 5A). Additionally, the possibility that the anti-CD44v6 MAb functionally blocks E-selectin binding sites was countered by treatment of anti-CD44 MAb (515)-captured antigens with anti-CD44v6 MAb. That is, anti-CD44v6 MAb did not reduce CHO-E cell adhesion to anti-CD44 MAb (515)-captured antigens. In contrast to CD44v isolated from BT-20 cells, the adhesion of CHO-E cells to CD44 isoforms captured from MDA-MB-231 cell lysate was limited and equivalent to isotype control (Fig. 5A), as expected. These results collectively demonstrate that CD44v isoforms, particularly CD44v3 and CD44v4/5, expressed by BT-20 cells possess sufficient E-selectin ligand activity to mediate cell adhesion under physiological flow conditions.

Fig. 5.

CD44v isoforms on BT-20 cells possess sufficient E-selectin ligand activity to support cell adhesion under hematogenous flow conditions. A: CHO-E cells were perfused over antigens captured by specific CD44v MAbs or isotype controls at a bone marrow microvascular wall shear rate of 100 s⁻¹. Values are means ± SE; n = 5 independent experiments. *P < 0.05 vs. isotype control (mIgG1). $P < 0.05 vs. respective BT-20 cell CD44v. B: number of CHO-E cells adhering to individual CD44v antigens was normalized with respect to CHO-E cell adhesion to all forms of CD44 (515). “All variants” bar is the sum of CHO-E cell adhesion to CD44v3, CD44v4/5, and CD44v6, normalized to CHO-E cell adhesion to all forms of CD44 (515). Values are means ± SE; n = 5 independent experiments.

To estimate the relative E-selectin ligand activities of CD44v vs. CD44s, the adhesion data of each variant were normalized to the adhesion data for all CD44 isoforms. If it is assumed that the anti-CD44 MAb 515 captures all CD44 isoforms (25), the normalized values represent percent contributions of each variant isoform to E-selectin ligand activity. As shown in Fig. 5B, the adhesion attributed to CD44v3 and CD44v4/5 was >50% and 35%, respectively, indicating that the total contribution of CD44v was >50%. These data correspond to the staining intensities in BT-20 Western blots in Fig. 3B, in which the E-selectin ligand activity of CD44 at ∼150 kDa (most likely CD44v) is the highest and that at ∼95 kDa (most likely CD44s) is the lowest. These findings suggest greater contribution of CD44v than CD44s to E-selectin ligand activity.

BT-20 cell CD44 is HECA-452-negative, with E-selectin ligand activity primarily associated with N-linked glycans.

To examine whether the E-selectin ligand activity of CD44 expressed by breast cancer cell lines is associated with classical terminal sialofucosylated groups, indicating selectin binding sites (21, 52), CD44 immunoprecipitated from breast cancer cell lysate was subjected to Western blot analysis using HECA-452 MAb. Neither CD44 isolated from the BT-20 cell line nor CD44 isolated from the MDA-MB-231 cell line was reactive to HECA-452 MAb (Fig. 6A). In contrast, CD44 isolated from LS-174T colon cancer cells (positive control) (24) was reactive to HECA-452 MAb at ∼150 kDa, as expected (Fig. 6A). These observations strongly suggest that the E-selectin ligand activity of CD44 isolated from BT-20 cells is due to non-HECA-452-reactive antigens and, perhaps, novel terminal glycosylations.

Fig. 6.

E-selectin ligand activity on BT-20 cell CD44 may be conferred by HECA-452-negative glycans. A: CD44 immunoprecipitated from 1 × 10⁷ lysed cells (using anti-CD44 MAb 515) was tested by Western blotting with HECA-452 MAb. Data are representative of 3 independent experiments. B: E-Ig chimera immunoprecipitates from 1 × 10⁷ untreated or glycosylation inhibitor-treated BT-20 cells were subjected to Western blot analysis using anti-CD44 MAb (2C5) or isotype control. β-Actin loading controls are shown for lysates of treated and untreated BT-20 cells. Data are representative of 3 independent experiments. DMJ, deoxymannojirimycin; Bzl-GalNAc, benzyl-N-acetyl-α-galactosaminide. C: qRT-PCR was performed on untreated or glycosylation inhibitor-treated BT-20 cell mRNA. Data are presented as expression relative to the housekeeping gene GAPDH. Values are means ± SE; n = 4 independent experiments. No statistically significant difference was found among the means of untreated or treated BT-20 cells by one-way ANOVA coupled with Tukey's multiple-comparison test.

To further elucidate the glycan characteristics responsible for CD44 function as an E-selectin ligand, lysates of BT-20 cells cultured with N- and O-linked glycosylation inhibitors (DMJ and Bzl-GalNAc, respectively) were subjected to immunoprecipitation with E-Ig chimera and then to Western blot analysis using anti-CD44 MAb (2C5). The E-selectin ligand activity of CD44 from DMJ-treated cells was clearly diminished relative to that of CD44 from untreated and Bzl-GalNAc-treated cells (Fig. 6B). Neither of the treatments significantly reduced the expression levels of CD44 isoforms on the basis of qRT-PCR performed on mRNA extracted from treated and untreated cells (Fig. 6C). Therefore, glycosylation inhibitor treatments revealed that the E-selectin ligand activity of CD44 expressed by BT-20 cells is mainly associated with N-linked glycans.

Fucosyltransferase-3 and -6 are upregulated in BT-20 cells compared with MDA-MB-231 cells.

To unravel molecular mechanisms involved in the synthesis of E-selectin ligands, mRNA expression levels of α-(1,3)- and α-(1,4)-fucosyltransferases (FTs), which catalyze terminal fucosylation necessary for E-selectin ligand function (8), were analyzed. The qRT-PCR analysis (7) showed that the expression levels of FT-3 and FT-6 mRNA were significantly higher in BT-20 than MDA-MB-231 cells and the expression of FT-5 and FT-7 was not significantly different in the two breast cancer cell lines (Fig. 7). FT-4 was expressed by MDA-MB-231, but not BT-20, cells. Taken together, these data imply that FT-3 and FT-6 primarily regulate the E-selectin ligand expression in BT-20 cells.

Fig. 7.

Fucosyltransferases (FT-3 and FT-6) are upregulated in BT-20 cells compared with MDA-MB-231 cells. mRNA expression levels of fucosyltransferases were determined by qRT-PCR. Data are presented as expression relative to the housekeeping gene GAPDH. ND, not detected. Values are means ± SE; n = 3 independent experiments. *P < 0.05 vs. BT-20.

Breast cancer cell expression of epithelial and mesenchymal cell markers.

Recently, it has been shown that expression of E-selectin ligands in colon cancer cells is regulated by epithelial-to-mesenchymal transition (EMT) (43), a process believed to be critical for metastasis (36, 39). Also, it has been shown that expression of CD44 isoform switching, through downregulation of CD44v, is necessary for EMT (10). In light of these reports, we sought to uncover whether the differential expression and E-selectin ligand function of CD44 isoforms correlate with epithelial or mesenchymal phenotype of the breast cancer cell lines. A dramatically higher mRNA level of the epithelial marker E-cadherin, yet markedly lower mRNA levels of the mesenchymal markers N-cadherin and SLUG (Fig. 8A), the latter of which is a transcriptional repressor of E-cadherin, were expressed by BT-20 than by MDA-MB-231 cells. Similar to mRNA expression, BT-20 cells expressed higher protein levels of E-cadherin but almost no N-cadherin, while MDA-MB-231 cells expressed higher protein levels of N-cadherin but almost no E-cadherin (Fig. 8B). These data indicate that CD44 as an E-selectin ligand may be expressed preferentially by breast cancer cells in an epithelial-like state.

Fig. 8.

BT-20 cells are epithelial-like cells, and MDA-MB-231 cells are mesenchymal-like cells. A: qRT-PCR was performed on mRNA extracted from breast cancer cells. Data are presented as expression relative to the housekeeping gene GAPDH. Values are means ± SE; n = 4 independent experiments. *P < 0.05 vs. BT-20. B: breast cancer cells grown on tissue culture slides were labeled with anti-E-cadherin or N-cadherin MAb or isotype control followed by isotype-matched Alexa Fluor 488-conjugated secondary antibodies (green). Slides were counterstained with DAPI (blue) and then imaged using epifluorescence microscopy. Scale bar = 25 μm.

DISCUSSION

Glycoforms of CD44 are prevalent E-selectin ligands promoting shear-resistant adhesion of circulating cells to endothelium in a variety of physiological and pathological processes. For instance, E-selectin-CD44 interactions mediate homing of hematopoietic stem cells to bone marrow (18) and are suggested to be involved in the osteotropism of prostate cancer cells (15). Notably, breast cancer frequently metastasizes to bone marrow (53), the endothelium of which constitutively expresses E-selectin (44), and certain breast cancer cell lines adhere to endothelium via E-selectin (23, 37, 46, 47). For these reasons, we were compelled to perform a detailed investigation of E-selectin ligand function of breast cancer cell CD44 under bone microvascular flow conditions. Our results project a novel perspective on the CD44-E-selectin pathway by demonstrating that CD44v isoforms expressed by breast cancer cells are more relevant E-selectin ligands than CD44s under physiological flow conditions.

Consistent with previous studies (1, 38, 45), each of the four breast cancer cell lines used in the present study strongly expressed CD44 (Fig. 1A) detected by an anti-CD44 MAb (515; recognizes all isoforms) (25). The distinguishing feature, however, was the difference in expression levels of specific CD44v isoforms. While BT-20 cells expressed high levels of mRNA for most variant isoforms, MDA-MB-231 cells expressed conspicuously low levels of CD44v mRNA (Fig. 1B). These expression patterns were even more convincing for cell surface proteins (Fig. 2). Western blot analysis further confirmed these patterns of expression (Fig. 3A): BT-20 and MDA-MB-231 cells expressed the ∼95-kDa form, most likely CD44s (18), but only BT-20 cells expressed ∼110- and ∼150-kDa forms, which are most likely CD44v isoforms (24). These cell lines provided a system to compare the E-selectin ligand activity of highly and weakly CD44v-positive breast cancer cells.

It is believed that since CD44v isoforms possess additional glycosylation sites and may extend farther from the cell surface than CD44s, CD44v may have higher functional selectin ligand activity (12, 24, 25). In accordance with this notion, the level of E-selectin ligand activity was higher for CD44 on BT-20 (CD44v-expressing) cells than MDA-MB-231 (predominantly CD44s-expressing) cells (Fig. 3, B and D). Specifically, CD44 from BT-20 cells was sufficient to engage flowing CHO-E cells (Fig. 3D), was necessary for stabilizing E-selectin-mediated cell rolling (Fig. 4B), and appeared essential for high-avidity binding (Fig. 4C). Furthermore, antigen capture assays clearly suggest that the major E-selectin ligand activity of breast cancer cell CD44 is associated with CD44v, particularly CD44v3 and CD44v4/5 (Fig. 5). Notably, solid cancer cells with strong E-selectin ligand activity, such as colon (12, 24) and breast cancer cells (present data), are associated with high levels of CD44v (25, 32, 38). Thus the expression of CD44v as an E-selectin ligand could be a potential predictive metastasis marker, at least in certain cancer types.

Since E-selectin binds to carbohydrate epitopes on a core molecule, such as CD44, appropriate glycosylations are requisite for E-selectin ligand function (25, 31). In this regard, the HECA-452 MAb recognizing sialofucosylated groups has been classically used to detect E-selectin-reactive carbohydrates (18, 24). However, HECA-452-negative E-selectin ligands have been reported (53, 54), and HECA-452-negative molecules have been predicted to be the principal E-selectin ligands on the MDA-MB-468 breast cancer cell line (29, 46). Similarly, the E-selectin ligand activity of BT-20 cell CD44v was associated with HECA-452-negative glycans (Fig. 6A), suggesting a novel glycoform on the BT-20 breast cancer cell line. These E-selectin-reactive epitopes were predominantly carried by complex N-linked glycans, rather than by O-linked glycans (Fig. 6B). Previously, the selectin ligand activities of hematopoietic stem/progenitor cells, AML cells, and the KG1a AML cell line have been attributed to HECA-452-reactive, N-glycosylated CD44s and, to a lesser extent, CD44v (17, 18, 42) and that of LS-174T colon cancer cells has been attributed to HECA-452-reactive, O-glycosylated CD44v (12, 24, 25). Taken together, these multiple CD44v glycoforms in different cell types suggest significant adaptability of CD44 as a selectin ligand.

A recent investigation linked CD44v with EMT by showing that expression of CD44v is downregulated as epithelial cells become mesenchymal cells (10). In line with this report, our data showed relatively higher expression of CD44v on epithelial BT-20 cells than on mesenchymal MDA-MB-231 cells. Moreover, CD44v expressed on BT-20 cells possessed E-selectin ligand activity, most likely rendered by the action of α-(1,3)/α-(1,4)-fucosyltransferase (FT-3) and/or α-(1,3)-fucosyltransferase (FT-6) (Fig. 7). Therefore, these data suggest an association between EMT and CD44v as an E-selectin ligand and, perhaps, E-selectin ligand activity in general.

Insight into the negative relationship of E-selectin ligand activity of CD44 with the mesenchymal state of breast cancer cells can be obtained from other studies. During metastasis, tumor cells undergo EMT (36, 39), as well as the reverse, mesenchymal-to-epithelial transition (MET) (35). While the mesenchymal state may be beneficial for cell invasion and migration in tissue (28), expression of functional E-selectin ligands is presumably unnecessary for these processes. However, the epithelial state with corresponding upregulated E-selectin ligands (such as glycoforms of CD44v) may be required for the stable, shear-resistant adhesion of blood-borne CTCs to vascular endothelium expressing E-selectin. Of note, CTCs have been found to express markers of epithelial and mesenchymal phenotypes (2, 5, 9, 27). Thus we hypothesize that successful metastasis is partially attributable to regulation of E-selectin ligand expression by cancer cells in these complementary states (13) and that targeting E-selectin ligand activity of CD44 at an opportune time may be necessary to reduce metastasis in vivo. For instance, in line with a previous report by Dallas et al. (15), it is possible that knockdown of CD44v in breast cancer cells may not reduce, but rather increase, their metastatic potential in vivo, if these cancer cells express the glycoforms of CD44v that possess E-selectin ligand activity. This is because CD44v knockdown may facilitate EMT (10), while redundant E-selectin ligands expressed on CTCs (46, 47), which are less sensitive to EMT regulation, may serve as the critical shear-resistant adhesion molecules mediating adhesion to the vascular wall. The mechanistic underpinnings of E-selectin ligand expression due to EMT and MET may reveal a new paradigm for breast cancer metastasis, which we continue to investigate in our laboratory (13).

In summary, the present investigation demonstrates that CD44, particularly CD44v3 and CD44v4/5, expressed on the epithelial-like BT-20 breast cancer cell line are E-selectin ligands under physiological flow conditions. Furthermore, the E-selectin ligand activity of CD44v is due to HECA-452-negative N-glycans generated most likely by FT-3 and FT-6. In contrast, the predominant CD44 expressed by the mesenchymal-like MDA-MB-231 breast cancer cell line has very limited ability to function as an E-selectin ligand. Further investigation of glycoforms of CD44 as shear-resistant E-selectin ligands associated with epithelial and mesenchymal states may ultimately lead to the development of new diagnostics, prognostics, and therapeutics against cancer.

GRANTS

This work was supported by Grant CBET-1106118 (to D. F. J. Tees, F. Benencia, and M. M. Burdick) and Major Research Instrumentation Grant CBET-1039869 (to D. F. J. Tees, F. Benencia, D. J. Goetz, and M. M. Burdick) from the National Science Foundation and National Cancer Institute Grant 1R15 CA-161830-01 (to M. M. Burdick).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

V.S.S., T.L., F.B., D.J.G., and M.M.B. are responsible for conception and design of the research; V.S.S., T.L., L.F.D., and C.M.C. performed the experiments; V.S.S., T.L., L.F.D., C.M.C., D.F.J.T., F.B., D.J.G., and M.M.B. analyzed the data; V.S.S., D.F.J.T., F.B., D.J.G., and M.M.B. interpreted the results of the experiments; V.S.S., T.L., and M.M.B. prepared the figures; V.S.S. and M.M.B. drafted the manuscript; V.S.S. and M.M.B. edited and revised the manuscript; V.S.S., T.L., L.F.D., C.M.C., D.F.J.T., F.B., D.J.G., and M.M.B. approved the final version of the manuscript.