Cell surface-expressed moesin-like HDL/apoA-I binding protein promotes cholesterol efflux from human macrophages.

HDL and its major component, apolipoprotein A-I (apoA-I), play a central role in reverse cholesterol transport. We recently reported the involvement of a glycosylphosphatidylinositol anchor (GPI anchor) in the binding of HDL and apoA-I on human macrophages, and purified an 80 kDa HDL/apoA-I binding protein. In the present study, we characterized the GPI-anchored HDL/apoA-I binding protein from macrophages. The HDL/apoA-I binding protein was purified from macrophages and digested with endopeptidase, and the resultant fragments were sequenced. Cholesterol efflux, flow cytometry, immunoblotting, and immunohistochemical analyses were performed to characterize the HDL/apoA-I binding protein. Two parts of seven amino acid sequences completely matched those of moesin. Flow cytometry, immunoblotting, and immunohistochemistry using anti-moesin antibody showed that the HDL/apoA-I binding protein was N-glycosylated and expressed on the cell surface. It was termed moesin-like protein. Treatment of macrophages with anti-moesin antibody blocked the binding of HDL/apoA-I and suppressed cholesterol efflux. The moesin-like protein was exclusively expressed on macrophages and was upregulated by cholesterol loading and cell differentiation. Our results indicate that the moesin-like HDL/apoA-I binding protein is specifically expressed on the surface of human macrophages and promotes cholesterol efflux from macrophages.—Matsuyama, A, N. Sakai, H. Hiraoka, K-i. Hirano, and S. Yamashita. Cell surface-expressed moesin-like HDL/apoA-I binding protein promotes cholesterol efflux from human macrophages.

Anti-moesin antibody decreased the binding and subsequent cholesterol efflux of HDL/apoA-I from macrophages, suggesting that the moesin-like protein is involved in the efflux of cholesterol from human macrophages.

Amino acid sequencing
The purified 80 kDa apoA-I/HDL binding protein (11) was applied for sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and extracted directly from the gel by an in-gel lysil-endopeptidase digestion method. The digested fragments were fractionated by high-performance liquid chromatography, and seven peaks were individually collected. They were subjected to amino acid sequencing in a Perkin-Elmer sequencer, and two amino acid sequences were identified.

Isolation of human monocyte-derived macrophages
Mononuclear cells were isolated from the buffy coats of blood collected from healthy volunteers using density gradient centrifugation with Lymphoprep (Nycomed; Oslo, Norway) (12) and cultured as described previously (11). After a 6 day culture, macrophages were loaded with cholesterol by incubation with 50 mg/ml of acetylated LDL (AcLDL), and then used for the experiments.

Cell culture
HepG2 and HEK293 cells, human fibroblasts, and THP-1 cells were obtained from American Type Cell Collection. HepG2 and HEK293 cells and human fibroblasts were cultured in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) heatinactivated fetal calf serum (FCS). THP-1 cells were cultured in RPMI 1640 supplemented with 10% FCS and were transformed to macrophages by incubation for 24 h with 50 ng/ml of phorbol 12-myristate 13-acetate (PMA) (Nacalai Tesque; Kyoto, Japan).
Lipoprotein isolation, modification, and labeling HDL3 was isolated from human plasma by ultracentrifugation at a density of 1.125-1.210 g/ml (13). Protein concentration was measured following the procedure of Lowry et al. (14). ApoA-I was purchased from Sigma (St. Louis, MO). HDL3 and apoA-I were labeled with fluorescein isothiocyanate (FITC) (Molecular Probes; Eugene, OR) as reported previously (11,15). DiI-AcLDL was purchased from Molecular Probes. DiI-oxidized LDL was from Biomedical Technologies, Inc. (Stoughton, MA).

Immunoblotting analysis
Plasma membranes were prepared from human monocytederived macrophages, HepG2 cells, HEK293 cells, and human fibroblasts as described previously (11), and were applied for immunoblotting analysis with anti-moesin antibody (TK88, kind gift from S. Tsukita, Kyoto University Graduate School of Medicine). For two-dimensional immunoblotting analysis, the membrane proteins were separated using isoelectric focusing (IEF) gels (Amersham Pharmacia Biotech; Buckinghamshire, UK) according to the protocol supplied by the manufacturer and 4-20% polyacrylamide gradient gels (Daiichi Pure Chemicals; Tokyo, Japan), transferred onto nitrocellulose membranes, and blotted with antimoesin antibody (TK88). After incubation with peroxidaseconjugated anti-rabbit IgG antibody, the blots were visualized with an ECL kit (Amersham Pharmacia Biotech).

PI-PLC treatment
To determine the effect of PI-PLC (Sigma), the macrophages were incubated at 378C with 1 U/ml of PI-PLC for 1 h and applied for experiments.

Enzymatic deglycosylation
Concentrated PI-PLC-treated media (10 ml) were incubated for 24 h at 378C with or without 50 units of N-glycosidase F (Sigma) in potassium buffer (200 mM K 3 PO 4 , 20 mM EDTA, pH 7.2) as indicated by the manufacturer, followed by the addition of 1 vol of 23 Laemmli sample buffer, and applied for immunoblotting analysis.

Flow cytometry
Cell surface expression of moesin-like protein was confirmed by flow cytometric analysis using anti-moesin antibody (TK88 and TK89, a kind gift from S. Tsukita). Competition analyses with anti-moesin antibody or control IgG were performed by incubating cells for 1 h at 48C with FITC-labeled HDL3 or apoA-I. All fluorescence measurements were performed by flow cytometry using a FACScan (Becton-Dickinson; Mountain View, CA). Cell-bound fluorescence intensity was expressed as the mean intensity of fluorescence (MIF).

Immunohistochemistry
For immunohistochemical analysis to detect moesin, cultured human macrophages were incubated in a 2-well glass chamber slide (Falcon) and fixed for 30 min with 4% paraformaldehyde. The cells were blocked with 5% normal horse serum and incubated with anti-moesin antibodies (TK88 and TK89) in PBS containing 3% BSA for 1 h under detergent-free conditions, followed by sequential incubation with FITC-labeled anti-rabbit IgG antibody. The treated cells were observed using confocal fluorescence microscopy (Floview FV1000; Olympus, Tokyo).

Cell surface biotinylation
The cholesterol-laden macrophages and fibroblasts were washed with ice-cold phosphate buffer (10 mM, pH 7.4). They were subsequently treated with 1 ml of 0.8 mM NHS-biotin (watersoluble biotin; Pierce, Rockford, IL) in phosphate buffer for 15 min on ice, and washed three times with 0.192 M glycine-25 mM Tris (pH 8.3) to quench the reaction. The cells were lysed in MES buffer containing a cocktail of protease inhibitors (0.5 mg/ ml leupeptin, 1 mg/ml aprotinin, 1 mg/ml pepstatin A; Roche Applied Science), and membrane fractions were obtained from the materials as described previously (11). The membrane pellets were dissolved and applied for immunoprecipitation using antimoesin antibody and control IgG. Immunoprecipitates were applied for SDS-PAGE and visualized with streptavidin-horseradish peroxidase (HRP) and the ECL kit.

Measurement of cellular cholesterol efflux from macrophages
Cellular cholesterol efflux via HDL3 or apoA-I was determined as described previously (11,16). The cells were incubated with 20 mg/ml of HDL3 or 5 mg/ml of apoA-I diluted in RPMI with 5 mg/ml of BSA with 100 mg/ml of anti-moesin antibody or control IgG (Transduction Laboratories; San Jose, CA). After incubation for the indicated time, radioactivities in both the medium and the cells were measured separately. Fractional cholesterol efflux was calculated as the amount of radioactivity in the medium divided by the total radioactivity in each well and expressed as a percentage. The cellular cholesterol contents were measured using a cholesterol-measuring kit (Kyowa Medics, Tokyo) according to the instructions provided by the manufacturer.

Sequencing of the 80 kDa HDL/apoA-I binding protein of macrophages
A single band of the 80 kDa apoA-I/HDL binding protein (5 pmol) was obtained from human macrophages derived from 25 healthy volunteers (10 l of whole blood) as reported previously (11), and applied for lysil-endopeptidase digestion followed by fractionation using HPLC. Seven peaks were individually collected and subjected to amino acid sequencing in a Perkin-Elmer sequencer. Two of them were successfully sequenced. The sequences of the peptides (DQWEERIQVWH and TANDEMHAEN) were the same as the internal sequences of human moesin (amino acid 164-174 and 540-549, respectively) (17).

Designation of the 80 kDa HDL/apoA-I binding protein as moesin-like protein
Previous studies reported that moesin is localized in the sub-membranous cytoskeleton, filopodia, and other membranous protrusions (17), but to our knowledge, cell surface expression of moesin has not been reported previously. To determine whether the HDL/apoA-I binding protein is identical to moesin, we first performed immunoblotting analyses using anti-moesin antibody (Fig. 1). The 80 kDa moesin-like protein was recovered in the conditioned media after PI-PLC treatment, whereas no immunoreactive mass was observed without PI-PLC treatment (Fig. 1Aa). It was noted that the molecular weight of the moesin-like protein was about 2 kDa larger than that located in the cell membrane. To confirm the loss of the moesin-like protein from the membrane after treatment with PI-PLC, membrane fractions of macrophages with or without PI-PLC treatment were applied for immunoblotting with anti-moesin antibody (Fig. 1Ab). The larger bands of two moesin-like proteins were reduced after PI-PLC treatment. These results suggest that the protein is a glycosylphosphatidylinositol (GPI)anchored type and not identical to moesin.
To confirm that the 80 kDa HDL/apoA-I binding protein is not identical to moesin, two-dimensional immunoblotting was performed (Fig. 1B). The membrane fraction of human macrophages and the conditioned media after PI-PLC treatment were applied for two-dimensional immunoblotting with anti-moesin antibody. A massive amount of moesin at pH 6.0 (asterisk), as predicted by the amino acid content of moesin, and a small amount of protein with isoelectric points of pH 7.2 to 7.6 (arrowheads) were observed in membrane fraction proteins. The conditioned media with PI-PLC treatment showed moesin-like protein (arrowheads) but not moesin.
To further confirm that the 80 kDa HDL/apoA-I binding protein was not identical to moesin, enzymatic deglycosylation with N-glycosidase F was performed. After deglycosylation with the enzyme, the molecular weight of the 80 kDa HDL/apoA-I binding protein was shifted down compared with that of the nondigested protein (Fig. 1Ca), indicating that the protein was N-glycosylated. Next, the change in isoelectric point of the protein after deglycosylation was examined. If the protein was N-linked, charged glucans would be detached after deglycosylation and the isoelectric point would change. After N-glycosidase F treatment (Fig. 1Cb), the isoelectric point of the protein (pH 7.2 to 7.6) changed to pH 6.0. This result reinforced the notion that the protein was N-glycosylated. Because moesin is not expressed on the cell surface, as reported previously (17), the 80 kDa HDL/apoA-I binding protein was confirmed to be different from moesin and thus designated as moesin-like protein.

Expression of moesin-like protein on the cell surface of human macrophages
To determine whether the moesin-like protein was expressed on the cell surface of human macrophages, flow cytometry was performed using anti-moesin antibodies under detergent-free conditions. As shown in Fig. 2Aa, moesin-like protein was detected on the surface of macrophages by anti-moesin antibodies (TK88 and TK89). To confirm the loss of moesin-like protein from the cell surface, flow cytometric analysis was performed after PI-PLC treatment (Fig. 2Ab). PI-PLC treatment shifted the distribution to the left, indicating that the moesin-like protein was GPIanchored and expressed on the cell surface.
To observe the distribution of the cell surface-expressed moesin-like protein, confocal microscopic analysis was performed using anti-moesin antibody (TK88) under detergent-free conditions (Fig. 2B). En face sections (xy) images were taken off the macrophage from the top to the bottom, and computer-reconstructed vertical section (xz and yz) images taken through the full thickness of the macrophage were obtained. Moesin-like immunoreactivities were observed only on the cell surface of the macrophages. No fluorescence signal was observed using control rabbit IgG (data not shown).
To further confirm that human monocyte-derived macrophages express moesin-like protein on their cell surface, cell surface biotinylation analysis was performed (Fig. 2C). After a monolayer of macrophages was treated with NHSbiotin (water-soluble biotin), the membrane fraction of the cells was collected and immunoprecipitated with antimoesin antibody (TK88). The immunoprecipitates were applied for SDS-PAGE, transferred onto nitrocellulose membrane, and then visualized with avidin-HRP. As indicated by the arrow (Fig. 2C), the 80 kDa protein was labeled with NHS-biotin, suggesting that the moesin-like protein was expressed on the cell surface of macrophages. However, the biotinylated protein was not visualized on the fibroblasts, indicating that the expression of moesinlike protein is specific for macrophages.
Effect of anti-moesin antibody on HDL/apoA-I binding and HDL/apoA-I-mediated cholesterol efflux from macrophages To determine whether moesin-like protein is an HDL/ apoA-I binding protein, competition studies with anti-moesin antibody were performed using flow cytometry (Fig. 3A). Compared with control IgG, anti-moesin monoclonal antibody reduced the MIF of FITC-labeled HDL3 from 46.0 to 21.9 (Fig. 3Aa) and that of FITC-labeled apoA-I from 203.2 to 80.0 (Fig. 3Ab). To confirm that anti-moesin antibody could specifically reduce HDL3 or apoA-I binding, we examined whether the antibody could suppress the bindings of DiI-oxidized LDL and DiI-AcLDL (Fig. 3Ac,   Fig. 1. Moesin-like protein is an 80 kDa HDL/apolipoprotein A-I (apoA-I) binding protein. A: Formation of moesin-like protein by phosphatidylinositol-specific phospholipase C (PI-PLC) treatment of human macrophages. After treatment of macrophages with or without 0.5 U/ml PI-PLC, the conditioned media were applied for immunoblotting analysis with anti-moesin antibody (TK88) (a). PI-PLC treatment resulted in the formation of the moesin-like protein from human macrophages. In control experiments (incubation without PI-PLC), no band was observed. After treatment with or without PI-PLC, the macrophage membranes were subjected to immunoblotting with anti-moesin antibody (b). The larger band of two moesin-like proteins was reduced after PI-PLC treatment. B: Two-dimensional immunoblotting using anti-moesin antibody. The membrane fraction of human macrophages (a) and the conditioned media after PI-PLC treatment with macrophages (b) were applied for two-dimensional immunoblotting analysis with anti-moesin antibody (TK88). A massive amount of moesin at pH 6.0 (asterisk), as predicted by the amino acid content of moesin, and a small amount of protein with an isoelectric point of pH 7.2 to 7.6 (arrowheads) were observed in membrane fraction proteins. The conditioned media with PI-PLC treatment showed moesin-like protein (arrowheads) but not moesin. C: Enzymatic deglycosylation of moesin-like protein with N-glycosidase F. The conditioned media, after treatment with 0.5 U/ml PI-PLC for 24 h at 378C, were incubated with or without N-glycosidase F. After digestion with the enzyme, the treated or nontreated conditioned media were applied for immunoblotting with anti-moesin antibody (TK88). The immunoreactive protein of digested media was shifted down compared with that of nondigested media (a), and the isoelectric point of the protein (pH 7.2 to 7.6) changed to pH 6.0 (b), indicating that the protein was N-linked. After fixation with 4% paraformaldehyde, the lipid-laden macrophages were preincubated with 5% normal horse serum and then with control rabbit IgG or anti-moesin antibody (TK88 or TK89) (a). After washing, the cells were incubated with FITC-labeled goat anti-rabbit IgG antibody. Cell-bound fluorescence was detected by FACScan. After treatment with or without PI-PLC, the macrophages were applied for FACScan analyses with anti-moesin antibody (TK88) (b). PI-PLC treatment reduced the fluorescence of the moesin-like moiety. Red peak, PI-PLC nontreated macrophages; blue peak, PI-PLC-treated; black peak, control. B: Confocal microscopic analysis of moesin-like protein. For immunohistochemical analysis of moesin-like protein, cultured human macrophages were fixed with 4% paraformaldehyde, blocked with 5% normal horse serum, and then incubated with anti-moesin antibody (TK88) for 1 h, followed by sequential incubation with FITC-labeled anti-rabbit IgG antibody, and finally confocal microscopic analysis. Vertical sections (xz and yz) were projected. En face sections (xy) images were taken off the macrophage from the top to the bottom and computer-reconstructed vertical section (xz and yz) images were taken through the full thickness of the macrophage. C: Immunoprecipitation after cell surface biotinylation. The cultured macrophages and human fibroblasts were treated with NHS-biotin (water-soluble biotin) for 15 min at 48C and then washed with 0.192 M glycine/25 mM Tris, pH 8.3, for 15 min at 48C to quench the reaction. The membrane fractions were subjected to immunoprecipitation using anti-moesin antibody (TK88) and control IgG. Immunoprecipitates were applied for SDS-PAGE, transferred onto nitrocellulose membrane, and then visualized with streptavidin-horseradish peroxidase. Ad). Neither the distribution of DiI-oxidized LDL nor that of DiI-AcLDL-incorporated cells was affected by incubation with anti-moesin antibody. Next, to address the percentage of apoA-I or HDL binding due to this specific receptor, binding studies were performed using PI-PLCtreated cells (Fig. 3Ae and f). After treatment with PI-PLC, the binding capacity of HDL3 diminished from 121.5 to 39.5 (67% reduction) and that of apoA-I from 229.1 to 55.4 (76% reduction).
To confirm the physiological function of moesin-like protein on the cell surface, we next examined whether it was involved in the HDL/apoA-I-mediated cholesterol efflux from macrophages. As shown in Fig. 3B, treatment of macrophages with anti-moesin antibody reduced HDLand apoA-I-mediated cholesterol efflux from macrophages throughout the experiment (up to 10 h). Next, we examined the effect of anti-moesin antibody on the residual cholesterol contents of macrophages after HDL-or apoA-Imediated cholesterol efflux (Fig. 3C). Anti-moesin antibody lessened the reduction of residual cholesterol after cholesterol efflux. These results suggest that the moesinlike protein is involved in HDL-and apoA-I-mediated cholesterol efflux from human monocyte-derived macrophages.

Specific expression of moesin-like protein in monocytes/macrophages
To determine whether the cell surface-expressed moesinlike protein is specifically expressed on macrophages, twodimensional immunoblotting was performed (Fig. 4A). Membrane fractions from human monocyte-derived macrophages, HepG2 cells, HEK293 cells, and human fibroblasts were collected and applied for two-dimensional immunoblotting with anti-moesin antibody (TK88). Human macrophages expressed moesin with an isoelectric point of pH 6.0 and moesin-like protein with an isoelectric point of pH 7.2 to 7.6. Cells other than monocyte macrophages expressed moesin (pH 6.0) but no moesin-like protein.
The moesin-like protein on the cell surface was further examined by flow cytometry, and its amount was determined as MIF in human macrophages, HepG2 cells, HEK293 cells, and human fibroblasts (Fig. 4B). The results confirmed the lack of moesin-like protein on the surface of nonmonocyte cell lines (HepG2, HEK293, and human fibroblasts) and its presence on human macrophages.

Effect of cholesterol loading and differentiation on the expression of moesin-like protein
To determine the functional role of the cell surfaceexpressed moesin-like protein in the reverse cholesterol transport system, we examined the effect of cholesterol loading. Six-day-cultured human macrophages were incubated for 24 h with or without AcLDL (50 mg/ml), and then subjected to flow cytometric analysis. The distribution of cell surface moesin-positive macrophages was shifted to the right by Ac-LDL loading (Fig. 5Aa), and the MIF of AcLDL-laden macrophages was upregulated about twofold to that of unladen macrophages (Fig. 5Ab). To confirm that the moesin-like protein was increased after cholesterol loading, membrane fractions derived from Ac-LDL-laden and -unladen macrophages were applied for immunoblotting analyses using anti-moesin antibody (Fig. 5B). The larger band of two moesin-like proteins, which was shown as the cell surface-expressed moesin-like protein, was increased by cholesterol loading. These results suggest that the expression of cell surface-expressed moesin-like protein is enhanced in association with foam cell formation. Finally, we examined whether moesin-like protein increases with cell differentiation. For this purpose, THP-1 cells were treated with PMA and subjected to flow cytometric analysis (Fig. 5C). The MIF of moesin-like protein after PMA treatment was expressed twice as much as that before treatment, indicating that the cell surfaceexpressed moesin-like protein was upregulated with differentiation of monocytes into macrophages.

DISCUSSION
In the present study, we identified the protein recognized by anti-human moesin antibodies, and termed it moesin-like protein. The protein is a GPI-anchored type, is expressed on the cell surface of human macrophages, is induced by differentiation and cholesterol loading of macrophages, and is involved in the binding and cholesterol efflux of HDL/apoA-I.
Internal sequencing of the previously purified apoA-I/ HDL binding protein matched the sequences of the two internal fragments of moesin. Moesin belongs to the ERM family of proteins (along with ezrin and radixin) that are known to be involved in cell adhesion and membrane dynamics, probably because of their ability to link plasma membrane components with the actin cytoskeleton (17)(18)(19). Although the functions of ERM proteins have not yet been fully delineated, these proteins are widely expressed in different tissues and cells. They have been identified in filopodia and are associated with other membranous proteins that are important for ligand recognition, signal transduction, and cell motility (17)(18)(19)(20). Most previous reports have focused on moesin as a linking protein of the submembranous cytoskeleton.
Although moesin has a cell surface domain involved in the binding of lipopolysaccharides (21), IL-2 fragments (22), and human immunodeficiency virus type 1 envelope protein gp 120 (23), moesin has no signal peptide, is not expressed on the cell surface, and is not a GPI-anchored type protein. In this study, we demonstrated that the newly purified apoA-I/HDL binding protein is expressed on the cell surface, exclusively on macrophages, and is a GPIanchored type. Two-dimensional immunoblotting analysis also revealed that the IEF of the protein (pH 7.2-7.6) is different from that of moesin (pH 6.0). We also showed that the protein underwent N-glycosylation. These characteristics are totally different from those of moesin, indicating that the protein is not moesin.
There are no reports on isoforms of moesin. The protein is recognized by two kinds of antibodies raised by human moesin. The epitopes of these antibodies are located between two internal sequences that matched the fragments sequences of the moesin-like protein. The RT-PCR product, with primers located on two matched sequences using RNAs from human macrophages, yielded a single band (data not shown). These data suggest that the protein prified in our study is an isoform of moesin. The isoform of moesin is thought to be transported by the Golgi-ER Six day-cultured human macrophages were incubated for 24 h with or without AcLDL (50 mg/ml), and then subjected to flow cytometry. The MIF of AcLDL-laden macrophages was upregulated about two-fold relative to that of unladen macrophages. B: After loading cholesterol with Ac-LDL, the macrophage membranes were subjected to immunoblotting with anti-moesin antibody. The intensity of the larger band of two moesin-like proteins was increased on Ac-LDL-laden macrophages compared with nonladen. C: Effect of differentiation on moesin-like protein expression. THP-1 cells were treated with phorbol 12-myristate 13-acetate (PMA) (50 ng/ml) for 24 h and then subjected to flow cytometry. The MIF of the moesin-like protein after PMA treatment was double that before treatment. Data are mean 6 SEM of n experiments. pathway and expressed on the cell surface of human macrophages. Gelsolin, for example, also has two isoforms; one functions as an actin binding protein intracellularly, and the other has a signal peptide and is secreted by the liver. The pattern of expression is also different for the two gelsolin isoforms. This alternative splicing mechanism of gelsolin resembles that of moesin. One of two isoforms of moesin can be a cell surface binding site of HDL and apoA-I.
In conclusion, our data demonstrate that the cell surface-expressed moesin-like protein is a binding protein for apoA-I/HDL on the surface of human monocyte-derived macrophages and can stimulate apoA-I-/HDL-mediated cholesterol efflux. Further studies are in progress to clone the cell surface-expressed moesin-like protein, which may enhance our understanding of the initial step of reverse cholesterol transport.