Identification of mRNA splicing factors as the endothelial receptor for carbohydrate-dependent lung colonization of cancer cells

Abstract

Cell surfaces of epithelial cancer are covered by complex carbohydrates, whose structures function in malignancy and metastasis. However, the mechanism underlying carbohydrate-dependent cancer metastasis has not been defined. Previously, we identified a carbohydrate-mimicry peptide designated I-peptide, which inhibits carbohydrate-dependent lung colonization of sialyl Lewis X-expressing B16-FTIII-M cells in E/P-selectin doubly-deficient mice. We hypothesized that lung endothelial cells express an unknown carbohydrate receptor, designated as I-peptide receptor (IPR), responsible for lung colonization of B16-FTIII-M cells. Here, we visualized IPR by in vivo biotinylation, which revealed that the major IPR is a group of 35-kDa proteins. IPR proteins isolated by I-peptide affinity chromatography were identified by proteomics as Ser/Arg-rich alternative pre-mRNA splicing factors or Sfrs1, Sfrs2, Sfrs5, and Sfrs7 gene products. Bacterially expressed Sfrs1 protein bound to B16-FTIII-M cells but not to parental B16 cells. Recombinant Sfrs1 protein bound to a series of fucosylated oligosaccharides in glycan array and plate-binding assays. When anti-Sfrs antibodies were injected intravenously into mice, antibodies labeled a subset of lung capillaries. Anti-Sfrs antibodies inhibited homing of I-peptide-displaying phage to the lung colonization of B16-FTIII-M cells in vivo in the mouse. These results strongly suggest that Sfrs proteins are responsible for fucosylated carbohydrate-dependent lung metastasis of epithelial cancers.

The apical cell surface of epithelia is covered by numerous carbohydrates attached to membrane proteins and lipids. When epithelial cells are transformed, the structure of these carbohydrates changes (1). Many investigators have suggested a correlation between cancer-associated carbohydrate antigens and poor patient prognosis, including metastasis (2–8). Despite extensive structural analysis and clinical observations using monoclonal antibodies against cancer-associated carbohydrate antigens, mechanisms underlying carbohydrate-dependent cancer metastasis through the circulation remain elusive. Because E-selectin is expressed on the inflammatory endothelial cell surface and binds to cancer-associated carbohydrate antigens such as sialyl Lewis A (sLeA) and sialyl Lewis X (sLeX), E-selectin provides a mechanism for sLeA and/or sLeX antigen-expressing cancer cells to be metastasized through the hemato genous route (4–6). Involvement of P-selectin and L-selectin in cancer metastasis has been also described (7, 8). However, substantial clinical data indicate the existence of selectin-independent but carbohydrate-dependent cancer metastasis (6, 10, 11).

In our previous studies, we screened a peptide-displaying phage library using a monoclonal anti-Lewis A antibody (clone 7LE) and identified a short peptide IELLQAR, designated I-peptide (12–14). When I-peptide was injected intravenously into a mouse, it bound to a specific subset of lung capillary endothelial cells but not to blood vessels of other organs including the liver (12). I-peptide administration inhibited carbohydrate-dependent lung colonization of fucosyltransferase-3-transfected sialyl Lewis X antigen-positive B16 (B16-FTIII-M) cells (15) in wild-type mice (12) and in E- and P-selectin doubly-deficient mutant mice (13), suggesting that the lung endothelial surface expresses unknown carbohydrate-binding receptor(s) enabling colonization of B16-FTIII-M cells. We designated this presumptive receptor I-peptide receptor (IPR) (13).

In this study, we isolated IPR candidate proteins by I-peptide-affinity chromatography and identified them by proteomics. The results showed that IPRs are Ser/Arg-rich alternative mRNA splicing factors or Sfrs gene products. Here, we present data suggesting that Sfrs proteins expressed by a subset of lung capillaries are responsible for carbohydrate-dependent cancer cell colonization of the lung in the mouse.

Results

Identification of IPRs.

We visualized endothelial surface proteins that bind to I-peptide by in vivo biotinylation, followed by I-peptide-displaying phage binding, a method used previously to identify endothelial receptors for organ-targeting peptides (16). A mouse was injected with a biotinylation reagent through the tail vein, so that all proteins exposed to the luminal surface of blood vessels were biotinylated. Fifteen minutes later, the mouse was perfused with PBS through the heart. I-peptide displaying phage (12) was injected through the heart, allowing phage to bind to biotinylated IPRs. IPR/phage complexes in lung tissue were then solubilized and immunoprecipitated by using anti-phage antibody. Biotinylated IPRs in immunoprecipitates were detected by peroxidase-avidin blot. This experiment revealed 2 bands: a major band at 35 kDa and a minor band at 15 kDa (Fig. 1A).

Fig. 1.

Identification of IPRs expressed on the mouse lung endothelial cell surface. (A) Visualization of IPRs by in vivo biotinylation. Mice were injected intravenously either by PBS (lane 1) or a biotinylation reagent (lanes 2–6), followed by i.v. injection of I-peptide displaying phage (lanes 3 and 4) or control phage (lanes 5 and 6). After perfusion with PBS, lungs were isolated, and phage was immunoprecipitated with rabbit anti-phage antibody (lanes 4 and 6) or rabbit IgG (lanes 3 and 5). Biotinylated proteins were resolved by SDS/PAGE and detected by a peroxidase-avidin blot. Lanes 1 and 2 each contain 1/500 of the lung tissue lysate, and lanes 3–6 each contain immunoprecipitated material from1/10 of the lung tissue lysate. (B) SDS/PAGE of affinity-purified IPR proteins visualized by Coomassie blue staining. Microsomal membrane proteins from rat lung were bound to I-peptide agarose beads in the presence of 1 mM CaCl₂. Bound materials were eluted in 1 mM EDTA, which was applied to a second aliquot of I-peptide agarose beads. After washing the column with buffer containing an irrelevant peptide (1 mg/mL), the bound materials were eluted by I-peptide (1 mg/mL) and subjected to gel electrophoresis. In A and B, 9% gels were used.

To identify these proteins, a membrane protein fraction was prepared from rat lung, and proteins were solubilized by detergent and applied to an I-peptide-conjugated agarose column in the presence of CaCl₂. Because a previous study indicated that I-peptide binding to its receptor is calcium-dependent (13), I-peptide beads binding proteins in the presence of calcium were eluted with EDTA. Partially purified proteins were applied to a second I-peptide-affinity column. The material that bound to the I-peptide column was eluted with irrelevant control peptide and yielded no visible band on an SDS/PAGE. Proteomics analysis of the corresponding 35-kDa region showed no signals indicating the splicing factor (data not shown). IPRs were then eluted from the I-peptide column with buffer containing I-peptide. SDS/PAGE gel electrophoresis stained by Coomassie blue showed 35- and 15-kDa proteins (Fig. 1B), consistent with results obtained by in vivo biotinylation (Fig. 1A).

Proteomics analysis of 35-kDa proteins showed that the band was composed of more than one protein and identified components as serine/arginine-rich alternative splicing factors or Sfrs1, 2, 5, and 7 gene products [supporting information (SI) Fig. S1]. Sfrs protein molecular masses are 25–28 kDa, but Sfrs proteins reportedly migrate at 35 kDa on SDS/PAGE gels (17). Proteomics analysis identified the 15-kDa protein as a fragment of annexin 1.

Expression of Sfrs on the Mouse Lung Endothelial Cell Surface.

Sfrs proteins are RNA-binding proteins and therefore localize to the nucleus and cytoplasm. There are no reports of cell surface expression of Sfrs. However, it is known that endothelial cells often express cytoplasmic and/or nuclear proteins on their surfaces (16, 18, 19). When in vivo biotinylated lung endothelial surface proteins were subjected to immunoprecipitation, these 35-kDa proteins were immunoprecipitated by anti-Sfrs antibodies (Fig. 2A).

Fig. 2.

Cell surface expression of Sfrs proteins on mouse lung endothelial cells. (A) In vivo biotinylated lung endothelial surface proteins and their immunoprecipitated bands detected by anti-Sfrs antibodies. Total cell lysates (lane 1) were immunoprecipitated with rabbit IgG (lane 2), with rabbit anti-phage antibodies (lane 3), with goat IgG (lane 4), with goat anti-Sfrs antibody E16 (lane 5), and with goat anti-Sfrs antibody P-15 (lane 6). A gradient 6–18% gel was used. Biotinylated cell surface proteins were detected by a peroxidase-avidin blot. Lane 1 contains 1/500 of the lung tissue lysate, and lanes 2–6 each contain immunoprecipitated material from 1/10 of the lung tissue lysate. A 15-kDa band visible in Fig. 1A is not detected, likely because this small protein was not retained on the nitrocellulose filter during electrotransfer. (B) Immunohistochemistry of lung tissue sections from a mouse intravenously injected with FITC-conjugated LEA (tomato) lectin and with irrelevant goat antibody (Upper) or with anti-Sfrs antibody (Lower). Frozen lung sections were stained with Alexa Fluor 560-conjugated F(ab′)₂ fragment of anti-goat IgG antibody. (Left) Sfrs (red). (Center) Blood vessels by FITC-LEA (green). (Right) Merged with DAPI nuclear staining (blue). (Scale bar, 200 μm.)

Lung tissue sections prepared from mice injected through the tail vein with anti-Sfrs antibody were analyzed with fluorescence-conjugated secondary antibody. A few numbers of capillaries reacting with the secondary antibody was observed (Fig. 2B Lower). Immunohistochemistry of similar sections from mice injected with irrelevant first antibody showed no staining (Fig. 2B Upper). These results support the hypothesis that Sfrs proteins are expressed on the luminal cell surface of a subset of lung capillaries.

I-Peptide-Binding Activity by Sfrs1.

Recombinant Sfrs proteins were produced as bacterial GST-His fusion proteins. Each protein was spotted on a nitrocellulose filter, and biotinylated I-peptide was overlayed to the filter. Biotinylated I-peptide binding was detected by peroxidase avidin and peroxidase color reaction (Fig. S2A). Because Sfrs1, 2, 5, and 7 are highly similar proteins, only experiments with Sfrs1 are described hereafter.

I-peptide-binding activity to Sfrs1 was evaluated by a plate assay using fluorescence-conjugated I-peptide (Fig. S2B). In this assay, GST-His-Sfrs1 protein was coated in wells of a plastic plate, and the fluorophore Alexa Fluor 488-conjugated I-peptide was added to wells. The result showed that I-peptide bound to the plate in an Sfrs1 dose-dependent manner.

Carbohydrate-Binding Activities of Sfrs.

Because I-peptide is a carbohydrate mimic and inhibited colonization of B16-FTIII-M cells to the mouse lung (12), we tested whether Sfrs1 binds to the cell surface of B16-FTIII-M cells. GST-His-Sfrs1 was conjugated with fluorescent nanoparticle Qdots and added to B16-FTIII-M cells. Binding of Qdot-Sfrs1 to B16-FTIII-M cells (Fig. 3Aa) was inhibited by I-peptide (Fig. 3Aa Inset), but the same Qdot-Sfrs1 did not bind to mock-transfected B16 cells (Fig. 3Ab).

Fig. 3.

Carbohydrate-binding activity of Sfrs proteins. (A) Binding of Qdot-conjugated Sfrs1 to B16-FTIII-M (a) in the presence of 1 mM I-peptide (Inset) and mock-transfected B16 (b) cells. (B) Binding of synthetic biotinylated PAA-conjugated carbohydrates to GST-His-Sfrs1. Binding of each carbohydrate was determined by peroxidase avidin and a peroxidase color reaction. PAA carbohydrates added to GST-His-coated plates showed absorbance of <0.1. LacNAc (1); Type 1H (2); Type 2H (3); Le^a (4); Le^x (5); sLe^a (6); sLe^x (7); LeY (8); LeB (9); biotinylated (10) PAA without carbohydrate. (C) Glycan arrays by GST-His-Sfrs1. Glycan structures listed starting from the strongest binder are 13: α-L-Rhα–Sp8; 76: Fucα1–3GlcNAcβ-Sp8; 58: Fucα1–2Galβ1–3GalNAcα–Sp8; 68: Fucα1–2Galβ1–4(Fucα1–3)GlcNAcβ–Sp8; 67: Fucα1–2Galβ1–4(Fucα1–3)GlcNAcβ–Sp8; 74: Fucα1–2Galβ–Sp8; 66: Fucα1–2Galβ1–4(Fucα1–3)GlcNAcβ1–3Galβ1–4(Fucα1–3)GlcNAcβ1–3Galβ1–4(Fucα1–3)GlcNAcβ-Sp0; 65: Fucα1–2Galβ1–4(Fucα1–3)GlcNAcβ1–3Galβ1–4(Fucα1–3)GlcNAcβ-Sp0; 73: Fucα1–2Galβ1–4Glcβ–Sp0; 72: Fucα1–2Galβ1–4GlcNAcβ–Sp8; 70:. Fucα1–2Galβ1–4GlcNAcβ1–3Galβ1–4GlcNAcβ1–3Galβ1–4GlcNAcβ-Sp0; and 69:. Fucα1–2Galβ1–4GlcNAcβ1–3Galβ1–4GlcNAc–Sp0.

When recombinant Sfrs1 protein was immobilized in wells of a plastic plate and synthetic polyacrylamide (PAA) carbohydrates added to wells, PAA-carbohydrates bound to Sfrs1 (Fig. 3B).

To determine carbohydrate-binding activity and binding specificity of Sfrs1, we used a glycan array analysis, in which structure-defined oligosaccharides presented as an array (20) were overlayed with Sfrs1. Recombinant Sfrs1 protein bound to a series of fucosylated oligosaccharides, including type 1H, type 2H, LeX-, dimeric LeX-, LeY-, and LeB- structures (Fig. 3C). Although LacNAc binds well to Sfrs in a PAA assay (Fig. 3B), LacNAc is not a strong binder in the glycan array assay (bars 152/153 in Fig. 3C). We therefore consider that the LacNAc result is biased by use of the PAA assay. Strong binding to α-rhamnose (6-deoxy-mannose) may be linked to the structural similarity of this carbohydrate to fucose (6-deoxy-galactose).

Inhibition of I-Phage Targeting to Lung Vasculature by Anti-Sfrs Antibodies.

We next asked whether anti-Sfrs antibodies inhibit I-phage targeting to the mouse lung in vivo. Mice were injected intravenously with anti-Sfrs antibodies and then injected with I-peptide-displaying phage (12). Phage-counting assays showed that I-phage targets the lung but not the liver, and lung targeting of I-phage was completely inhibited by anti-Sfrs antibody (Fig. 4A).

Fig. 4.

Inhibition of I-phage binding and B16-FTIII-M cell colonization to lung vasculature by anti-Sfrs antibody. (A) Effect of anti-Sfrs antibody on I-phage targeting to mouse lung. A wild type mouse was intravenously injected with normal goat IgG (pink) or with anti-Sfrs antibody (blue), followed by i.v. injection of I-peptide displaying phage. I-phage targeting lung or liver was determined by counting transformant bacterial colonies. Graph shows results obtained from triplicate analysis. (B) Effect of anti-Sfrs antibody on lung colonization of B16-FTIII-M cells in vivo. Mice were injected with normal goat IgG (1) or with goat anti-Sfrs antibodies (2) through the tail vein, followed by i.v. injection with B16-FTIII-M cells. Two weeks later, numbers of melanoma foci (mean ± SE) formed in the lungs were determined. Numbers of foci in 1 are 307 ± 34, n = 6, and those in 2 are 32 ± 9, n = 6. Differences indicated by asterisks are statistically significant (P < 0.0001). Photo at the left shows representative images of control lung (Upper) and that from anti-Sfrs antibody -treated mice (Lower). Swelling of the lung with many melanoma foci (Upper) is likely caused by the large amount of cancer cells colonized to this organ.

Because our results suggest that anti-Sfrs antibody can block carbohydrate-binding activity on endothelial surfaces in lung, we asked whether anti-Sfrs antibody inhibits lung colonization of B16-FTIII-M cells in vivo. Mice were injected intravenously with anti-Sfrs or irrelevant goat IgG antibodies. Fifteen minutes later, B16-FTIII-M cells were injected intravenously, and 2 weeks later, lungs were inspected for melanoma foci. We found that anti-Sfrs antibody completely blocked lung colonization of B16-FTIII-M cells (Fig. 4B), supporting the hypothesis that Sfrs expressed on the lung endothelial surface is responsible for lung colonization of B16-FTIII-M cells.

Targeted Apoptosis of Lung Endothelial Cells Expressing Sfrs.

The above results suggest that a subset of lung endothelial cells express Sfrs proteins on their surface, providing the site for carbohydrate-dependent cancer cell colonization. We hypothesized that if such sites could be removed from the lung, B16-FTIII-M cells would no longer be able colonize the lung. Because the acidic ganglioside GD3 induces apoptosis (21), we prepared liposomes containing palmitoyl I-peptide and GD3 to induce targeted apoptosis of Sfrs-expressing lung endothelial cells. When liposomes containing irrelevant palmitoyl peptide and GD3 were injected intravenously and a TUNEL assay performed 1 day later, no apoptotic cells in the lung were seen (Fig. 5Aa). By contrast, when mice were injected with liposomes containing palmitoyl I-peptide and GD3, apoptotic nuclei were detected by TUNEL assay (Fig. 5Ab).

Fig. 5.

I-peptide-targeted apoptosis of lung endothelial cells and blocked lung colonization of B16-FTIII-M cells. (A) Mice were injected with liposomes containing irrelevant palmitoyl peptide plus GD3 ganglioside (a) or with palmitoyl I-peptide plus GD3 (b). One day later, lung tissue sections were prepared and subjected to a TUNEL assay. Green fluorescence indicates apoptotic nuclei. (B) Numbers of melanoma foci in mouse lung. Mice (n = 6 for each group) were treated with irrelevant peptide/GD3 (column 1) or with I-peptide/GD3 (column 2) as described above. One day later, B16-FTIII-M cells were injected through the tail vein. Numbers of melanoma foci in lungs 2 weeks after injection were determined. Differences indicated by asterisks are statistically significant (P = 0.0007). Photo at the left shows representative images of control (Upper) and that from I-peptide/GD3-treated mice (Lower).

To determine whether induced apoptosis could remove the sites for the colonization of B16-FTIII-M cells in the lung, mice were injected intravenously with I-peptide/GD3 or control peptide/GD3, and 1 day later, B16-FTIII-M cells were injected intravenously. Two weeks later, melanoma foci formed in these mice were counted. Lungs from mice treated with control peptide/GD3 showed many melanoma foci, whereas few melanoma foci were detected in lungs from I-peptide/GD3-treated mice (Fig. 5B). These results support the hypothesis that Sfrs-expressing lung endothelial cells serve as the sites for carbohydrate-dependent cancer colonization.

Discussion

In this study, we identified presumptive IPRs, whose existence we proposed in a previous study, using B16-FTIII-M cell lung colonization in E- and P-selectin null mice (13). We identified a 35-kDa IPR as Sfrs1, 2, 5, and 7 (Figs. 1 and 2). It remains to be determined whether other Sfrs proteins can bind carbohydrates. Identification of Sfrs proteins as carbohydrate-binding proteins was unexpected, but this finding parallels observations on galectins and annexins. Galectins are well-established carbohydrate-binding proteins (22), but several galectin proteins also localize to the nucleus and exhibit pre-mRNA splicing activity (23–25). The annexins also form a large protein family, and some members localize to the nucleus and bind RNA (25–27). Annexin 4 binds to N-glycan (28), and Annexins 4, 5, and 6 bind sulfated glycosaminoglycans (28, 29). Although Sfrs, annexins, and galectins are not homologous, their activities suggest that carbohydrate-binding capacity overlaps with RNA-binding activity.

Many carbohydrate-binding proteins are C-type lectins that share a calcium-dependent carbohydrate-binding domain structure (30, 31). Sfrs requires calcium for binding to carbohydrates but are not classified as C-type lectins. Sfrs proteins contain an RNA recognition motif and a Ser/Arg-rich domain. Currently, no information is available on the structural basis of Sfrs carbohydrate-binding activity.

We found that Sfrs1 protein binds fucosylated oligosaccharides (Fig. 3 B and C). This suggests that cancer cells expressing highly fucosylated carbohydrates on their surface have a high risk of lung metastasis. The binding specificity of Sfrs provides a rationale for why tumor cells expressing LeY antigen, LeB antigen, or type 1H antigen often metastasize to the lung (6, 11, 32). LeY antigen is considered a primary target of cancer therapy using humanized monoclonal antibodies (33–35) and by cancer vaccine with LeY mimicry peptide (36, 37). Our study supports strategies targeting LeY antigen in breast, lung, ovary, and colon cancers (6, 38, 39).

In addition to fucosylated carbohydrate antigens, sialylated complex carbohydrate structures, such as GD1α, are closely related to metastasis (40). Not only these terminal carbohydrate structures, but internal structures, particularly those forming polylactosamines, are closely related to cancer malignancy (41, 42). To understand mechanisms underlying carbohydrate-dependent cancer malignancy and metastasis, carbohydrate-binding proteins functioning in these processes must be identified. Previous studies suggested functions of cell surface carbohydrates in normal and pathological conditions (reviewed in ref. 43). These studies nevertheless suggest the existence of carbohydrate-binding proteins, which may play a specific role in development and pathogenesis. Carbohydrate mimicry peptides identified from peptide-displaying phage libraries (44–47) may serve as useful reagents for future studies of as yet unknown carbohydrate-binding proteins.

Although we cannot exclude the possibility that another carbohydrate-binding receptor mediates carbohydrate-dependent cancer colonization in the lung, we believe that Sfrs proteins provide the major mechanism for fucosylated carbohydrate-dependent cancer colonization to the lung, because anti-Sfrs antibodies almost completely inhibited lung colonization of B16-FTIII-M cells (Fig. 4). Overall, the present study identified Sfrs proteins as carbohydrate-binding receptors in lung endothelial cells. The carbohydrate-binding activity exhibited by Sfrs provides a basis for fucosylated carbohydrate-dependent hematogenous metastasis of epithelial cancers to the lung.

Materials and Methods

Materials.

Peptides, including CDQENQDPRR used as an irrelevant peptide in this study, were synthesized by GenScript. Goat anti-Sfrs (SF2/ASF) antibodies (P-15 and E-16) were from Santa Cruz Biotechnology. Anti-phage (fd) antibody was from Sigma. Phage clones displaying I-peptide and control phage without displaying peptide have been described (12).

Use of Vertebrate Animals.

Mouse protocols were approved by the Institutional Review Committees at Burnham Institute for Medical Research.

In Vivo Biotinylation.

The method described by Rajotte and Ruoslahti (16) was followed. Briefly, a C57/BL6 mouse was anesthetized with avertin. One hundred microliters of HNS-LC-biotin (1 mg/mL; Pierce) in PBS was injected intravenously through the tail vein. Fifteen minutes later, the animal was perfused with 15 mL of PBS through the heart. I-peptide-displaying phage (2 × 10⁶ pfu in 0.5 mL of PBS) was injected through the heart. Control mice received phage without displaying peptide. Five minutes later, the mouse was perfused with 15 mL of PBS. The mouse was killed by cervical dislocation, the lung isolated, and tissue lysates prepared by homogenization after solubilization with PBS containing 1% Nonidet P-40 and protease inhibitors (Roche). A biotinylated IPR/I-peptide-displaying phage complex was immunoprecipitated by using rabbit anti-phage fd antibody (Sigma) or with rabbit IgG. Immunoprecipitates were subjected to a peroxidase-avidin blot. In vivo biotinylated lung lysates were also subjected to immunoprecipitation with anti-splicing factor antibodies, followed by a peroxidase avidin blot.

Peptide-Affinity Chromatography and Proteomics.

I-peptide-conjugated agarose beads were prepared by conjugating IELLQARC peptide to sulfo-link agarose gel (Pierce), according to the manufacturer's instructions. Twenty-five pairs of rat lungs (Pel-Freeze Biologicals) were homogenized with 20 mM Tris·HCl buffer (pH7.4) containing protease inhibitors (Roche). The membrane fraction (100,000 × g pellet) was prepared by ultracentrifugation, and membrane proteins were dissolved in 50 mL of 20 mM Tris·HCl buffer (pH 7.4) containing 10 mM octyl-thio-glucoside (TBSO), 1 mM CaCl₂, and 2 mL of I-peptide-conjugated agarose beads. After washing beads with TBSO with 1 mM CaCl₂, bound materials were eluted with 2 mL of TBSO containing 1 mM EDTA. The eluate was adjusted with 2 mM CaCl₂, and the sample incubated with a new 0.2 mL of I-peptide-conjugated beads. Bead-bound materials were eluted first by 100 μL of TBSO with 1 mM CaCl₂ containing an irrelevant peptide, followed by 100 μL of TBSO with 1 mM CaCl₂ containing I-peptide (1 mg/mL). Proteins were resolved on SDS/PAGE and stained with Coomassie blue. IPR candidate bands at 35 kDa were cut from the gel, digested with trypsin, and analyzed by capillary column chromatography followed by mass spectrometry at the Harvard Microchemistry Laboratory (Boston, MA) and the proteomics facility at Burnham Institute for Medical Research.

Immunohistochemistry of Lung Vasculature.

A mouse was anesthetized with avertin and injected intravenously through the tail vein with goat anti-Sfrs antibodies (20 μg in 200 μL) together with 15 μg of FITC-conjugated tomato lectin (Vector Laboratories). Fifteen minutes later, the mouse was killed and perfused through the heart with PBS. Lung tissue was isolated and fixed with 4% paraformaldehyde in PBS, and frozen sections were cut. Sections on glass slides were stained with Alexa Fluor 560-conjugated rabbit anti-goat IgG antibody (Fab′)₂ fragment (Invitrogen). The sections were covered by Vectashield with DAPI (Vector Laboratories).

Inhibition of in Vivo Phage Targeting by Antibodies.

A mouse was anesthetized with avertin and injected intravenously with goat anti-Sfrs antibody (E-16, Santa Cruz Biotechnology) or with 100 μL or 20 μg of goat IgG. Fifteen minutes later, I-peptide displaying phage clone (1 × 10⁵ pfu) in 100 μL of PBS was injected intravenously. The mouse was perfused with TBS containing 1 mM CaCl₂, and lung and liver tissues were isolated. Competent K91 bacteria were incubated with the tissue homogenate and plated on LB agar plates containing tetracycline (10 μg/mL) and kanamycin (100 μg/mL). After culturing at 37 °C for 20 h, colonies were counted.

Binding of PAA-Carbohydrates to GST-His-Sfrs.

Full-length cDNAs encoding mouse Sfrs1, Sfrs2, Sfrs 5, and Sfrs7 were obtained from Invitrogen. Each Sfrs cDNA was subcloned into the pET41a bacterial expression system (Novagen) to produce GST-His-Sfrs fusion proteins. Wells of an ELISA 96-well-plate were coated with each recombinant protein, and binding of biotinylated PAA-oligosaccharide was assayed as described (12).

Glycan Array.

Glycan microarrays were printed as described (20). v2.1 (www.functionalglycomics.org/static/consortium/resources/resourcecoreh8.shtml) of the printed microarray from the Consortium for Functional Glycomics was used for analysis. GST-His-Sfrs1 was diluted to 0.2 mg/mL in TBSCM [20 mM Tris, 150 mM NaCl, 2 mM CaCl₂, 2 mM MgCl₂ (pH 7.4)] containing 1% BSA and 0.05% Tween-20. An aliquot (70 μL) was applied to a microarray slide and incubated under a coverslip for 60 min in a humidified chamber at room temperature. Coverslips were then gently removed in a solution of TBSCM/0.05% Tween-20 and subjected to the TBS standard wash procedure. To detect protein binding, the slide was similarly incubated in the dark with FITC-labeled, goat-anti-GST (Abcam) at 5 μg/mL, washed as described above, and washed with deionized water to remove salts. The slide was then spun in a slide centrifuge for ≈15 sec to dry and immediately scanned in a PerkinElmer ProScanArray Scanner using an excitation wavelength of 488 nm and ImaGene software (BioDiscovery, Inc.) to quantify fluorescence. Data are reported as average RFU of 4 of 6 replicates (after removal of highest and lowest values) for each glycan represented on the array.

Binding of Qdot-Conjugated GST-Sfrs1 to B16 Cells.

GST-Sfrs1 (1 mg) described above was bound to fluorescent nanoparticle Qdot (125 μL at 4 μM, Qdot antibody conjugation kit; Quantum Dot) in accordance with the manufacturer's protocol. B16-FTIII-M cells and mock-transfected B16 cells (5 × 10⁴ cells) were grown on glass coverslips placed in a 6-well plate. After 3 washes with PBS, the cells were fixed with 1% paraformaldehyde in PBS. After blocking cells with 3% BSA in PBS, Sfrs1-Qdots at 50 nM dissolved in 20 mM Tris·HCl buffer (pH 7.4) containing 1 mM CaCl₂ were added. Ten minutes later, the cells were washed 3 times with the same buffer and inspected under a fluorescence microscope.

I-Peptide/GD3 Treatment and TUNEL Assay.

GD3 was prepared from buttermilk as described (48). N-terminally modified palmitoyl I-peptide was synthesized at GenScript. Palmitoyl I-peptide and GD3 were mixed in the liposome as described (49). I-peptide-coated liposomes containing 20 μg of equivalent GD3 in 100 μL of PBS were injected through the tail vein. One day later, the mouse was killed and lung tissue fixed in 4% paraformaldehyde in PBS. Frozen tissue sections were subjected to a TUNEL assay using an apoptosis detection kit (ApopTag; Chemicon). Tissue sections were covered with Vectashield containing DAPI (Vector Laboratories).

Statistical Analysis.

Statistical analyses were performed by using SPSS and Microsoft Excel programs. All values in figures and text are expressed as means ± SD of n observations, where n is the number of animals analyzed. Datasets were compared with Student's unpaired t test or Mann–Whitney's U test. A P value ≤0.05 was considered significant.