Recombinant Glycoproteins That Inhibit Complement Activation and Also Bind the Selectin Adhesion Molecules*

Soluble human complement receptor type 1 (sCR1, TP10) has been expressed in Chinese hamster ovary (CHO) DUKX-B11 cells and shown to inhibit the classical and alternative complement pathwaysin vitro and in vivo. A truncated version of sCR1 lacking the long homologous repeat-A domain (LHR-A) containing the C4b binding site has similarly been expressed and designated sCR1[desLHR-A]. sCR1[desLHR-A] was shown to be a selective inhibitor of the alternative complement pathway in vitroand to function in vivo. In this study, sCR1 and sCR1[desLHR-A] were expressed in CHO LEC11 cells with an active α(1,3)-fucosyltransferase, which makes possible the biosynthesis of the sialyl-Lewisx (sLex) tetrasaccharide (NeuNAcα2–3Galβ1–4(Fucα1–3)GlcNAc) during post-translational glycosylation. The resulting glycoproteins, designated sCR1sLex and sCR1[desLHR-A]sLex, respectively, retained the complement regulatory activities of their DUKX B11 counterparts, which lack α(1–3)-fucose. Carbohydrate analysis of purified sCR1sLex and sCR1[desLHR-A]sLex indicated an average incorporation of 10 and 8 mol of sLex/mol of glycoprotein, respectively. sLex is a carbohydrate ligand for the selectin adhesion molecules. sCR1sLex was shown to specifically bind CHO cells expressing cell surface E-selectin. sCR1[desLHR-A]sLex inhibited the binding of the monocytic cell line U937 to human aortic endothelial cells, which had been activated with tumor necrosis factor-α to up-regulate the expression of E-selectin. sCR1sLex inhibited the binding of U937 cells to surface-adsorbed P-selectin-IgG. sCR1sLex and sCR1[desLHR-A]sLex have thus demonstrated both complement regulatory activity and the capacity to bind selectins and to inhibit selectin-mediated cell adhesion in vitro.

has been shown to inhibit the effects of complement activation in a variety of disease models, including ischemia-reperfusion injury (1)(2)(3), burns and acute lung injury (4), hyperacute rejection of transplanted organs (5), autoimmune diseases (6,7), and many others. sCR1[desLHR-A], a truncated version of sCR1 lacking the C4b binding site, was shown to be a selective inhibitor of the alternative complement pathway (8) and was protective in several models of myocardial injury (9 -11).
The selectin adhesion molecules, in particular P-and Eselectin, have been shown to mediate cellular interactions in the required rolling phase, which precedes the adherence and extravasation of leukocytes from the vasculature at sites of inflammation. The selectins, and in particular P-selectin, also mediate the adherence of platelets to lymphocytes (12) as well as the adherence of platelets to neutrophils.
The significance of selectin-mediated processes to various immune and inflammatory responses in vivo has been demonstrated using selectin-deficient animals, anti-selectin antibodies, soluble bivalent selectin-IgG chimeric proteins, and the selectin ligand sLe x tetrasaccharide as well as its analogs. Many of the same animal disease models, which have been shown to be complement-dependent using sCR1 as described above, have also been shown to be selectin-dependent using the sLe x tetrasaccharide. For example, myocardial ischemia-reperfusion injury (13), cobra venom factor-induced rat lung injury (14), and IgG immune complex-induced rat lung injury (15) have all been ameliorated by the use of the sLe x tetrasaccharide.
Sialated, fucosylated oligosaccharides, including the sLe x tetrasaccharide, are carbohydrate ligands for the P-, E-, and L-selectin adhesion molecules (16) and as such represent a potential means to combine anti-selectin activity with the anticomplement activity of a glycoprotein such as sCR1. Furthermore, sCR1 and sCR1[desLHR-A] expressing the sLe x tetrasaccharide should have the potential not only to inhibit complement activation and selectin-mediated cellular interactions, but also to localize these activities at sites of inflammation where activated endothelium has up-regulated the expression of P-and E-selectin. microtiter plate format using two mouse monoclonal antibodies (mAb) specific for CR1. Microtiter plates were coated with antibody (mAb 6B1.H12 (17)), and detection employed a horseradish peroxidase-conjugated antibody (mAb 4D6.1 (17)). Alternatively, concentrations of the purified proteins were determined by absorbance at 280 nm using a molar extinction coefficient of 2.5 ϫ 10 5 liter mol Ϫ1 cm Ϫ1 for sCR1 and sCR1sLe x and of 2.0 ϫ 10 5 liter mol Ϫ1 cm Ϫ1 for sCR1[desLHR-A] and sCR1[desLHR-A]sLe x , obtained from quantitative amino acid analysis (Cornell Biotechnology Program, Ithaca, NY). The human histiocytic lymphoma cell line U937 (18) was obtained from ATCC. The cloned cell line PRO Ϫ LEC11.E7, referred throughout this paper simply as the LEC11 cell line (19), was obtained from Dr. Pamela Stanley of the Albert Einstein College of Medicine.
CHO Cell Lines Expressing Soluble P-selectin-IgG Chimera and Cell Surface E-selectin-The plasmid CDM7 B (20), which codes for a chimeric protein of amino acids 1-292 of human P-selectin fused with the H-CH2-CH3 heavy chain region of human IgG1, was a gift from Dr. Brian Seed of the Massachusetts General Hospital. The chimeric gene was cloned into the eukaryotic expression vector pTCSLneo, which is pTCSgpt (21) modified to carry the neomycin resistance marker, and cotransfected with pTCS-DHFR, a plasmid containing the dihydrofolate reductase gene, into CHO DUKX-B11 cells. Single clones that secreted 3-5 g/ml P-selectin-IgG protein were selected for subsequent experiments using a commercial immunoassay for human IgG1 (The Binding Site, San Diego, CA).
The plasmid containing cDNA for human E-selectin (22) was also provided by Dr. Brian Seed. The E-selectin cDNA fragment was excised with XbaI. Utilization of the single, internal XbaI restriction site at position 2717-2722 resulted in the truncation of 1136 base pairs of the 3Ј-untranslated region, thus deleting several potential mRNA-destabilizing sequence motifs such as ATTTA (six of seven motifs). Cotransfections were performed as described above. E-selectin expression was monitored by flow cytometry, and single clones were amplified in methotrexate.
LEC11 Cells Expressing sCR1sLe x -The expression plasmid coding for sCR1 was prepared as described previously (2). The plasmid pTC-SLdhfr* coding for a mutant mouse dihydrofolate reductase with an abnormally low affinity for methotrexate (23) was derived from pSV2-DHFR* (BamHI-HindIII, ϳ1.5-kilobase fragment) cloned into pTCSLneo by direct substitution of the neomycin gene. LEC11 cells were cotransfected with pTCSLdhfr* and the plasmid coding for sCR1. Clones producing high concentrations of sCR1sLe x were selected in growth medium containing methotrexate.
Western Blot Analysis-Each of the purified glycoproteins (0.5 g) was subjected to nonreducing SDS-polyacrylamide gel electrophoresis followed by transfer to polyvinylidene fluoride membranes (Millipore Corp., Bedford MA) using a semi-dry blotting system. The membranes were blocked with 5% dry milk in Dulbecco's PBS (phosphate-buffered saline without Ca 2ϩ and Mg 2ϩ , Life Technologies, Inc.) and incubated with a mAb specific for CR1 (mAb 3C6.D11 (17)) or specific for sLe x (mAb 2F3, PharMingen or mAb KM93, Kamiya Biomedical Co., Thousand Oaks, CA) at 5 g/ml in 3% bovine serum albumin followed by washing and detection using horseradish peroxidase-conjugated goat anti-mouse IgM antibody (Southern Biotechnology Associates) or goat anti-mouse Ig (Cappel) as appropriate.
Monosaccharide Analysis by HPLC-The neutral and amino sugar composition of the glycoproteins was determined by trifluoroacetic acid hydrolysis, reductive amination with anthranilic acid, separation by C18 reverse phase HPLC, and fluorescence detection (25). Sialic acid content was determined by sodium bisulfate hydrolysis, reaction with o-phenylenediamine, separation by C18 reverse phase HPLC, and fluorescence detection (26).
Oligosaccharide Characterization by Mass Spectrometry-Glycoprotein samples were prepared and analyzed by electrospray ionization mass spectrometry as described previously (27)(28)(29)(30) and as summarized below.
A 25-g aliquot of glycoprotein was reduced, denatured, and digested with N-glycosidase-F (Boehringer Mannheim) for 16 h at 37°C, in the presence of 0.15% SDS (Bio-Rad) and 0.3% Nonidet P-40 (Sigma). The SDS and protein were removed after digestion by elution of the released glycans from a C18 Sep-Pak. The Sep-Pak cartridge (Waters number 20515) was attached to a 5-ml syringe (Hamilton number 1005) and rinsed with 7-8 ml of acetonitrile followed by 7-8 ml of 90% H 2 O, 10% AcCN by use of gentle air pressure. The digest was loaded onto the cartridge, and the glycans were eluted with 4 ml of 90% H 2 O, 10% AcCN. The volume was reduced on a Speed Vac Concentrator (Savant), and the sample transferred to a 1.5-ml glass Reacti-Vial, dried, and vacuum-desiccated prior to permethylation.
For permethylation, desiccated samples were dissolved in 100 l of a NaOH/Me 2 SO suspension, prepared by vortexing Me 2 SO and powdered sodium hydroxide. After 1 h at room temperature, 30 l of methyl iodide was added, and the solution was held for 1 h at room temperature with occasional vortexing. Glycans were partitioned by adding 100 l of chloroform, and the suspension was back-extracted six times with approximately 0.5 ml of H 2 O and the chloroform layer taken to dryness. Permethylation was repeated. All samples were stored at Ϫ20°C.
Periodate oxidation was performed using a 9 mM solution of NaIO 4 buffered with 0.1 M sodium acetate at pH 5.5 in a dark cold room (4°C) for 3 days. The reaction was quenched with 3 l of ethylene glycol and allowed to stand overnight under the same conditions. The sample was neutralized with 0.1 M NaOH and reduced by the direct addition of 5 mg of solid NaB 1 H 4 or NaB 2 H 4 (which remained at room temperature for an additional 16 h). Excess reducing agent was destroyed by the addition of 5 l of acetic acid and the solution dried in a vacuum centrifuge. Borate was removed as its ester by repeated addition and drying with methanol. The sample was vacuum-desiccated overnight prior to methylation.
Electrospray ionization mass spectrometry (ES-MS) was performed on a Finnigan-MAT TSQ-700 (Finnigan-MAT Corp., San Jose, CA) instrument equipped with an electrospray ion source. Methylated glycans were dissolved in water/methanol solutions (40:60 v/v) containing 0.25 mM NaOH and analyzed by syringe pump flow injected at a rate of 0.85 l/min directly into the electrospray chamber through a stainless steel hypodermic needle. The voltage difference between the needle tip and the source electrode was Ϫ3.7 kV. A unique feature of ES-MS is the generation of multiply charged ions from a single molecular species. Unlike fast atom bombardment mass spectroscopy, which has difficulty in obtaining spectra of high molecular weight glycans, multiple charging brings the instrumentally measured mass (m/z) of large glycopeptides and glycans to an easily determined lower value. ES-MS data were analyzed directly or deconvoluted by an algorithm designed to extract molecular weight information from the peak spacing exhibited by parent ion multiply charging. This computational approach was accomplished by pattern recognition using a relative, or level-2, entropy algorithm (28). Treatment of the data in this manner suppresses the artifacts associated with alternative deconvolution algorithms.
Carbohydrate Analysis by Fluorophore-assisted Carbohydrate Electrophoresis (FACE®, Glyko Inc. Novato, CA)-The N-linked oligosaccharides were released intact from the glycoprotein by digestion with the recombinant enzyme peptide N-glycosidase F (Glyko). The glycoprotein was denatured by boiling for 5 min in 0.1% SDS, 50 mM ␤-mercaptoethanol. Nonionic detergent Nonidet P-40 was added to 0.75% followed by incubation with 0.5 milliunit of peptide N-glycosidase F overnight at 37°C. The protein was removed by ethanol precipitation, and the supernatant containing the oligosaccharides was dried prior to labeling. The released oligosaccharides were fluorescently labeled by the addition of 0.15 M disodium 8-aminonapthalene-1,3-disulfonate in 15% acetic acid followed by an equal volume of 1.0 M NaBH 3 CN. The 8-aminonapthalene-1,3-disulfonate-labeled oligosaccharides were then separated by electrophoresis in a 21% polyacrylamide gel and imaged and analyzed using the FACE® Imager and software.
Assays of Complement Regulatory Activity-The inhibition of complement-mediated lysis of antibody-sensitized sheep erythrocytes (classical pathway) or of guinea pig erythrocytes in EGTA-Mg 2ϩ buffer (alternative pathway) was assessed as described previously (8).
Radioligand Binding of sCR1sLe x to CHO Cells Expressing E-selec- were harvested using PBS with 1 mM EDTA and washed with PBS containing 1% bovine serum albumin (BSA), 1 mM CaCl 2 . E-selectin expression was monitored with a fluorescein-labeled anti-ELAM1 mAb (Southern Biotechnology, Birmingham, AL). sCR1sLe x and sCR1 were radiolabeled with 125 I using the Bolton-Hunter reagent (NEN Life Science Products) to yield a specific activity of 1-2 ϫ 10 5 cpm/g of labeled protein. Mixtures of labeled and unlabeled protein (1:5) over a concentration range of 0.2-3 M were incubated for 3 h at room temperature with 5 ϫ 10 6 CHO cells expressing E-selectin in a sample volume of 0.25 ml. Bound and free ligand were separated by centrifuging the samples through a 0.2-ml layer of dibutyl/ dioctyl phthalate (7/4 volume ratio) and quantitated using a ␥-counter. Samples were run in triplicate. Nonspecific binding was determined using sCR1, and the apparent dissociation constant (K d (app)) for specific binding of sCR1sLe x was estimated as the concentration yielding half-maximal binding.
Binding of sCR1sLe x to CHO Cells Expressing E-selectin by Flow Cytometry-CHO cells were removed from flasks using trypsin and washed. The buffer used for flow cytometry was 0.2% BSA, 0.02% NaN 3 in PBS, which included 0.9 mM Ca 2ϩ and 0.5 mM Mg 2ϩ . Fluorescein isothiocyanate (FITC)-labeled glycoproteins were incubated at 500 g/ml for 60 min at 4°C with CHO/E-selectin cells (3 ϫ 10 6 cells/ml), washed, and resuspended for analysis on a Becton Dickinson FACScan flow cytometer. Negative controls included untransfected CHO cells and sCR1-FITC. Expression of E-selectin was confirmed using a mAbspecific for human E-selectin (Serotec) and an isotype-matched irrelevant control mAb, which were incubated with the transfected CHO cells at 2 g/ml for 30 min at 4°C followed by a FITC-labeled goat antimouse Ig antibody (Becton Dickinson) for an additional 30 min.
Inhibition of U937 Cell Adhesion to Activated Endothelial Cells-Human aortic endothelial cells (Clonetics), confluent in 96-well microtiter plates, were stimulated with tumor necrosis factor-␣ (100 units/ml) and phorbol 12-myristate 13-acetate (20 ng/ml) for 4 h at 37°C to up-regulate E-selectin. The cells were then washed twice with Dulbecco's modified Eagle's medium containing 1% fetal bovine serum. Serial dilutions of sCR1[desLHR-A]sLe x or sCR1[desLHR-A] were added to achieve final concentrations up to 2.7 M. To each well, 5 ϫ 10 5 U937 cells in Dulbecco's modified Eagle's medium were added and incubated for 20 min at 37°C. The wells were filled with media, sealed, and centrifuged inverted at low speed for 5 min. The seal was removed, the plates blotted, and the number of bound cells in three microscope fields was determined and reported as a function of glycoprotein concentration with S.E. values.
Inhibition of U937 Cell Adhesion to Immobilized P-selectin-IgG Chimera-Flat bottom, black 96-well microtiter plates were coated with 2 g/ml goat anti-human IgG␥, incubated with P-selectin-IgG-containing supernatants, and blocked using 2.5% BSA in PBS. U937 cells at 2 ϫ 10 6 cells/ml in PBS were incubated 15 min at 37°C with 2 g/ml calcein-acetoxymethyl ester and washed three times with Dulbecco's PBS with 0.9 mM Ca 2ϩ and 0.5 mM Mg 2ϩ . Varying concentrations of sCR1 or sCR1sLe x were added to the P-selectin-IgG-coated microtiter plates followed by the fluorescently labeled U937 cells at 2 ϫ 10 5 cells/100 l in each well. The plates were then centrifuged very slowly (Ͻ600 rpm) for 5 min followed by a 15-30 min incubation at room temperature. The wells were completely filled with PBS with Ca 2ϩ and Mg 2ϩ , sealed, inverted, and centrifuged at low speed as before. The plate sealer was removed to empty the wells. Adherent cells remaining behind in the microtiter wells were lysed with 100 l of TRAx® lysis buffer (Avant) and the fluorescence was measured ( ex ϭ 485 nm, em ϭ 530 nm; Cytofluor II, Perseptive Biosystems) and reported as means and S.E. The control mAb CSLEX1 which is specific for sLe x was obtained as a hybridoma line (ATCC) and used as ascites fluid.

RESULTS
Western Blot Analysis-The presence of sLe x on the purified glycoproteins derived from LEC11 cell expression is clearly evident by staining with two different anti-sLe x mAb (Fig. 1,  lanes 2, 4, 10, 12). The glycoproteins derived from both the LEC11 and DUKX B11 cells show comparable reactivity using anti-sCR1 mAb (lanes 5-8).
Monosaccharide Analysis- Table I summarizes the monosaccharide content for typical lots of each glycoprotein. As might be expected, the most striking difference between the LEC11-and DUKX B11-derived proteins was found in the fucose content, which was more than doubled for the LEC11derived proteins, most likely due to ␣(1,3)-fucosyltransferase activity, which is a primary distinction between these cell lines. In general, the LEC11 versions of the proteins also had higher sialic acid content. This higher sialic acid content may reflect the fact that, in addition to sCR1 expression, the LEC11 clones were selected for high sLe x expression and hence high sialic acid. Both versions of sCR1 tended to have higher glucosamine, galactose, and mannose than did the versions of sCR1[desLHR-A], reflecting a greater number of complex oligosaccharide chains on the larger protein.
To the left of the blot are the apparent molecular masses from markers run and blotted in the left-most unlabeled lanes of each blot.
Significant quantities of high mannose oligosaccharides were not observed. No evidence of unusual oligosaccharide types or sulfated forms was observed. While measurable differences were obtained between separate lots of the same protein (data not shown) or between each of the four different proteins (Table  II), certain general trends were clearly evident. For example, the predominant species of complex N-linked oligosaccharide was typically biantennary (70 -80%), followed by triantennary (12-25%) and tetraantennary (1-10%). Approximately 97% of oligosaccharides had core ␣(1-6)-fucose.
If it is assumed that the sialic acid and ␣(1-3)-fucose distribute randomly between branches, the mole percent of sLe x species can be estimated from these data. To test the assumption of a random sialic acid and fucose distribution, the oligosaccharides from sCR1[desLHR-A]sLe x were treated with a fucosi-dase that removes ␣(1-3)-fucose only from branches that lack terminal sialic acid. ES-MS analysis of the biantennary glycans with a single sialic acid (BiNA 1 , BiNA 1 F 1 , and BiNA 1 F 2 ) after such fucosidase treatment yielded quantities for each form that were within 10% of the expected values, based on the quantities prior to treatment and an assumption of random distribution (data not shown). Using the assumption of random sialic acid and ␣(1-3)-fucose distribution for the sample lots reported here, both sCR1sLe x and sCR1[desLHR-A]sLe x yielded an sLe x mole percent of 60% (Table II).
Carbohydrate Analysis by Fluorescence-assisted Carbohydrate Electrophoresis-Typical profiles of the intact oligosaccharide chains for sCR1[desLHR-A] (Fig. 2, lane 1) and sCR1 (lane 3) were dominated by three major bands corresponding to biantennary structures, which have zero (BiNA 0 ), one (BiNA 1 ), or two (BiNA 2 ) terminal sialic acid moieties. This was consistent with ES-MS results described above in which biantennary structures were the predominant species (Table II). The sCR1[desLHR-A]sLe x (Fig. 2, lane 2) and sCR1sLe x (lane 4) FACE® profiles included these same bands at lower relative intensities, but had additional bands from variously fucosylated forms of these structures. The lower intensities of the nonfucosylated bands for the LEC11 glycans occur because the predominant biantennary species have been apportioned among various fucosylated versions having zero, one, or two branched ␣(1-3)-fucoses. Thus the FACE® oligosaccharide profiles clearly demonstrated the striking differences in oligosaccharide structures obtained from glycoproteins expressed by DUKX-B11 cells (lanes 1 and 3) as compared with those derived from LEC11 cells (lanes 2 and 4). These differences can be attributed to ␣(1,3)-fucosyltransferase activity.
In addition, the individual bands from the oligosaccharide profile were cut out and subjected to further digestion and separation in order to sequence and quantitate the various species. By such methods, the mole percent of each oligosaccharide species, as well as the mole percent of sLe x per oligosaccharide chain, were determined and shown to be in good agreement with data obtained by ES-MS for these glycoproteins (31).
The sialic acid linkage was determined by digestion with the neuraminidases NANase I (␣2-3-specific) and NANase III (␣2-3,6,8-specific). The banding patterns of both digestions were identical (31), and the changes in mobility were consistent with the loss of NeuNAc, indicating that the NeuNAc linkage is ␣2-3.
Fucose linkage was also determined using a combination of fucosidase and galactosidase digestions. After neuraminidase treatment, fucose was removed using FUCase III, indicating the linkage to be either ␣1-3 or ␣1-4 to GlcNAc. After removing Fuc, galactosidases were used to determine the position of the remaining Gal and, by inference, the linkage of Fuc (31). Digestion with GALase III (Gal␤1-4-specific) changed the band migration pattern, indicating that Gal is linked ␤1-4, consistent with the presence of Fuc␣1-3. Digestion with ␣GALase (Gal␣1-3-specific) did not change the banding pattern, indicating that Gal is not ␣1-3-linked, also consistent with the presence of Fuc␣1-3 (31).
Inhibition of Complement Activation in Vitro-Having established the presence of the sLe x tetrasaccharide on the LEC11 glycoproteins, it was important to examine the effects of such glycosylation on complement inhibitory function. In general, the concentrations required to inhibit human complement-mediated lysis of erythrocytes were similar for glycoproteins expressed by LEC11 cells (sLe x versions, open squares and circles,  Fig. 3) compared with those expressed by DUKX-B11 cells (non-sLe x version, closed squares and circles). A representative  Table II are as follows. Bi, Tri, and Tetra refer to biantennary, triantennary, and tetraantennary, respectively. NA 0 , NA 1 , NA 2 , NA 3 , and NA 4 refer to no, a single, two, three, and four terminal N-acetylneuraminic acid residues, respectively. F 1 and F 2 refer to one and two branched ␣(1-3)-fucose residues, respectively. Core ␣(1-6)-fucosylation was 97% and is assumed in the structures but not indicated in the abbreviation. BiNA 0 -(Gal) and BiNA 1 -(Gal) refer to biantennary complex oligosaccharides with no or a single neuraminic acid lacking a terminal galactose. titration of each of the four glycoproteins is depicted in Fig. 3 in both an assay of classical pathway mediated lysis (Fig. 3, A and C) as well as in an assay of alternative pathway mediated lysis (Fig. 3, B and D). As expected, both versions of sCR1 (squares) and sCR1[desLHR-A] (circles) were similar in their capacity to inhibit alternative pathway mediated lysis, but the sCR1 versions were much more effective than the sCR1[desLHR-A] versions in the inhibition of classical pathway lysis.
A single such titration can be characterized by the concentration of glycoprotein required for 50% inhibition of maximal lysis (IC 50 ) so that a lower IC 50 value would be indicative of a more effective inhibitor in these assays. From an examination of the means and S.D. values of the IC 50 values from numerous such titrations, the mean concentration of sCR1sLe x required to yield half-maximal lysis (IC 50 ϭ 0.27 Ϯ 0.082 nM, n ϭ 31) of sensitized sheep erythrocytes was somewhat higher than required for sCR1 (IC 50 ϭ 0.21 Ϯ 0.060 nM, n ϭ 65). The difference in mean IC 50 values between sCR1sLe x and sCR1 is statistically significant (p ϭ 10 Ϫ5 by analysis of variance), indicating a possible effect of differing glycosylation in these assays. The actual difference, however, is smaller than the differences observed between lots of either glycoprotein or even between multiple assays of a single lot. The most important aspect of these results is that both sCR1sLe x and sCR1 are potent inhibitors of classical complement activation as indicated by the subnanomolar IC 50 values in these in vitro assays. Similarly, in the assay of alternative pathway lysis of guinea pig erythrocytes, sCR1sLe x (IC 50 ϭ 38 Ϯ 16 nM, n ϭ 37) appeared somewhat less effective than sCR1 (IC 50 ϭ 19 Ϯ 6.6 nM, n ϭ 10), which again was statistically significant (p ϭ 0.0012 by analysis of variance). The relatively higher IC 50 values obtained in the alternative versus the classical pathway assays reflect the higher concentrations of serum complement used in the alternative pathway assay. Nevertheless, both sCR1sLe x and sCR1 were shown to be potent inhibitors of alternative pathway activation as reflected in IC 50 values of approximately 10 Ϫ8 M.
A general observation was that in all of these assays, the sLe x versions were somewhat less effective inhibitors of complement mediated cell lysis than were the non-sLe x versions, but that these differences in mean IC 50 values were in all cases modest (less than 60%). The greatest difference was observed in the assays of classical pathway mediated lysis for the sLe x and non-sLe x versions of sCR1[desLHR-A], a selective inhibitor of alternative complement activation. These assays have large inherent variability, especially when comparing different lots of serum complement, different lots of target erythrocytes, or different lots of complement inhibitory glycoproteins. Most important, despite measurable differences in IC 50 values, both sCR1sLe x and sCR1 are potent inhibitors of classical and alternative complement activation, and both sCR1[desLHR-A]sLe x and sCR1[desLHR-A] are potent inhibitors of alternative complement activation.
Radioligand Binding of sCR1sLe x to CHO Cells Expressing E-selectin-Having established the presence of sLe x in the Nlinked complex oligosaccharide of sCR1sLe x , it was important to demonstrate its capacity to function as a selectin ligand. The binding of 125 I-sCR1sLe x and 125 I-sCR1 to CHO/E-selectin cells as a function of glycoprotein concentration is shown in Fig. 4. Specific sCR1sLe x binding was obtained by subtracting sCR1 binding, which was assumed to be nonspecific, from the observed sCR1sLe x binding. A K d (app) for the binding of sCR1sLe x was estimated from the concentration yielding halfmaximal specific binding and found to be 1.4 M.
Binding of sCR1sLe x to CHO Cells Expressing E-selectin by Flow Cytometry-Because the low affinities of sCR1sLe x were near the limits of typical radioligand binding methods, a second method was used to confirm the binding of sCR1sLe x to cell surface E-selectin on transfected CHO cells. Nearly identical staining by sCR1sLe x -FITC and sCR1-FITC indicated no binding of either labeled protein to the control CHO cells (Fig. 5A). Similarly, anti-E-selectin mAb-FITC and an irrelevant control mAb-FITC showed no evidence of binding to the control CHO cells (not shown). As expected, the anti-E-selectin mAb-FITC stained the CHO/E-selectin cells compared with the control mAb-FITC (not shown), confirming the presence of cell surface E-selectin. Significantly greater staining of CHO/E-selectin cells was observed for sCR1sLe x -FITC as compared with sCR1-FITC (Fig. 5B), demonstrating specific binding of sCR1sLe x -FITC to cell surface E-selectin. The addition of 1.0 mM EDTA essentially eliminated the selective staining of CHO/E-selectin by sCR1sLe x -FITC (data not shown), which is consistent with Ca 2ϩ -dependent binding as expected for the selectins. In similar experiments, sCR1[desLHR-A]sLe x -FITC, but not sCR1 [desLHR-A]-FITC, was also shown to bind CHO/E-selectin. In

Inhibition of U937 Cell Adhesion to Activated Endothelial
Cells-In addition to binding cell surface E-selectin, the blocking of E-selectin-mediated cell-cell interactions was examined in static cell binding assays. The capacity of sCR1[desLHR-A]sLe x to inhibit the binding of the monocyte-like human lymphoma line U937 to activated human aortic endothelial cells was examined in static binding assays. Activation of endothelial cells leads to the up-regulation of E-selectin after 4 h, which was confirmed by flow cytometry using anti-E-selectin mAb-FITC. As shown in Fig. 6, sCR1[desLHR-A]sLe x , but not sCR1[desLHR-A], inhibited the binding of U937 cells in a concentration-dependent manner (IC 50 ϭ 0.7 M).
Inhibition of U937 Cell Adhesion to Immobilized P-selectin-IgG Chimera-The capacity of sCR1sLe x to inhibit P-selectinmediated cell adhesion was examined in static binding assays. As shown in Fig. 7, the increased fluorescence obtained upon lysing the adherent U937 cells, which had been incubated with PBS buffer or with 12 M sCR1, reflects binding to the immobilized P-selectin-IgG. As expected, this U937 binding was inhibited by removing Ca 2ϩ using 20 mM EDTA or by blocking with the sLe x -specific mAb CSLEX1. Similar inhibition of U937 binding to immobilized P-selectin-IgG was obtained using 12 M sCR1sLe x (Fig. 7). Inhibition of U937 binding was dependent on sCR1sLe x concentration with an IC 50 ϭ 3.1 M (not shown). DISCUSSION We have expressed the recombinant human protein sCR1, and its deletion mutant sCR1[desLHR-A], in the LEC11 cell line in order to incorporate sLe x glycosylation in the resulting mature glycoproteins. Cotransfection with a plasmid coding for a mutant mouse dihydrofolate reductase enabled gene amplification to increase protein expression using methotrexate. This approach is a practical method enabling the large scale production of recombinant glycoproteins bearing a desired, specific carbohydrate moiety, namely, the sLe x tetrasaccharide. Alternative approaches, such as chemical (32) or enzymatic (33) glycosylation of purified proteins, may not be as suitable for the preparation of biological drugs. The general applicability of the approach has been exemplified using two different, although related, glycoproteins. sCR1 appears especially well suited for the incorporation of multiple sLe x moieties because of its elongated, flexible structure bearing 25 potential N-linked glycosylation sites, which should facilitate multivalent binding to cell surface selectins. This elongated flexible structure is characteristic of proteins comprised of short consensus repeat domains such as CR1, CR2, membrane cofactor protein (MCP, CD46), decay accelerating factor (DAF, CD55), C4-binding protein, and even the selectins themselves.
Because only N-linked, complex-type oligosaccharides were observed in the mass spectroscopy analysis, an estimate of the number of oligosaccharide chains per glycoprotein can be made from the mannose content, assuming a trimannosyl core structure for each chain. Such a calculation yields 12.1 and 13.7 oligosaccharide chains on sCR1[desLHR-A] and sCR1[desLHR-A]sLe x , respectively, and 20.3 and 16.2 chains on sCR1 and sCR1sLe x . Combined with the mole percent sLe x (Table II), the average incorporation is 8 and 10 moles of sLe x per sCR1[desLHR-A]sLe x and sCR1sLe x , respectively. sCR1[desLHR-A]sLe x inhibited the E-selectin-dependent binding of U937 cells to tumor necrosis factor-␣-activated human aortic endothelial cells in a concentration-dependent manner with an IC 50 of 1.3 M. These results are similar to the IC 50 values of 1 M reported for the inhibition of HL60 cell binding to immobilized E-selectin by SLeX 16 BSA and by SLeX 11 BSA, bovine serum albumin conjugates with 16 and 11 mol of sLe x tetrasaccharide/mol of protein, respectively (32). It should be noted, however, that the sLe x on these serum albumin conjugates was not part of a natural oligosaccharide. The IC 50 values obtained for blocking E-selectin mediated cell adhesion reflect dissociation constants for multivalent sLe x binding to E-selectin, which was directly confirmed by radioligand binding studies of sCR1sLe x (K d (app) ϭ 1.4 M). sCR1sLe x also inhibited the binding of U937 cells to surface-adsorbed P-selectin-IgG chimera in a concentration-dependent manner (IC 50 ϭ 3.1 M).
In summary, sCR1sLe x and sCR1[desLHR-A]sLe x have been expressed in LEC11 cells and purified from culture supernatants. Carbohydrate analysis presented here and elsewhere (31) confirmed the presence of the desired sLe x tetrasaccharide. These sLe x -containing glycoproteins retain potent complement inhibitory activity that is comparable with that of the non-sLe x versions expressed by DUKX-B11 cells. The sLe x -containing glycoproteins functioned as selectin ligands in vitro by binding to cells expressing surface E-selectin and by blocking selectinmediated cell adhesion in a static system. The in vivo targeting and anti-inflammatory activities of these glycoproteins are currently being examined in animal models of lung injury, stroke, and myocardial ischemia-reperfusion injury.
Note Added in Proof-A manuscript comparing the four glycoproteins described here in a rat lung injury model by Mulligan, M. S., Warner, R. L., Rittershaus, C. W., Thomas, L. J., Ryan, U. S., Foreman, K. E., Crouch, L. D., Till, G. O., and Ward, P. A. and entitled "Endo- thelial Targeting and Enhanced Anti-Inflammatory Effects of Complement Inhibitors Possessing sLe x Moieties" has been accepted for publication in The Journal of Immunology.