The Soluble Form of E-selectin Is an Asymmetric Monomer EXPRESSION, PURIFICATION, AND CHARACTERIZATION OF THE RECOMBINANT PROTEIN*

The gene coding for a soluble form of human E-selec- tin (sE-selectin) has been expressed in Chinese hamster ovary (CHO) cells. Cells seeded into a hollow fiber reac- tor secreted protein at a level of 160 mgfliter. The protein was purified to >95% pure and low endotoxin (e2 ng/mg), using physiological pH and buffers. The amino acid composition and N-terminal sequence were as predicted from the cDNA sequence. HL-60 cells bound to sE-selec-tin-coated plates in a dose-dependent manner, and this binding could be blocked up to 100% by pretreatment of HMO cells with sE-selectin. The concentration of sE-se- lectin required for 50% inhibition was 1 VI. This value puts an upper limit for the affinity of E-selectin for its natural receptor. sE-selectin also inhibited inflammatory migration of neutrophils in a selective fashion. Purified sE-selectin exhibited a broad band of M,

The gene coding for a soluble form of human E-selectin (sE-selectin) has been expressed in Chinese hamster ovary (CHO) cells. Cells seeded into a hollow fiber reactor secreted protein at a level of 160 mgfliter. The protein was purified to >95% pure and low endotoxin (e2 ng/mg), using physiological pH and buffers. The amino acid composition and N-terminal sequence were as predicted from the cDNA sequence. HL-60 cells bound to sE-selectin-coated plates in a dose-dependent manner, and this binding could be blocked up to 100% by pretreatment of HMO cells with sE-selectin. The concentration of sE-selectin required for 50% inhibition was 1 VI. This value puts an upper limit for the affinity of E-selectin for its natural receptor. sE-selectin also inhibited inflammatory migration of neutrophils in a selective fashion. Purified sE-selectin exhibited a broad band of M, -75,000 on nonreducing SDS-PAGE. sE-selectin eluted with M, -310,000 from size exclusion chromatography at physiological pH and buffers, suggesting an oligomeric state. Matrix-assisted laser-desorption MS gave a molecular weight of 80,000, while the minimum monomer molecular weight from the gene sequence should be 58,571, demonstrating that the monomeric molecule thus expressed had 27% carbohydrate. Equilibrium analytical ultracentrifugation gave an average solution molecular weight of 81,600 (* 4,500). Velocity ultracentrifugation gave a sedimentation coefficient of 4.3 S and, from this, an apparent axial ratio of 10.5:1, assuming a prolate ellipsoid of revolution. An analysis of the NMR NOESY spectra of sE-selectin, sialyl-Lewis X, and sE-selectin with sialyl-Lewis X demonstrates that the recombinant protein binds sialyl-Lewis X productively. Hence, in solution, sE-selectin is a functional elongated monomer.
The specific recruitment of leukocytes from the blood through endothelial cells in postcapillary venules is the first step in the pathophysiology of a wide range of acute and chronic inflammatory conditions, as well as wound healivg and tissue repair. At sites of inflammation, leukocytes, which have been shown to roll along the surface of the venules, become attached to the endothelium, flatten, and pass between adjacent cells into the subendothelium in a process called diapedesis (Ather-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "uduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 8 To whom correspondence and reprint requests should be addressed: Dept. of Macromolecular Sciences (UE-0447B1, SmithKline Beecham Pharmaceuticals, 709 Swedeland Rd., King of Prussia, PA 19406-0939. Tel.: 215-270-7887; E-mail: HENSLEYCP%PHVAX.DNET@SB.COM. ton and Born, 1972). This weak attachment is stimulated by cytokines, such as tumor necrosis factor-a and interleukin-1, as well as by lipopolysaccharide, which induce the expression of cell adhesion molecules (Wellicome et al., 1990).
Selectins are a class of cell adhesion molecules which mediate this rolling behavior via low-avidity transient interactions between leukocytes and endothelial cells (Lawrence and Springer, 1991;von Andrian et al., 1991). To date, three classes of selectins, E-selectins, P-selectins, and L-selectins have been identified. They are distinguished by the cell type with which they are primarily associated. E-selectin is found on endothelial cells, P-selectin is found on platelets, and L-selectin is found on leukocytes (Bevilacqua et al., 1991). All mature selectins share similar structural motifs. These consist of an -120amino acid calcium-dependent mammalian lectin domain at the N termini, followed by an epidermal motif, and are followed by a variable number of short consensus repeats found in complement regulatory proteins (six for E-selectin), a transmembrane region, and a C-terminal cytoplasmic domain. The selectins are predicted to be glycosylated, and the sites of Nlinked glycosylation vary among the three classes (Bevilacqua et al., 1989;Johnston et al., 1989;Siegelman et al., 1989;Hession et al., 1990). E-selectin is expressed on the endothelial cells within 4-6 h following induction by cytokines and helps to mediate the initial rolling and adhering of resting neutrophils (Bevilacqua et al., 1987,19891, eosinophils (Kyan-Aung et al., 1991, basophils (Bochner et al., 19911, a subpopulation of T-lymphocytes (Graber et al., 1990;Picker et al., 1991;Shimizu et al., 1991), and monocytes (Carlos et al., 1991). The ligand for E-selectin on leukocytes has been determined to be a blood group antigen, sialyl-Lewis X (sLe")' (Goelz et al., 1990;Lowe et al., 1990;Phillips et al., 1990;Walz et al., 1990) and is found on many leukocyte cell surface molecules. Because of the central role that E-selectin plays in the initial stages of inflammation, a significant effort has been undertaken to identify analogs of sLe" that will antagonize this transient interaction.

23949
Soluble E-selectin Is a n Elongated Monomer stroke, acute respiratory distress syndrome, and potentially other inflammatory conditions. As a first step in this drug discovery effort, a soluble form of human E-selectin has been produced. Such a molecule will be a valuable tool in establishing screens to identify sLe" antagonists, and, once they are identified, it will play a role in highresolution structure-function studies. However, an important question in the study of selectin-cell surface receptor or selectin-ligand interactions, even with purified macromolecules, is the question of valency. It has been suggested that P-selectin (GMP 140) is a tetramer on the basis of its behavior in gel permeation chromatography experiments (Skinner et al., 1991). sE-selectin has also been proposed to be multivalent (Lobb et al., 1991) based on gel permeation chromatography experiments as chemical cross-linking studies. This potentially complicates the interpretation of ligand binding data. However, it has been recently reported that soluble P-selectin is an elongated rod-like monomer (Ushiyama et al., 1993) as determined by hydrodynamic and thermodynamic methods as well as electron microscopy. To establish the valency of sE-selectin and to facilitate purification, a recombinant secreted form of E-selectin lacking both transmembrane and cytoplasmic regions has been expressed in Chinese hamster ovary (CHO) cells. The present study describes the expression, large scale purification, biological, biochemical, and biophysical characterization of a recombinant soluble form of human E-selectin.

MATERIALS AND METHODS
cDNA Cloning and Mutagenesis-The cDNA for full-length E-selectin, was obtained from British Biotechnology Limited (Oxford, UK) in the pCDM8 expression vector (here denoted pCMD8ELAM). The gene was excised on a XbaI fragment and cloned into pSelectTMl (Promega).
A stop codon and unique EcoRV site were introduced by site-directed mutagenesis (Altered Sites System, Promega) into the boundary between the sixth consensus repeat and the transmembrane domain, using the following oligonucleotide, which starts at nucleotide 1776: 5"CC The sE-selectin cDNA was inserted into the mammalian expression vector, RLDN? as a KpnIIEcoRV fragment to yield the vector rsEL-AMdn (Fig. 1, A and B ) . In addition to 80 base pairs of 5"untranslated E-selectin DNA, this fragment also contains 55 base pairs of DNA from the p-Selectml vector.
COS Expression-COS-1 cells were transfected with 10 pg of DNA from either rsELAMdn or pCMD8ELAM as described previously (Caltabiano et al., 1989) with the following modification. After the cells were shocked with a 10% dimethyl sulfoxide in PBS solution (20 n" sodium phosphate, 150 m M NaCl, pH 7.4) and washed, the cells were refed with MR1 MOD3 medium (an in-house proprietary serum-free medium, here called maintenance medium) containing nucleosides and then incubated for 72 h at 37 "C in 5% CO,, 95% humidified air.
CHO Expression-The CHO cell line, DG-44 (Urlaub et al., 1983), was adapted for suspension growth in maintenance medium. The cells were grown at 37 "C in a 5% CO,, 95% air-humidified incubator and were passaged at 4 x lo5 celldml per 100-ml spinner flask twice weekly.
The rsELAMdn expression vector was linearized by digestion withNotI and electroporated, using a Bio-Rad Gene Pulser (Bio-Rad Laboratories), into DG-44 cells as described (Trill et al., 1990) with several modifications, i.e. 1 x lo7 cells were suspended into 0.8 ml of ice cold PBSucrose (272 m M sucrose, 7 m M sodium phosphate, pH 7.4, 1 m M MgCl,). The Gene Pulser was set at 380 V and 25 microfarads. After a 10-min incubation on ice, the cells were plated into 96-well plates at 6 x lo3 celldwell in maintenance medium for 48 h prior to selection. Cells were selected for resistance to 400 pg/ml G418 (Geneticinm, Life Technologies, Inc.) in maintenance medium. 24 h prior to assay, the cells were fed with the selective medium. Conditioned medium from individual colonies was assayed by standard ELISA assay (see below).
Individual colonies stably expressing sE-selectin were selected for resistance to methotrexate (MTX, see J. Trill, manuscript in preparation.
individual colony was scaled into a 100-ml spinner at 4 x IO5 cells/ml in medium containing 100 n~ MTX and passaged twice weekly to determine the growth and productivity parameters of this cell line.
To generate conditioned medium for purification, CHO cells were seeded at a density of 3 x lo8 cells into the extracapillary space of a BR1970 (M, = 70,000 cutoff, 19 f t ' ) Hollow Fiber Bioreactor in a Cell Pharm Cell Culture System I1 (Unisyn Fibertec, San Diego, CA). Four days postinoculation, 50-200 ml of medium was collected daily from the extracapillary side of the reactor. Cells were removed by centrifugation, and the medium was pooled and frozen at -70 "C. Protein Concentration-Protein concentrations in crude extracts were determined by the bicinchoninic acid method (Pierce Chemical Co.) using bovine serum albumin as a reference. Protein samples (10 p1) were mixed with the 250 pl of BCA reagent in microtiter wells and incubated for 20 min at 45 "C. Absorbance at 562 nm was measured in a Molecular Devices vMax microplate reader. Concentrations of purified sE-selectin were determined by absorbance at 280 nm = 102,000 M -~ cm"). Relative amounts of sE-selectin present in CHO media and at each stage of purification were determined by densitometric scanning of silver-stained SDS-PAGE gels of various amounts of samples compared to various amounts of purified protein, in an LKB Model 2222 Scanning Densitometer. The values were compared with the results of ELISA.
Quantitation of sE-selectin by ELISA-An ELISA was established to identify COS and CHO cells expressing sE-selectin and to quantitate sE-selectin during purification. Immunolon-2 plates were coated overnight at 4 "C with 50 pVwell goat anti-mouse IgG (Boehringer Mannheim) diluted to 5 pg/ml in 0.1 M NaHCO,, pH 9.1. Wells were aspirated and blocked with 3% bovine serum albumin in PBST (10 nm sodium phosphate, 150 m M NaC1, and 0.05% Triton X-100, pH 7.5) for 1 h at room temperature on a microplate shaker. Plates were washed three times with 200 pVwell PBST between each 1-h incubation which were 50 pl of 500 ng/ml anti-E-selectin monoclonal 1.2B6, 100 pl of culture supernatant, diluted sE-selectin column fractions or purified sE-selectin standard, 50 pl of a 1:500 dilution of rabbit polyclonal antibody raised against sE-selectin, 100 pl of horseradish peroxidase-conjugated goat anti-rabbit IgG (Zymed Laboratories), and, finally, plates were developed for 20 min with 100 pVwell ABTS substrate (2,2-azino-di-3ethylbenzthiazoline sulfonate, Kirkegaard and Perry Laboratories, Inc.) and read at 410 nM.
Purification-Frozen media from a hollow fiber reactor (420 ml) were thawed and clarified by centrifugation for 20 min at 4,500 x g to remove any precipitate formed upon freezing and thawing. The supernatant was concentrated 10-fold in a pressurized Amicon stirred cell using Y"30 membrane ( M , = 30,000 cut off), diluted back to 1 liter with cold deionized water, and concentrated to 50 ml.
Chelating Sepharose fast flow (Pharmacia Biotech) was packed in a column (7.5 x 10.5 cm) and charged with zinc by the following method. The column was first flushed with 2 column volumes of water, washed with 5 column volumes of 50 nm ammonium bicarbonate, pH 10.4, 10 column volumes of water, 10 column volumes of zinc chloride solution at 1 g/liter, 5 column volumes of water, and finally with 10 column volumes of equilibration buffer, 20 m M Tris, 1 M NaC1, pH 7.5. 50 ml of concentrated and diafiltered media were applied to the column at a flow rate of 30 cm/h, washed with 900 ml of equilibration buffer, 800 ml of 0.1 M ammonium acetate, pH 6.0, followed by 700 ml of 50 m M EDTA.
The majority of sE-selectin found in the Zn2+-chelate flow through fractions (445 ml) was diafiltered to 0.2 M NaCl and concentrated to 21 ml in an Amicon stirred cell. Superose 12 prep grade column (4.4 x 62 cm) was equilibrated with 25 m M Tris HC1, pH 7.8, containing 0.15 M NaCl and 2 m M CaC1, prepared in pyrogen-free water. Concentrated Zn2+-chelate pool was then injected to the column, and sE-selectin was eluted with same buffer at 20 cmh. Fractions containing >95% pure sE-selectin were pooled, concentrated to 0.5-1 mg/ml, and stored at -70 "C. The apparent molecular weight was determined by calibrating the Superose 12 column with known standards (Whitaker, 1963;Andrews, 1964).
Amino Acid Analysis-sE-selectin was hydrolyzed in vacuo under 6 N HC1 for 24-72 h at 110 "C. 1-2 pg of the resulting hydrolysates were analyzed by ion exchange amino acid analysis using post-column ninhydrin detection on a Beckman 6300 analyzer equipped with a System Gold data acquisition system (Beckman Instruments).
N-terminal Sequence Analysis-Sequence analysis was performed on an Applied Biosystems 470A gas-phase protein sequencer equipped

Soluble E-selectin Is an Elongated Monomer 23951
with a Beckman 126/166 system for on-line phenylthiohydantoin analysis. Data were acquired using System Gold chromatography software. Samples were spotted directly onto Polybrene-coated GFIC filters (Applied Biosystems, Inc.), and standard AB1 sequencing cycles were used. SDS-PAGE and Western Immunoblotting-Purification was monitored by silver-staining of SDS-PAGE and Western blot analysis. SDS-PAGE was performed using 10% acrylamide gels (Laemmli, 1970). Gels were fixed with formaldehyde (Steck et al., 1980) and stained with silver nitrate (Morrissey, 1981). Western blotting was performed according to Towbin in a Bio-Rad mini-transblot transfer cell (Towbin et al., 1979). Blots were probed with rabbit polyclonal antisera directed to Escherichia coli-expressed sE-selectin, which was purified from SDS-PAGE. They were stained with protein A horseradish peroxidase using 4-chloro-1-naphthol substrate.
Endotoxin Assay-Levels of endotoxin contamination in sE-selectin samples were measured using a Limulus amoebocyte lysate gelation assay (Associates of Cape Cod Inc., Woods Hole, MA). The concentration of endotoxin in a sample was calculated from the absorbance values of solutions containing known amounts of endotoxin standards. When the level was >2 ng of endotoxidmg of protein, the sE-selectin sample was rechromatographed on Superose 6. Endotoxin eluted in void fractions separated from sE-selectin.
Coated Plate Cell Binding Assay-To assess biological activity of recombinant sE-selectin, a coated plate assay modified from a previously described method  was employed. Microtiter plates (96-well Immunolon-2, Dynatech, Chantilly, VA) were coated with purified sE-selectin for 16 h at 4 "C. Typically, sE-selectin was diluted to 5 pg/ml in PBS (10 m M sodium phosphate, 150 m M NaC1, pH 7.4), and 50 pl was added per well. To reduce nonspecific binding, plates were blocked with 2% heat-inactivated bovine serum albumin in PBS for 1 h at room temperature. HL60 cells, grown in RPMI 1640 media containing 10% fetal bovine serum, were washed one time in PBS and resuspended to 5 x lo6 celldml. BCECFIAM (2',7'-bis-(2-carboxy- Prior to assay, coated plates were washed once with 0.2 ml of HBSS', then 0.1 ml of dye-labeled cell suspension was added to coated wells. Cells were allowed to settle and adhere to plates for 30 min at room temperature. Following adhesion, plates were washed twice with 0.2 ml of HBSS' by vacuum aspirating wells through a 20-gauge needle and gently adding HBSS' with an 8-channel multipipette. To quantitate adhesion, adherent cells were lysed for 10 min in 0.1 ml of 0.1 M NaOH containing 0.1% SDS, and plates were read in a Fluoroskan I1 plate reader (Labsystems, Helsinki, Finland) at excitation wavelength 485 nm, emission wavelength 538 nm. In some experiments, sE-selectin-coated wells were pretreated with 0.1 ml of monoclonal antibody diluted in HBSS' for 30 min at room temperature. Labeled cells were added directly to wells containing antibody. Hybridomas producing anti-E-selectin were obtained from D. Haskard (Wellicome et al., 1990), and F(ab'), fragments were prepared and purified by Maine Biotechnology Service (Scarborough, ME). Data were analyzed directly using nonlinear least squares methods and the equation, % input cells bound = (B, X)/(K,,, + X ) + NS (Johnson and Frasier, 1985;Johnson and Faunt, 1992). Where B , is the maximum fraction bound, X is the concentration of sE-selectin in mg/ml, Ks0 is the concentration of sE-selectin added to the well producing 50% cell adhesion, and NS is the proportion of cells that are nonspecifically bound. To demonstrate inhibition of adhesion with sE-selectin, dye-labeled HL60 cells were pretreated with increasing concentrations of purified protein for 30 min at room temperature prior to plating onto sE-selectin-coated plates.
Arachidonic Acid-induced Inflammation-The method used was as described previously (Griswold et al., 1993) with minor modifications. Arachidonic acid (Sigma) in acetone (1 mg/20 p1) was applied to the inner and outer surfaces of the left ear. The thickness of both ears was measured with a thickness gauge (Mitutoyo, Japan) 1 h after application of arachidonic acid, and the data were expressed as the change in thickness cm) between treated and untreated ears. The application of acetone does not cause an appreciable response, and, therefore, the difference in ear thickness represented the response to arachidonic acid. The sE-selectin was administered intravenously in a volume of 0.15 mumouse just prior to the application of arachidonic acid. The total amount of sE-selectin in this volume is given in Fig. 7.
Assay of Myeloperoxiduse Activity in Inflamed Ilssue-The method of Bradley (Bradley et al., 1982) was used with modification. Inflamed ear tissue was minced and homogenized (10% w/v) with a Tissumizer (Tekmar Co.) in 50 m M phosphate buffer (pH 6) containing 0.5% hexadecyltrimethylammonium bromide. The tissue homogenates were taken through three cycles of freeze-thaw, followed by brief sonication (10 SI. The appearance of a colored product from the myeloperoxidase-dependent reaction of o-dianisidine (0.167 mg/ml; Sigma) and hydrogen peroxide (0.0005%; Sigma) was measured spectrophotometrically at 460 nm. Supernatant myeloperoxidase activity was quantified kinetically (change in absorbance measured over 3 min, sampled at 15-9 intervals) using a Beckman DU-7 spectrophotometer and a kinetics analysis package (Beckman Instruments). One unit of myeloperoxidase activity is defined as that degrading 1 pmol of peroxide per min at 25 "C.
Mass Spectrometry-Matrix-assisted laser desorption mass spectrometry (MALD-MS) data for sE-selectin were obtained on a Vestec Model 2000 laser desorption time-of-flight mass spectrometer. Samples were prepared for analysis by mixing 1 pl of a 6 pmoupl solution of sE-selectin in 25 m M Tris, 150 m M sodium chloride, 2 m~ CaCI,, pH 7.8, with 1 pl of the matrix, sinapinic acid, on the stainless steel target. The 355 nm line from a Lumonics model HY 400 pulsed Nd:YAG laser (10-ns pulse width, 10-Hz repetition rate) was used for desorptiodionization of the sample. The spectrum is the sum of approximately 25 laser shots. Spectra were calibrated externally using the BSA and BSA dimer peaks.
Analytical Ultracentrifugation-Equilibrium sedimentation and equilibrium velocity experiments were performed with a Beckman XL-A analytical ultracentrifuge (Beckman Instruments, Spinco Division). Double sector cells with charcoal-filled epon centerpieces and sapphire windows were used. Sedimentation experiments were performed at 25 "C in a buffer of 25 m M Tris, 150 m M NaCl, 2 m M CaCI,, pH 7.8. The initial protein concentrations were 0.13 mg/ml for equilibrium experiments and 0.36 mg/ml for velocity experiments. Similar sedimentation equilibrium experiments were performed with the addition of 6 m M sLe".
Equilibrium sedimentation data were analyzed using nonlinear least squares methods (Hensley et al., 1986;Johnson and Faunt, 1992;Brooks et al., 1994a, 199413) under the control of a modified version of IGOR (WaveMetrics, Lake Oswego, OR) running on an MAC ~omputer.~ Data sets were collected after reaching equilibrium, 18-24 h, at a rotor speed of 15,000 rpm. Equilbrium was established by determining that scans taken 4 h apart were superimposable.
At equilibrium, the concentration distribution of a single, homogeneous species is given by, Here, c, and co are the concentrations of the protein at a radial position, r, and at a reference position, usually the meniscus, respectively. M is the protein molecular weight, 6, is the partial specific volume of the protein, p is the solvent density, o is the angular velocity, r is the distance in cm from the center of rotation, r, is the radial position of the reference position in cm, R is the universal gas constant, T i s the absolute temperature, and base is a baseline term to describe nonsedimenting absorbance. The data were tit to Equation 1, with c,, M, and base as the fitted parameters. The partial specific volume, 6, was calculated from the weight percent average of the partial specific volumes of the component amino acids (Zamyatnin, 1972) and the weight percent average of the partial specific volumes of the sugars, assuming a ratio of hexose:hexosamine:sialic acid of 5:4:2 (biantennary, typical for CHO cells), with 21,474 Da due to carbohydrate (from above). This yields a value of U = 0.688 ml g". The solvent density, p, was estimated at 1.006 g ml-'. As as extra constraint on the fit, data obtained at two rotor speeds were analyzed simultaneously (10,000 and 15,000 rprn).
Sedimentation velocity data were analyzed using the program VEL-GAMMA (Beckman Instruments). FiReen data sets were collected, 200 s apart, starting from the time the rotor reached 60,000 rpm. The sedimentation coefficient, s, is determined from the slope of a plot of In r, uersus d t . Where rm is the radial position of the meniscus, o angular velocity in radians per second and t is time in seconds. The sedimentation coefficient may also be expressed in terms describing the hydrodynamic properties of the molecule, i.e. s = M (1 -Up)/Nf (Eq. 2) where N is Avogadro's number and f is the frictional coefficient. Shape information is described in terms of a normalized frictional factor, termed the frictional ratio, i.e. I. Brooks, K. K. Soneson, and P. Hensley, manuscript in preparation.

Soluble E-selectin Is an Elongated Monomer
Here, f , is the frictional factor for a sphere of the equivalent molecular weight and partial specific volume, i.e.
f , = 6 q ( 3 Mfi14wNP (Eq. 4) flf, is the component of the frictional ratio contributed by hydration, which is defined as, flL = (1 + w/Up)'" (Eq. 5) where w is the grams of H,O bound per g of protein.
This leaves f i f , , which is the component of the frictional ratio contributed by shape. The value of f)L, can now be determined from Equation 2. The component of the frictional ratio due to shape can then be related to the axial ratio, p = b / a (assuming a prolate ellipsoid), where b is the short semi-axis and a is the long semi-axis, from the following equation (Cantor and Schimmel, 1980) f./f, = (1 -p2)'r21(L/(pm1n ((1 + (1 -p2)In)/p)) (Eq. 6) The reason why prolate and not oblate ellipsoids were considered is discussed below (see "Results and Discussion").

NMR-A derivative of sialyl-Lewis X (Neu5Ac a-(2-3)-Galp-(1-4)-
[Fuca-(l-3)1GlcNAc~-l-NAc (sLe") was purchased from Oxford Glycosystems (Abington, UK). NMR samples consisted of 1-2 mg of sLe", and of 5 mg of sE-selectin in 0.5 ml of deuterated Tris buffer, pH 7.8, and 1 mM CaCI,. All NMR experiments were carried out on a Bruker AMX (Karlsruhe, Germany) spectrometer operating a t a frequency of 500.14 MHz. Phase-sensitive NOESY experiments were performed usingstandard pulse sequence (Kumar et al., 1980) using time proportional phase incrementation (TPPI) for quadrature detection along w1 direction. Residual HDO signal was suppressed by presaturation during relaxation delay and mixing times. All the data were processed off-line on a SUN Sparcstation I (Sunnyvale, CA) using FELIX program (Hare Research Inc., Seattle, WA). Final matrix size of each spectrum was 1024 x 1024 points.

RESULTS AND DISCUSSION
Construction of Expression Vector-The construction of sEselectin is described in Fig. L4 and the expression vector in Fig.   1B. The insertion of a stop codon a t amino acid 537 (leucine) creates a soluble molecule which lacks the transmembrane and cytoplasmic domains of the protein. These results were confirmed by transfection of the rsELAMdn (producing soluble E-selectin) and pCDM8ELAM (producing membrane-bound Eselectin) plasmids into COS-l cells and the subsequent analysis of both the conditioned medium and the cells. FACS analysis, using a fluorescent monoclonal antibody to sE-selectin, indicates that the rsELAMdn transfected cells have a mean channel fluorescence equal to that of the untransfected cells. Finally, the ELISA assay results indicate that the conditioned medium from the rsELAMdn transfected cells contains six times more sE-selectin than either the untransfected or pCDM8ELAM transfected cells (data not shown).
CHO Expression-After the initial G418 selection and two rounds of amplification, a 100 nM MTX cell line (here called Acc-2351, which secreted in excess of 100 pg/ml in a 96-well plate, was scaled up for growth in spinner flasks and eventually grown in a hollow fiber reactor. Over a 5-day period in spinner flasks, this cell line secreted in excess of 100 mg/liter sE-selectin. However, to produce sufficient quantities of the protein for purification purposes, the hollow fiber was used to generate the conditioned medium. Over a 17-day period, the reactor was fed initially a t a rate of 50 ml of medium per day for days 4-8 to allow the cells to establish and populate the reactor. This rate was expanded to 100 ml per day for days 8-19 and finally a t 200 muday for the last 4 days of the run. In total, 2 liters of medium were collected a t a concentration range of 90-220 pg/ml/day with a peak concentration of 273 pg/mVday. Over the 23-day reactor run, the concentration of sE-selectin produced averaged 200 pg/ml/day, based on ELISA assays. conditioned media and purified sE-selectin whose concentration was determined by quantitative amino acid analysis were subjected to SDS-PAGE under identical conditions and were silver-stained. Lanes 1-4,38, 75, 150, and 300 ng of purified sE-selectin; lanes 5-7, 1,0.5, and 0.25 pl of media; lune 8, molecular weight standards. The relative intensities of the sE-selectin bands in the gel were measured with a scanning densitometer (LKB Model 2222) and converted to mass (mg) by cutting and weighing the scanner output peaks. Using the linear portion of the graph (40-300 ng), the concentration of sE-selectin in the media was estimated as 157 mgfiiter.
achieve the maximum efficiency from the Zn2+-chelate column, the conditioned media were diafiltered to remove the majority (>go%) of low molecular weight media components, which bound to and lowered the capacity of the column. Alternatively, sE-selectin was captured on a Q-Sepharose column; however, the recovery was markedly reduced compared to that from diafiltration.
Inclusion of 1.0 M NaCl in the sample and Zn2+ chelate column equilibration buffer prevented the nonspecific binding of sE-selectin to the column. The major protein component in the flow-through fraction was sE-selectin based on SDS-PAGE. To the pooled sE-selectin fractions, 2 mM CaCl, was quickly added to maintain active lectin domain. These were then concentrated for gel filtration on Superose 12 (Fig. 4). On the basis of silver staining, the fractions containing >95% pure sE-selectin were pooled. The level of endotoxin was <2 ng/mg of protein. sE- selectin appeared to be a sticky protein as there was a 50% loss of protein in each chromatographic step, and it was not possible to recover the protein anywhere in the fractions. Overall yield was >25% as estimated by densitometric scanning of silver staining and ELISA. 25 mg of sE-selectin were purified from 420 ml of hollow fiber media. The purification is summarized in Table I. Fig. 3B). Both bands were present in the original media and were positive in Western blot analysis. Therefore, broad banding pattern and an extra band in the gel may be due to heterogeneity in carbohydrate composition. The analytical reverse-phase HPLC profile suggests that one major protein component was present in the purified fraction (Fig. 5).

SDS-PAGE and Western immunoblot of purified protein showed a broad band and an extra band underneath the major band (lanes 6 and 7 in
Determination of Covalent Structure-One N-terminal sequence (15 cycles) was detected, and it confirmed the sequence predicted for the first 15 residues of the mature protein. The absence of an Asn residue in the fourth cycle may indicate N-linked glycosylation. This was seen previously (Lobb et al., 1991). The sequence is given, The amino acid composition is given in Table I1 and agrees well with the published sequence.
In Vitro Activity ofsE-selectin-HL6O cells bind to sE-selectin-coated plates in a dose-dependent manner (Fig. 6A ). Up to 91% of input cells will adhere under assay conditions used here.
Background levels in uncoated wells were -14%. Half-saturation was achieved at -2.8 pg/ml or 35 nM sE-selectin. The binding of HL60 cells shows a plateau a t 100 pglwell, with as little as 10 ng/well supporting adhesion of cells at levels above background. Pretreatment of coated wells with F(ab), fragments of monoclonal antibody to E-selectin will block binding to 52% of control levels (Fig. 6B). An anti-ICAM F(ab), fragment had no effect. The reason for the lack of complete inhibition of binding in the presence of anti-E-selectin monoclonal antibody (1.2B6) is not clear. Concentrations of up to 400 pg/ml F(ab), did not significantly increase inhibition. In contrast, the binding of HL60 cells was completely inhibited by the pretreatment of cells with purified sE-selectin. The midpoint for inhibition Soluble E-selectin Is a n Elongated Monomer Protein concentrations in crude samples were determined by using Pierce BCA reagent. The concentration of pure sE-selectin was determined by measuring the absorbance at 280 nm = 102,000 "' cm"). silver-stained SDS-PAGE gels and ELISA. tions from Superose 12 (20 pg) were applied to a Brownlee Aquapore C3 column (4.6 x 30 mm) and eluted with a 15-ml linear gradient of acetonitrile (-) in 0.1% trifluoroacetic acid. Absorbance was monitored at 214 nm. * Values not determined due to co-elution of hexosamine sugars. was 1 VM and reached completion. It is worth noting that Ushiyama et al. (1993) have reported that the apparent affinity of sP-selectin for HL60 cells is 70 nM. These experiments were done by direct titration of cells with radiolabeled sP-selectin. This number is only 14-fold lower than the number reported here, which is determined by indirect measurements. This value (1 p~) is probably an underestimate of the intrinsic affinity of sE-selectin for its true receptor as the competitive affinity of HL60 cells for sE-selectin coating the well surface is not considered in the analysis. The instrinsic affinity of sE- Data are expressed as percent of cells binding compared to wells pretreated with buffer only and represent the mean of three independent experiments. The binding of HL60 cells showed no dependence on anti-ICAM F(ab'), fragments. The transition for the anti-sE-selectin F(ab'), fragments was half-maximal at 0.4 pg/ml and saturated at 54%. In contrast, the midpoint for inhibition of the binding of HL-60 cells by free soluble sE-selectin was 1 and reached completion. This value is an underestimate of the intrinsic affinity of sE-selectin for its true receptor as the competitive affinity of HL60 cells for sE-selectin coating the well surface is not considered.
selectin for its natural ligand may well be closer to the value for sP-selectin for its natural ligand. Another group has reported only partial inhibition of binding when HL60 cells are pretreated with sE-selectin and then allowed to adhere to sEselectin-coated plates. They observed maximum inhibition of up to 60% of control (Lobb et al., 1991). Effect of sE-selectin on the Inflammatory Response, in Vivo-In order to evaluate the in vivo functional activity of recombinant sE-selectin, the protein's ability to interfere with a n inflammatory response was evaluated. As seen in Fig. 7, the administration of sE-selectin markedly blunted the influx of neutrophils into tissue inflamed by the application of arachi-Soluble E-selectin Is a n Elongated Monomer were administered sE-selectin a t the doses indicated intravenously via tail vein just prior to the application of arachidonic acid (1 mg/ear). One hour later, the edematous response was measured using a thickness gauge (edema values represent the difference between the untreated ear and the inflamed ear). The animals were then sacrificed and myeloperoxidase (MPO) was extracted from the inflamed tissue, the activity measured was spectrophotometrically, and the values were used as a quantification of neutrophil infiltration. The statistical significance was judged using Student's t test. The saline control values were 0.1578 + 0.009 unitdear for myeloperoxidase activity and 18 -c 0.5 x cm for ear edema. Superose 12 column as described in Fig. 3. Data were analyzed by standard methods (Whitaker, 1963;Andrews, 1964). donic acid, as measured by the inhibition of myeloperoxidase activity. This effect was observed at low concentrations of sEselectin which had little effect upon the edematous response. In this model, the edematous response should not be affected by sE-selectin as it is mediated largely by vasoactive amines and peptidoleukotriene production (Crummey et al., 1987). In contrast, the higher concentration (10 pg of sE-selectidmouse) did not inhibit neutrophil influx, but did blunt the edematous response. This reciprocal relationship may suggest a nonspecific effect on the edematous response at higher sE-selectin concentrations. In multiple experiments using a wide range of concentrations, the predominant effect of sE-selectin has been the inhibition of neutrophil infiltration with minimal effect upon edema. The evanescent inflammatory response induced by arachidonic acid is well-characterized, and the neutrophil influx appears to be mediated by leukotriene B, production (Griswold et al., 1991). From other studies, it is not clear, at present, what adhesion molecules mediate leukotriene B,-induced inflammatory cell infiltration, although CD18 and CD54 have been implicated in vitro (Plamblad and Lerner, 1992). These new data suggest a n additional role for sE-selectin, in vivo. ms. All spectra were processed in an identical manner and plotted just above the noise floor. Panel B, the cross-peak labels are: 3a-3e, NOE between H3, , and H3, , , , , , , ; 3a-4, NOE between H3,i, and H4; 3a-5, NOE between H3, , and H5 of the sialic acid. Panel C , the NOE between the fucose H1 and GlcNAc H3 is labeled as a; the NOE between the fucose methyl group and Gal H2 is shown as b. Weak NOE peak between the H3,,, proton of the sialic acid and Gal H3 proton is labeled as c. Panel D , schematic representation of the inter-residue NOE contacts in sLe' in the presence of sE-selectin. Only relevant protons are labeled, and the arrows correspond to the NOE cross-peaks observed in the transferred NOE experiment (see panel C ) .
Molecular Mass and Hydrodynamic Properties-SDS-PAGE of purified sE-selectin under nonreducing conditions showed a band of M, -75,000 (Fig. 3C, lane 8), which is bigger than the value predicted from the cDNA sequence (58,600). Analysis of the protein after pretreated with dithiothreitol gave a single band of M, -100,000 (Fig. 3C, lane 9). During purification, the protein was eluted from a Superose 12 column as M, -331,000 (Fig. 8).
Mass Spectrometry-The measured M, of sE-selectin based on the (M + H' ) peak is 80,000 (Fig. 9). The molecular mass of sE-selectin based on its peptide sequence alone is 58,826 Da. The mass difference of 21,474 Da suggests that sE-selectin is 27% carbohydrate. The width of this peak suggests significant carbohydrate heterogeneity.
Analytical Ultracentrifugation-The results of equilibrium analytical ultracentrifugation of sE-selectin is shown in Fig. 10. Simultaneous analysis of data from two rotor speeds in terms of Equation 1 fit best to a single sedimenting species yielding an average apparent molecular weight of 81,600 (2 4,500). Weis et al. (1991) solved the structure of the lectin domain of a man-nose binding protein which has similarity to the lectin domain of sE-selectin. Inspection of the structure yielded no apparent sugar binding site. A possible explanation is that the sugar binds between two lectin domains, one from each to two molecules of sE-selectin. To test this hypothesis, the molecular weight of sE-selectin was also determined in the presence of sLex. The best fit of the data obtained from four repeats of sE-selectin in the presence of 6 lll~ sLe" is to a single species with a molecular weight of 79,500. Hence, sLe" does not promote self-association, and the hypothesis is not supported.
Velocity sedimentation experiments gave a sedimentation coefficient for sE-selectin of szo,+, = 4.3 and a diffusion coefficient of 4.1 x cm2 s-l (data not shown, Muramatsu and Minton (1988)). Using a value of 0.35 g of H,O/g of protein for the hydration (Kuntz and Kauzman, 1974) and assuming the shape to be a prolate ellipsoid, sE-selectin was shown to have an axi ! l ratio of 1 0 5 1 and dimensions of approximately 25 A by 270 A (see Tinoco et al. (1978) and Cantor and Schimmel (1980)). From these measurements alone, prolate and oblate ellipsoids cannot be distinguished. However, Ushiyama et al. (1993) have shown by electron microscopy that sP-selectin, a highly homologous protein, is a n extended rod-like molecule, so this is not an unreasonable assumption. Additionally, these calculations assume the ellipsoid is smooth and that the relative decrease in sedimentation rate, compared to a sphere, can be interpreted solely in terms of extended dimensions. However, sedimentation properties of the proteins are also a sensitive function of the rugosity (roughness) of the protein surface (Teller et al., 1979). Hence, interpretation of the sedimentation properties of sE-selectin in terms of a smooth ellipsoid will give an upper limit t o the dimensions.
NMR-Binding of sialyl-Lewis X (sLe") to sE-selectin was studied by the two-dimensional version of transferred nuclear Overhauser effect (Clore and Gronenborn, 1982;Campbell and Sykes, 1993). The NOE information in the bound state is transferred to the free ligand by chemical exchange if binding is weak, and the transferred NOE spectrum thereby provides structural information of bound ligand Sykes, 1991, 1993). Fig. 1lA shows the NOE spectrum of sLe" at 5 "C. NOE cross-peaks in the spectrum under these conditions are very weak, suggesting that O T~ is close to 1.1. Fig. 1l.B shows the NOESY spectrum of M sE-selectin under similar conditions. Normally, protein resonances in a large molecule of the size of sE-selectin are expected to be too broad to be observed. However, several well resolved cross-peaks corresponding to the sialic acid moiety are observed in the NOESY spectrum of sE-selectin. The cross-peak 3a-3e (Fig. 1lB) corresponds to the NOE between the geminal protons on carbon-3 of sialic acid, and both these protons show NOE contacts with the H4 and H5 protons of the sialic acid. These NOE cross-peaks are relatively intense and reflect high abundance and/or mobility of the sialyl group associated with glycosylation of sE-selectin. These results are consistent with mass spectral data, which show that the carbohydrate content in the sE-selectin is 27%. The twodimensional version of transferred NOE spectrum of sLe" in the presence of 0.12 m~ sE-selectin is shown in Fig. 11C. It shows several cross-peaks which correspond to "negative" NOE and are relatively more intense than the sE-selectin or sLe" spectra. Since free sLe" does not contribute significantly to the NOE spectrum (Fig. lU), these cross-peaks essentially result from the transferred NOE and arise from the conformation of sLe" when bound to sE-selectin. In addition to several intraresidue contacts, NOE cross-peaks between protons belonging to different monosaccharide residues were observed. Anomeric protons (Hls) of fucose and galactose residues show "sequential" NOEs to H4 and H3 protons, respectively, of the GlcNAc residue (see Fig. 1LD). Furthermore, H5 and CH, protons of the fucose residue show NOEs to HZ proton of the galactose residue. These "contacts" suggest a compact conformation of the Le" trisaccharide in which the fucose moiety is packed in close proximity to the galactose residue (Fig. llD). Wormald et al. (1991) andMiller et al. (1992) showed that the Le" trisaccharide in LNFP-3 has a rigid structure which shows these interproton contacts. In addition to these NOE contacts, a weak NOE between the axial proton H3a of sialic acid and the galactose-H3 proton was observed in our studies. This is the only contact observed between the sialic acid moiety and the Le" trisaccharide. A number of independent investigations (Ball et al., 1992;Lin et al., 1992;Miller et al., 1992;Mukhopadhyay et al., 1994) reported similar results in their studies on the structure of sialyl-Lewis X in solution. Our results imply that the bound conformation of sLe" is similar to its structure in solution.
Summary-In summary, sE-selectin has been expressed at high levels in CHO cells and an efficient large-scale purification protocol, using only two chromatography steps under mild aqueous conditions, has been developed. The protein thus produced is active in vitro, in that HL-60 cells will bind to surfaces coated with it and monoclonal antibodies to sE-selectin or pretreatment of the cells with sE-selectin will compete for this adhesion. This competitive effect puts an upper limit on the affinity of sE-selectin for its natural receptor of -1 VM. sEselectin will inhibit neutrophil influx in a mouse inflammation model, demonstrating that is also active in uiuo. The protein has been shown to have a monomer molecular weight of 80,000 by mass spectroscopy and the same molecular weight in solution as determined by analytical ultracentrifugation. Hydrodynamic experiments show that the molecule is an extended monomer, as was found for sP-selectin, with an axial ratio of 10.5:l. Finally, NMR data show that this protein productively binds its normal carbohydrate ligand, sLe". This molecule is well behaved and is suitable for configuration in screens for the identification of sLe" antagonists.