AU7863591A

AU7863591A - Production of multimeric glycoproteins through chemical coupling

Info

Publication number: AU7863591A
Application number: AU78635/91A
Authority: AU
Inventors: Ling-Ling Chen; R. Blake Pepinsky; Joseph J Rosa
Original assignee: Biogen Inc
Current assignee: Biogen Inc
Priority date: 1990-09-14
Filing date: 1991-05-03
Publication date: 1992-04-15
Also published as: WO1992005188A1; CA2068651A1; EP0502131A1

Description

PRODUCTION OF MULTIMERIC GLYCOPROTEINS THROUGH CHEMICAL COUPLING

TECHNICAL FIELD OF THE INVENTION This invention relates to the field of protein biochemistry and, in particular, multimeric glycoproteins and methods for producing them.

BACKGROUND OF THE INVENTION

Advances in biotechnology have vastly broadened the variety and quantity of pharmaceuticals and other biologically useful molecules available to medicine and industry. Bioengineers are now able to produce in quantity many important proteins such as human insulin and interferons which, only a few years ago, were rare and expensive.

One current area of progress in biotechnology involves the development of methods to increase the biological activity of these newly available bio- molecules so that they can be administered in smaller amounts, with fewer side effects and with less expense. To this end, scientists are now investigating multimeric forms of proteins. Certain kinds of proteins are likely to exhibit greater biological activity in multimeric form because their individual active sites have an opportunity to act in concert with a multiplicity of identical active sites. One such class of proteins is receptor and ligand proteins and, in particular, those proteins involved in virus-cell or cell-cell interactions. For example, the attachment of an invading AIDS virus to target cells (i.e., lymphocytes), or the interaction between different white blood cells that characterizes the immune response to infection, involves the interaction of receptor and ligand proteins on the surfaces of the viruses and cells. The binding affinity of any single receptor for its ligand may be low. However, viruses and cells normally exhibit hundreds to thousands of copies of a particular surface molecule, and many receptor-ligand interactions take place simultaneously. When many surface molecules become involved in binding, they act synergistically so that the affinity of a virus for a cell, or of one cell for another, is greater than the mere sum of the binding affinities of the individual molecules. The result is effective virus-cell or cell- cell adhesion.

By contrast, when receptor or ligand proteins are removed from the cell surface and purified, or isolated by recombinant DNA techniques, for use, e.g., as therapeutics, they act as monomers and lose the advantage of acting in concert with many other copies of the same protein closely associated on the surface of a cell. Thus isolated, the low affinity of a given protein for its ligand or receptor may become a serious drawback to its effectiveness. For example, if a soluble adhesion protein having low binding affinity for its ligand is administered as a therapeutic to block a particular binding pathway (in the hope of preventing a binding-dependent pathology) , the protein will be readily cleared from the patient's system and/or will not compete effectively for the ligand with a pathogen able to present multiple receptors. Effective treatment will therefore require constant administration and very high doses of the protein.

Such drawbacks might be avoided, however, if a means could be found to provide multimeric forms of an isolated protein. The multimeric forms would have higher overall binding affinities for their receptors or ligands than the monomeric form. Accordingly, there is a need in the art to develop means for chemically coupling proteins or polypeptides together to obtain multimeric forms of the protein or polypeptide.

Scientists have devised several methods of chemically coupling proteins. However, most chemical methods are non-specific, i.e., they do not direct the point of coupling to any particular site on the protein. As a result, conventional coupling agents may attack functional sites or sterically block active sites, rendering the coupled proteins inactive. Furthermore, the chemically coupled products may be oriented so that the active sites cannot act synergistically, thereby rendering the products no more effective than the monomeric protein alone. Cross- linking with glutaraldehyde is one example of non¬ specific coupling. (Reichlin, 1980.) Heterobifunctional cross-linking agents have also been used to chemically couple proteins. Such agents have two different functional groups, for example, an amine-reactive group and a thiol-reactive group, that will cross-link two proteins having free amines and thiols, respectively. One example is SPDP (N-succinimidyl-3-(2-pyridyldithio)propionate) . However, since most proteins have many free amines, this type of cross-linking also tends to be non¬ specific. One approach to increasing coupling specificity is to direct chemical coupling to an amino acid found only a few times in the polypeptide chain, for example, to the thiol group of a cysteine residue. Raso and Basala (1984) cross-linked cysteine-containing Ricin A chain to SPDP-linked transferrin. Of course, this requires that the protein have cysteine residues in areas of the molecule that are not involved in activity, so that coupling at that particular site will not interfere with function. Techniques exist for genetically engineering a cysteine residue into the polypeptide*s primary sequence, e.g., near the amino- or carboxy-terminus. The thiol-functional side chains of the cysteines on two or more polypeptides may then be oxidized to form disulfide cross-links. However, this technique requires that the cysteines be sterically accessible for linking and requires that the conformations of the polypeptides permit cross-links between polypeptides to form without eliminating activity.

Thiol or thiol-reactive groups have also been introduced directly into the peptide using iminothiolane or heterobifunctional cross-linking agents such as SPDP. Coupling of the proteins has then been accomplished through disulfide bonds formed either directly or through homobifunctional cross-linking agents. (See, e.g., Srinivasachar and Neville, Jr., (1989) ; Ra akrishnan and Houston, (1984) ; and Lambert et al., (1985) .) Another approach to increasing coupling specificity is to direct chemical coupling to the carbohydrate groups of glycoproteins. Chemistries have been developed for oxidizing the sugar moieties of glycoproteins into reactive aldehydes (Liao et al., 1973) , thereby generating active sites for coupling. However, the currently developed chemistries tend to be inefficient and result in low product yield.

Moroney et al. (1987) describe coupling a disaccharide, lactose, to polyacrylamide beads for an affinity column. They reacted α-lactose with cystamine diacetic acid salt to create N-(2'-Mercaptoethyl) lactamine, which was then coupled to a prepared polyacrylamide through a disulfide linkage.

Although basic coupling chemistries have been explored, better methods are needed for chemically coupling proteins or polypeptides to yield functional multimeric products. In particular, there is a need for methods directed at coupling of the carbohydrate groups on a wide range of glycoproteins that are more specific and efficient than existing methods of protein coupling, and for methods that produce multimeric glycoproteins having increased activity, e.g., increased affinity for the receptor/ligand of the glycoprotein, in comparison to the glycoprotein (or glycosylated polypeptide) in its monomeric or uncoupled form.

SUMMARY OF THE INVENTION

The present invention addresses the aforementioned needs by providing a process for chemically coupling glycoproteins to produce multimeric forms. The multimeric glycoproteins produced according to the present invention are coupled specifically at glycosylation sites, can be engineered for steric flexibility and demonstrate increased biological activity as compared with monomeric proteins. The methods for producing these glycoproteins are more efficient than conventional methods. In addition, they permit the production of homo-multimers. Thiol- functional glycoprotein intermediates of the process of this invention are themselves novel molecules. The process of the present invention comprises the steps of: (1) oxidizing a solution including glycosylated polypeptides to produce aldehyde- functional glycoproteins;

(2) reacting the aldehyde-functional glycoproteins with an aminothiol compound to produce thiol-functional glycoproteins; and

(3) chemically coupling two or more thiol- functional glycoproteins by oxidizing to form disulfide bonds therebetween or by reacting with a homobifunctional cross-linking agent. As used herein, the term "glycoprotein" is meant to include glycoproteins, glycopolypeptides (which may correspond to a fragment or a derivative of a glycoprotein) and aggregated glycoproteins or glycopolypeptides (which may comprise several monomeric units, i.e., dimers, trimers, etc.). The multimeric glycoproteins of this invention comprise two or more "glycoprotein" units crosslinked via an organic aminothiol bridge to produce multimers, i.e., dimers, trimers, tetramers, etc. , of such units. Preferably the multimeric glycoproteins of the present invention will be homomultimers, i.e., multimers formed by cross- linking two or more of the same glycoprotein.

Most preferably, the multimeric glycoproteins of the present invention will be homopolymers of adhesion proteins or portions thereof (e.g., the extracellular region of a transmembrane adhesion protein or a portion thereof) . As used herein, the term "adhesion protein" means a cellular, viral or extracellular matrix-associated protein whose biological activity includes binding to another cellular, viral or extracellular matrix-associated protein. Such proteins, thus, will include surface receptors and ligands involved in cell-cell and virus- cell interactions (e.g., CD4, LFA3, VCAM-lb) as well as involved in anchoring cells to the extracellular matrix (e.g., fibronectin) . Also contemplated are multimers formed from soluble forms of monomeric receptor/ligand proteins, e.g., polypeptides derived from transmembrane or surface-bound receptors/ligands by eliminating the transmembrane region or isolating an extracellular portion (e.g., by recombinant means). Reference herein to specific adhesion proteins, such as CD4, LFA3 and VCAM-lb, includes the soluble form of these proteins. Specifically contemplated herein are homopolymers produced by cross-linking recombinant soluble CD4 protein ("rsCD4") , recombinant soluble LFA3 ("rsLFA3") and recombinant soluble VCAM-lb ("rsVCAM-lb") . Recombinant soluble forms of CD4, LFA3 and VCAM-lb are known. (See PCT patent applications WO 89/01940, WO 88/09820, and WO 90/1330, incorporated herein by reference) .

Finally, in the cases of biologically active sugar complexes or glycoproteins in which the biological activity resides exclusively in the attached carbohydrates, we contemplate coupling sugar complexes without any associated protein to produce multimeric sugar complexes. In these cases, the process of the invention comprises the steps of oxidizing a solution including sugar complexes to produce aldehyde- functional sugar complexes; reacting the aldehyde- functional sugar comple. _s to produce thiol-functional sugar complexes; and chemically coupling two or more thiol-reactive sugar complexes through a direct disulfide bond or through a homobifunctional cross- linking agent. This process is limited to cases in which the coupling of the sugar complexes does not interfere with carbohydrate function.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 depicts an SDS-polyacrylamide gel on which disulfide-linked and BMH cross-linked multimeric CD4 were separated according to molecular weight.

Lanes (a) and (f) show prestained BRL® molecular weight markers.

Lane (b) shows 5 μg recombinant soluble CD4 (rsCD4) in non-reducing sample buffer.

Lane (c) shows 5 μg disulfide-linked rsCD4 (cysteamine) in non-reducing sample buffer.

Lane (d) shows 5 μg rsCD4 in 2-mercaptoethanol. Lane (e) shows 5 μg disulfide-linked rsCD4 (cysteamine) after treatment with 2-mercaptoethanol. The multimers have been reduced back to monomeric thiol-functional rsCD4.

Lane (g) shows 5 μg thiol-functional rsCD4 (cysteamine) . Lanes (h) to (k) show 5 μg BMH cross-linked rsCD4 (cysteamine) .

Figure 2 (top) is a graph showing the absorbance at 280 nm of soluble BMH cross-linked CD4 fractionated on a SUPEROSE 6^® FPLC gel filtration column, according to molecular weight.

Figure 2 (bottom) depicts an SDS-polyacrylamide gel stained with Coomassie Blue on which protein from column fractions 24-39 were separated according to molecular weight. The lanes correspond to specific fraction numbers.

Figure 3 depicts a 10% SDS polyacrylamide gel on which various LFA3 preparations were separated according to molecular weight. Lane (a) shows prestained BRL® molecular weight markers.

Lane (b) shows rsLFA3 stained with Coomassie blue. Lanes (c) and (d) show different concentrations of disulfide-linked rsLFA3 detected by silver staining.

Lanes (e) and (f) show different concentrations of disulfide-linked rsLFA3 reduced with 2% 2- mercaptoethanol and detected by silver staining.

Lane (g) shows erythrocyte PI-LFA3 detected by fluorography.

Lane (h) shows erythrocyte PI-LFA3, treated with N-glycanase and detected by fluorography.

Lane (i) shows BMH cross-linked rsLFA3 detected by fluorography. Lar.e (j) shows BMH cross-linked rsLFA3, treated with N-glycanase and detected by fluorography. Figures 4A, 4B, and 4C depict three chromatographs showing absorbance at 280 nm of rsLFA3, human erythrocyte PI-LFA3, and BMH cross-linked rsLFA3 separated on a SUPEROSE 6 FPLC gel filtration column. Figure 4D depicts a silver stained SDS-PAGE gel of samples from selected column fractions from the BMH cross-linked rsLFA3. Lane (a) shows BRL® prestained high molecular weight markers. Lanes (b)-(p) correspond to column fractions 22-36, respectively. Figure 5 is a graph showing the binding curves of I-labeled multimeric LFA3 dimer, trimer, tetramer, PI-LFA3 and TS2/18 antibody (all adjusted to a final activity of 6000 cpm/ng protein) incubated with 5 x 10⁴ Jurkat cells. Binding curves were plotted as counts bound versus input counts. To assess non¬ specific binding, parallel samples were processed with cells that had been treated with TS2/18 α-LFA3 antibody prior to incubation with LFA3. All of the preparations produced similar values of background binding. Figure 6 is a graph showing the activity of BMH cross-linked rsLFA3 dimer, tri er, tetramer, and rsLFA3 monomer in a rosetting inhibition assay. Values are reported as percent inhibition per concentration of LFA3. Open squares represent rsLFA3, open triangles represent dimers, closed circles represent trimers and closed diamonds represent tetra ers.

Figure 7 is a graph showing the incorporation of ³H-thymidine by T-cells co-stimulated with α-Tll antibody and also PI-LFA3, multimeric LFA3 dimers, trimers or tetramers. Values are reported as total counts per concentration LFA3. Cells were cultured for three days and then labeled overnight with ³H- thymidine. Figure 8 depicts an SDS-polyacrylamide gel of samples of thiol-functional rsVCAM-lb incubated with varying concentrations of BMH. Lane (a) shows prestained BRL® molecular weight markers. The concentration of BMH used in lanes (b)-(e) was as follows: (b) O μM BMH, (c) 50 μM BMH, (d) 100 μM BMH, (e) 200μM BMH, (f) 400 μM BMH.

DETAILED DESCRIPTION OF THE INVENTION

This invention provides a process for chemically coupling glycoproteins through the carbohydrate moiety rather than through the amino acid sequence of the glycoprotein.

Polypeptides may have several glycosylation sites, either N-linked or O-linked, or both. Polypeptides may become N-glycosylated at specific tripeptide sites having the amino acid sequence Asn-X- Ser or Asn-X-Thr, where X can be any amino acid except proline. For example, CD4 protein has two N- glycosylation sites at amino acids 269-271 and 298-300 of the amino acid sequence. (Carr et al. , 1989.) LFA3 has six N-glycosylation sites. VCAM-lb has seven potential N-glycosylation sites. (One is cryptic, having proline in the second position) . No consensus amino acid sequence for O-glycosylation sites has yet been determined.

The specificity of chemical coupling ac rding to this invention may be increased by deleting one or more glycosylation sites. Sites may be deleted by producing truncated versions of the polypeptide in which one or more sites are eliminated. Alternatively, site-directed mutagenesis can be employed to eliminate a glycosylation site from a DNA sequence encoding the polypeptide. (Sambrook et al.. Chap. 15, 1989). Also, a suitable glycoslyation site may be inserted in a polypeptide de novo by engineering the site into a gene encoding the polypeptide. In this way, the proposed site(s) of cross-linking according to this invention can be directed away from the region of the polypeptide involved in its biological activity, and the cross-linking can be tailored so as not to interfere with the biological properties of the glycoprotein.

It is preferred that multimeric glycoproteins according to this invention will exhibit improved activity in multimeric form in comparison to the monomeric or non-coupled form. For example, viral or cell receptors and ligands are useful because they typically bind to particles or cells exhibiting many copies of the receptor. Multimeric proteins formed from them are useful in therapies that involve inhibiting virus-cell or cell-cell binding. Useful virus-cell receptors include ICAM1, a rhinovirus receptor; the polio virus receptor (White and Littman, 1989); CD2, the Epstein-Barr Virus receptor (Nemerow et al., 1990); and CD4, a receptor for human immunodeficiency virus (HIV) . Cell-cell receptors or ligands include members of the vascular cell adhesion molecule family, such as ICAMl, ELAM-1, and VCAM-1 and VCAM-lb and their lymphocyte counterparts (ligands) ; the lymphocyte associated antigens, LFAl, LFA2 (CD2) and LFA3 (CD58) , CD59 (a second ligand of CD2) , members of the CD11/CD18 family, and VLA4. These molecules are involved in pathologic inflammation. (Bevilacqua et al., 1987; Osborn et al., 1989.) Useful receptor and ligand glycoproteins will include not only the naturally occurring forms of those glycoproteins but also derivative forms such as rsCD4, rsLFA3 or phosphatidylinositol-linked LFA3 (PI-LFA3) or other truncated or modified versions of such proteins. Particular mention is made of rsCD4, rsLFA3 and rsVCAM-lb.

CD4 is the glycoprotein receptor on T-lymphocytes (T-cells) that is recognized by HIV, the virus that causes AIDS and ARC. (Maddon et al., 1986.) Specifically, CD4 recognizes and binds to the HIV surface protein complex of gpl20/gpl60. Fisher et al. (1988 and PCT patent application WO 89/01940) have shown how to produce recombinant soluble CD4 (rsCD4) and demonstrated in vitro that rsCD4 blocks the infection of T-cells by HIV. Chao et al. (1989) have shown that the biological activity of CD4 resides in the first 113 amino acids. Soluble CD4 is attractive as an agent for AIDS therapy, to act as a decoy to block T-cell binding to gpl20/gpl60 on the invading HIV. Initial studies suggest that blocking the virus will require high doses of the monomeric protein. (Yarchoan, 1990.) However, multimeric forms of CD4 according to this invention may be effective at lower doses than the monomeric protein. Lymphocyte Function-Associated Antigen 3 ("LFA3") is a surface glycoprotein expressed on many cell types. It binds to the T-cell surface glycoprotein, CD2. This interaction is important in mediating thymocyte interactions with thymic epithelial cells, in antigen-independent and -dependent interactions of T-cells with target cells and antigen presenting cells, and in the T-cell rosetting with erythrocytes. (Waliner et al. , 1987.) LFA3 is found in two forms, i.e., transmembrane and "PI-1inke ", which differ by their method of attachment to the cell surface. (Dustin et al., 1987.) In the transmembrane form, a hydrophobic stretch of amino acids (the transmembrane region) , penetrates through the lipid bilayer. In the Pi-linked form, LFA3 is anchored to the membrane by means of phosphatidylinositol ("PI")-containing glycolipid that is covalently attached to the C- terminus of the protein. Currently, investigators believe that the two forms may result from differential processing of a single mRNA precursor. (Wallner et al., PCT patent application WO 90/02181.) Wallner et al. have isolated a cDNA encoding the PI form of LFA and have shown how to express it in CHO cells. The protein may be processed to give a non-covalently aggregated octomer. These investigators also showed how to express a recombinant soluble form of LFA3 (rsLFA3) by removing portions of the hydrophobic transmembrane region of the Pi-linked form of LFA3. (Wallner et al., PCT patent application WO 88/09820.) LFA3 is an important regulator of T-cell activity, and therefore a multimeric form having high binding affinity may be employed advantageously as an immunomodulator. Vascular Cell Adhesion Molecules 1 and lb ("VCAM-1" and "VCAM-lb") are glycoproteins expressed on the surface of endothelial cells during inflammation. Human umbilical vein endothelial cells in vitro express VCAM-1 and VCAM-lb when induced by the inflammatory cytokines, interleukin-1 and Tumor Necrosis Factor α. (Osborn et al., 1989.) VCAM-1 and VCAM-lb are adhesion proteins involved in the binding of T and B lymphocytes to endothelial cells through their ligand, VLA-4. (Elices et al., 1990). Since the presence of activated T cells is the hallmark of numerous inflammatory and autoimmune diseases, such as arthritis, lupus and multiple sclerosis, inappropriate expression of VCAM-1 and VCAM-lb might be the fundamental pathochemical characteristic of these diseases.

The amino acid sequences of VCAM-1 and VCAM-lb are similar, but not identical. VCAM-lb is longer and contains an additional polypeptide domain. (Hession et al. , PCT patent application WO 90/13300.) The difference may be the result of differential mRNA splicing or a separate VCAM allele. The most recent evidence indicates that VCAM-lb is more prominently expressed on the cell surface than VCAM-1. (Hession et al., 1991.) Although we use VCAM-lb in our example, one could also use VCAM-1.

Hession et al. (PCT patent application WO 90/13300) proposed inhibiting T cell binding to the endothelium by blocking the VCAM-1 binding pathway. They described the construction of an expression vector encoding soluble VCAM-lb and its expression in CHO cells. A multimeric sVCAM-lb, having a higher binding affinity for VLA-4 than monomeric VCAM-lb, may be a more effective therapeutic agent.

The methods of this invention also may be applied to bacterial immunogens, parasitic immunogens and viral immunogens to produce multimeric proteins useful as vaccines. Bacterial sources of these immunogens include those responsible for bacterial pneumonia and pneumocystis pneumonia. Parasitic sources include malarial parasites, such as Plasmodium. It will be clear that these immunogens include non- glycosylated proteins. Therefore, in order to produce multimeric glycoproteins of this invention it is necessary to glycosylate them, for example, by expression of recombinant proteins in eukaryotic expression systems.

It is an aim of this invention to chemically couple glycoproteins in such a way as to increase their biological activity. Therefore, glycoproteins in which th attached carbohydrate: are necessary for activity an chemical coupling interferes with this activity are not likely to be useful in this invention. One can easily identify which proteins include essential carbohydrates, by removing the carbohydrate with a glycosidase by growing cells in the presence of a glycosylation inhibitor, or constructing deletion mutants using recombinant techniques and testing the de-glycosylated protein for activity in an appropriate assay. The first step in the process of this invention involves oxidizing sugars on the carbohydrate moiety of a glycoprotein to aldehydes to produce aldehyde-functional glycoproteins. This is conveniently done by contacting the glycoprotein with a suitable oxidizing agent, most preferably sodium periodate. Typically, a molar excess (e.g., 10-fold) of sodium periodate is added to the glycoprotein in, e.g., 0.1 M sodium acetate (pH 5.0) and then the mixture is incubated at 23°C for 1 hour. This oxidizes the terminal sugar residue of the carbohydrate (typically sialic acid) to produce a reactive aldehyde. Excess oxidant may be removed by standard means, e.g., on a desalting column or by dialysis.

The second step of the process involves reductive amination of the aldehyde-functional glycoproteins with aminothiol compounds to produce intermediates having a free thiol group, called thiol- functional glycoproteins. The thiol-functional glycoprotein intermediates are conveniently prepared, for example, by incubating the glycoprotein with a large molar excess (e.g., 1000-fold) of the aminothiol compound and 4-morpholinoethane sulfonic acid (MES) at physiological pH at room temperature.

Aminothiol compounds are organic molecules having at least one functional amino group (-NH ) and also containing at least one sulfur moiety (-S-) that will give at least one free thiol group (-SH) under reducing conditions. Cysteamine (HSCH CH₂NH ) is the most preferred aminothiol compound for the reaction to produce a thiol-functional glycoprotein. However, we found that reacting cysteamine without protecting the thiol group often did not yield a suitably reactive thiol-functional glycoprotein. Therefore, the most preferred aminothiol reagent for this step of the process was the oxidized form of cysteamine, namely, cystamine, employed in the form of a hydrochloride salt, i.e., cystamine dihydrochloride

(NH₂CH₂CH₂SSCH.CH₂NH₂«2 HC1) . Once the conjugation was complete, we generated the free thiol group by reduction with DTT. The technique of employing aminothiol compounds in oxidized form or compounds in which the sulfur is protected during the reaction will be useful for other aminothiol compounds as well. Alternatively, the conditions and reagents used in this step should be selected so that the aldehyde moieties on the glycoprotein react preferentially with the amino-functional group of the aminothiol reagent. Other suitable aminothiol compounds include cysteine, glutathione, short peptides of up to twenty amino acids containing at least one cysteine residue, C . alkanes having amino and thiol (sulfhydryl) substituents, C , organic hydrazines containing reducible sulfur, and the like.

This second step reaction may be carried out by incubating an amount of aldehyde-functional glycoprotein (e.g., 0.5 mg/ml) in, e.g., 50 mM MES at pH 6.5, 20 mM aminothiol compound and 5 mM sodium cyanoborohydride for 6-20 hours at ambient temperatures. The reaction product is advantageously reduced to produce the free thiol by treatment with, e.g., 40 mM dithiothreitol (DTT) for, e.g., 40 minutes at 23°C, and dialyzed into a storage buffer, such as 10 mM sodium acetate (pH 5.0), 100 mM NaCl. These parameters may be modified as appropriate to the particular reactants involved.

This procedure typically creates a free thiol group on most of the reactive aldehyde sites on the glycoprotein created in the first step reaction. Where more than one glycosylation site on a given glycoprotein has a reactive aldehyde group, the resultant thiol-functional glycoprotein will enable the production not only of dimers, but also of higher order multimers such as trimers, tetramers, etc. because each thiol-functional glycoprotein will be capable of cross- linking with more than o other conjugate.

This conjugatio.i step is important for the improved efficiency of this process over other cross- linking methods. For example, we have attempted to chemically couple proteins by linking the aldehyde groups of aldehyde-functional glycoproteins to the amine groups of lysines on other proteins. This reaction was found to be inefficient, requiring a large molar excess of the lysine-containing protein and producing low yields. However, according to this invention thiol-functional glycoproteins may be prepared more efficiently by using low molecular weight aminothiol compounds, such as cystamine, which can be employed at high concentrations to drive the reaction. The resultant thiol-functional glycoproteins, in turn, are efficiently coupled by reacting the free thiols. These thiol-based cross-linking reactions are specifically directed to thiol groups and produce chemically coupled products that are stable in aqueous solutions. Two or more thiol-functional glycoproteins may be chemically coupled by direct disulfide bonds formed between free thiol groups. This results in disulfide-linked glycoproteins. Air oxidation is the preferred method to accomplish such disulfide coupling. Preferably, air oxidation is performed via buffer exchange to raise the pH, e.g., to between pH 7.0 and pH 8.5. For example, we performed buffer exchange from pH 6.5 storage buffer into PBS pH 7.2 on a P6DG (Biorad) or SEPHADEX G25 (Pharmacia) desalting gel and incubated the samples at 4°C for 16 hours.

Thiol-functional glycoproteins may also be chemically coupled using a homobifunctional cross- linking agent. These cross-linking agents have two thiol-reactive groups that form a bridging group between the thiol sites of thiol-functional glycoproteins.

The chemical cross-linking agents lend steric or spatial flexibility to the multimer and thus avoid a common drawback of the cysteine-cysteine and aldehyde- lysine methods of coupling polypeptides mentioned previously. A wide range of bifunctional cross-linkers is available from which molecules of various lengths and flexibility can be selected. The Pierce Co. Immunotechnology Catalogue and Handbook, Volume 1, §§ E31-E48 describes many useful homobifunctional cross-linkers and reaction conditions for their use, and those sections are incorporated herein by reference. The ortho- and para-phenylene dimaleimides, o-PDM or p-PDM, from Sigma Chemical Co. (St. Louis, Missouri) or Aldrich Chemical Co. (Milwaukee,

Wisconsin) , are also useful. Bis-maleimidohexane ("BMH") , available from Pierce Co. (Rockford, Illinois) , is the most preferred cross-linking agent. Heterobifunctional cross-linking agents that contain amine and thiol-reactive groups can be transformed into useful homobifunctional agents by reacting them with, e.g., diamines and purifying the species that is bifunctional for thiol-reactive groups. By varying the size of the diamines, one can create cross-linkers of varying lengths. Useful diamines for this purpose include ethylenediamine, 1,3-diaminopropane, 1,4-diaminobutane, 1,5-diaminopentane, 1,6-hexanediamine, p-phenylenediamine, etc. (Aldrich Chemical Co.). The cross-linking reaction may be carried out in known ways, e.g., by incubating the thiol-functional glycoprotein with a sufficient concentration of the cross-linking agent in a buffer adjusted to physiological pH. We typically performed this cross-linking by adjusting a solution of thiol-functional glycoprotein (5 mg/ml) in 50 mM MES to about pH 6.5 by adding 0.5 M MES. Then the mixture was incubated for 1 hour at "oom temperature with 50 μM to 300 μM of a homobifunctional cross-linking agent. This created a non-reducible chemical cross-link. Again, these parameters may be altered to suit the particular reactants involved.

Preferred multimeric glycoproteins according to this invention may be described by the following formula:

P.G. - N₁ - R₁ - S - X - S - R₂ - N₂ - G^, wherein: P and P represent the same or different polypeptides;

G and G represent the same or different glycosyl moieties on the polypeptides P and P , respectively;

N and N represent, independently, a linkage between G and R or G and R , respectively, through a secondary or tertiary amine nitrogen atom, i.e., N and N may either represent an imino (-NH-) or an aza (=N- or -N=) linkage between the glycosyl moieties (G , G ) and the R and R moieties;

R and R represent the same or different organic radicals of 1-16 carbon atoms or polypeptides of up to 20 amino acids;

S is sulfur; and

X is a direct bond or is a divalent organic radical of from 1-16 carbon atoms.

In multimeric glycoproteins of the foregoing formula, the exact nature of the bonding between the component moieties of the multimer (P , G , P , G , N , R_λ , N₂, R₂, S, X) will depend on the exact nature of the components and how the multimeric glycoprotein products are formed. For instance, if the multimeric product is formed according to the process described above and the terminal sugar of a glycosyl moiety is oxidized to produce two aldehyde groups, the following reaction with an aminothiol compound may result in cyclization of the sugar residue on the amino nitrogen of the aminothiol compound, resulting in a product in which the aminothiol compound forms two covalent bonds to the glycosyl moiety. Alternatively, where the amino nitrogen reacts with only one aldehyde group, a single covalent bond to the glycosyl moiety will be formed. Thus, in preferred multimeric glycoproteins, the following specific bonding schemes may occur and are specifically contemplated:

P₁G.-NH-R₁-S-X-S-R₂-NH-G₂P₂, P₁G.=N-R₁-S-X-S-R₂-N=G₂P₂,

P₁G₁-N=R.-S-X-S-R₂=N-G₂P₂, P₁G₁-NH-R.-S-X-S-R₂-N=G₂P₂, P₁G₁-NH-R.-S-X-S-R₂=N-G₂P₂, P₁G.-N=R₁-S-X-S-R₂-N=G₂P₂,

P₁G₁=N-R₁-S-X-S-R₂-NH-G₂P₂, P.G₁=N-R₁-S-X-S-R₂=N-G₂P₂ , wherein P_χ, G₁, P₂, G₂, R^ R₂, and X are as defined above, except having the appropriate valencies. Where N -R_, or N₂-R₂ represents the residue of an organic hydrazine, the N. or N. nitrogen will be bound directly to another nitrogen atom in ^ or R₂, respectively.

It will be immediately understood that where P and/or P₂ in any of the foregoing formulas are polypeptides having more than one glycosylation site along their respective amino acid chains, additional cross-linking structures of the formula -N -R -S-X-S- R₂-N₂- (where N_χ, N₂, R₂, R_, and X are as defined above) may bridge either the glycopolypeptides P and P or either of those glycopolypeptides and one or more different glycopolypeptides (e.g., P G , ^P ₄ ^G _4/ etc.).

In multimeric glycoproteins prepared according to the process of this invention, the G -N and G -N₂ bonds are formed by reacting an aldehyde group on the terminal sugar of a glycosyl moiety with the amino group of an aminothiol compound, as described above. Similarly, the S-X-S bond(s) will be either disulfide (-S-S-) linkages formed by reacting thiol- functional glycoproteins under oxidizing conditions (e.g., air oxidation), or -X- will represent the residue of a homobifunctional cross-linking agent after reacting with two thiol-functional glycoproteins through the free sulfhydryl groups.

Preferred groups for R. and R include: -CH₂~CH - (derived from cysteamine) ,

-CH-CH„- (derived from cysteine) , and C IOOH

O

II

-CH-(CH₂)₂-C-NH-CH-CH₂- (derived from glutathione)

Preferred multimeric glycoproteins, which will preferably be obtained according to the process of this invention, have the formula

(GP)-NH-CH₂CH₂-S-S-CH₂CH₂-NH-(GP) , (GP)=N-CH CH--S-S-CH_CH.-N=(GP) ,

nd (GP) S - CH₂-CH₂ - N = (GP) wherein (GP) are glycoproteins and most preferably the same glycoprotein. Preferred multimeric glycoproteins of the above formulas -vill include dimers, trimers, tetramers, etc. where (GP) represents sCD4, sLFA3 or sVCAM-lb.

Although homo-multimeric glycoproteins, i.e. multimerics comprising only one type of glycoprotein, are preferred, it is clear that the methods disclosed are equally suitable to produce heteromultimeric glycoproteins in which the cross-linked glycoproteins are different, e.g., different forms of the same glycoprotein or different glycoproteins altogether. If the coupling is performed with two different glycoproteins, it will be appreciated that more than one species of multimeric glycoprotein may result: for example, among the dimeric glycoproteins produced, two homodimers and one heterodimer will be produced. The number of distinct species will rise as the degree of multimerization rises or as the number of distinct glycoprotein "monomers" is increased. These different multimeric species may be separated by conventional means.

EXAMPLE I — PRODUCTION OF MULTIMERIC CD4

THROUGH CHEMICAL CROSS-LINKING

Aldehyde-functional CD4

A form of recombinant soluble human CD4 was obtained from Biogen, Inc. (Cambridge, Massachusetts) . The protein, corresponding to amino acids 1-375 in the primary sequence of natural CD4, was engineered by deleting sequences from the C-terminus of the protein, which encode the transmembrane sequence and cytoplasmic domain, rendering a secreted CD4 in a soluble form. (Fisher et al., 1989; PCT patent application WO 89/01940.)

One hundred μM CD4 was dialyzed against 0.1 M sodium acetate (pH 5.0) at 4°C. The preparation was incubated at 23°C for 1 hour with 1 mM aqueous sodium periodate and immediately desalted on a P6DG (Biorad) column that was equilibrated with 10 mM sodium acetate (pH 5.0), 100 mM NaCl. Aldehyde-functional CD4 was eluted from the column and stored at -70°C. The extent of oxidation was monitored by measuring incorporation of tritiated sodium borohydride. Typically 8-10 aldehydes per rsCD4 molecule were generated.

In subsequent studies, we determined that dialysis into 10 mM sodium acetate (pH 5) , 100 mM NaCl improved recovery of the protein.

To confirm that oxidation did not diminish CD4 function, the ability of the oxidized protein to bind gpl20 was tested in an ELISA format. IMMULON-2® (Dynatech Laboratories, Chantilly, Virginia) plates were coated with gpl20 obtained from American Bio- Technologies, Inc. (Cambridge, Massachusetts), and treated with CD4 or oxidized CD4. Then CD4 binding to gpl20 was determined with a reporter system using 0KT4 antibody (available from Ortho Diagnostics Systems, Raritan, New Jersey) that was conjugated with horseradish peroxidase. There was no difference in binding of CD4 or oxidized CD4 to gpl20. Finally, the oxidized CD4 was subjected to an amino acid analysis to see whether any structural alterations might have been generated by oxidation. There were no apparent differences between the amino acid contents of oxidized and untreated CD4. Construction of Thiol-functional CD4 with Cysteamine

Aldehyde-functional CD4 (0.5 mg/ml) was incubated overnight at 23°C in 50 mM MES, pH 6.5, with 20 mM cystamine dihydrochloride plus 5 mM sodium cyanoborohydride. Samples were reduced with 40 mM DTT for 40 minutes and dialyzed into storage buffer (10 mM sodium acetate pH 5.0, 100 mM NaCl). The extent of modification was monitored with Ellman's reagent. The samples were diluted into 100 μl of 100 mM sodium phosphate pH 8.0, 0.5 mM Ellman's reagent (DTNB) .

After 5 minutes, the absorbance was measured at 410 nm. Samples were calibrated against a standard curve that was developed with reduced glutathione. Both cystamine and glutathione treatments resulted in 3-5 sulfhydryl groups per CD4.

The thiol-functional CD4 preparations were concentrated to 5 mg/ml using a CENTRICON-10 filtration unit and stored at -70°C.

Cross-linking of the Thiol-functional CD4 Multimeric CD4 was generated by chemical coupling with a homobifunctional reagent bis- maleimidohexane (Pierce) . The thiol-functional CD4 (5 mg/ml) was adjusted to pH 6.5 by adding 0.5 M MES to 50 mM. Then it was incubated for 1 hour at room temperature with 100 μM bis-maleimidohexane BMH. This produced BMH cross-linked CD4.

Multimeric forms of the protein were also generated directly by inducing disulfide bond formation by air oxidation. Thiol-functional CD4 as prepared above was subjected to buffer exchange into PBS pH 7.2 on a P6DG desalting gel column or elevated to pH 7.2 with MES and the collected samples were incubated at the elevated pH at 4°C for 16 hours. This produced disulfide-linked CD4.

The products of these reactions were analyzed by SDS-PAGE. In these examples, unless otherwise noted, SDS-PAGE was performed using the Laemmli gel system. Proteins were detected by staining with Coomassie brilliant blue R250, silver staining (Wray et al., 1981) or fluorography as indicated. Prior to electrophoresis, samples were heated at 65°C for 10 minutes in electrophoresis sample buffer (50 mM tris HC1, pH 6.8, 2% SDS, 12.5% glycerol, 0.1% bromophenyl blue) .

Figure 1 shows the separation by SDS-PAGE of crude mixtures of CD4 coupled by air oxidation or by BMH cross-linkers. In both disulfide-linked CD4 (lane c) and BMH cross-linked CD4 (lanes h and i) , multimeric products account for over 50% of the input protein. While we attempted to drive the extent of the reaction by regulating the final protein concentration during the coupling reaction and by regulating BMH concentration, the maximum level of coupling never exceeded 70%.

BMH cross-linked CD4 was fractionated on a SUPEROSE 6 ® gel filtration column (20 ml per hour in Dulbecco's PBS.) Fractions of 1.5 minutes were collected. The results are shown in Figure 2. The top panel (Figure 2A) shows absorbance values at 280 nm for fractions containing CD4. Over 50% of the protein is shifted from CD4 monomer (fractions 33-36) to higher molecular weight multimers (fractions 26-33). The bottom panel (Figure 2B) shows results when aliquots from samples were separated by SDS-PAGE. Peak fractions for monomer (33-35) , dimer (30-31) , and tetramer (28-29) are readily apparent on SDS-PAGE. Under the best conditions we have observed that about 30% of the protein was converted into dimer, and 10% into tetramer. With CD4 there were no trimers formed by coupling, suggesting that the interactive unit of CD4 may be the dimer.

Biological Activity of Multimeric CD4

We isolated dimers and tetramers from the column fractions described above and assayed them for activity in a virus replication assay using HTLV IIIB obtained from Dr. Robert Gallo, NIH. Forty microliters of test sample was mixed with 40 μl of the HIV virus strain IIIB at a titer of 5000 TCID50 units/ml and incubated at 37°C for 1 hour. Twenty microliters of C8166 cells (4 x 10⁴ cells) were added and triplicate 30 μl aliquots were assayed for residual infectious virus. The aliquots were assayed by first diluting the samples with 200 μl of RPMI-1640 medium containing 20% serum in 96-well tissue culture plates and then growing the cells at 37°C. Samples were scored for cytopathic inhibition between days 4-10. The longer incubations ensured that infections had reached a plateau. By setting up CD4 samples that correspond to base-two dilutions of each test preparation, we were able to establish cross-over concentrations and therefore established the minimal concentration at which test proteins block infection.

The results obtained are shown in Table I. The dimer was about four times as active as monomer on a mass basis, while the tetramer was fifteen times as active. On a molar is, these increases correspond to eight and sixty times for the dimer and tetramer, respectively. Table I - ACTIVITY OF CD4 CONSTRUCTS IN HIV-IIIB REPLICATION ASSAY

CD4 TYPICAL BLOCKING CONCENTRATION rsCD4 (monomer) 1.25 μg/ml 25.0 nM rsCD4-dimer -0.3 μg/ml 3.0 nM rsCD4-tetramer 0.07 μg/ml 0.35 nM

EXAMPLE II — PRODUCTION OF MULTIMERIC LFA3 THROUGH CHEMICAL COUPLING

Purification of Soluble and Phosphatidylinositol-Linked LFA3

Stable cell lines producing recombinant soluble LFA3, obtained from Biogen, Inc., were established by transfecting Chinese Hamster Ovary cells with a clone encoding the extracellular domain of human LFA3, i.e., amino acid residues 1-184 of its primary sequence. (Wallner et al., 1987.) CHO transformants expressed 1 mg rsLFA3 per liter of medium. To produce rsLFA3, transfected CHO cells were grown at 37°C in en- Minimum Essential Medium without ribonucleosides or deoxyribonucleosides and supplemented with 10% dialyzed fetal bovine serum. Cells were grown to confluency in a 2.0 L cell factory in complete medium and then shifted into medium containing 2% serum. We collected 1.8 L batches of conditioned medium every third day. Cellular debris were removed by centrifugation and the medium was filtered, concentrated 20-fold by ultrafiltration, and stored at -70°C.

Recombinant soluble LFA3 was affinity purified from the concentrated medium in a single step by immunoaffinity chromatography. rsLFA3 from 900 ml of concentrate was batch loaded overnight onto 5 ml of a 7A6 immunoaffinity resin. The resin was prepared by coupling the 7A6 antibody (α-LFA3, prepared at Biogen, Inc.) to CNBr SEPHAROSE 4B® at 10 mg antibody per ml resin. The resin was collected by centrifugation and used to prepare a chromatography column (1.2 cm i.d. x 3 cm) . The resin was washed sequentially with 30 ml of PBS, 24 ml of triethylamine pH 10.0, 250 mM NaCl and 124 ml of PBS. The rsLFA3 was eluted in twenty fractions of 1 ml each, containing 50 mM glycine (pH 3.0), 250 mM NaCl, and neutralized with HEPES. Peak fractions were identified by SDS-PAGE, combined and stored at -70°C. The purified protein migrated on SDS-gels as a single band with an apparent molecular weight of 50 kD (Figure 3, lane b) . Chromatography on a SUPEROSE 6 gel filtration column revealed a single species with apparent molecular weight of 70 kD. Pi-linked LFA3 was purified from human erythrocytes and from transfected CHO cells expressing the recombinant protein previously described (Wallner and Hession, PCT patent application WO 90/02181, 1990) using a TS2/9 monoclonal antibody affinity column. (Dustin et al., 1989.) Both preparations were subjected to two cycles of binding/elution, first in the presence of 1% NP-40 and second in 1% octyl glucoside. Octylglucoside was removed from the protein by multiple cycles of ultrafiltration in an AMICON® stirred cell and the final product was divided into aliquots and stored at -70°C. Purified PI-LFA3 migrated on SDS-gels as a single band with an apparent molecular weight of 60 kD (Figure 3, lane g) .

The PI-LFA3 was deglycosylated using N-glycanase (Genzyme, Boston, Massachusetts) as follows: The PI-LFA3 was denatured by heating at 65°C for 5 minutes in 0.2% SDS, 10 mM 2-mercaptoethanol, diluted into digestion buffer (130 mM sodium phosphate, pH 8.6, 1% NP-40, 10 mM phenanthroline hydrate, 10% methanol) and incubated overnight at 37°C in the presence or absence of 6 units per ml N-glycanase. The digestion was terminated with SDS-PAGE sample buffer. - 31 -

Multimeric forms of the protein were also generated by inducing disulfide bond formation between thiol-functional rsLFA3 monomers. Samples of thiol- functional rsLFA3 as prepared above were subjected to buffer exchange into PBS pH 7.2 on a P6DG desalting gel column or elevated to pH 7.2 with MES and the collected samples were incubated at the elevated pH at 4°C for 16 hours. This produced disulfide-linked rsLFA3.

In Figure 3, lanes (c) and (i) show the separation by SDS-PAGE of crude mixtures of disulfide-linked rsLFA3 and BMH cross-linked rsLFA3, respectively. In both instances, multimeric rsLFA3 accounted for over 50% of the input protein. As with rsCD4, the reaction yield could be controlled somewhat by regulating the protein concentration during the coupling reaction and by regulating BMH concentration, however, the maximum level of coupling never exceeded 70%.

The following experiments were performed to confirm that the coupling reactions had proceeded through the sugar groups on LFA3. For oxidized samples, proteins were analyzed by SDS-PAGE before (Figure 3, lane (d) ) and after (Figure 3, lanes (e) and (f) ) treatment with 2-mercaptoethanol. Upon reduction with 2-mercaptoethanol all of the cross-linked forms separated into monomers. For BMH cross-linked proteins, we treated the preparation with N-glycanase, assuming that the activity of the glycosidase would not be affected by modifications on carbohydrate structures. As shown in Figure 3 lane (i) (before glycosidase treatment) and lane (j) (after glycosidase treatment) , glycosidase converted all of the cross- linked products into a single form with molecular weight of 25,000 daltons. The same product was - 32 -

generated when erythrocyte-derived PI-LFA3 was treated with N-glycanase (Figure 3, lanes (g) and (h)) .

We also examined multimeric forms of LFA3 size-fractionated on a SUPEROSE 6 FPLC gel filtration column. By quantitating the recovery of counts using ¹²⁵I-labeled BMH cross-linked LFA3, we estimated that after coupling about 90% of the protein was recovered from the gel filtration step. Figures 4A and 4B show elution profiles for soluble LFA3 and PI-LFA3, which exhibit apparent sizes of 70 kD and 600 kD, respectively. This is in agreement with previously reported mass estimates based on sedimentation centrifugation. (Dustin et al., 1989.) Figure 4C shows the elution profile of BMH cross-linked rsLFA3. In the cross-linked sample a series of products was observed that spanned the region detected for the soluble and Pl-forms. Figure 4D shows an SDS-PAGE profile of column fractions across the peak. The positions of monomers (fractions 30-32, lanes 1-n) , dimers (fractions 27-29, lanes i-k), trimers (fractions 25-27, lanes g-i) , and tetramers (fractions 23-25, lanes e-g) are readily apparent from the SDS-gel profile. Higher molecular weight aggregates of undefined size, which were retarded in the stacking gel, were observed in fractions 20-22. Similar results were obtained with air oxidized rsLFA3 (data not shown) .

To confirm that the chemical coupling procedure had not altered the specificity of LFA3 for CD2, we performed binding studies with iodinated LFA3 where each of the species was tested for binding to Jurkat cells. Dimeric, trimeric, and tetrameric rsLFA3 all showed saturable binding that could be blocked with the TS2/18 anti-CD2 monoclonal antibody (a gift from - 33 -

Dr. Tim Springer, Dana Farber Cancer Institute, Boston, Massachusetts). (See also, Dustin et al., 1989.)

Recombinant soluble LFA3 or BHM cross-linked rsLFA3 (5 μg protein) was iodinated with 200 μCi of Nal¹²⁵ (New England Nuclear) using IODOGEN^® (Pierce) according to the supplier's instructions. The iodinated proteins were separated from unincorporated label on a P6DG column that had been equilibrated in PBS with 1 mg/ml bovine serum albumin, aliquoted, and stored at -70°C. I-labeled monomer, dimer, trimer, and tetramer from the labeled cross-linked rsLFA3 (12,000 cpm/ng), were fractionated on a SUPEROSE 6 ® column that was run in PBS, 1 mg/ml bovine serum albumin. Jurkat cells (CD2-expressing) were washed in

RPMI 1640 medium containing 10% fetal bovine serum and counted. Cells were stored on ice for 30 minutes in the presence or absence of TS2/18 antibody (1 μg/10⁶ cells) . Fifty-thousand cells in 400 μl of growth medium were incubated with various samples of ¹²⁵ι- labeled LFA3 for 1 hour at 4°C, washed three times each with 2 ml of growth medium, and counted in a gamma counter. Total counts were determined by counting a 50 μl aliquot of the incubation mix prior to the washing step.

Figure 5 shows the results of binding when 5 x 10⁴ Jurkat cells were incubated with varying concentrations of LFA3 and counts bound determined by pelleting the cells, washing them three times with growth medium, and counting the final pellets in a gamma counter. In each instance, binding was blocked by pretreating the Jurkat cells with TS2/18 antibody to CD2. Under these conditions, binding with soluble LFA3 could not be detected over background. - 34 -

Since the coupled moieties were iodinated at a constant activity per μg protein, the preparations differ in specific activity at a ratio proportional to their valency. One would predict that the binding curves would plateau at heights proportional to the valency. This trend is apparent from the results shown in Figure 5 where dimer, trimer, tetramer, and PI-LFA3 reached saturation at 500, 1000, 2300, and 2800 cpm, respectively. In addition to investigating saturable binding, several sets of binding studies were performed in an attempt to generate binding constants through a Scatchard analysis. However, the plots were nonlinear. Whether the observed deviations are real or represent heterogeneity in the preparations remains to be determined.

Biological Activity of Multimeric rsLFA3 A. Rosetting Assay

To test whether the cross-linked forms of LFA3 were functional, the preparations were evaluated for activity in a rosetting assay using sheep erythrocytes and Jurkat cells. Rosetting in this system is solely dependent on CD2-LFA3 interactions and can be blocked with neutralizing antibodies recognizing CD2 or LFA3, or by soluble LFA3.

The rosetting assay was performed as follows. Jurkat cells were grown in RPMI 1640 medium plus 10% fetal bovine serum. A new stock of cells was used each week to insure c istency in the analyses. The cells were washed three _άmes in growth medium, suspended in growth medium containing 1% PEG-8000 and counted. Sheep red blood cells (Colorado Serum Company, Denver, Colorado) were subjected to the same procedure and stored on ice. Samples containing rsLFA3, multimeric - 35 -

LFA3 or antibody were incubated at 23°C for 20 minutes with Jurkat cells. Red blood cells were added and the samples incubated on ice for 5 minutes. The cells were pelleted by centrifugation (1000 rpm, 1 minute) and further incubated on ice for 2 hours. The samples were diluted with RPMI medium and counted with a phase microscope. Rosetted Jurkats were quantitated as a function of total Jurkats where a value for 100% blocking was taken from controls containing only Jurkats and sheep erythrocytes. For the studies presented, samples in a total volume of 100 μl received 1 x 10⁵ Jurkat cells and 1 x 10⁷ erythrocytes.

Upon testing, we obtained the striking results shown in Figure 6. Each sample showed dose- dependent inhibition of rosetting with the dimer (open triangles) exhibiting an IC_₀ of 2 μg/ml, the trimer (closed circles) with 0.2 μg/ml, and the tetramer (closed diamonds) with 0.07 μg/ml. These measurements on a molar basis represent 30-, 450-, and 1800-fold increases in activity. Recombinant soluble LFA3, recombinant PI-LFA3 and PI-LFA3 from human erythrocytes showed IC₅₀ values in the rosetting assay of 30 μg/ml, 0.07 μg/ml, and 0.07 μg/ml, respectively (data not shown) . The inhibition results demonstrate that the cross-linked adducts not only are functional, but show enhanced activity that is dependent on the valency of the multimer. Interestingly, the trimer was as active as PI-LFA3 octomer in the assay, suggesting that the orientation of the LFA3 in the PI aggregate is not ideal for blocking. Similar results were obtained with multimeric forms of LFA3 produced by air oxidation.

B. T-Cell Activation Assays

One mechanism through which peripheral blood lymphocytes can be stimulated to proliferate is through - 36 -

the activation of the CD2 pathway. We have found that recombinant PI-LFA3 together with o∑Tll. antibody can be used as a trigger for this event, thus providing a biological system for evaluating relative activities of preparations of LFA3.

We performed the assay as follows: Human blood was overlayed onto Ficoll-Plaque (Pharmacia) to separate peripheral blood lymphocytes ("PBLs") from plasma. Cells were washed in growth medium and 1 x 10⁵ cells incubated in 96-well round bottom tissue culture plates. Test solutions were added to give the indicated concentrations in a final volume of 150 μl. Control wells contained only PBLs and medium. Test wells contained anti-Til (an anti-CD2 mouse monoclonal antibody; gift of E sinnerz, Dana Farber Cancer

Institute, Boston, _sachusetts) with various amounts of specific LFA3 preparations. Cells were incubated at 37°C for 3 days. We then added 50 μl of ³H-thymidine (20 μCi/ml) and incubated the cells for 12 hours at 37°C. Cells were lysed and harvested on a SKATRON^® harvester (Skatron, Inc. , Sherering, Virginia) . We quantitated thymidine incorporation in a SKATRON® betaplate scintillation counter.

When PI-LFA3 and BMH cross-linked rsLFA3 dimer, trimer, and tetramer were tested for activity we obtained the results shown in Figure 7. The LFA3 tetramer produced a strong dose-dependent response that produced a maximum level of stimulation of about 20-fold over unstimulated cells (i.e., about 4,000 cpm versus 85,000 cpm). The trimer displayed low level activity, which required over 100 times as much protein as the tetramer to affect T-cell activation. Soluble LFA3 (not shown) and dimer were inactive over the concentration tested. No activation was observed in - 37 -

the absence of α-Tll, or when ct-Tll was substituted for

0-TII₃.

Disulfide-linked LFA3 was inactive. We do not know if the difference in activity results from increased flexibility of the BMH cross-linked material or from spacing arrangements that are a result of the site of BMH attachment.

EXAMPLE III — PRODUCTION OF MULTIMERIC sVCAM-lb

THROUGH CHEMICAL CROSS-LINKING Purification of Recombinant Soluble VCAM-lb

Stable cell lines producing rsVCAM-lb were provided by Biogen, Inc. These cell lines were established by transfecting Chinese Hamster Ovary (CHO- DHFR^") cells with a plasmid encoding an sVCAM-lb, having amino acid residues 1-698 of its primary sequence. (Hession et al., PCT patent application WO 90/13300.) The cells were grown at 37°C in α- Minimum Essential Medium without ribonucleosides or deoxyribonucleosides and supplemented with 10% dialyzed fetal bovine serum. Cells were grown to confluency in a 2.0 L cell factory in complete medium and then shifted into medium containing 2% serum. We collected 1.8 L batches of conditioned medium. Cellular debris were removed by centrifugation and the medium was filtered, concentrated 20-fold by ultrafiltration, and stored at -70°C.

Recombinant sVCAM-lb was purified from the concentrated conditioned medium by immunoaffinity chromatography on a 4B9 affinity column. 4B9 (a gift of Biogen, Inc.) is a monoclonal antibody that recognizes VCAM-lb. The protein migrated by SDS- polyacrylamide gel electrophoresis as a single diffuse band with an apparent mass of 100 kD. - 38 -

Recombinant DNA expression vectors encoding VCAM-1 and monoclonal antibodies recognizing VCAM-1 are available from British Bio-technology, Abingdon, Oxon, England.

Preparation of Aldehyde-Functional rsVCAM-lb

Recombinant sVCAM-lb was concentrated to 1 mg/ml in a CENTRICON-10® filtration unit (Amicon) . The concentrate was dialyzed overnight at 4°C against 100 mM sodium acetate (pH 5.0). Sodium periodate was added to 0.12 mM and the sample was incubated at room temperature for 1 hour. After oxidation, the sample was dialyzed against 10 mM sodium acetate (pH 5.0), 100 mM NaCl. Aldehyde-functional sVCAM-lb was incubated overnight at room temperature with 50 mM MES (pH 6.5), 20 mM cystamine dihydrochloride and 5 mM sodium cyanoborohydride. The samples were treated with 40 mM DTT for 1 hour at room temperature. The reaction mixture was dialyzed against 10 mM sodium acetate (pH 5.0), 100 mM sodium chloride with three buffer changes. The preparation was concentrated to a final protein concentration of 2.5 mg/ml in a CENTRICON-10® filtration unit, aliquoted and stored at -70°C.

Cross-linking of the Thiol-functional sVCAM-lb

We generated multimeric sVCAM-lb by chemical coupling with the homo-bifunctional cross-linking agent, BMH. Thiol-functional sVCAM-lb was adjusted to pH 6.5 by adding MES to 50 mM, and then incubated for 2 hours at room temperature with 300 μM BMH. This produced BMH cross-linked sVCAM-lb. For trial reactions, cross-linking was stopped by adding electrophoresis sample buffer and the products were analyzed by SDS-PAGE. For preparative runs, excess BMH was removed on a P6DG desalting column - 39 -

that was equilibrated in PBS. Samples were aliquoted and stored at -70°C.

Figure 8 shows results from a trial reaction in which we cross-linked thiol-functional sVCAM-lb with varying concentrations of BMH. This figure demonstrates that in the presence of BMH, thiol- functional sVCAM-lb was cross-linked into multimeric sVCAM-lb that, by molecular weight, represent sVCAM-lb dimers ("(VCAM)₂") and (^«'(VCAM)₄") tetramers. The optimal concentration of BMH to produce multimeric sVCAM-lb was 200 μM. This concentration converted approximately 80% of the starting material into higher molecular weight adducts.

We have compared the binding affinity of BMH cross-linked sVCAM-lb and rsVCAM-lb for leukocytes as a function of their ability to block the binding of Ramos cells (which express VLA-4) to microtiter plates coated with rsVCAM-lb. Preliminary results indicate that multimeric sVCAM-lb may be about four times as effective as monomeric rsVCAM-lb in blocking Ramos cell binding.

While we have hereinbefore described a number of embodiments of this invention, it is apparent that our basic embodiments can be altered to provide other embodiments which utilize the processes and compositions of this invention. Therefore, it will be appreciated that the scope of this invention includes all alternative embodiments and variations which are defined in the foregoing specification and by the claims appended hereto; and the invention is not to be limited by the specific embodiments which have been presented herein by way of example. - 40 -

REFERENCES CITED

ARTICLES

Bevilacqua, M.P., et al., "Identification of an Inducible Endothelial-Leukocyte Adhesion Molecule", Proc. Natl. Acad. Sci.. USA. 84. pp. 9238-42 (1987)

Carr, S.A. et al., "Protein and Carbohydrate Structural Analysis of a Recombinant Soluble CD4 Receptor by Mass Spectrometery", J. Biol. Chem. , 264, pp. 21286-95 (1989)

Chao, B.H., et al., "A 113-amino Acid Fragment of CD4 Produced in Escherichia coli Blocks Human Immunodeficiency Virus-induced Cell Fusion", J. Biol. Chem.. 264. pp. 5812-17 (1989)

Dustin, M.L., et al., "Anchoring Mechanisms for LFA-3 Cell Adhesion Glycoprotein at Membrane Surfaces", Nature. 329. pp. 846-48 (1987)

Dustin, M.L., et al., "Correlation of CD2 Binding and Function Properties of Multimeric and Monomeric Lymphocyte Function-Associated Antigen 3", J. Exp. Med.. 169. pp. 503-17 (1989)

Elices, M.J., et al., "The Integrin VLA-4 Mediates Leukocyte Adhesion to VCAM-1 on Activated Endothelium", Cell. 60_f P . 577-84 (1990)

Fisher, R.A. , et al., "HIV Infection Is Blocked in Vitro by Recombinant Soluble CD4", Nature. 331. pp. 76-78 (1988)

Hession, CA. , et al., "Cloning of an Alternate Form of Vascular Cell Adhesion Molecule 1 (VCAMl)," J. Biol. Chem.. 266. pp. 6682-85 (1991) . ^•

Lambert, J.M. , et al., "Purified Immunotoxins That Are Reactive with Human Lymphoid Cells", J. Biol. Chem.. 260. pp 12035-41 (1985)

Liao, T.H., et al., Modification of Sialyl Residues of Sialoglycoprotein(s) of the Human Erythrocyte Surface", J. Biol. Chem.. 248. pp. 8247-53 (1973)

Maddon, P.J., et al., "The T4 Gene Encodes the AIDS Virus Receptor and Is Expressed in the Immune System and the Brain", Cell, 47, pp. 333-48 (1986)

Moroney, S.E- , et al., "Modification of the Binding Site(s) of I _tins by an Affinity Column Carrying an - 41 -

Activated Galactose-terminated Ligand", Biochem.. 26. pp. 8390-98 (1987)

Nemerow, G.R. , et al. , "Soluble Recombinant CR2 (CD21) Inhibits Epstein-Barr Virus Infection", J. Virol.. 64₍ pp. 1348-52 (1990)

Osborn, L. , et al., "Direct Expression Cloning of Vascular Cell Adhesion Molecule 1, A Cytokine-Induced Endothelial Protein that Binds to Lymphocytes", Cell. 5£, pp. 1203-11 (1989)

Ramakrishnan, S., and L.L. Houston, "Comparison of the Selective Cytotoxic Effects of Immunotoxins Containing Ricin A Chain or Pokeweed Antiviral Protein and Anti- Thy 1.1 Monoclonal Antibodies", Cancer Res.. 44. pp. 201-08 (1984)

Raso, V. and M. Basala, "A Highly Cytotoxic Human Transferin-Ricin A Chain Conjugate Used to Select Receptor-modified Cells", J. Biol. Chem.. 259., pp. 1143-49 (1984)

Reichlin, M. , "Use of Glutaraldehyde as a Coupling Agent for Proteins and Peptides", Meth. in Enz. , 70. pp. 159-65 (1980)

Sambrook, J. , et al., Molecular Cloning. A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York (1989)

Sanchez-Madrid, F., et al. "Three Distinct Antigens Associated with Human T Lymphocyte-mediated Cytolysis: LFA-1, LFA2, and LFA3", Proc. Natl. Acad. Sci.. USA. 79. 7489- (1982) .

Srinivasachar, K. and D.M. Neville, Jr., "New Protein Cross-linking Reagents That Are Cleaved by Mild Acid", Biochem.. 28. pp. 2501-09 (1989)

Wallner, B.P., et al., "Primary Structure of Lymphocyte Function-Associated Antigen 3 (LFA-3) the Ligand of the T Lymphocyte CD2 Glycoprotein", J. Exp. Med.. 166. pp. 923-32 (1987)

White, J.M. , and D.R. Littman, "Viral Receptors of the Immunoglobulin Superfamily", Cell, 56, pp. 725-28 (1989)

Wray et al., "Silver Staining of Proteins in Polyacrylamide Gels", Anal. Biochem. _f 118, pp. 197-203 (1981) - 42 -

Yarchoan, R. , et al., "Phase I Study of rCD4-IgG Administered by Continuous Intravenous (IV) Infusion to Patients with AIDS or ARC", Abstracts: Sixth International Conference on AIDS. San Francisco, California, 3 , Abstr. # SB479 (1990)

Patent Publications

Hession, CA. , et al., "Endothelial Cell-Leukocyte Adhesion Molecules (ELAMs) and Molecules Involved in Leukocyte Adhesion (MILAs)", PCT patent application WO 90/13300 (28 November 1990)

Wallner, B.P. and C Hession, "DNA Sequences, Recombinant DNA Molecules and Processes for Producing PI-Linked Lymphocyte Function Associated Antigen-3", PCT patent application WO 90/02181 (8 March 1990)

Fisher, R.A., et al., "DNA Sequences, Recombinant DNA Molecules and Processes for Producing Soluble T4 Proteins", PCT patent application WO 89/01940 (9 March 1989)

Wallner, B.P., et al., DNA Sequences, "Recombinant DNA Molecules and Processes for Producing Lymphocyte Function Associated Antigen-3", WO 88/09820 (15 December 1988)

The pertinent disclosures of the foregoing PCT publications relatinc, to CD4, LFA3, PI-LFA3, VCAM-l and VCAM-lb are incorporated herein by reference.

Claims

- 43 -CLAIMS We claim:

1. A process of producing multimeric glycoproteins comprising the steps of:

(1) oxidizing a solution including glycosylated polypeptides to produce aldehyde- functional glycoproteins;

(3) chemically coupling two or more thiol-functional glycoproteins by oxidizing the glycoproteins to form disulfide linkages or by reacting the glycoproteins with a homobifunctional cross- linking agent.

2. The process according to claim 1, wherein the homobifunctional cross-linking agent is bis-maleimidohexane.

3. The process according to claim 1 or 2, wherein the aminothiol compound is cystamine dihydrochloride.

4. The process according to claim 3, wherein the glycoprotein is an adhesion protein or a portion thereof.

5. The process according to claim 4, wherein the glycoprotein is selected from the group consisting of viral receptors and cell receptors.

6. The process according to claim 5, wherein the glycoprotein is CD4. - 44 -

7. The process according to claim 6, wherein the CD4 is an rsCD4.

8. The process according to claim 5, wherein the glycoprotein is LFA3.

9. The process according to claim 8, wherein the LFA3 is an rsLFA3.

10. The process according to claim 5, wherein the glycoprotein is VCAM-1 or VCAM-lb.

11. The process according to claim 10, wherein the glycoprotein is an rsVCAM-lb.

12. The process according to claim 1 or 2, wherein the glycoprotein is selected from the group consisting of glycosylated bacterial immunogens, viral immunogens and parasitic immunogens.

13. The process according to claim 1 or 2, wherein the aminothiol compound is selected from the group consisting of cysteine, glutathione, peptides of up to 20 amino acid residues containing at least one cysteine residue, C, , alkanes having amino and thiol groups and organic hydrazines containing reducible sulfur.

14. A multimeric glycoprotein prepared according to the process of claim 1.

15. A multimeric glycoprotein of the formula ^■ R₁ - S - X - S - R₂ - N₂ - G_{2 2}, wherein:

P and P represent the same or different polypeptides; - 45 -

N and N represent, independently, a linkage between G. and R or G, and R₂, respectively, through a secondary or tertiary amine nitrogen atom, i.e., N and N_ may either represent an imino (-NH-) or an aza (=N- or -N=) linkage between the glycosyl moieties (G , G ) and the R and R moieties;

S is sulfur; and

X is a direct bond or is a divalent organic radical of from 1-16 carbon atoms.

16. A multimeric glycoprotein according to claim 15, wherein X is a direct bond.

17. A multimeric glycoprotein according to claim 15, wherein X is a divalent organic radical derived from a homobifunctional cross-linking agent.

18. A multimeric glycoprotein according to claim 17, wherein X is

•CH-CH₂- 46 -

19. A multimeric glycoprotein according to any one of claims 15-18, wherein R_λ and R₂ are -CH₂CH₂~.

20. A multimeric glycoprotein according to claim 19, wherein P and P are adhesion proteins or portions thereof.

21. A multimeric glycoprotein according to claim 20, wherein P and P₂ are selected from the group consisting of viral receptors and cell receptors.

22. A multimeric glycoprotein according to claim 21, wherein the glycoprotein is CD4.

23. A multimeric glycoprotein according to claim 22, wherein the CD4 is an rsCD4.

24. A multimeric glycoprotein according to claim 20, wherein the glycoprotein is LFA3.

25. A multimeric glycoprotein according to claim 24, wherein the LFA3 is an rsLFA3.

26. A multimeric glycoprotein according to claim 21, wherein the glycoprotein is VCAM-1 or VCAM-lb.

27. A multimeric glycoprotein according to claim 26, wherein the glycoprotein is an rsVCAM-lb.

28. A multimeric glycoprotein according to claim 15, wherein P. and P₂ are selected from the group consisting of glycosylated bacterial immunogens, viral immunogens and parasitic immunogens. - 47 -

29. A thiol-functional glycoprotein of the formula P G -N -R -SH, said glycoprotein being capable of cross-linking under oxidizing conditions to form disulfide cross-linked multimeric glycoproteins, wherein:

P represents a polypeptide;

G represents glycosyl moiety on the polypeptide P^*

N represents a linkage between G and R through a secondary or tertiary amine nitrogen atom, i.e., N. may either represent an imino (-NH-) or an aza (=N- or -N=) linkage between the glycosyl moiety G and the R moiety;

R represents organic radicals of 1-16 carbon atoms or polypeptides of up to 20 amino acids; and

S is sulfur.

30. The thiol-functional glycoprotein according to claim 29, wherein R is -CH CH -.

31. A process for producing multimeric sugar complexes comprising the steps of:

(1) oxidizing a solution including sugar complexes to produce aldehyde-functional sugar complexes;

(2) reacting the aldehyde-functional sugar complexes with an aminothiol compound to produce thiol-functional sugar complexes; and

(3) chemically coupling two or more thiol-functional sugar complexes by oxidizing the sugar complexes to form disulfide bonds or by reacting the sugar complexes with a homobifunctional cross-linking agent.