CA2006475A1 - Method for preparing water soluble polypeptides - Google Patents

Abstract

METHOD FOR PREPARING WATER SOLUBLE POLYPEPTIDES
Abstract of the Disclosure Methods are provided for the preparation in recombinant host cells of biologically active soluble variants of discretely encoded, heteromultimer polypeptide receptors. Such variants are synthesized by the secretion from recombinant transformants of transmembrane-modified heteromultimer receptors. Preferred receptors are extracellular matrix, cell surface, or plasma protein-binding receptors such as GPIIb-IIIa.

Description

7~
, hlETHOD FOR PRE~ARING WAT~R SOLUBLE POLYP~PTIDES
This invention is concerned with the preparation of complex soluble receptors. ~n particular it is directed to the synthesis of recombinant receptors for cell matrix or plasma proteins.
Cellular membranes contain polypeptides which are lodged in the lipid bilayer. Such polypeptides colltain a domain which anchors the protein in the cell membrane, ahydrophobic transmembrane domain, together in many instances with a C-terminal cytoplasmic sequence. In general, these polypeptides are single chain molecules or are multiple chain molecules derived from an ancestral sin~le chain expression product by post-translational proteolytic processing. Such multiple chain polypeptides usually are covalently linked by disulfide bonds. However, some of these polypeptides are noncovalentlyassociated with one another by salt bridges, Van der Waals forces, hydrophobic interactions and the like, and in such cases this association of polypeptide subunits into a larger aggregate is a prerequisite for biological activity.
The biological activity of such membrane-bound, multiple subunit molecules is varied, but in general reflects a receptor or binding function. Receptors serve to signal the cell regarding a condition or substance in the exterior environment of the cell, they serve to in~ernalize an extracellular substance, or they function to attach cells ~o one another, to extracellular matrix substances, cell surface or plasma proteins.
A further subclass of membrane bound multiple subunit polypeptides are those in which each subunit is different, i.e. is not substantially homologous, and is encoded by a discrete gene. Such polypeptides are termed "MSP (multiple subunit polypeptides) for the purposes of this invention. Numerous examples of such polypeptides or receptors are known, but the most substantial group is the class of cell surface receptors for extracellular matrix molecules, some of which have currently been identified and DNA encoding them ~ cloned (see for example, E~uclc et al., "Ann. Rev. Cell Biol. ~:179 11987] ar,d Ruoslahti et al., "Science 23~: 491 [1987].) Of particular interest is the platelet glycoprotein Ilb-llla, a platelet mernbrane-bound receptor involved in platelet aggregation and which binds to fibrinogen, fibronectin, vi~ronectin and von Willebrand factor. The two subunits constituting this receptor have been cloned (Fitz~erald et al. Biochemistry~ ~:8158 11987] and Fit7gerald ~t al. J. Biol. Chem.
~62(9):3936 [ 1987~). Bennett et al. reported expressisn of the GPIIlb subunit in Cos- I cells, but the subuni~ was not found on the cell membrane (AHA 61st Scientific Sessions, Nov.
15, 1988). Bennett et al. suggested that membrane localization might require the formation of the Ilb-Illa eomplex. There was no teaching or suggestion that a recombinant,membrane-bound GPllb-llIa, even if it could be made, ~vould bind to its proper ligands, e.g., fibrinogen. In addition, an oral disclosure by Frelinger et al. at the same meeting purported to describe the transient express;on of full length GPIlb-llla on an unidentified recombinant cell surface; no other information was provided relating to the manner in which expression was allegedly obtained.

Corbi et al. orally reported the transient e~pression of functional full length LFA-I in COS cells in September 1988 at the Titisee Symposium sponsored by Boehringer Ingelheim.
Membrane-bound MSPs present difficulties in pur;fication and stability since the5 hydrophobic domains tend to induce the MSPs to micelles or aggregates. A form of these receptors is needed that ;s soluble, particularly in body fluids such as blood and in pharmacological excipients such as saline, without forming multiple molecular aggregates beyond proper heterodimer assembly. Accordingly, it is an object herein to synthesize such MSP forms.
It is another object to produce soluble forms of the GPIIb-lIla receptor which are capable of properly binding their normal ligands.
It is a further objecc to express &Pllla in recombinant cell culture.
It is an additional object to produce high yields of GPllb-lTla from recombinant cell culture.
These and other objects will be apparent from consideration of this application as a whole.
In accordance with this invention, a method is provided for the preparation of asecreted analogue of a cell membrane-bound multiple subunit polypeptide (MSP), each subunit of which is encoded by a discrete gene, comprising 1) introducing into the nucleic 20 acid encoding each of the subunits a mutation encoding an amino acid sequence variant of the MSP that renders the MSP no longer capable of becoming lodged in a lipid bilayer, and 2) transfecting a host cell with the nucleic acid of step 1, 3) culturing the host cell of step 2 and 4) recovering from the host cell culture biologically active soluble MSP. Also in accordance with this invention, nucleic acid and expression vectors are provided which 25 encode an amino acid sequence variant of an integrin chain, in particular a variant in which the transmembrane domain of the inte~rin chain is modified so that it is no longer capable of becoming lodged in the cell membrane.
Also provided is a method for the preparation of GPIlb-llla comprising transforming a permissivç host cell with nucleic acid encoding GPllb-llla and culturin~ the host cell until 30 GPllb-llla accumulates in the cell membrane.
In specific embodiments, the objects of this ;nvention are accomplished by providing a biologically active MSP amino acid sequence variant selected from the group consisting of (a) an MSP amino acid sequence variant having an inactivated membrar~e anchor dornain and (b) a polypeptide comprising an h~SP extracellular domain fused to the sequence of a 3S polypeptide which is different from the MSP, this latter, for example, selected from an immuno~en or a protein with a lon8 plasma half life such as an immunoglobulin constant domain.
In another embodiment, MSP amino acid residues or carbohydrate substituents of MSPs or MSP analogues otherwise described herein are derivatized by covalçnt modification 40 or are conjugated to nonproteinaseous polymers such as polyethylene glycol to produce an MSP derivative which exhibits improved circulatory half life.

,J~ 36~75 In particular embodimen~s a polypeptide comprising a biologically active extracellul3r domain of an integrin is fused at its C-terminus to an immunoglobulin constant domain, or is linked to an immunogenic polypeptide.
The MSP variants provided herein are purified and formulated in pharmacologically S acceptable vehicles for diagnostic or ;: reparatory utility or in vivo use in the modulation of cell adhesion.
Figs. Ia- If depict the amino acid and nucleotide sequence of a secreted form of the GPllb subunit of the MSP GPllb-llla. The signal processing site for the heavy and light forrns of this subunit are designated, respectively, with arrow-H and arrow-L.
Figs. 2a-2d depict the amino acid and nucleotide sequence of a secreted form of the GPllla subunit of the MSP GPIlb-llla. The si~nal processing site is designa~ed with an arrow Fig. 3 depicts a comparison of the native and redesigned nucleic acid sequences at the 5' end of the C;PIIla gene.
An MSP is defined herein to be a multichain polypeptide, at least one chain of which is ordinarily anchored in a cell membrane and at least two chains of which are discretely encoded. MSPs ordinarily contain at least two distinct chains, two of which are lodged directly in the cell membrane. One or more additional chains may be covalently or noncovalently bound to the MSP chains ordinarily lodged in the cell membrane, but the ~0 additional chains may not themselves be anchored in the membrane. Such chains typically result from the post-translational processing of a single chain that becomes membrane anchored. Discretely encoded subunits are those which do not result from the posttrans-lational processing of a single translated protein, and their amino acid sequences are not homologous (i.e. the sequences of the subunits are not the same, and they do not assemble 25 in nature into dimers or multimers of the same polypeptide). Instead, they are produced " by the translation of independent mRNAs or polycistronic messages. Thus, the nucleic acids encoding MSP polypeptides ordinarily are found in nature under the control of different promoters and osher transcription control sequences. MSPs include principally cell surface receptors for extracellular matrix molecules, also defined as cellular adhesion 30 receptors. Many of these receptors and their ligands, such ligands including the e~tracellular matrix molecules and plasma proteins such as fibrinogen as well as cell surface proteins such as l-CAh~l, are central to cellular adhesion phenomena involved in wound healing, morphogenic mobility, developmentally unrelated cellular migrations, hemostasis and metastasis. These cellular adhesiorl receptors are identified by functional and structural 3~ features. Functionally, they typically bind to polypeptides incorporatin~ the sequence RGD, from which they are dissociated by competition with other polypeptides containing the RGD sequence such as the peptides RGDS or RGDV. Also, they frequently require a divalent cation such as calcium for ligand binding. MSPs may or may not include members of the imrnunoglobulin superfamily such as the T cell receptor. A ~roup of MSPs involved 40 in cell surface intracellular adhesive interactions have been designated integrins (see Buck et al., rAnn. Rev. Cell Biol." 3:179-205 [1987]).

Structurally, such cellular adhesion receptors belong to a supergene family of multimers in ~vhich a first sin21e-chain polypeptide or disulfide cross-linked multi-chain polypep~ide (a-chain) is noncovalently associated with a second and different polypeptide (designated a B-chain), thereby for,ning a heseromultimer. The a-chains of these receptors S are quite diverse in terms of their amino acid sequence, and include the a~ subunit of avian integrin (band 1 ); ~ 2. and ~ of VLA I, 2 and 4; D~3 of VLA 3 and avian integrin (band 2); C~F f VLA 5 and the fibronectin receptor; C~L of LFA-I; c~ of Mac-l; ~XX f plS0,95;
L of GPllb; and aV of vitronectin. The û-chains typically fall into three classes, B1 (avian integrin [band 3]; fibronectin receptor and VLA), ~ (LFA-I/Mac-l; pl50,95) and 0 ~3 (GPllb-llla and vitronectin receptor), the members of each ~-class being substantially homologous or identical. It is preferred that the MSP selected contain the two (or more) chains which ordinarily associate with one another in nature since non-naturally occurring heteromers may not form complexes.
Each chain of an MSP is expressed in its native environment as a preprotein comprising a secretion signal which is processed during the extracellular orientation of the receptor. Also, at least one chain of each subunit will have a hydrophobic anchor containing a polypeptide sequence serving as a site for covalent addition of lipid, e.g.
phospholipid, or a domain located in the C-terminal portion of the polypeptide and containing about from 10 to 30 predominantly hydrophobic residues such as phe, leu, ile, val, met, gly and ala. Such membrane anchoring sequences or domains will be collectively referred to herein as membrane anchor domains. A short hydrophilic cytoplasmic domain, on the order of 10 to 100 residues, usually is found C-terminal to transmembrane domains.
The t0rm subunit should be understood to mean polypeptide chain; it does not refer to domains or functional subregions of a given polypeptide chain.
(~ertain MSPs share o~her structural features, îor example, wherein one subunit of . the receptor contains cysteine-rich tandem amino acid sequence repeats in which greater than about 80% of the cysteine residues are alignable within about two residues of the cysteine residues of the tandem repeats of GPllla, wherein one subunit has the consensus N-terminal sequence Tyr/Phe/Leu-Asn-Leu-Asp, or one subunit contains an amino acid domain having substantial sequence homolo~y to the calmodulis~ calcium binding site.
Also included within the scope of MSPs are those receptors which are homologous to the above-described men-bess of the integrin superfamily. Homologous, as defined herein, msans having the sequence of a polypeptide of a member of the integrin superfamily which at least has substantially the san-e amino acid sequence homology to a 3~ known member of the saperfamily as any pressntly known member has to any other known member. l`ypically, homologous means haYiY~g greaeer than about 40% amino acid homology after alignin8 sequences for maximum homology, but not taking into account conservative substitutions.
This invention in part is based upon the discovery that discretely encoded MSPs,when modified to eliminate their ability to insert into the host cell membrane, nonetheless are fully assembled and secreted in biologically active form by recombinant host cells.

'~3(~6~75 Recombinant host cells secrete the subunits in correct association with one another such that the assembly exhibits the biological activity oî the e:~tracellular domain of the native MSP, despite the fact that proper association of the subunits is no longer facilitated by juxtaposition in the cell membrane. Further, proper assembly has been obtained even when the MSP sequences have not been fused to multimer-forming polypeptides, i.e. it has been found that MSPs will properly associate even without the aid of extraneous cross-linking polypeptides such as immunoglobulin chains.
Biolo~ical activity is clefined in terms of the ability of the secreted MSP to qualitatively bind the ligand ordinarily bound by the MSP in its native environment, although it will be appreciated that the kinetics or other quantitative characteristics of ligand binding by the secreted MSP may vary from those of the native cell bound MSP. While secreted MSP most likely will retain many functional immune epitopes capable of cross-reacting with antibody raised against the native MSP, this alone is not enough for the secreted MSP to exhibil biological activity as defined herein; ~biologically active" secreted MSP must exhibit the ability to bind to its ligand as well. However, it will be understood that not all MSP
produced in accord with this invention need to exhibit biological activity in the sense defined here. Such biologically inactive but, for example, immunologically active MSP
analogues find use in diagnostic assays, in raising antibodies against MSP, or in the purification of antibodies to MSP.
This invention is particularly concerned with amino acid sequence variants of MSPs.
Amino acid s0quence variants of MSPs are prepared with various objectives in mind, including increasing the affinity of the MSP for its binding partner, facilitating the stability, puriricatiosl and preparation of the MSP (including enhanced water solubility and reduced membrane affinity), increasing its plasma half life, improving therapeutic efficacy as described above, introducing additional functionalities and lessening the severity or s occurrence of side effects during therapeutic use of the MSP. Amino acid sequence Yariants of MSPs fall into one or a combination of the following classes: insertional, substitutional or deletional ~ariants. Each MSP variant or analogue will have one inactivated rnembrane anchor domain, and this will be accomplished by insertion, substitution or deletion, but these variants optionally comprise additiorlal mutations that are involved in other than inactivating the membrane anchor domain of one chain of the native MSP.
Insertional amino acid sequence variants are those in which one or more amino acid residues extraneous to the MSP are introduced into a predetermined site in the MSP
including the C or N termini. Such Yariants are referred to as fusions of the MSP and a polypeptide containin8 a sequence which is other than that which is normally found in the MSP at the inserted position. S~veral ~roups of fusions are contemplated herein.Immunologically active M5P fusions comprise an MSP and a polypeptide containing a non-MSP epitope. The non-MSP epitope is any immunolo~ically competent polypeptide, i.e., any polypeptide which is capable of eliciting an immune response in the animal to which the fusion is to be adrninistered or which is capable of being bound by an antibody raised against the non-MSP polypeptide. Typical non-MSP epitopes will be those which ~30~75 are borne by allergens, autoimmune epitopes, or other potent immunogens or antigens recognized by pre-existing antibodies in the fusion recipient, including bacterial polypeptides such as trpLE, beta-galEIctosidase, viral polypeptides such as herpes gD
protein, and the like. Immunogenic fusions are procluced by cross-linking in Yi~rO or by 5 recombinant cell culture transformed with DNA encoding an immunogenic polypeptide.
It is preferable that the immunogenic fusion be one in which the immunogenic sequence is joined to or inserted into the MSP or fragment thereof by a peptide bond(s). These products therefore consist of a linear polypeptide chain containing MSP epitopes and at least one epitope foreign to the MSP. It will be understood that it is within the scope of this 10 invention to introduce the epitopes anywhere within the MSP molecule or fragment thereof.
Such fusions are conveniently made in recombinant host cells or by the use of bifunctional cross-linking agents. The use of a cross-linking agent to fuse the MSP to the immunogenic polypeptide is not as desirable as a linear fusion because the cross-linked products are not as easily synthesized in stTu~turally homogeneous form.
These immunogenic insertions are particularly useful when formulated into a pharmacologically accep~able carrier and administered to a subject in order to raise antibodies against the MSP, which antibodies in turn are useful in diagnostics or in purification of MSP by immunoaffinity techniques known per se. Alternatively, in the purification of MSPs, binding partners for the fused non-MSP polypeptide, e.g. antibodies, 20 receptors or ligands, are used to adsorb the fusion from impure admixtures, after which the fusion is eluted and, if desired, the MSP is recovered from the fusion, e.g. by enzymatic cleava~e.
Other îusions, which may or may not also be immunologically active, include fusions of the mature MSP sequence with a signal sequence heterologous ~o the MSP, fusions of 25 transmembrane-modified MSPs (including se~uence deletions or modifications so that the MSP could not lodge in the cell membrane), for example, to polypeptides having enhanced plasma half life (ordinarily >about 20 hours) such as immunoglobulin chains or fragments thereof which confer enhanced plasma half life.
Signal sequence fus;ons are employed in order to more e~peditiously direct the 30 secretion of the MSP. The heterologous si~nal replaces the nati~e MSP signal, and when the resulting fusion is recopnized, i.e. processed and cleaved by the host cell, the MSP is secreted. Signals are selected based on the intended host cell, and may include bacterial yeast, mammalian and viral sequetlces. The native MSP sipnal or the herpes gD
glycoprotein signal is suitable f~r ~Ise in mammalian expression systems.
Plasma proteins which have enhanced plasma half-life longer than that of solubleforms of MSPs having modified membrane anchor domains include serum albumin, immunoglobulins, apolipoproteirs, and transferrin. Preferably, the MSP-plasma protein used for the fusion is not significantly immunogenic in the animal in which it is used (i.e., it is homologous to the therapeutic target) and the plasma protein does not cause undesirable 49 side effects in patients by virtue of its normal biological ac~ ity.

36~5 In a specific embodiment the MSP extracellulas domain is conjupated with an immunoglobulin cons~ant region sequence. Immunoglobulins and certain variants thereof are known and many have been prepared in recombinant cell culture. For example, see U.S.
Patent 4,74~,055; EP 256,654; Faulkner e~ al., Nature 298:286 (1982); EP 120,694; EP
125,023; Morrison, J. lmmun. ~:793 (19791; Kohler er al., P.N.A.S. USA 77:2197 (1980);
Raso e~ al., Cancer Res. 41:2073 (1981); Morrison et al., Ann. Rev. Immunol. 2:239 (1984);
Morrison, Science 229:1202 (1985); Morrison et al., P.N.A.S. USA 81:6851 (1984); EP
255,694; EP 266,663; and WO 88/03559. Reassorted isnmunoglobulin chains also are known.
See for e~sample U.S. patent 4,444,878; WO 88/03565; and EP 68,763 and references cited therein. See also Gascoigne e~ al., P.N.A.S. USA 84:2936-2940 (May, 1987), EP 325,224, and Thesis of Andrew Scott Peterson (Harvard University; de8ree awarded November 22, 1988).
Ordinarily, the extracellular domains of MSPs are fused C-terminally to the N-terminus of the constant region of an immunoglobulin in place of the variable region(s) thereof, retaining at least functionally active hinge, CH2 and CH3 domains of the constant region of an immunoglobulin heavy chain. Two forms of such fusions are embraced herein.
In one, the extracellular domains of two or more ordinarily membrane-bound MSP chains are fused N or C terminally to immunoglobulin constant regions ~heterofusion), while in the other form only one chain of the MSP is fused to a constant region (monofusion). The heterofusions include fusions with either light or heavy chain constant regions, or both.
The heterofusion is produced by transforming a host cell with DNA encoding the light chain fusions, the heavy chain fusions or both. For example, transfection with DNA
encoding one MSP chain fused to a heavy chain constant region and the other MSP chain fused to a light chain constant region will result in heterotetramers or heterodimers bearing light and heavy chain fusions with MSP chains. These are not as desirable as monofusions `~ since they are not as likely to be biologically active. Note that monofusions may contain more than one fused chain, but in these cases the MSP chain will always originate with the same subunit.
Monofusions are immunoglobulin variants in which one chain of an MSP is fused toa heavy or light chain (or constant domain thereof), while the remaining chain(s) of the MSP are not fused to an immunoglobulin but rather are associated with the fused chain in substantially the fashion as is normally the case with the native MSP. Typically, both the fused and unfused MSP chains in monofusions will be variants in which the membrane anchor domains are msdified so as to not lodge in the membrane, most commonly where the membrane anchor domain of one MSP chain is deleted, and ;n the other the membrane anchor domain is deleted and then the remaining e~tracellular region fused at its N-terminus to the C-terminlls of an immuno~lobulin constant domain. The MSP chain or its fragment is fused to either a light chain or a heavy chain, but preferably a heavy chain.
If the MSP only contained one membrane anchored chain then the remaining chain(s) will typically have their native sequence.

It may be desirable to produce mono-or polyfusions having immunoglobulin antigenbinding capability as well as the capacity to bind the MSP ligand. Such products are made by transforming the host cells with DNA erlcoding light and heavy chain capable of binding an antigen (or are selected to already produce light chain) together with the light and/or 5 heavy chain MSP fusion and the unfused MSP chain(s) (in the case of monofusions). This will yield constructs, for e1tample, havin~ the normal structures of immunoglobulins except that one or both light-heavy arms of the immunoglobulin will comprise a fusion with one chain of the MSP which in turn is assembled (covalently or noncovalen~ly) with the remaining chain(s) of the MSP.
In those instances in which the fusion transformants also produce (or are transformed to produce) immunoKlobulin chains not fussd to an MSP subunit, the immunoglobulin variable domains may have unknown or known specificity for a ~iven antigen. It is preferred that the host cells not be constitutively capable of making undetermined antibody, but rather that if they are to produce antibody that it be by transformation with DNA
15 encoding a known immuno~lobulin. Such immunoglobulins (which may include both hea-/y as well as light chains) exhibit specificity for a known antigen. Alternatively, these companion immunoglobulin chains will be devoid of functional variable or hypervariable domains (so as to be capable of multimer assembly but not antigen binding activity). For example, a product MSP fusion secreted and recoverable from host cells capable o~
20 expressing an intact heavy and light chain companion immunoglobulin will bear an antigen binding functionality as well as an MSP functionality. Such products will facilitate the crosslinking of MSP ligand with any desired antigen. Host cells may make more than one immunoglobulin product in such multiple transformations, and accordingly it may be necessary to recover one multimer form from another. This, however, will be a routine 25 matter requiring separation on a ~el or other chromatographic procedure, or by affinity chromatography based on the MSP Jigand, the antigen or both.
Other proteins having extended plasma half life are fused to the MSP in similar fashion, except that instead.of an immunoglobulin chain a transferrin, albumin, apolipoprotein or other sequence is ~mployed. Monofusions are preferred when MSP chains are fused to 30 single chain plasma proteins which do not ordinarily assemble into multimers.The boundary for an MSP ei~tracellular domain generally is at, or within about 20 residues N-terminal from, the N-terminus of the membrane anchor domain, and are readily identified from an inspection of the MSP sequence. It is not necessary to use the entire MSP e~tracellular domain, however, since smaller segments are commonJy found to be 35 adequate for ligand binding. Such segments are routinely identified by makin~ deletional mutants or enzymatic di~ests and screening for ligand binding to identify active fra~ments, and fall within the scope of the term "MSPn.
The MSP extracellular domain generally is fused at its C-terminus to the N-terminus of the immunoglobulin constant region or other stable plasma protein. The precise site at 4û which the fusion is made is not critical; other sites neighboring or within the extracellular re~ion or C-terminal to the mature N-terminus of the plasma protein may be selected in Z'~6~75 g order to op~imize the secretion or binding characteristics of the soluble MSP. The optimal site will be determined by routine experimentation.
Exemplary hetero-and chimeric MSP-immunoglobulin variants produced in accordance with this invention are schematically diagrammed below. "A~ means at least a 5 portion of the e~tracellular domain of an MSP containing its ligand binding site; A~, A2, A3, etc. represent individual subunit chains of A; VL~ V~l~ CL and Cll represent light or heavy chain variable or constant domains of an immunoglobulin; n is an integer; and Y
designates a covalent cross-linking moiety.
(a) ACL;
10 (b) ACL-ACL;
(c) AC~ [AC~, ACL-AC~, ACL V8C~. VLCL-AC~, or VLCL-VEIC~];
(d) ACL ACH [AC~, ACL-AC8, A~L VE~C~. VLCL-AC~, or VLCL-V~C~];
(e) ACL V~CE [AC,~, ACL ~C~, ACL V~C~, VLCL-AC~, or VLCL-V~C13];
( ) LCL AC~ [ACH. ACL-AC~, ACL V~CE~. YLCL AC~, or VLCL-V~C}I~; or 15 (g) 1A Y] ~-1VLCL VLCB]2 The structures shown in this table show only key features, e. g. they do not show disulfide bonds. These are omitted in the interests o~!' brevity. However, where such domains ar0 required for binding ac~ivity they shall be cons~rued as being present in the ordinary locations which they occupy in the immunoglobulin domain. These examples are 20 representative of divalent antibodies; more complex structures would result by employing immunoglobulin heavy chain sequences from other classes, e.g. IgM. The immunoglobulin VLV~ antibody combining site, also designated as the companion immunoglobulin, preferably is capable of binding to a predetermined antigen.
Exemplary immunoglobulin constructs are described schematically below. Vertical 25 lines indicate a noncovalent or covalent associative relationship.
~, ~-lo-~h) A2 (m) D _ ~= CH r A~ CH LVL Cl~ n CH

L CL whe~e n ~ S and CH' iS the A2 s~reted heaYy chain of IgM

~j) A2 C~ (n) V}~
~ CH ~ CH
A2~ H
A2¦, ) Al ~ C~H
~lc~ (O) ~, ~2 ~ ~L
CH

CH ~ CH

CH ~2 ~H ~1 ~L
PrOdUCt ~(O)U ~bC CH 'Y domains have been deleted.

~0~)6~75 "
Suitable companion immuno~lobulin combining sites and fusion partners are obtained from human IgC;-I, -2, -3, or -4 subtypes, IgA, IgE, IgD or IgM, but preferably IgG-I.
It is preferred to use the soluble form of IgM, or one in which the IgM membrane anchor domain has been modified so that it no longer lodges in the membrane.
A preferred embodiment is a fusion of an N-terminal portion of an MSP with a sequence beginning in the hin~e region just upstream of the papain cleavage site which defines lgG Fc chemically (residue 216, taking the first residue of heavy chain constant re~ion to be 114 [Kabat e~ al., "Sequences of Proteins of Immunological Interest" 4t}: Ed., 1587], or analogous sites of other immunoglobulins).
The immunoglobulin nr other plasma-stable polypeptide is fused to the C-termini of one or more of the MSP subunits, typically in place of at least one transmembrane and cytoplasmic domain of an MSP chain, although ordinarily only one of the subunits is substituted. In the case of GPllb-lIla this would be the beta subunit. The immunoglobulin domain such as a heavy chain also can be associated in normal fashion with a truncated or 15 intact immunoglobulin heavy chain.
Variants in which an MSP extracellular domain is substituted for the variable region of an immunoglobulin chain are believed to exhibit improved in vivo plasma half life.
These chimeras are constructed in a fashion similar to chimeric antibodies in which a variable domain from an antibody of one species is substituted for the variable domain of 20 another species. See, for example, EP 0 125 023; Munro, Nature 312: (13 December 1984);
Neuberger e~ al., Nature 312: (13 December 1984); Sharon e~ al., Nature 309: (24 May 1984);
Morrison et al., Proc. Natl. Acad. Sci. USA ~1:6851-6855 (1984); Morrison et al. Science 229:1202-1207 (1985); and Boulianne et al., Nature 312:643-646 (13 December 1984). The DNA encoding the MSP e~tracellular domain is cleaved by a restriction enzyme at or 25 proximal to the 3' end of the DNA encoding the domain and at a point at or near the DNA
. encoding the N-terminal end of the mature MSP polypeptide (where use of a different leader is contemplated) or at or proximal to the N-terminal coding region for the MSP
(where the native MSP signal is employed). This DNA fra~ment then is readily inserted into DNA encoding e.g.an immunoglobulin light or heavy chain constant region and, if 30 necessary, tailored by deletional mutagenesis. Preferably, this is a human immunoglobulin.
DNA encodi~g immunoglobulin light or heavy cl~ain constant regions is known or readily available from cDNA libraries or is synthesized. ~e for example, Adams et ~1., Biochemistry ~2:2711-2719 (1980); Gough el al., Biochemistry 19:2702-2710 (1980); Dolby e~ al., P.N.A.S. USA, 77:6027-6031 (1980); Rice et al., P.N.A.S. USA 79:7862-7865 (1982);
35 Falkner et al., Nature ~2~:286-288 (1982); and Morrison et al., Ann. Rev. Immunol. 2:239-256 (1984).
DNA encoding the chimeric chain(s) is transfected into a host ceil for expression.
If the host cell is producin~ an immunoglobulin prior to transfection then one need only transfect with the MSP fused to light or to heavy chain to produce a heteroantibody. The 40 aforementioned immunoglobulins }laving one or more arms bearing the MSP domain and one or more arms bearing companion variable regions result in dual specificity for MSP

Z~3~ 7~

ligand and for an antigen. These are produced by the above-described recombinantmethods or by in vitro procedures. ~n the latter case, for example, F(ab')2 fragments of the MSP fusion and an immunoglobulin are prepared, the F(ab')2 fragments converted to Fab' fragments by reduction under mild reducing conditions, and then reoxidized in each other's 5 presence under acidic conditions in accord with methods known per se. See also U.S. Patent 4,444,878.
Additionally, procedures are known for producing intact heteroantibodies from immunoglobulins having different specificities. These procedures are adopted for the i~l vifro production of hetero~himeric antibodies by simply substituting the MSP fusions for 10 one of the previously employed immunoglobulins.
In an alternative method for producing a heterofunctional antibody, host cells producing an MSP-immunoglobulin fusion, e.g. transfected myelomas, also are fused with B cells or hybridomas which secrete an~ibody having the desired companion specificity for an antigen. Heterobifunctional antibody is recovered from the culture medium of such 15 hybridomas, and thus may ~e produced somewhat rnore conveniently than by conventional in vi~ro resortin& methods (EP 68,763).
ADother class of MSP variants are deletional variants. Deletions are characterized by the removal of one or more amino acid residues from an MSP sequence. Typically, the membrane anchor and cytoplasmic domains of all MSP subunits are deleted. However, any 20 other suitable site N-terminal to the transmernbrane which preserves the matrix protein or ligand-binding capability of the MSP is suitable. Excluded from the scope of deletional variants are the protein digestion fragments that may have heretofore been obtained in the course of elueidating amino acid sequences of MSPs.
Substitutional variants are those in which at least one residue in the MSP sequence 25 has been remov~d and a different residue inserted in its place. l[able I below describes substitutions which in general will result in fine modulation of the characteristics of an MSP.
TABLE I
Ori~inal Residue _ ExemDlarv ~ubs~itutions ~la ser Arg lys Asn gln; his Asp glu Cys ser; ala Gln asn Glu asp Gly pro His asn; gln Ile leu, val Leu ile; val Lys ar8; ~In; glu Met leu; ile Phe met; leu; tyr Ser thr Thr ser Trp tyr Tyr trp; phe Val ile; leu ~3~ 75 Substantial changes in function or immunological identity are made by selecting substitutions that are iess conservative than those in Tsble 2, i.e., selecting residues that differ more significantly in their effect on maintaining (a) the structure of the polypeptide backbone in the area of the substih~tion, for example as a sheet or helical conformation, (b) the charge or hydrophobicity of the molecule at the target site or (c) the bulk of the side chain. The substitutions which in general are expected to produce the greatest changes in MSP properties will be those in which (a) a hydrophilic residue, e.g. seryl or threonyl, is substituted for (or by) a hydrophobic residue, e.g. leucyl, isoleucyl, phenylalanyl, valyl or alanyl; (b) a cysteinyl or prolyl is substituted for (or by) any other residue; (c) a residue having an electropositive side chain, e.g., Iysyl, arginyl, or histidyl, is substituted for (or by) an electronegative residue, e.8., giutamyl or aspartyl; or (d) a residue having a bulky side chain, e.~., phenylalanyl, is substituted for (or by) one not having a side chain, e.g., glycyl.
A preferred class of substieutional or deletional variants are those involving amembrane anchor region of the MSP. Transmernbrane regions of MSP subunits are highly hydrophobic or lipophilic domains that are the proper size to span the lipid bilayer of the cellular membrane. They are believed to anchor the ~SP in the cell membrane. Other cell surface molecules are anchored by lipid modification, as by phospholipid anchors.
Deletion or substitution of the membrane anchor domain will facilitate recovery and provide a soluble form of the MSP by reducing its cellular or membrane lipid affinity and improving its water solubility. If the membrane anchor domains are deleted one avoids the introduction of potentially immunogenic epitopes, either by exposure of otherwise intracellular polypeptides that might be recogni~ed by the body as foreign or by insertion of heterologous polypeptides that are potentially immunogenic. A principal advantage of 2~ the membrane ~nchor domain-deleted MSP is that it is secreted into the culture medium of recombinant hosts. This variant is soluble in body fluids such as blood and does not have an appreciable affinity for cell membrane lipids, thus considerably simplifying its recovery from recombinant cell culture. Surprisingly, MSPs in which membrane inserted chains have been modified so as to be no ionger capable of stable insertion ;nto cell membranes are capable of proper association and secretion from recombinant host cells even if the MSP
chains are not fused to a multimer-~orming sequence such as an immunoglobulin. Amultimer-f`orming sequence is a multichain polypeptide that contains that postion of a multiple chain polypeptide that, when in the ~nfused form in nature, ~orms covalently or noncovalently associated multiple chain structures.
It will be amply apparent from ~he foregoin~ discussion that substitutions, deletions, insertions or any combination thereof are introduced to arrive at a final construct. None of the variants will have a fuDctional membrane anchor domain and preferably will not have a functioDal cytoplasmic sequence. This is generally accomplished by deletion of the relevant domain, although adequaee insertional or substitutional variants also are effective for this purpose. For example, the transmembrane domain is substituted by any amino acid sequence, e.g. a random or predetermined sequence of about 5 to 50 serine, threonine, ~S~ 75 Iysine, arginine, glutamine, aspartic acid and like hydrophilic residues, which altogether exhibit a hydrophilic hydropathy profile. Like the deletional (truncated) MSPs, these variants are secre~ed into the culture medium of recombinant hosts MSP variants are prepared conveniently by site specific mutagenesis of nucleotides in the DNA encoding the MSP, thereby producing DNA encoding the variant, and thereafter expressin~ the DNA in recombinant cell culture. Obviously, changes in the DNA
encoding the variant MSPs must not place the sequence out of readin8 frame and preferably will not create complementary regions ~hat could produce secondary mRNA structure deleterious to expression (EP 75,444A). The MSP variants typically exhibit the same matrix or ligand bindin~ activity as does the naturally-occurring prototype, although variants also are selected in order to modify the characteristics of the MSP as indicated above.
While the site for introducing an amino acid sequence variation is predetermined, the mutation per se need not be predetermined. For exa~nple, in order to optimize the performance of a mutation at a given site, random or saturation mutagenesis (where all 20 possible residues are inserted) may be conducted at the target codon or region and the expressed MSP variants screened for the o~timal combination of desired activities.
MSP variants that are not capable of binding to their matrix proteins or ligands are useful nonetheless as immunogens for raising antibodies to the MSP or as immunoassay kit components (labelled, as a competitive rea8ent for native MSP, or unlabelled as a standard for an MSP assay) so long as at least one MSP epitope remains active.
Contemplated herein are MSPs or MSP amino acid sequence or glycosylation variants (including those already described above) whereiD one or more MSP subunits are conjugated with a nonproteinaceous polymer. It will be understood that the nonproteinaceous polymer which is conjugated to MSP excludes oljgosaccharides that are present in the same positions in the native or starting MSP, i.e. the polymer is extraneous or heterologous to thç MSP.
n It is within the scope hereof to move, add or delete glycosylation sites by site-directed mutagenesis of MSP polypeptide in order to increase the number of or change the location of the carbohydrate substituents. The nature of the carbohydrate ss modified in conventional fashion by if ~ vitro enzymatic digestion or by selecting host cells that affix the selected carbohydrate (or do ~ot glycosylate at all).
The nonproteinaceous polymer ordinarily is a hydrophilic synthetic polymer, i.e., a polymer no~ otherwise found in nature. Howe~er, polymers which exist in nature and are produced by recombinant or methods are useful, as are polymers which are isolated from nature. Hydrophilic poly~inyl polymers fall within the scope of this invention, e.g.
polyvinylalcohol and polyvinylpyrrolidone. Particolarly useful are ~olyalkylene ethers such as polyethylene glycol, polypropylene ~Iycol, polyoxyethylene esters nr methoxy spolyethylene glycol; polyoxyalkylenes such as polyoxyethylene, polyoxypropylene, and block copolymess of polyoxyethylene and polyoxypropylene (Pluronics); polymethacrylates;
carbomers; branched or unbranched polysaccharides which comprise the saccharide monomers D-mannose9 D- and L-~alactose, fucose, fructose, D-xylose, L-arabinose, D-glucuronic acid, sialic acid, D-galacturonic acid, D-rnannuronic acid (e.g. polymannus;onic '~S313~7~

acid, or alginic acid), D-~lucosamine, D-galactosamine, D-glucose and neuraminic acid including homopolysaccharides and heteropolysaccharides such as lactose, amylopectin, starch, hydroxyethyl starch, amylose, de~ttran sulfate, dextran, dextrins, ~Iycogen, or the polysaccharide subunit of acid mucopolysaccharides, e.g. hyaluronic acid; polymers of sugar 5 alcohols such as polysorbitol and polymannitol; and heparin. Where the polysaccharide is the native glycosylation or the glycosylation attendant on recombinant z%pression of MSP, the site of substitution ordinarily is located at other than an N or O-linked glycosylation site of the MSP or the MSP variant is an amino acid sequence variant in which an additional or substitute N or O-linked si~e has been introduced into the molecule.
Mixtures of such polymers are employed, or the polymer may be homogeneous. The polymer prior to crosslinking need not be, but preferably is, water soluble, but the final conjugate must be soluble in biological fluids such as blood. In addition, for therapeutic uses the polymer should not be highly immunogenic when conjugated to the MSP, nor should it possess viscosity that is incompatible with intravenous infusion or injection if it 15 is intended to be administered by such routes.
Preîerably the polymer contains only a single group which is reactive with MSP. This helps to avoid cross-linking of MSP molecules. However, it is within the scope herein to optimize reaction conditions to reduce cross-linking, or to purify the reaction products through ~el filtration or chromatographic sieves to recover substantially homogeneous 20 derivatives.
The molecular weight of the polymer ranges about from 100 to 500,000, and preferably is about from 1,000 to 20,000. The molecular weight chosen will depend upon the nature of the polymer and the degree of substitution. In general, the 8reater the hydrophilicity of the polymer and the ~reater the de8ree of substitution, the lower the 25 molecular weight that can be employed. Optimal molecular weights will be determined by roufine experimentation. Ordinarily, the molecular weight of the MSP-polymer conjugate will exceed about 70,000 although molecules having lesser molecular weights are suitable.
The polymer generally is covalently linked to MSP through a multifunctional crosslinking agent which reacts with the polymer and one or more amino acid or sugar 30 residues of MSP. However, it is within the scope of this invention to directly crosslink the polymer to the MSP by reacting a derivatized polymer with MSP, or vice versa A suitable MSP covalent crosslinlcing site is the N-terminal amino group and epsilon amino ~roups found on Iysine residues, although other amino, imino, carboxyl, sulfydryl, hydro~cyl or other hydrophilic groups serve as useful sites of substitutiorl. The polymer may 35 be co~alently bonded directly to MSP without the use of a multifunctional ~ordinarily bifunctional) crosslinking agent. Examples of such crosslinking agents include I, I -bis(diazoacetyl)-2-phenylethane,glutaraldehyde,N-hydroxysuccinimideesters,forexample esters with 4-azidosalicylic acid, homobifunctional imidoesters including disuccinimidyl esters such as 3,3'-dithiobis (succinimidyl-propionate), and bifunctional maleimides such 40 as bis-N-maleimido-1,8-octane. Derivatizing agents such as methyl-3-[(p-azido-phenyl)dithio] propioimidate yield photoactivatable intermediates which are capable of forming cross-links in the presence of li~ht. Alternatively, reactive water soluble matrices such as cyanogen bromide activated carbohydrates and the systems described in U.S. patents 3,9~,0~0; 3,969,287; 3,691,016; ~,195,128; 4,247,642; 4,229,537; 4,055,635 and 4,330,440 are suitably m~dified for cross-linking the polymer and MSP. Covalent bonding to MSP
5 amino groups is accomplished by known chemistries based upon cyanuric chloride, carbonyl diimidazole, aldehyde reactive groups (P~G alkoxtde plus diethyl acetal of bromoacetaldehyde; PEG plus DMSO and acetic anhydride, or PEG chloride plus the phenoxide of 4-hydroxybenzaldehyde, succinimidyl active esters, activated dithiocarbonate PEG, 2,4,5-trichlorophenylchloroformate or p-nitrophenylchloroformate activated PEG.
10 Carbo~yl groups are derivatized by coupling PE~i-amine using carbodiimide.
Polymers are conjugated to the oligosaccharide substituents by chemical, e.g.
metaperiodate, or enzymatic oxidation, e.8. glucose or galactose oxidase, (to produce the aldehyde derivative of the carbohydrate), followed by reaction with hydrazide or amino-derivatized polymers, in the same fashion as is described by Heitzmann et al., P.N.A.S., 71:3537-3541 ~1974) or Bayer e~ al., Methods in Enzymology, 62:310 (1979), for the labeling of oligosaccharides with biotin or avidin. Further, other chemical or enzymatic methods which have been used heretofore to link oligosaccharides and polymers may be suitable.
Substituted oligosaccharides are particularly advantageous since there ase fewercarbohydrate substitutions than amino acid sites for derivatization, thus improving the 20 stability, activity and homogeneity of the conjugate. Finally, the MSP oligosaccharide substituents are enzyrnatically modified to remove sugars, e.g. by neuraminidase digestion, as a final product or prior to polymer derivatization.
The polymer will bear a group which is directly reactive with an amino acid sidechain, or the N- o~ C- terminus of MSP, or which is reactive with the multifunctional 25 cross-linking agent. In general, polymers bearing such reactive groups are known for the `- preparation of immobilized proteins. In order to use such chemistries here, one should employ a water soluble polymer otherwise derivatized in the same fashion as insoluble polymers heretofore employed for protein immo~ilization. Cyanogen bromide activation is a particularly useful procedure to employ in crosslinking polysaccharides to MSP.
"Water soluble~ in reference to the startin~ polymer means th~t the polymer or its reac~ive intermediate used for conjugation is sufficiently water soluble to participate in a derivatization reaction with ~ISP.
The degree of substitution of MSP will vary dependin~ upon the number of reactive sites on the protein, whether intact or truncate~ MSP is used, whether the MSP is a fusion 35 with a protein heterologous to ~SP, the molecular weight, hydrophilicity and other characteristics of the polymer, and the particular sit0s chosen. In ~eneral, the MSP portion of the conjugate is substituted with about from I to 10 polymer molecules, while any heterologous sequence which is fused to MSP may be substituted with an essentially unlimited number of polymer molecules so long as the activity of the MSP moiety is not 40 significantly adversely af~ected. The optimal degree of crosslinking is easily determined by an experimental matrix in which the time, temperature and other reaction conditions are 7~

varied to change the de~ree of substitution, after which the ability of the conjugates to bind matrix protein or ligand is determined.
The polymer, e.g., PEG is crosslinked to MSP by a wide variety of methods known per se for the covalent modification of proteins with nonproteinaceous polymers such as PEG. Certain of these methods, however, are not preferred for the purposes herein.
Cyanuric chloride chemistry leads to many side reactions, including protein cross-linking.
In addition, it may be particularly likely to lead to inactivation of proteins containing sulfhydryl groups. Carbonyl diimidazole chemistry (Beauchamp el al., "Anal. Biochem."
~L:2.~-33 11983]) requires high pH (>8.S), which can inactivate proteins. Moreover, since the ~activated PEG" intermediate can react with water, a veJy large molar excess of "activated PEG" over protein is required. In general, aldehyde chemistry (Royer, U.S.
Patent 4,002,531) is preferred since it requires only a 40 fold molar excess of PEG and a 1-2 hr incubation. However, the manganese dioxide su~ested by Royer for preparation of the PEG aldehyde is problematic "because of the pronounced tendency of PEG to form complexes with metal-based oxidi~ing agents~ (Harris e~ al., "J. Polym. Sci., Polym. Chem.
Ed." 22:341-352 [1984]). Use of a moffatt oxidation, utilizing DMSO and acetic anhydride, obviates this problem. In addition, the sodium borohydride su~gested by Royer must be used at a high pH and has a significant tendency to reduce disulphide bonds. In contrast, sodium cyanoborohydride, which is effective at neutral pH, has very little tendency to recluce disulphide bonds.
The MSP conjugates of this invention typically are separated from unreacted starting materials by gel filtration. Most conveniently, MSP conjugates are eluted from hydrophobic interaction chromatography medium, e.g. alkyl Sepharose, by the use of a decreasing salt gradient. This, as well as the gel filtration approach described above, resolves conjugates on the basis of the degree of substitution.
The DNA encoding an MSP is obtained by known procedures, in most instances by reference to publications describing DNA encoding the MSP. In general, prokaryotes are used for clonin~ of MSP variant DNA sequences. For e2~ample, a -resistant strain of E.
coli JM 101 for propagating M13 phage; Messing e~ ucl. Acids. Res. 2(2~:309-321 ~1981]); and E. coli K12 s~rain 294 (ATCC No. 31446) are particularly useful. O~her microbial strains which may be used include E. coli B, or UMI01. These examples a}e illustrative rather than limiting. Nucleic acid also is cloned using various well known in vitro amplification processes.
DNA encoding the variant MSPs are inserted for expression into vectors containing promoters and control sequences which are derived from species compatible with the intended host cell. The vector ordinarily, but need not, carry a replication site as well as one or more marker seQuences which are capable of providing phenotypic selection in transformed cells. For 0xample, E. coli is typically transformed using a derivative of pBR322 which is a plasmid derived from sn E. coli species (Bolivar, e~ al., Gene 2: 95 [1977]). pBR322 contains genes for ampicillin and tetracycline resistance and thus provides easy means for identifying transformed cells. The pBR322 plasmid, or other microbial ~f~6475 plasmid must also contain or be modified to cDntain promoters and other control elements commonly used in recombinant DNA constructions.
Promoters suitable for use with prokaryotic hosts illustratively include the n-lactamase and lactose promoter systems (Chang et al., Nature, 275: 615 [1978]; and Goeddel e~ al., Nature 281: 544 [1979]), alkaline phosphatase, the tryptophan (trp) promoter system (Goeddel, Nucleic Acids Res. ~: 4057 11980] and EPO Appln. Publ. No. 36,776) and hybrid promoters such as the tac promoter (H. de Boer e~ al., Proc. Natl. Acad. Sci. USA 80: 21-25 I19831). However, other functional bacterial promoters are suitable. Their nucleotide sequences are generally known, thereby enabling a skilled worker operably to ligate them 10 to DNA encoding the MSP variant using linkers or adaptors to supply any required restriction sites (Siebenlist et al., Cell 20: 269 [1980]). Promoters for use in bacterial systems also will contain a Shine-Dalgarno (S.D.) sequence operably linked to the DNA encoding the antigen.
In addition to prokaryotes, eukaryotic microbes such as yeast cultures also are useful 15 as cloning or expression hosts. Saccharomvces cerevisiae, or common baker's yeast is the most commonly used eukaryotic microorganism, although a number of othe} strains are commonly available. For expression in Saccharomyces, the plasmid YRp7, for example, (Stinchcomb, et al., Nature 282. 39 [1979]; Kingsman et al, Gene 7: 141 11979]; Tschemper et al., Gene 10: 157 [1980]) is commonly used. This plasmid already contains the trpl gene 20 which provides a selection marker for a mutant strain of yeast lacking the ability to grow in tryptophan, for example ATCC no.44076 or PEP4- 1 (Jones, Genetics 85: 12 [1977]). The presence of the trpl lesion as a characteristic of the yeast host cell genome then provides an effective means of selection by growth in the absence of tryptophan.
Suitable promoting se4uences for use with yeast hosts include the promoters for 3-25 phosphoglycerate kinase (Hit~eman et al., 1. Biol. Chem. 255: 2073 [1980]) or other glycolytic enzymes (Hess e~ al., J. Adv. Enzyme Reg. 7: 149 [1968]; and Holland,Biochemistry 17: 4900 ~ 1978]), such as enolase, glyceraldehyde-3-phosphate dehydrogenase, hexokinase, pyruvate decarboxylase, phosphofsuc;okiDase, glucose-6-phosphate isomerase, 3-phosphoglycerate mutase, pyruvate kinase, triosephosphate isomerase, phosphoglucose 30 isornerase9 and glucokinase.
Other yeast promoters, which are inducible promoters having the additional advantage of transcription controlled by gro~th conditions, are the promo~er re&ions for alcohol dehydrogenase 2, isocytochrom~ C, acid phosphatase, degradative enzymes associated with nitrogen metabolism, metallothionein, glyceraldehyde-3-phosphate35 dehydrogenase, and enzymes responsible for maltose and galactose utilization. Suitable vectors and promotsrs for use in yeast expression are further described in R. Hit2ernan et al., European Patent Publication No. 73,657A. Yeast enhancers also are adYantageously used with yeast promoters.
Promoters for sontrolling transcription from vectors in mamsnalian host cells may be 40 obtained from various sources, for example, the genomes of viruses such as polyoma, Simian Yirus 40 (SV40), adenovirus, retroYiruses, hepatitis-B virus and most preferably ;475 cytomegalovirus, or from heterologous mammalian promoters, e p. the beta actin promoter.
The early and late promoters of the SV40 virus are conveniently obtained as an SY40 res~riction fra8ment which also contains the SV40 viral origin of replication. Fiers et al., Nature, 273: 113 (1978). The immediate early promoter of the human cytomegalovirus is S conveniently obtained as a Hindlll E restriction fra8ment. Greenaway, P.J. et al., Gene 18:
35~-360 (1982). Of course, promoters from the host cell or related species also are useful.
DNA transcription in higher eukaryotes is increased by inserting an enhancer sequence into the vector. Enhancers are cis-actin8 elements of DNA, usually from a~out 10 to 300bp, that act to increase the transcription initiation capability of a promoter.
10 Enhancers are relatively orientation and position independent having been found 5' (Laimins~ L. e~ al., Proc.Natl.Acad.Sci. 78: 993 [1981~) and 3' (Lusky, M.L., et al., Mol. Cell Bio. ~: 1108 [1983]) to the transcription unit, within an intron (Banerji, J.L. el al., Cell 33:
729 [1983~) as well as within the coding seq-~ence itself (Osborne, T.F., e~ al., Mol. Cell Bio.
4: 1293 11984]). Many enhancer sequences are now known from mammalian genes (globin, 15 elastase, albumin, ~-fetoprotein and insulin). Typically, however, one will use an enhancer from a eukaryotic cell virus. Examples include the SV40 enhancer on the late side of the replication origin ~bp 100-270), the cytomegalovirus early promoter enhancer, the polyoma enhancer on the late side of ~he replication origin, and adenovirus enhancers.
Expression vectors used in eukaryotic host cells (yeast, fungi, insect, plant, animal, 20 human or nucleated cells) may also contain sequences necessary for the termination of transcription which may affect mRNA expression. These regions are transcribed aspolyadenylated segments in the untranslated portion of the mRNA encoding the MSP.
Expression vector systems generally will contain a selection gene, also termed aselectable marker. Examples of suitable selectable markers for mammalian cells are 25 dihydrofolate redllctase (DHFR), thymidine kinase or neomycin. When such selectable . markers are successfully transferred into a mammalian host cell, the transformed mammalian host cell can survive if placed under selective pressure. There are two widely used distinct çategories of selective regirnes. The first cate~ory is based on a cell's metabolism and the use of a mutant cell line which lacks the ability to grow independent 30 of a supplemented medium. Two examples are: CHO DHFR- cells and mouse LTK- cells.
These cells lack the ability to ~row without t3~e addition of such nutrients as thymidine or hypo~anthine. ~ecause these cells lack certain genes necessary for a complete nucleotide synthesis pathway, they cannot survive unless the missing nucleotides are provided in 8 supplemented medium. An alternative to supplementing the medium is to introduce an 35 intact DHFR or TK gene into cells lacking the respective genes, thus altering theis growth requirements. Individual cells which were not transformed with the DHFR or TK Bene will not be capable of survival in non supplemented media.
The second category is dominant selection which refers to a selection scheme used in any cell type and does not require the use of a mu~ant cell line. These schemes typically 40 use a dru~ to arrest growth of a host cell. Those cells which have a novel gene would express a protein conveying drug resistance and would survive the selection. Examples of ~0~)~47~

sueh dominant selection use the dru~s neomycin, Southern P. and Berg, P., J. Molec. Appl.
&enet. 1: 327 (1982), mycophenolic acid, Mulligan, R.C. and Berg, P. S~ience 209: 1422 (1980) or hygromycin, Sugden, B. et al., Mol. Cell. Biol. 5: 410-413 (1985). The three examples ~iven above employ bacterial genes under eukaryotic control to convey resistance to the appropriate drug G418 or neomycin (geneticin), ~gpt (mycophenolic acid) or hygromycin, respectively.
"Amplificationn refers to the increase or replication of an isolated r~gion within a cell's chromosomal DNA. Amplification is achieved using a selection a8ent e.g.
methotre~ate (MTX) wh;ch is inactivated by DHFR. Arnplification or the making ofsuccessive copies of the DHFR ~ene results in 8reater amounts of DHFR being produced in the face of ~reater amounts of MTX. Amplif;cation pressure is applied notwithstanding the presence of endogenous DHFR~ by adding ever 8reater amounts of MTX to the media.
Amplification of a desired 8ene can be achieved by cotransfecting a mammalian host cell with a plasmid having a DNA encoding a desired protein and the DHFR or amplification 8ene permitting cointegration. One ensures that the cell requires more DHFR, ~hich requirement is met by replication of the selection gene, by selecting only for cells that can grow in the presence of ever-greater MTX concentration. So long as the gene encoding a desired heterologous protein has cointegrated with the selection Bene replication of this 8ene gives rise to replication of the gene encoding the desired proeein. The result is that increased copies of the gene, i.e. an amplified ~ene, encoding the desired he~erologous protein express more of the desired heterolo~ous protein.
Preferred host cells for expressing the MSP variants of this invention are mammalian host-vector systems, examples of suitable hosts including: monkey kidney CVl line transformed by SV40 (t:OS-7, ATCC CRL 1651); human embryQnlc kidney line (293, Graham, F.L. e~ al., J. Gen Virol. 3~: 59 119771 and 293S cells, either of which are equally . satisfactory); baby hamster kidney cells (BHK, ATCC CCL 10); chinese hamster ovary-cells DHFR (CHO, Urlaub and Chasin, Proc.Natl.Acad.Sci. (USA) 77: 4216, [1980]); mouse sertoli cells (TM4, Mather, J.P., Biol. Reprod. 23: 243-251 [1980]); monkey kidney cells (CVI ATCC CCL 70); african 8reen monkey kidney cells (VERO-76, ATCC CRL-1587);
human cervical carcinoma ce31s (HELA, ATCC CCL 2); canine kidney cells (MDCK, ATCC
CCL 34); buffalo rat liYer cells (BRL 3A, ATCC CRI, 1442); human lung cells (W138, ATCC CCL 75); human liver cells ~Hep G2, HB 8065); mouse mammary tumor (MMT
060562, ATC(: CCL51 cells); and TRI cells (Mather, J.P. et al., Annals N.Y. Acad. Sci. 383:
44-68 [1982]).
"Transformation~ means introducing DNA ;nto a~ organism 30 that the DNA is replicable, either as an extrachromosomal element or by chromosomal integration. One suitable for transformation of the host cells is the method of Graham, !F. and van der Eb, A., Virology 52: 456-457 (1973). However, other methods for introducing DNA into cells ~uch as by nuclear injection or by protopiast fusion may also be used. If prokaryotic cells 10 or cells which contain substantial cell walls are used as hosts, the preferred method of transfection is calcium treatment usin~ calcium chloride as described by Cohen, F.~. et al., Proc. Natl. Acad. Sci. (USA), 69: 2110 (1972).
Construction of suitable vectors con~aining the desirecJ coding and control sequences employ standard and manipulative li~ation techniques. Isolated plasmids or DNA fragments are cleaved, tailored, and religated in the form desired to form the plasmids required.
Suitable procedures are well known for the construction described herein. See, for example, (Maniatis, T. et a/., MQIÇÇU~ . 133-134 Cold Spring Harbor, [1982]; "Current Protocols in Molecular Biologyn, edited by Ausubel et al., [1987], pub. by Greene Publishing Associates ~: Wiley-lnterscience).
Ordinarily, DNA encoding each subunit of a given MSP (or transmembrane modified 10 variant) is simultaneously cotransfected in~o the host cell, although such transfections can be done sequentially. MSP variants in which one subunit is exchanged for the analogous subunit of another MSP (to produce heterologous heterodimers) are produced by cotransforming a recombinant host (typically mammalian cell) with each of the heterologous subunits, for example, exchanging the fibronectin c~ subunit for the ~ subunit of GPllb-15 lIla (an c~ subunit exchange), or the fibronectin n subunit for the 13 subunit of GPIlb-llla (a n subunit exchan~e).
Correct plasmid sequences are confirmed by transforming E. coli K12 strain 294 (ATCC 3144S) with ligation mixtures, successful transformants selected by ampicillin or tetracycline resistance where appropriate, plasmids from the transformants prepared, and 20 then analyzed by restriction enzyme digestion and/or sequenced by the method of Messing e~ ~1.., Nucleic Acids Res. 2: 309 (1981) or by the meehod of Maxam et al., Methods in Enzymolo~y 65: 499 (1980).
Host cells are transformed wi~h the expression vectors of this invention. Thereafter they are cultured in appropriate culture media, e.g. containir~ substsnces for inducing 25 promoters, selecting transformants or amplifying genes. The culture conditions, such as ~`- temperature, pH and the like, are those previously used with the host cell selected for expression, and will be apparent to the ordinarily skilled artisan. For expression of C;PIlb-IIIa it is preferable that the culture medium contain calcium and ma~nesium salts since divalent cations are needed to enhance the stability of secreted &Pllb-lIIa and other calcium 30 dependent MSPs.
The secreted MSP variants are recovered and purified f}om the cultuse supernatants or lysates of recombinant hosts. Typically, the supernatants are concentrated byu1trafiltration, contacted with a ligand (e.g. RGD~ or matrix protein affinity or immunoaffinity resin so ~s to adsorb the MSP variant, and eloted from the adsorbent.
35 Optionally, the MSP is purified by HPLC, lectin columns, gel exclusion, hydrophobic interaction or ion exchange chromatography.
The purified MSP is formulated into conventional pharmacologically acceptable e~cipients.
The soluble MSP vasiants of this invention are useful in therapeutics, diagnostics and 40 preparatiYe procedures. In diagnostics, the soluble MSPs are employed in place of membrane extracts as standards or controls, or are labelled with a radioisotope or other ~0~647~ii detectable group for use in competitive^type radioimmuno- or radioreceptor assays for the MSP or its antibodies.
The soluble MSPs are crosslinked to insoluble supports by the methods described herein and employed for the purification of their ligands or matrix proteins, e.g.
5 fibronectin, fibrinogen and the like. Alternatively, the soluble MSPs are used to adsorb ligand or matrix protein in solution, followed by precipitation by an~isera, ammonium sulfate or the like in order to recover the ligand or matrix protein complex. The complex is then dissociated by HPLC, electrophoresis, 8el chromatography or other conventional methods.
Therapeutic uses of soluble MSPs will be a îun_tion of the biological activity of each MSP, and will be apparent therefrom. The soluble MSP variants herein may act as agonists or antagonists of the corresponding native, membrane-bound receptors. The soluble GPlIb-llla receptor, for example, is useful as an anticoagulant and for the treatment of disorders associated with platelet aggregation, particularly in the prevention of reocclusion following 15 thrombolytic therapy. Soluble matrix receptors, especially soluble GPllb-llla, also are useful as antagonists to matrix-adhesion dependent neoplastic metastasis. Soluble LF~-I variants are an antagonist of T-lymphocyte function, thereby being efficacious as immunosuppressive or anti-inflammatory agents, particularly in reperfusion injury. Soluble Mac-l variants may find use in the treatment of complement activation disorders.In order to facilitate understanding of the following examples certain frequently occurring methods and/or terms will be described.
"Plasmids" are designated by a lower case p preceded and/or followed by capital letters and/or numbers. The starting plasmids herein are either commercially available, publicly available on an unrestricted basis, or can be constructed from available plasmids 2~ in accord with published procedures. In addition, equivalent plasmids to those described `- are known in the art and will be apparent to the ordinarily skilled artisan.
"Digestion" of DNA refers to catalytic cleavage of the ~NA with a restriction enzyme that acts only at certain sequences in the DNA. The various restriction enzymes used herein are commercially available and their reaction conditions, cofactors and other requirements 30 were used as would be known to the ordinarily skilled artisan. For analytical purposes, typically I ~Lg of plasmid or DNA fragment is used with about 2 units of enzyme in about 20 ,ul of buffer solutiom For the purpose of isolating DNA fragments for plasmideonstruction, typically S to S0 ~ g of DNA are digested with 20 to 250 units of enzyme in a lar~er volume. Appropriate bllffers and substrate amounts for particular restriction 35 enzymes are specified by the manufacturer. Incubation times of about I hour at 37C are ordinarily used, but may vary in accordance with the supplier's instructions. After digestion the reaction is electrophoresed directly on a polyacrylamide or agarose gel to isolate the desired fragment.
"Recovery" or Uisolation" of a given fragment of DNA from a restriction digest means 40 separation of tl~e digest on polyacrylamide or agarose gel by electrophoresis, identification of the fra8ment of interest by comparison of its mobility versus that of marker DNA

7~

fragments of known molecular weight, removal of the gel section containing the d~sired fragment, and separation of the gel from DNA. This procedure is known generally (Lawn, R. et a/., Nucleic Acids Res. 2: 6103-6114 119BI], and Goeddel, D. et al., Nucleic Acids Res. 8: 405711980]).
"Li~ation~ refers to the process of forming phosphodiester bonds between two double stranded nucleic acid fra~ments (Maniatis, T. e~ ~., ~. at 146). Unless otherwise provided, ligation may be accomplished using known buffers and conditions w;th 10 units of T4 DNA
ligase (nligase") per 0.5 ~L~ of approximately equimolar amounts of the DNA fragments to be ligated.
The following examples merely illustrate the best mode now contemplated for practicing the invention, but should not be construed to limit the invention.
Example I
Cloninl! of GlYcoDrotein llb (C;PII~D~,~
Messenger RNA was prepared from cultured human erythroleukemia cells (HEL, ATCC TIB 180). An oligo(dT)-primed cDNA library was prepared using this mRNA in the bacteriophage lambda ZAP (Stratagene Cloning Systems). The lambda ZAP library was screened with a 45-mer oligonucleotide (2bl~ deri~ed from the 5' end of the published cDNA sequence for GPllb from HEL cells (Poncz e~ al., "J. Biol. Chem." 262(18):8476-8482 [1987]). Several positively-hybridi~ing phage were purified, and the cDNA inserts they contained were sub.iect0d to restriction enzyme di~estion analysis. From these results a phage which appeared to contain a full-length coding insert for GPllb was selected for further analysis. DNA sequencing of this phage insert D~A gave over 300 bases which corresponded exactly with the published cDNA sequ~nce from the ~' end of the mRNA
(Poncz et al.) except having 4 additional bases 051 its 5' end. The cDNA insert was digested with EcoRI (this site being derived from the linkers ligated to the ends of the cDNAs during production of the library) and ~indlll, which cuts the GPlIb insert uni~uely downstream of the end of the coding sequence. This EcoRI to HindllJ restriction fra~rnent, containing the entire codin~ re~ion l or GPllb was ligated into mammalian cell expression vec~or pRK5 (European Appln. Pub. No. 307,247 published IS March 1989) which had been digested with EcoRI and Hindlll, and e~pression vector GPIlb-pRK5 was reco~tered.
Collstruction of Full-Le~th C;l~rotein Illa (GPllla) cDNA
A cDNA clone for GPllla, incomplete at its 5' end, was obtained ~Rosa et al., "Blood"
7~(2):593 [1988]). The cDNA was provided as an EcoRI (site derived from the cDNAlibrary construction linker) to Pstl (site downstream of the end of the coding sequense) insert i~ ~he plasmid vector plB120 (International Biotechnologies, llnc.) This plasmid was digested with Hindlll to cut the plasmid at the u~ique Hindlll site in plB120 downstream of the terminal Pstl site in the cDNA insert, and incompletely with Apal, to give a sDNA
fragment bounded by the Apal site at the 5' end of the sequence and HindIII from the plasmid vector. The relevant domain for the construction is shown below.

'~(3~ '75 -7 -l I 6 L A G V G V G G P N I C T
.... CTG GC& G{iC CiTT GGC GTA GGA GGG CCC AAC ATC TGT ACC ...
.... GAC CGC CCG CAA CC~ CAT CCT CCC GGG TTG TAG ACA TGG ....
S EcoRI Apal Hindlll Synthetic complementary oli~onucleotides were used to reconstruct a full-length coding construct for GPIIIa based on the published cloned cDNA sequence (Fitzgerald et al., "J. Riol. Chem." ~(9):3936 ~1987]). The oligonucleotide sequence, ending in ApaI, was ligated to the Apal site of the above ~paI-~lindlll fragment, to 2ive a DNA fragment now bounded by EcoRI and HindlII. This EcoRI to Hindlll fragment, containing the entire coding region for GPllla was ligated into pRK5 which had been digested with EcoRI and HindllI, and expression vector GPllla-pRK5 was recovered. The relevant oligonucleotide sequences are shown below.

M R A R P R P R P L W
AAT TCT AGA GCC GCC ATG AGA GCA CGT CCT CGA CCA CGT CCT CTC TGG -EcoRI
XbaI
-I
A T V L A L G A L A G V G V G G
GCG ACT GTG CTG GCA CTG GGA GCA CTG GCT GGT GTT GGA GTA GGA GGG CC
CGC TGA CAC GAC CGT GAC CCT CGT GAC CGA CCA CAA CCT CAT CCT C
Apal The synthetic oligonucleotides were designed such that the amino acids encoded were identical to those predicted îrom the published cloned cDNAs (Fitzgerald ef al., Rosa et al.), but the codons were not always identical with the naturally-occurring cloned cDNA. Fig 3 compares the codin~ strands of the synthetic and natural sequences. As~erisks between 35 each sequence indicate which nucleotides are identical. These changes were introduced for three reasons.
1. In light of difficulties encountered in sequencing the cDNA, we concluded that the DNA could contain seconda}y structure adver~e to translational efficiency. To minimize possible secondary structure in the mRNA produced from e~tpression constructs, the 40 percenta~e of G and C bases il~ the natural coding sequence was lessened by changing some codons to others which had a lower G and/or C content, but which coded for the same amino acid. These altered codons were chosen such tha~ only codons used frequen~ly in the remainder of the cDNA were substituted. Karnick et al., "1. Biol. Chem. 262(5):9255 ~1987); Devlin et al., ~Gene" 65:13 (1988).
45 2. The codon for arginine (R, amino acid -25), immediately following the initiator methionine codon (M -26), was changed from CGA to AGA. Kozak, "Nucl. Acids Res.
15(20):8125 [1987] and Kozak, "J. Mol. Biol.U 196:947 119871.
3. The DNA sequence upstream of the initiator methionine codon was not based on :, , ;,~()~3~475 ~25 -the natural D~IA sequence. The synthetic complementary oligonucleotides were such that an EcoRI site was present a~ one end, followed by an Xbal recognition sequence, and ~hen followed by a GCC GCC motif immediately upstream of the initiator methionine. Kozak, ~. Mol. Biol." Id.
S The plasmids encoding GPllb and GPllla (GPllb-pRK5 and GPllla-pRK5) were transfected in 293S cells and culturcd under conventional conditions for transient e~pression as described below. The cells were harvested and analyzed for ~PIIb-l~la expression.
E~tpression was confirmed by the presence of correctly sized bands on a Western gel, immunologically visualized by FACS sorting, and immunoprecipitation of intact cells labeled rnetabolicaily with S3s or by 1231 surface labelling.
Example 2 Construction of çDNA E codinQ Truncated GPllk The starting point for the construction of ~he GPllb truncated form was the full-len~th coding construction for GPllb described in Example 1. The relevant domain for this construction is shown below.
putative transmembrane region L R A L E E R A
..... CTC CGG GCC TTG GAG GAG AGG GCC ATT
..... GAG GCC CGG AAC CTC CTC TCC CGG TAA
EcoRI StyI
The DNA fragment from ~he EcoRI site ~upstream of the initiator ATG codon) to the S~yl site indicated above was isolated and ligated to complementary synthetic oligonucleotides such that the DNA sequence thus obtained coded for the natural GPlIb sequence up to amino acid residue 962 (arginine) and was then followed by a TGA stop codon.
A L E E R Stop ". C TTG GAG GAG AGG TGA TGA A
CTC CTC TCC ACT ACT TTC GA
StyI ~indIII
In the na~ural sequence, arginine 962 is followed by an approximately 26 amino acid putative hydrophobic transmembrane domain and a cytopiasmic domain (Poncz et al.).
Thus, in this constsuction both of these domains have been deleted from the coding region 35 of the construction. The end of the synthetic fragment terminated in ~ Hindlll restriction site. The entire DNA fragment bounded by EcoRI and Hindlll restriction sites w~s ligated into pRK5 which had been digested with EcoRI and HindllI. Expression vector GPllbtrunc-pRK5 was recovered.
The EcoRI to Hindlll fragment outlined above was rescued from GPllbtrunc-pRK5 40 and subjected to analysis by DNA sequencing. Over 250 bases from each end of the insert were sequenced and corresponded exactly to that which was predicted.

s Constru~tion of cDNA Enç~dinR Truncated GPlTla The startin~ point for the construction of the GPllla truncated form was the full-length codin~ construction for GPllla described in Example I . The relevant domain for this construction is shown below.
putative trsnsmem~rane region P K G P D I L L
~.~.... .........CCC AAG GGC CCT GAC ATC CTG GTG
..... GGG TTC CCG GGA CTG TAG GAC CAC
Xb~I ApaI

The DNA fsagment from the Xbal site (upstream of the initiator ATG codon) to theAp~al site indica~ed below was isolated and ligated to compl~mentary synthetic 15 oligonucleotides such that the DNA sequence thus obtained coded for the natural GPIlla sequence up to amino acid residue 692 (aspartic acid) and was then followed by a TGA
stop codon.
G P D Stop CT GAC TGA TGA GAT CTA
CCG GGA CTG ACT ACT CTA GAT TCG A
ApaI ~indIII

In the natural sequence, aspartic acid 692 is followed by an approximately 29 arnino acid putative hydrophobic transmembrane domain and a cytoplasmic domain (Fitzgerald el al.) Thus, in this construction both of these domains have been deleted from the coding region of the construction. The end of the synthetic fragment terminated in a Hindlll restriction site. The entire fragment bounded by Xbal and HindIII JestrictioD sites, was 30 liga~ed into pRK5 previously digested with Xba5 and ~indlll and trunc expression vector GPlllatrunc-pRK5 was recovered.
The Xbal to Hindlll fragment outlilaed above was rescued from GPlllatrunc-pRK5, and subjected to analysis by DNA se~uencing. Over 200 bases from each es~d of the insert were sequenced and corresponded exactly to thar which was predicted.
35 E~res~n_f Truncated Human (~PII~-llla R~ceptor in a EukarYoticHost Human embryonic kidney cells (293S) wer¢ cotransf~cted with the expression vectors GPllbtrunc-pRKS and GPlllatrunc-pR~C5 using CaPO~, (Graham et al., "Virology" ~456 11973]) using the host system described in EP 260,148.
Transient ExDression High levels of transient e~tpression were obtained when 2g3s cells were cotransfected with GPllbtrunc-pRK5, GPlllatrunc pRK5 and adenovirus VA RNA-I:INA (European Appln. Pub. No. 309,237 published 29 March 1989; Akusjarvietal, "Mol. Cell. Biol." 7:549 11987]~ and grown in standard ~rowth media (S0% Dulbeccos Modified Eagle Media, 50%
F12 mi~ture, 2 mM L-glutan)ine and 10~ fe~al bovine serum). 16 hours after glycerol 45 shock cells were transferred to serum free media (Dulbeccos Modified Eagle Media, 0.1%

6~7~

glucose, 10 ~g/ml insulin) and grown for a further 48 hours, a~ whish time cells and culture media were harvested. Conditioned cell culture fluid was centrifuged to remove contaminating cell debris and then quick frozen in dry ice-ethanol and stored at -70C until analyzed. Cells were removed irom 6 cm plates by suspension in 0.6 ml of 1~0 mM NaCI, 10 rnM Tris ~pH 7.5), lqb Triton X-100, 2 mM PMSF, 0.5 ~g/ml leupeptin and 2 ~g/ml pepstatin A followed by extraction for 30 minutes on ice with vortexing. Cellular debris was removed by centrifugation at 10,000 8 and samples stored at -70C. The soluble GPllb-llla was recovered by Q-Sepharose (fast-flow) chromatography with 10 column volumes of 20 mM MES buffer/lmM CaCI2 pH 6.5 and gradient elution over 0-400 mM
NaC1. The peak soluble GPllb-llla tended ~o elute at about 200-250 mM NaCI. The eluate was concentrated to 3% of the column volume of an S-300 column, after which the concentrate was exclusion chromatographed on the a-350 column using 10 mM Tris/l SOmM
NaCI/lmM CaC12 pH 7.5. Some of the full length GPllb transfected into 293S cellsassociated with endogenous ~v. The secretion of soluble GPllb with soluble GPlIla avoided the need to purify BPllb-llla from the ~I~,B3 vitronectin receptor, as would have been th~
case if the full length subuni~s had been used. See Bodary e~ al., J. Biol. Chem. 32:18859 (November 15, 1989).
Stable Expression Stable 293S clones expressing truncated GPllb-llla were established by co-transfection of GPllbtrunc-pRK~ and GPlllatrunc-pRK5 with pRSVneo (Gorman el al., "Science" 221:5~1-552 11983~). Forty eight hours after transfection cells were passaged into standard growth media containing 800 ~g/ml of G418. Two weeks later, G418 resis~ant clones were picked and grown in standard growth media containing 400 ~g/ml of G4]8.
Clones were grown for 48 hours in serum free medium and the conditioned culture medium assayed for the expression of secreted forms of GPllb-llla by Western blot analysis.
Analvsis of Expressed TruncatedQPllb-llla Transiently transfected cells were assayed for expression by pulse-chase analysis followed by immunoprecipitation using a panel of monoclonal antibodies generated against purified platelet GPllb-Illa. S3s-cysteine and -methionine metabolically labeled psoteins were recovered from the culture fluid of cells cotransfected with both GPllbtrunc-pRK5 and GPlllatrunc-pRK5 as described above. Truncated GPIlb-Illa was immunoprecipitated from cell culture fluid with a panel of mouse monoclonal antibodies (AP2 [Montgomery e~
al., "]. Clin. Invest." 71:385 (1983~], 2D2, 3A8, 4B12, and AP3 INewman et al., "Blood"
65:227 (1985)]) by incubation with Protein A Sepharose CL4B (Pharmacia), bound to rabbit IgG antibodies directed against mouse IgG. Electrophoresis of the immunoprecipitated proteins demonstrated the secretion of recombinant truncated GPIlb-llla whose size was in agreement with the molecular weights e~pected of the modified cDNAs. Monoclonal antibodies specific to the GPllb-Tlla complex (AP2), GPllb (2D2, 3A~) and GPllla (4B12, AP3) all immunoprecipitate both the GPllb and GPllla trurwated proteins, demonstrating that the recombinant secreted proteins are present in the form of a complex. Cells which received no DNA or the GPllbtrunc-pRK5 alone or GPlllatrunc-pRK5 alone do not secrete proteins at levels which a~e detectable by monoclonal antibodies to GPllb or GPllla.
The expression of individual subunits of GPllb or GPllla in transiently transfected cells was demonstrated usin~ YVestern blot analysis. Cells were extracted as described above and culture media (recovered as above~ were concentrated 2-fold by ultrafiltration and analyzed by electrophoresis on polyacrylamide gels (Laemmli, U.K., "Nature~ 227:680-6BS [19701) and by Western Blotting (Towbin et al., Proc.Natl.Acad.Sci.USA 76:4350-4354 [1979]). Mouse monoclonal antibodies specific for GPllb and GPlIla were used in this analysis. Horse radish pero~idase-conjugated antibodies directed against the murine monoclonals were used to visualize the individual GPllb~runc and GPllIatrunc proteins in the e~tracts. ~
The stable clones e~pressing the GPllb-llla truncated constructs were shown to secrete the recombinant pro~eins of the expected sizes using Western blot analysis.
That the GPllb-llla trunc proteins secreted from stable clones were p~esent as acomplex was demonstrated by their detection, after direct transfer of culture medium to nitrocellulose by aspiration, with monoclonal antibody AP2.
The truncated GPllb or GPllla proteins were not detected in culture media when expressed as individual subunits: either they are not secreted or the efficiency of secretion is reduced to levels which preclude detection by immunoprecipitation or by Western blot analysis.
Example 3 Demonstration of Fibrino~en Bindin~ of Secreted Hurnan GPllb-llla Polv~eDtide ComPlex The functional activity of the secre~ed truncated GPllb-llla is shown by its specific absorption to an affinity matrix containin~ the natural ligand, fjbrinogen, for the GPlIb-llla recep~or.
A stable clone from Example 2 which was expressing the GPIIb-llla truncated ` ~. polypeptide complex was gsown for 20 hours under serum free conditions (DMEM culture medium, 0.1% glucose, 10 ,ug/ml insulin, 1.5 ,ug/ml L-cysteine, 2.4 ~g/ml L-methionine, 200 ~Ci/ml S35 methionine and 200 ~Ci/ml S3s cysteine). The conditioned cell culture fluid was first concentrated by ultrafiltration then purified by fibrinogen affinity chromatography. The fibrinogen affinity column was produced by coupling highly purified human fibrinogen to CNBr-actiYated Sepharose 4B (Pharmacia) using the manufacturer's recommended procedure. The concentrated cell cu1ture fluid was applied first to a control Tris/ethanolamine reacted CNBr-activated Sepharose 4B column and the unbound material applied directly to the fibrinogen-Sepharose column. The contaminating proteins were washed away at ~oom temperature with phosphate buffered saline solution containing I mM
Ca2', ImM Mg2~, 25 mM octy1glucoside (OG) and 2 mM phenylmethylsulfonylfluoride (PMSF). The bound GPIlb-lIla was eluted from the column at room temperature withphosphate buffered saline containing 15 mM EDTA, 25 mM OG and 2 mM PMSF. The eluted GPllb-llla was then concentrated by ultrafiltracion and the subunits of expected molecular weight identified by autoradiography and by Western blot analysis using monoclonal antibodies specific to GPllb (3A8) and GPllla ~4Bl2). The specificity of the 7~

binding to the fibrinogen column is shown by the absence of the protein in the eluate from the control column determined by both methods.
Example 4 ExDression of LFA-l and Mac-l ~runca~ions LFA-J and Mac-l are inteRrins having identical beta chains (beta-2) and distinctalpha chains (alpha L and alpha M, respectively). In this study the full length chains were transformed into host cells. In addition, the DNA encoding the transmembrane domains of the alpha and beta chains of each of these integrins was deleted and the truncated DNAs transformed into host cells for coe~pression.
Transformants with full length LFA- I alphaL chain did not e~press any detectable cell bound alphaL, but cotranformation with truncated alpha~ and truncated beta-2, or with truncated alphaM and truncated beta-2, resulted in the secretion of the truncated heterodimers. Interestingly, transformation with the full length alphaM chain of Mac-1 alone did yield cell surface alphaM. It has not been confirmed that this product represents a stable alphaM monomer since it is conceivable that the recombinant alphaM chain became associated with a beta chain endogenous to the host cell.

Claims (20)

1. A method for the preparation of a soluble analogue of a multiple subunit polypeptide (MSP) comprising 1) introducing into the nucleic acid encoding one of the subunits a mutation such that the mutated nucleic acid encodes an amino acid sequence variant of the MSP that renders the MSP no longer capable of becoming lodged in a cell membrane, 2) transforming a host cell with the nucleic acid of step 1, 3) culturing the host cell of step 2 and 4) recovering from the host cell culture biologically active soluble MSP.

2. The method of claim 1 wherein the subunit in nature contained a membrane anchor domain and the mutation serves to delete or modify sufficient of the membrane anchor domain to render the anchor domain no longer sufficiently hydrophobic to anchor the MSP
analogue in a cell membrane.

3. The method of claim 1 wherein the subunits are noncovavalently associated.

4. The method of claim 1 wherein the MSP is a heterodimeric receptor that participates directly in intercellular adhesion or adhesion of cells to extracellular matrix proteins.

5. The method of claim 4 wherein the MSP is p150,95, Macl, LFA-1 or GPIIb-IIIa.

6. The method of claim 2 wherein the mutation is the substitution of sufficient of a membrane anchor domain of a first subunit with a sufficiently hydrophilic amino acid sequence that the MSP no longer is capable of becoming lodged in a cell membrane.

7. The method of claim 6 wherein the hydrophillic amino acid sequence comprises an immunoglobulin constant domain.

8. The method of claim 1 wherein the mutation does not comprise the introductionof DNA encoding a multimer forming polypeptide into the DNA encoding the MSP subunit.

9. The method of claim 7 wherein the membrane anchor domain is a transmembrane domain.

10. The method of claim 4 wherein the MSP is a leukocyte adhesion receptor.

11. The method of claim 4 wherein the MSP is a member of the VLA family.

12. The method of claim 5 wherein the MSP is GPIIb-IIIa and the recovered GPIIb-IIIa is capable of binding fibrinogen, fibronectin, vitronectin or von Willebrand factor.

13. The method of claim 1 wherein the MSP is capable of binding to polypeptides containing the sequence RGD.

14. The method of claim 2 wherein the mutation comprises inserting an amino acidinto, substituting or deleting an amino acid from the membrane anchor domain of at least one subunit of the MSP.

15. The method of claim 12 wherein the membrane anchor domain is deleted.

16. The method of claim 15 wherein the cytoplasmic domain is deleted.

17. The method of claim 1 wherein the MSP comprises a subunit having the consensus N-terminal sequence Tyr/Phe/Leu-Asn-Leu-Asp, requires a divalent cation for ligand binding, or contains an amino acid domain having substantial sequence homology to the calmodulin calcium binding domain.

18. The method of claim 9 further comprising transforming the host cell with DNAencoding a second subunit wherein the DNA encoding the second subunit mutated to delete a membrane anchor domain.

19. The method of claim 1 wherein the MSP contains multiple subunits, and DNA
encoding all of those subunits which in nature comprise membrane anchor domains are mutated as provided in step 1.

20. The method of claim 14 wherein one of the MSP subunits contains two disulfide bonded polypeptide chains, only one of which comprises a membrane anchor domain.21. The method of claim 7 wherein the constant domain is a heavy chain constant domain.
22. The method of claim 7 wherein the host cell is transformed with DNA encoding(1) a fusion of a first MSP chain and an immunoglobulin heavy chain constant domain and (2) with DNA encoding a second MSP chain that is not substituted with immunoglobulin, this second MSP chain being capable of binding directly to the first MSP chain.
23. An MSP polypeptide amino acid sequence variant that is not capable of becoming lodged in a cell membrane, provided that the MSP polypeptide comprises two variant chains and does not comprise an immunoglobulin constant domain.
24. An integrin amino acid sequence variant that is not capable of becoming lodged in a cell membrane.
25. The variant of claim 24 which is free of detergent.
26. A sterile aqueous solution of the variant of claim 24.
27. Nucleic acid encoding an insegrin subunit amino acid sequence variant from which sufficient of the transmembrane domain has been deleted that the variant is not capable of transmembrane insertion into a cell membrane.
28. A vector comprising the nucleic acid of claim 27.
29. A method for the preparation of GPIIb-IIIa comprising transforming a permissive host cell with nucleic acid encoding GPIIb-IIIa and culturing the host cell until GPIIb-IIIa accumulates in the cell membrane.
30. The method of claim 29 wherein the domain GCC GCC is present immediately 5' of the initiation methionine codon for the nucleic acid encoding GPIIIa.
31. The method of claim 29 wherein the nucleic acid encoding GpIIIa encodes preGPIIIa and wherein the codon for arginine-25 in the preGPIIIa signal sequence is AGA.
32. The method of claim 29 wherein the percentage of G and C within the first about 100 bases of the 5' end of the nucleic acid encoding GPIIIa has been reduced below the percentage of G and C in the GPIIIa cDNA.
33. A soluble amino acid sequence analogue of an MSP not capable of being lodgedin a cell membrane, comprising (1) a first MSP chain fused at its C-terminus to a sequence comprising an immunoglobulin constant domain and (2) a second MSP chain which is not fused to an immunoglobulin constant domain.
34. The MSP of claim 33 wherein the MSP chains are disulfide bonded.
35. The MSP of claim 33 further comprising an unfused immunoglobulin chain.
36. The MSP of claim 35 wherein the unfused chain is a light chain having its variable domain deleted and the fused constant domain is a heavy chain constant domain.
37. The MSP of claim 33 wherein the transmembrane domain of the first MSP chain is deleted.
38. The secreted analogue of claim 35 wherein the unfused immunoglobulin chain comprises a variable domain capable of binding a known antigen, the variable domain-bearing immunoglobulin chain being disulfide bonded to the fusion of the immunoglobulin constant domain and the first MSP chain.
39. The secreted analogue of claim 38 comprising disulfide bonded immunoglobulinheavy and light chains capable of binding a known antigen in turn disulfide bonded to the fusion of the immunoglobulin constant domain and the first MSP chain.
40. The secreted analogue of claim 33 wherein the MSP chains are not members of the immunoglobulin superfamily.
41. A recombinant host cell transformed with DNA encoding a soluble amino acid sequence analogue of an MSP, comprising (1) a first MSP chain fused at its C-terminus to a sequence comprising an immunoglobulin constant domain and (2) a second MSP chain which is not fused to an immunoglobulin constant domain.
42. A method comprising culturing the host cell of claim 41 and recovering the soluble analogue from the host cell culture.