CA2054608A1 - Fusion proteins comprising gm-csf and il-3 - Google Patents

Abstract

TITLE
Fusion Proteins Comprising GM-CSF and IL-3 A fusion protein is disclosed which comprises GM-CSF and IL-3. Such fusion proteins more biologically active than GM-CSF or IL-3 alone or or GM-CSF and IL-3 combined.

Description

2 ~

IM ~nUNEX CO RF~RATIO N 0520-CA

Fusion Proteins Comprising GM-CSF and IL-3 R~CK(~OUND OF THE INVENTTON
The present invention relates generasly to analogs of GM-CSF and IL-3 proteins, and particularly to the construction of fusion proteins comprising GM-CSF and IL-3.
The differentiation and proliferation of hematopoietic cells is regulateds by secreted glycoproteins collectively known as colony-stimulating factors (CSFs). In humans, these proteins include granulocyte-macrophage CSF (GM-CSF), which promotes granulocyte and macrophage production from normal bone marrow, and which SlSSO appears to regulate the activity of mature, differentiated granulocytes and macrophages. IL-3 (arsso known as multi-CSF) also stimulates formation of a broad range of hematop~ietic cells, including granulocytes, macrophages, eosinophils, mast cells, megakaryocytes and erythroid cells. GM-CSF and IL-3 thus have considerable overlap in their broad range of biological activities. Other CSFs have a more res~ricted range of activity, macrophage CSF (M-CSF) stimulating almost exclusively macrophage colony formation, and granulocyte CSF (G-CSF) stimulslting primarily granulocyte colonies. Although GM-CSF and IL-3 have distinct amino acid sequences, 2() preclinical studies indicate that they may be useful to treat various cytopenias, and to potentiate immune responsiveness to infectious pathogens, and to assist in reconstituting normal blood cell populations following viral infection or radiation or chemotherapy-induced hemsltopoietic cell suppression. The genes encoding GM-CSF and IL-3 are located on the same chromosome in mouse and man and the expression of the genes is lillked in some cells, sllCIl as activlltcd r lymphocytes (Kelso et ul., .1. Iotmunol. J3~5:171X, I'~R~; Y;ln~ ct sll., Blo-)~l 71:95X, 19X8;
Bnrlow et al., EMBO .1. ~5:fil7, 1')X7).
Short-term experimcnts have dcmonstratcd thslt thc sirnulîancous combinatiQn of GM-CSF; und IL-3 was morc erfcctivc th;lll cithcr GM-CSF orll,-3 aloncillillcrc.lsingcyclill~r~l~es and numbers of m;lrlow hemsltopoictic progclliîor cclls in vitro in lactofcrrin-trellted mice 3() (Brc~xmeyer et al., l~roc. N~tl. ~c~l(l. 5'ci. USA ,Y4:3X71, t9~7). No sucll ~yncrgy hsls becn observed in vivo t`or simllltancolls adlllillistr;ltioll of GM-(:SF alld IL-3, althollgh clinical stlldieS hnve ShOWIl th;lt the consec~ive sldtninis~ration of reco mbin.lnt human IL-3 and recombinantlllltmlll(3M-CSF wsls more effective in raisin~ white blood cell counts in normal cynomolgus monkeysthalleither GM-CSF or IL-3 alone (Krumwieh et al., Behring In.st. Mi~t.
.Y3:250, 1988; Donahue et al., Science 241:1820, 1988).
The biological nctivities of GM-CSF and IL-3 are mediated by binding to specific cell surface receptors expressed on primary cells and in vitro cell lines. GM-CSF and IL-3 each bind to their respective receptor, resulting in transduction of a biological signal to various immune effector cells. Recenî studies of the characteristics and distribution of the receptor for 2a~6~8 IL-3 on the human myelogenous leukemia cell line KG-I and human pre-B cell line JM-I
indicate that a subclass of receptor exists w'nich also binds (:iM-CSF (Park et al., J. Biol.
Chem. 264:5420, 1989). These studies showed that human GM-CSF is capable of almost completely inhibiting t'he binding of 125I-IL-3 to KG-l cells and, conversely, that IL-3 is 5 capable of substantially inhibiting binding of 125I-GM-CSF to the same cells. This direct competition between GM-CSF and IL-3 for a single cell surface receptor indicates that a single receptor is capable of binding both GM-CSF and IL-3. Although is not yet clear whether the heterogeneity in IL-3 and GM-CSF binding is due to the existence of a receptor molecule which is distinct from that which binds IL-3 alone or GM-CSF alone, or whether IL-3 and 10 GM-CSF receptors may exist as multisubunit complexes, composed of different ratios of IL-3 and GM-CSF binding proteins, the receptor(s) will be referred to herein as the GM-CSF/IL-3 receptor.

SUMMARY OF THE TNyENTION
The present invention is a fusion protein comprising GM-CSF and IL-3. The fusionproteins have a forrnula selected from the group consisting of RI-R2~ R2-R1, RI-L-R2 and R2-L-R1 2() wherein Rl is GM-CSF; R2 is IL-3; and L is a linker peptide sequence. In preferreed aspects of the present invention, GM-CSF and IL-3 are linked together via a linker sequence which does not interfere with the folding of eithcr the GM-CSF or IL-3 domains.
The fusion proteins of the present invention are morc biolo~,ical1y active than GM-CSF
or IL-3 alone or in combination and, relative to IL-3, have a significantly hiFher affinity of 25 bindinF to cell lines which have GM-CSr/lL-3 rcccptors comparcd to ccll lines with only IL-3 or GM-CSF receptorx.

I~RIII;.F l)'~S(~R1r'1'1~N (~` T11F nRl~W11~1(.S
FIGURE~ I is a n1lcleotide sc(1llcncc and correspondil)&~, amino ackl se~luencc of a 3() hlllllall GM-CSF/11-3 IllSiOIl prOtCill. 'I`I1C C-terl11inus of humnn GM-C~SF tamino acids 1-127) is 1inked to the N-tcrmin1ls of hlllll;ln ll.-?i (llmino acids 13'~-271) via al linker sequence (lllllillO acids i 2X- 1 3X).
FlGUtRF. 2 is n nllcleotidc sc(111cllcc nnd correspondin~ amino acid sequence of a hulllall 1L-3/GM-CS'r l;!sioll protein. 'I'he C-terminus of hum.ln IL-3 (amino acids 1-133) is linked to the N-termin11s of hllm.lll G'lvl-CS'F (amino acids 14g-275 via a linker sequence (amillo acids 134-14~).
FIGURE 3A-3D are graphs showing that the GM-CSF/IL-3 fusion protein (O) enhances BFU-E (FIG. 3A), CFU-GM (FIGS. 3B and 3C) and CFU-GEMM (FIG. 3D) colony formation relative to IL-3 (--) or GM-CSF (O) alone, or to IL-3 and GM-CSF
combined (~).

DETAILE:D ~E~CRIPrION OF'rHE INVENTION
efinitions The term "GM-CSF" refers to proteins having amino acid sequences which are S substantially similar to the native human granulocyte-macrophage colony-stimulating factor amino acid sequences (e.g., ATCC 53157) and which are biologically active in that they are capable of binding to GM-CSF receptors, transducing a biological signal initiated by binding GM-CSF receptors, or cross-reacting with anti-GM-CSF antibodies raised against GM-CSF.
Such sequences are disclosed, for example, in Anderson et al. (Proc. Nat'l. Acad. Sci. USA
82:6250, 1985). The term "GM-CSF" also includes analogs of GM-CSF molecules which exhibit at least some biological activity in common with native human GM-CSF. Exemplary analogs of GM-CSF are disclosed in EP Publ. No. 212914, which describes GM-CSF analogs having KEX2 protease cleavage sites inactivated so as to increase expression of GM-CSF in yeast hosts, and in WO Publ. No. 89/03881, which describes GM-CSF analogs havingvarious glycosylation sites eliminated. Other GM-CSF analogs which are described herein may also be used to construct fusion proteins with IL-3. Furtherrnore, those skilled in the art of mutagenesis will appreciate that other analogs, as yet undisclosed or undiscovered, may be used to construct GM-CSF/IL-3 fusion proteins as described herein. A DNA sequence encoding a particularly preferred GM-CSF protein having potential glycosylation sites removed 2() has been deposited with the American Type Culture Collection under accession number ATCC
67231 (GM-CSFLLeu23Asp27Glu39l). The nomenclature used herein to specify amino acid sequences desi~nates amino acids that differ from the native form in brackets immediately following the protein name and designates the species with which the protein is associated immediately preceding the protein name. Thus, huGM-CSFlLeu23Asp27Glu29l refers to a humlm GM-CSF in which amino ucid 23 has been changcd to a leucine residue, amino acid 27 has been changed to all asparaginc rcsidue, and amino acid 2~) has been chall~ed to glutamic acid residue.
The term "lL-3" refers to protcins h.lving amillo llcid seg(lenccs which alre subst.lntially similar to the native h~ltnan lnterleukin-3 lltllillO llCid SCqllCllCCS alld which arc biologically 3() active in that they are capnble of bindill~ ~o IL-3 recel)tors or trallsdllcin~ ~l biological si~nal iniliated by bindin~ to IL-3 r~ceptors, or cross-rcaclin~ with anti-lL-3 wltibodies raised against IL 3. Sucll se(lllences ale discloscd, for exalllple, in EP Publ. Nos. 275,59R and 282,lXS.
The term "lL-3" also incllldes lln~llo~s of 1l,-3 molccl~les which exllibit at least some biological activity in commoll witll native IL-3. Exemplary an;llogs of IL-3 are also disclosed in EP Pllbl.
No. 2X2,185. Particlllarly preferred forms of IL-3 which may be fused to GM-CSF in accordance with tlle present invention include hulL-3LPro8Aspl5Asp70l, hulL-3lSer8Aspl5Asp70l, and hulL-3lSer8]. A DNA sequence encoding another IL-3 protein suitable for incorporation into fusion proteins as described herein is on deposit with ATCC
under accession number ATCC 67747.

2 ~ 8 As used herein, the terrn "fusion protein" refers to a C-terminal to N-terminal fusion of GM-CSF and IL-3. The fusion proteins of the present invention include constructs in which the C-terminal portion of GM-CSF is fused to the N-terminal portion of IL-3, and also constructs in which the C-terminal po~tion of IL-3 is fused to the N-terminal portion of GM-CSF. Specifically, the fusion proteins of the present invention have a formula selected from the group consisting of R1-R2, R2-Rl, Rl-L-R2 and R2-L-RI
wherein R1 is GM-CSF; R2 is IL-3; and L is a linker peptide sequence. GM-CSF is linked to IL-3 in such a manner as to produce a single protein which retains the biological activity of GM-CSF and IL-3. Specific fusion protein constructs are named by listing the GM-CSF and Il,-3 domains in the fusion protein in their order of occurrence twith the N terminal domain specified first, followed by the C-terminal domain). Thus, GM-CSF/IL-3 refers to a fusion protein eomprising GM-CSF followed by IL-3 (i.e., the C-terminus of GM-CSF is linked to the N-terminus of IL-3). Unless otherwise specified, the terms GM-CSF/IL-3 and IL-3/GM-CSF refer to fusion proteins with a linker sequence added. Similarly, huGM-CSF[Leu23Asp27Glu39l/hulL-3[Pro8Aspl5Asp701 refers to a fusion protein in which the N-terminal region of the fusion construct is huGM-CSFLLeu23Asp27Glu39l, and the C-terminal region is huIL-3[Pro8Aspl5Asp701.
The term "substantially identical," when used to define either amino acid or nucleic acid sequences, means that a particular subject sequenee, for example, a mutant sequence, is substantially full-length and varies from the seqllenee of Figures I or 2 by one or more substitutions, deletions, or additions, the net effect of whieh is to retain biologieal activity of the protein when derived as a GM-CSF/IL-3 or IL-3/GM-CSF fusion protein. Alternatively, DNA analog sequences are "substantially identical" to the speeifie DNA sequenees diselosed herein if: (a) the DNA smalog sequenee is derived from substantially the entire eoding regions of the native mammsllinn GM-CSP and 11.-3 genes; or (b) the DNA analog sequenee is eomparable in length with and capable of hybridi~ation ~o DNA sequences of (1) under 3() moderately strin~ent conditions and wllich encode biologieally aetive GM-CSF or IL-3 moleeules; or (e) DNA sequel1ees whicll ure degenerlte us a result of the ~enetie code to the DNA analog sequellces defined in (a) or (b) nlld whieh eneode biolo~ically aetive GM-CSF or IL-3 molecules. Substnl1tially identical analog proteins will be greater than ubout X0 pereent similar to the eorresponding sequence of the native protein. Sequenees having lesser degrees of similarity but comparable biologieal aetivity are eonsidered to be equivalents. In defining nucleic acid sequences, all subject nucleic acid sequences capable of encoding substantially similar amino acid seqllences are eonsidered substantially similar to a referenee nucleie aeid sequence.
Pereent similurity may be determined, for example, by comparing sequence information using the GAP computer program, version 6.0, available from the University of Wisconsin 2 ~ g Genetics Computer Group (UWGCG). The GAP program utilizes the alignment method of Needleman and Wunsch (J. Mol. Biol. 48:443, 1970), as revised by Smith and Watennan (Adv. Appl. Math. 2:482, 1981). Briefly, the ~AP program defines similarity as the number of aligned symbols (i.e., nucleotides or amino acids) which are similar, divided by the total 5 number of symbols in the shorter of the two sequences. The preferred default parameters for the GAP program include: (1) a unary comparison matrix (containing a value of 1 for identities and 0 for non-identities) for nucleotides, and the weighted comparison matrix of Gribskov and Burgess, Nucl. Acids Res. 14:6745, 1986, as described by Schwartz and Dayhoff, eds., A~las of Protein Seq~ence and Structure, National Biomedical Research Foundation, pp. 353-358, 1979; (2~ a penalty of 3.0 for each gap and an additional 0.10 penalty for each symbol in each gap; and (3) no penalty for end ~aps.
"Recombinant," as used herein, means that a protein is derived from recombinant (e.g., microbial or mammalian) expression systems. "Microbial" refers to recombinant proteins made in bacterial or fungal (e.g., yeast) expression systems. As a product, "recombinant microbial" defines a protein produced in a microbial expression system which is essentially free of native endogenous substances. Protein expressed in most bacterial cultures, e.g., E. coli, will be free of glycan. Protein expressed in yeast may have a glycosylation pattern different from that expressed in mammalian cells.
"Piologically active," as used throughout the specification means lhat a particular 2() molecule shares sufficient amino acid sequence similarity with the embodiments of the present invention disclosed herein to be capable of binding to GM-CSF receptor, IL-3 receptor or GM-CSF/IL-3 receptor (see, e.g., Park et al., J. Biol. Chem. 264:5420, 1989), transmitting a GM-CSF ~md/or IL-3 stimulus to a cell, or cross-reacting witll antibodies raised against GM-CSF or IL-3.
"DNA sequence" refers ~o a DNA polymer, in thc form of a separ.lte fragment or as a component of a largcr DNA constrllct. Prefcrably, the DNA se(lucllces are in a qllalltity or concentration enllblillg idenlificutioll, mllniplll;ltion~ and récovcry of thc sequencc and its component nucleotide scqllences by stallllard biochemiclll methods, for cx~lmplc, usin~ a cloning vector. Sucll se~lllel1ce~; are prefcrslbly providc(l hl thc forrn of lln open reading frnmc 3() ullinterrllpted by hlterll~ll nontr;lllslntcd sc~lllcllces, or introns, which are typicully prcscnt in euk:lryotic genes. ~enomic DNA contslillillg thc rclevan~ se(luences could also be used.
Sequellces of noll-~r(lllsl:lted l~NA m,ly hc prescllt 5' or 3' from the open reading frame, where the satne do not interfere with mnlliplll.ltion or expression of the codin~ regions.
"Nucleotide se(luence" refers to a heteropolymer of deoxyribonucleotides. DNA
sequences encoding the proteins provided of this invention can be assembled from cDNA
fragments and short oli~onucleotide linkers, or from u series of oligonucleotides, to provide a synthetic gene which is capable of being expressed in a recombinant transcripdonal unit.
"Recombinant expression vector" refers to a replicable DNA construct used either to amplify or to express DNA which encodes the fusion proteins of the present invention and 2 0 S Q ,~

which includes a transcriptional unit comprising an assembly of (I) a genetic element or elements having a regulatory role in gene expression, for example, promoters or enhancers, (2) a structural or coding sequence which is transcribed into mRNA and translated into protein, and (3) appropriate transcription and translation initiation and termination sequences.
5 Structural elements intended for use in yeast expression systems preferably include a leader sequence enabling extracellular secretion of transl~ted protein by a host cell. Alternatively, where recombinant protein is expressed without a leader or transport sequence, it may include an N-terminal methionine residue. This residue may optionally be subsequently cleaved from the expressed recombinant protein to provide a final product. "Recombinant microbial 10 expression system" means a substantially homogeneous monoculture of suitable host microorganisms, for example, bacteria such as E. coli or yeast such as S. cerevisiae, which have stably integrated a recombinant transcriptional unit into chromosomal DNA or carry the recombinant transcriptional unit as a component of a resident plasmid. Generally, cells constituting the system are the progeny of a single ancestral transformant. Recombinant 15 expression systems as defined herein will express heterologous protein upon induction of the regulatory elements linked to thc DNA sequence or synthetic gene to be expressed.

Construction of cDNA Sequences Encoding Fusion Proteins Com~rising GM-(: S~ and IL-3 A DNA sequence encoding ia fusion protein is constructed using recombinant DNA
2() techniques to assemble separate DNA fragments encoding GM-CSF and IL-3 into an appropriate expression vector. The 3' end of a DNA fragment encoding GM-CSF is ligated to the S' end of the DNA fragment encoding 11,-3, with the reilding frames of the sequences in pllase to permit mRNA lransl1ltion of the seqlIences into a sin~lc biolo~ically active fusion protein. The resultinL~ protein is hllGM-(:Sl~ll,eu23Asp27~ 3')l/lllIlL-3lPro8AsplsAsp7()l~
25 Alternatively, the 3' end of a ONA fragmcl)t cncoding IL,-3 m,ly be li~utcd to the 5' end of the DNA fragment encodill~ (JM-CSI~, wilh thc rcslding frslmcs of lhc scquences in phnsc to permit mRNA translIltioll of the scqllellces into a singlc biolo~ically active fusion protein, yieldin~ the prolein hulL-31PIa~Aspl~Asp7()l/hll~;M-CSl~lLcll2~Asp27Glu39l~ The regllliltory elements responsible for trallscription of l~NA intc) ml~NA arc rct;lined on the first of thc two 3() DNA sequellces, while binding siL~nuls or slOp c~xlolls~ which wolll(l prcvcnt read-throllgh to the second DNA seqllellce, l~re elimill1ltcd. Convcrscly, rc~ulatory clcmcllts are removcd from the second DNA scqllellce while stop codons rcqllircd to cnd trallsl;llion are retiaincd.
In preferred aspects of thc prcsent invcntioll, mcillls are provided for linking the GM-CSF and IL-3 domaills, preferably via a linker sequence. Thc linker sequence separates GM-35 CSF and IL-3 domslhls by a distance sufflcieIlt to ensure that each domain properly folds into its secondary and tertiary structures. Suitable linker sequences (I) will adopt a flexible extended conformation, ~2) will not exhibit a propensity for developing an ordered secondary structure which could interact with the functional GM-CSF and IL-3 domains, and (3) will have minimal hydrophobic or charged character which could promote interaction with the 2~5~
functional protein domains. Typical surface amino acids in flexible protein regions include Gly, Asn and Ser. Virtually any permutation of amino acid sequences containing Gly, Asn and Ser would be expected to satisfy the above criteria for a linker sequence. Other near neutral amino acids, such as Thr and Ala, may also be used in the linker sequence.
The length of the linker sequence may vary without significantly affecting the biological activity of the fusion protein. For example, the GM-CSF and IL-3 proteins may be directly fused without a linker sequence. Linker sequences are unnecessary where the proteins being fused have non-essential N- or C-terminal amino acid regions which can be used to separate the functional domains and prevent s~eric interference. In one preferred embodiment of the present invention, the C-terminus of GM-CSF may be directly fused to the N-terminus of IL-3. GM-CSF has six amino acids following the C-terminal cysteine residue, which is involved in disulfide bonding and is essential for proper folding of the protein. IL-3 has 15 amino acids preceding its N-terrninal cysteine residue. The combined terminal regions thus may provide sufficient separation to render the use of a linker sequence uilnecessary.
Generally, the two protein domains will be separated by a distance approximately equal to the small unit dimension of GM-CSF or IL-3 (i.e., approximately 0.38 nm, as determined by analogy with similar four-helix hormones). In a preferred aspect of the invention, a linker sequence len~th of abollt 11 amino acids is used to provide a suitable separation of functional protein domains, although longer linker sequences may also be used. The length of the linker 2() sequence separatinL~ GM-CSF and IL-3 is from I to 5()0 amino acids in length, or more preferably from I to I (X) amino acids in length. In the most preferred aspec~s of the present invention, the linker sequcnce is from about 1-2() amino acids in length. In the specific embodiments disclosed herein, the linkcr sequence is from about 5 to about 15 amino acids, and is advantageously from about 1() to about 15 .mlino acids. .~\mino acid sequences useful as linkers of GM-CS~ nn(l IL-3 inclllde, by way of example, (Gly4Ser)3 and Gly4SerGlysSer.
The linker se(luence is incorporated into thc filsion protein constr-lct by well known standnrd methods of mlllltgellesis as Iescribc(l below.

I`he present inventio1l provi(lcs a fusion pro~cin comprisill~ hl1m.ll1 GM-CSr~ and humllll lL-3. Derivatives of the fusioll protcins of lhc presellt invelltion also inclllde various structl1r~l1 forllls of thc prim,1ry proteill which retain biolo~ical aclivity. OlJC to the presence of ioni~.s1ble lmlino alld carboxyl groups, for cxample, a fusion protein may be in thc form of acidic or basic salts, or may be in nclltr:ll form. Individu~ll amino acid resid~les may also be modified by oxidation or reduction.
The primary amino acid stmctllre may be modified by forming covalent or aggregative conjllgates with other c~hemical moieties, such as glycosyl groups, lipids, phosphate, acetyl groups and the like, or by creating amino acid sequence mutants. Covalent derivatives are prepared by linking particular functional groups to amino acid side chains or at the N- or C-tennini. Other derivatives of the fusion protein within the scope of this invention include covalent or aggregative conjugates of the fusion protein with other proteins or polypeptides, such as by synthesis in recombinant culture as N- or C-terrninal fusions. For example, the conjugated peptide may be a signal (or leader) polypeptide sequence at the N-terminal region of 5 the protein which co-translationally or post-translationally directs transfer of the protein from its site of synthesis to its site of function inside or outside of the cell membrane or wall (e.g., the yeast a-factor leader). Peptides may also be added to faeilitate purification or identification of GM-CSF/IL-3 fusion proteins (e.g., poly-His). The amino acid sequence of the fusion protein can also be linked to the peptide Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys (DYKDI)DDK) (Hopp et al., BiolTechnology 6:1204, 1988.) The latter sequenee is highly antigenic and provides an epitope reversibly bound by a specific monoclonal antibody, enabling rapid assay and facile purification of expressed recombinant protein. This sequence is also specifieally cleaved by bovine mucosal enterokinase at the residue immediately following the Asp-Lys pairing. Fusion proteins capped with this peptide may also be resistant to intracellular 15 degradation in E. coli.
Fusion protein derivatives may also be used as immunogens, reagents in reeeptor-based immunoassays, or as binding agents for affinity purification procedures of binding ligands.
Derivatives may also be obtained by cross-linking agents, such as M-maleimidobenzoyl suceinimide ester and N-hydroxysuecinimide, at cysteine and Iysine residues. Fusion proteins 2() may also be covalently bound ~hrough reactive side groups to various insoluble substrates, such as cyanogen bromide-aetivated, bisoxirane-aetivated, earbonyldiimida~ole-aetivated or tosyl-aetivated agarose struetures, or by adsorbing to polyolefin surfaces (with or without glutaraldehyde eross-linking) The present invention also ineludes proteins with or without .Issociuted native-pattern 25 glyeosylation, Expression of DNAs encoding the fusion proteins in baeteria sueh as E. coli provides non-glyeosylated moleeules. Funetionul mlltant analogs huving inaetivated N-glyeosylation sites C~lll be produced by oligonueleotide syn~hesis and ligation or by site-speeifie mutagellesis teehni(llles~ 'I`hese ~malo~ proteins c~ln be produeed in a holllogelleous, redueed-earbohydralte form in good yield USill~ yeast expression systems. N-glyeosylation sites in 3() eukllryotie proteins are charaeteri~.e(l by Ihe llmhlo acid Iriplet Aso-AI-Z, where Al is any amino aeid exeept Pro, slnd Z is Ser or Tllr In this seqllence, asparagine provides a side ch,lin amillo group for covalellt slttachlllellt of carbohydr.lte Such a site can be eliminated by SUbStitUtillg anotl)er ;lmillO acid for Asn or for residue Z, deleting Asn or Z, or inserting a non-Z amino aeid between Al and Z, or an amino acid other than Asn between Asn and Al 35 Examples of human GM-CSF allalogs in which glycosylation sites have been removed include huGM-CSF[Lell23Asp27Glu391, huGM-CSF[Leu23J, huGM-CSFLLeu23Asp27], huGM-CSF[Glu39l, huGM-CSF[Asp27Glu39], huGM-CSF[Leu23Glu39J and huGM-CSF[Asp27]
l~xamples of human IL-3 analogs in which glyeosylation sites have been removed include 2 ~
huIL-3[Pro~AsplSAsp701, huiL-3[Asp70], hUII,-3[AsplsAsp7o]~ huIL-3[Pro~Aspl5], huIL-3lPro8Asp70], and huIL-3[Aspl5].
Derivatives and analogs may also be obtained by mutations of the fusion protein. A
derivative or analog, as referred to herein, is a polypeptide in which the GM-CSF and IL-3 domains are substantially homologous to full-length GM-CSF and IL-3 domains of the sequences disclosed in Figures 1 and 2 but which has an amino acid sequence difference attributable to a deletion, insertion or substitution.
Bioe4uivalent analogs of fusion proteins may be constructed by, for example, making valious substitutions of residues or sequences. For example, cysteine residues can be deleted or replaced with other amino acids to prevent formation of incorrect intramolecular disulfide bridges upon renaturation. Other approaches to mutagenesis involve modification of adjacent dibasic amino acid residues to enhance expression in yeast systems in which KEX2 protease activity is present. Generally, substitutions should be made conservatively; i.e., the most preferred substitute amino acids are those having physicochemical characteristics resembling those of the residue to be replaced. Similarly, when a deletion or insertion strategy is adopted, the potential effect of the deletion or insertion on biological activity should be considered.
Mutations in nueleotide sequenees construeted for expression of analogs must, ofeourse, preserve the reading frame phase of the coding sequenees and preferably will not create complementary regions that could hybridize to produce secondary mRNA stmctures such as 2() loops or hairpins which would adversely affect translation of the GM-CSF/IL-3 receptor mRNA. Although a mutation site may be predetermined, it is not necessary that the nature of the mutation per se be predetermined. For example, in order to select for optimum characteristics of mutilnts at a given site, random mutagenesis may be conducted at the target codon and the expressed mutants screened for the desired aetivity.
Not all mutations in nueleotide sequences whieh eneode fusion proteins comprising GM-CSF and IL-3 will be expressed in the final produc(, for example, nucleo~ide substitutions may be made to enllance expression, primllrily to avoid secondary strllctllre lc)ops in the transeribed mRNA (see EPA 75,444A, incorpor~lted herein by refcrence), or to provide codons tllat are more readily trlmslatcd by thc selected host, e.~., the well-known /:. co~i preferenee 3() codons for E. coti expression.
Mutations call be introduee(l ut particulllr loci by synthesi~ing oligonucleotides contllining u mutllllt scqllellce, flankcd by restrietion sites enabling ligation to rragments of the native sequence. I~ollowing ligatioll, the resulting reeonstrueted sequence encodes an analog having the desired alnino aeid insertion, substitution, or deletion.
Alternatively, oligonueleotide-directed sitè-speeifie mutagenesis procedures can be employed to provide an altered gene having particular codons altered according to the substitution, deletion, or insertion required. Exemplary methods of making the alterations set forth above are disclosed by Walder et al. (Gene 42:133, 1986); Bauer et al. (Gene 37:73, 1985); Craik (BioTechniques, January 1985, 12-19); Smith et al. (Genetic Engineering:

2 ~ 8 Principles and Methods, Plenum Press, 1981); and U.S. Patent Nos. 4,518,584 and 4,737,462, and are incorporated by reference herein.

Expression of Recombinant Fusion Proteins Comprisin~ GM-CSF and IL-3 ~he present invention provides recombinant expression vectors which include synthetic or cDNA-derived DNA fragments encoding human fusion proteins comprising GM-CSF and IL-3 or bioequivalent analogs operably linked to suitable transcriptional or translational regulatory elements derived from mammalian, microbial, viral or insect genes.
Such regulatory elements include a transcriptional promoter, an optional operator sequence to control transcription, a sequence encoding suitable mRNA ribosomal binding sites, and sequences which control the termination of transcription and translation, as described in detail below. The ability to replicate in a host, usually conferred by an origin of replication, and a selection gene to facilitate recognition of transformants may additionally be incorporated. DNA
regions are operably linked when they are functionally related to each other. For example, DNA for a signal peptide (secretory leader) is operably linked to DNA for a polypeptide if it is expressed as a precursor which participates in the secretion of the polypeptide; a promoter is operably linked to a coding sequence if it controls the transcription of the sequence; or a ribosome binding site is operably linked to a coding sequence if it is positioned so as to permit translation. Generally, operably linked means contiguous and, in the case of secretory leaders, contiguous and in reading frame.
Due to code degeneracy, there can be considerable variation in nucleotide sequences encoding the same lmlino acid sequence; exemplury DNA embodiments are those corresponding to the nucleotide se(luences shown in Figures I or 2. Other embodiments hlclude sequences commensurate in Icngth with and capable of hybridi~.ing to the sequellces of r;igures 1 or 2 under moderatcly stringcnt conditiolls (~()C, 2 ~ SSC) alld which encode biologically activc fusioll proteins.
Transformed host cells ~lre cells which have been tr~msforme(l or transfected witll fusiotl protein veclors collstrlleted using recombillllnt DNA techniqlles. 'I'rallsformed host cclls ordinarily express thc dcsired fusioll proleill, but host cells trunsformed for purposes ot 3() clonillg or amplifyillg DNA do not ncc(l to cxprcss lhc proteill. Expressed fusion protein will gellerally be secreted into the culture supcrn.ltant. Suit;lble host cells for expression of fusion protein include prokaryotes, yeast or higher cuk,lryotic cells uns~er the control of appropriate promoters. Prokaryotes include gram neg.ltive or gram positive organisms, for example ~. coli or bacilli. Higher euk.lryotic cells include established cell lines of mammalian origin as described below. Cell-free translation systems could also be employed to produce fusion protein using RNAs derived from the DNA constructs of the present invention. Appropriate cloning and expression vectors for use with bacterial, fungal, yeast, and mammalian cellular hosts are described by Pouwels et al. (Cloning Vec~ors: A Laboratory Manual, Elsevier, New York, 1985), the relevant disclosure of which is hereby incorporated by reference.

Prokaryotic expression hosts may be used for expression of fusion protein that do not re~uire extensive proteolytie and disulfide processing. Prokaryotic expression vectors generally comprise one or more phenotypic selectable markers, for example a gene encoding proteins confemng antibiotic resistance or supplying an autotrophic requirement, and an orsgin of 5 replication recognized by the host to ensure amplification within the host. Suitable prokaryotic hosts for transforrnation include E. coli, Bacillus suhtilis, Salmonella typhimurillm, and various species within the genera Pseudomonas, Streptomyces, and Staphyolococcus, a'sthough others may also be employed as a matter of choice.
Useful expression vectors for bacterial use can comprise a selectable marker and10 bacterial origin of replication derived from commercially available plasrnids comprising genetic elements of the well-known cloning vector pBR322 (ATCC 37017). Such commercial vectors include, for example, pKK223-3 (Pharrnacia Fine Chemicals, Uppsala, Sweden) and pGEM1 (Promega Biotec, Madison, WI, USA). These pBR322 "backbone" sections s3re combined with an appropriate promoter and the structural sequence to be expressed. E. coli is typically 15 transformed using derivatives of pBR322, a plasmid derived from an E. coli species (Bolivar et al., Gene 2:95, 1977). pBR322 contains genes for ampicillin and tetracycline resistance and thus provides simple means for identifying transformed cells.
Promoters commonly used in recombinant microbial expression vectors include the b-IslctaMase (penicillinase) and lactose promoter system (Chang et al., Nature 275:615, 1978; and 2() Goeddel et al., Nature 281:544, 1979), ~he tryptophan (trp) promoter system (Goeddel et al., Nucl. Acids Res. 8:4~)57, 198(); and EPA 36,776) and tac promoter (Maniatis, Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Lslboratory, p. 412, 1982). A particularly useflll bacterisll expression system employs the phage A Pl, promoter and c1857ts thermoinducible repressor. Plslsmid vectors av.lihlble from thc American Type Culture 25 Collection which incorporate derivsl~ivcs of the ~, P1, promotcr includc plasmid p~lUB2, resident in E. coli slrain JMB~ (A'lCC 37()')2) an(l pPl,c2X, rcsidcnt in E. coli RRI (/\l'CC
~3()X2).
Recombinslllt fusion proteills may sllso bc cxprcsscd in ycsu~;t hos~s, prcferably from îhe Sacehclro~lyces specics, such as ,S. cerevisia~n Ycslst of other gcncrsl such as Pichia or ~() Kluyver(~t~yces msly sllso be employed. Ycast vcctors will genersllly contain arl origin of replicsltioll from the ~m yeslst plasmi(l c-r 5ul alllollo1l1ollsly replicslting scquénce (I~RS), promoter, DNA ellco(lillg the fusion pro~cin, sequences for polysldenylsltion and transcription terminatiol1 su~d sl selection gene. Preferably, yeslst vcctors will include an origin of replication and selectslble mslrker perlnitting transformation of both yeast and E. coli, e.g., the ampicillin ~5 resistsmce gene of E. coli suld S. cer~visiae trpl gene, which provides a selec~ion marker for a mutsmt strslin of yeast lacking the ability to grow in tryptophan, and a promoter derived from a higllly expressed yeast gene to induce transcription of a structural sequence downstream. The presence of the trpl lesion in the yeast host cell genome then provides an effective environment for detecting transformation by growth in the absence of tryptophan.

2 ~ 8 Suitable promoter sequences in yeast vectors include the promoters for metallothionein, 3-phosphoglycerate kinase (Hitzeman et al.9 J. Biol. Chem. 255:2073, 1980) or other glycolytic enzymes (Hess et al., J. Adv. Enzyme Reg. 7:149, 196g; and Holland et al., Biochem. 17:4900, 19783, such as enolase, glyceraldehyde-3-phosphate dehydrogenase, 5 hexokinase, pyruvate decarboxylase, phosphofructokinase, glucose-6-phosphate isomerase, 3-phosphoglycerate mutase, pyruvate kinase, triosephosphate isomerase, phosphoglucose isomerase, and glucokinase. Suitable vectors and promoters for use in yeast expression are further described in R Hitzeman et al., EPA 73,657.
Preferred yeast vectors can be assembled using DNA sequences from pBR322 for 10 selection ~nd replication in E. coli (Ampr gene and origin of replication) and yeast DNA
sequences including a glucose-repressible ADH2 promoter and -factor secretion leader. The ADH2 promoter has been described by Russell et al. (J. Biol. Chem. 258:2674, 1982) and Beier et al. (Nature 300:724, 1982). The yeast o~-factor leader, which directs secretion of heterologous proteins, can be inserted between the prornoter and the struetural gene to be expressed. See, e.g., Kurjan et al., Cell 30:933, 1982; and Bitter et al., Proc. Natl. Acad. Sci.
USA 81:5330, 1984. The leader sequence may be modified to contain, near its 3' end, one or more useful restriction sites to facilitate fusion of the leader sequence to foreign genes.
Suitable yeast transformation protoeols are known to those of skill in the art; an exemplary technique is described by l~linnen et al., Proc. Natl. Acad. Sci. USA 75:1929, 2() 1978, selecting for Trp+ transformants in a selective medium consisting of 0.67% yeast nitrogen base, 0.5% casamino acids, 2% glucose, lO Ilg/ml adenine and 20 ~lg/ml uracil.
Host strains transformed by vectors comprising the AD~12 promoter may be grown for expression in a rich medium consisting of 1% yeast extract, 2% peptone, and 1% glucose supplemented with 8() llg/ml adenine and 80 ~lg/ml urncil. Derepression of the ADH2 promoter 25 occurs upon exhaustion of medium glucose. Crude yeast supernatants are harvested by filtration and held at ~C prior to further purificll~ion.
Various mammalian or insecl cell cultllre systems call be employed to express recombinatlt protein. Baclllovirlls systelns for production of heterologolls proteins in insect cells ure reviewed by Lllckow and Summers, Biol7'~ nolo~y (S:47 (1~)8N). E~xamples of 3() suitllble mammlllilln host cell lines incl(lde the COS-7 lines of monkey kidlley cells, described by Gluzman (Cell 23:175, 1~)81), alld other cell lines capable of expressin~ an appropriate vector including, for example,1~ cells, C127, 3T3, Chinese hamster ovary (CHO), HeLa and BHK cell lines. Mammalian expression vectors may comprise non-transcribed elements such as an origin of replication, a sllitable promoter and enhancer linked to the gene to be expressed, 35 and other 5' or 3' flanking nontranscribed sequences, and 5' or 3' nontranslated sequences, sllch as necessal~ ribosome binding sites, a poly-adenylation site, splice donor and acceptor sites, and transcriptional terrnination sequences.
The transcriptional and translational control sequences in expression vectors to be used in transforming vertebrate cells may be provided by viral sources. For example, commonly 6,~
used promoters and enhancers are derived from Polyoma, Adenovirus 2, Simian Virus 4 (SV40), and human cytomegalovirus. DNA sequences derived from the SV40 viral genome, for example, SV40 origin, early and late promoter, enhancer, splice, and polyadenylation sites may be used to provide the other genetic elements required for expression of a heterologous 5 DNA sequence. The early and late promoters are particularly useful because both are obtained easily from the virus as a fragment which also contains the SV40 viral origin of replication (Fiers et al., Na~ure 273:113, 1978). Smaller or larger SV40 fragments may also be used, provided the approximately 250 bp sequence extending from the Hind m site toward the Bgll site located in the viral origin of replication is included. Exemplary vectors can be constructed as disclosed by Okayama and Berg (Mol. Cell. Biol. 3:280, 1983).
A useful system for stable high level expression of mammalian receptor cDNAs in C127 murine mammary epithelial cells can be constructed substantially as described by Cosman et al. (Mol. Immunol. 23:935, 1986).
Particularly preferred eukaryotic vectors for expression of GM-CSF/IL-3 DNA include pIXY321 and plXY344, both of which are yeast expression vectors derived from pBC102.K22 (ATCC 67,255) and contain DNA sequences from pBR322 for selection andreplication in E. coli (~pr gene and origin of replication) and yeast, as described below in Examples 1 and 7.
Purified mammalian fusion proteins or analogs are prepared by culturing suitable2() host/vector systems to express the recombinant translation products of the DNAs of the present invention, which are then purified from culture media or cell extracts.
For example, supernatants from systems which secrete recombinant protein into culture media can be first concentrated using a commercinlly available protein concentration filter, for example, an ~\micon or Millipore Pellicon ultrafiltration unit. Following the 25 concentration step, the concentrate can be applicd to a suitable purificntion matrix. For example, a suitable affillity mutrix can comprise a ClM-CSF~ or IL-3 receptor or lectin or antibody molecule bollnci to n sllitu~lc sllppon, Alternatively, un allion cxchunge rcsin cnn be employed, for exalllplc~ a matrix or substrlltc having pcndalnt diethylamin~thyl tDEA1) groups, The matrices cnn be acrylumide, agslrosc, dcxtran, cellulose or other lypes comlnonly 3() employed in protein purificatioll. Alternativcly, a cation exchan~e step can be employed, Suitllble catioll exchall~els inclllde variolls insoluble mutrices comprising sulfopropyl or carboxymethyl grollps. Sulfopropyl groups are preferred.
Finally, one or more reversed-phase high performance liqllid chromatography (RP-HPLC) steps employing hydrophobic RP-~IPLC media, e.g., silica gel having pendant methyl 35 or other aliphatic groups, call be employed to further purify a fusion protein composition.
Some or all of the foregoing purification steps, in various combinations, can also be employed to provide a homogeneous recombinant protein.
Recombinant protein produced in bacterial culture is usually isolated by initial extraction from cell pellets, followed by one or more concentration, salting-out, aqueous ion exchange or 2 ~ 8 si~e exclusion chroma~ography steps. Finally, high performance liquid chromatography (HPLC) can be employed for final puri~lcation steps. Microbial cells employed in expression of recombinant fusion proteins can be disrupted by any convenient method, including freeze-thaw cycling, sonication, mechanical clisruption, or use of cell Iysing agents.
Felmentation of yeast which e~press fusion proteins as a secreted protein greatly simplifies purification. Secreted recombinant protein resulting from a large-scale fennentation can be purified by methods analogous to those disclosed by Urdal et al. ~J. Chromatog.
29~:171, 1984). This reference describes two sequential, reversecl-phase HPLC steps for purification of recombinant murine aM-CSF on a preparative HPLC column.
Fusion protein synthesized in recombinant culture is characterizecl by the presence of non-human cell components, including proteins, in amounts and of a character which depend upon the purification steps taken to recover the fusion protein from the eulture. These components ordinarily will be of yeast, prokaryotie or non-human higher eukaryotie origin and preferably are present in innocuous eontaminant quantities, on the order of less than about 5 pereent by scanning densitometry or chromatography. Further, reeombinant eell culture enables the production of the fusion protein free of proteins which may be normally associated with GM-CSF or IL-3 us they are found in nature in their respeetive speeies ~f origin, e.g., in eells, cell exudates or body fhlids.
Fusion protein compositions are prepared for administration by mixing fusion protein 2() having the desired degree of purity with physiologically acceptable earriers. Sueh earriers will be nontoxie to recipients at the dosages and eoneentrations employed. Ordinarily, the preparation of suell eompositions entails eombining the fusion protein with buffers, antioxidants sueh as aseorbie aeid, low moleeular weight (less than about ten residues) polypeptides, proteins, amino acids, carbohydrates including glucose, s-lcrose or dextrins, chelating agents sueh as EDTA, ~lutllthione und other stabilizers and excipients.
Fusion protein compositions muy be used to cnhance prolifemtion, differentiution und funetional uetivlltion of lle~nutol)oietic progenitor cells, sueh us bon~ murrow cells.
Speeifieally, eompositions contnillin~ the fusioll proteill mlly be used to increase peripheral blood leukocyte nullll)ers and increuse circulating granuloeyte COUlltS in myelosuppressed 3() patients. To uehieve this reslllt, a therlll)eutically effeetive qlIlllltity of a fusion protein composition is udltlillistered to u mummul, preferubly u hulllun, in assoeiation with u phllrmllcellliclll carl ier or dihlellt.
The followin~ examples are offered by way of illustration, und not by way of limitution.

2~S~
~L~

Example 1 Svnthesis of cDNAs Encoding GM-CSF/IL-3 Fusion Protein s A. Isolation of cDNA encodin~ huIL-3. Peripheral blood Iymphocytes were isolatedfrom buffy coats prepared from whole blood (Portland Red Cross, Portland, Oregon, USA) by Ficoll hypaque density centrifugation. T cells were isolated by roset~ing with 2-amino-ethylthiouronium bromide-treated sheep red blood cells. Cells were cultured in 175 cm2 flasks at 5 X 106 cells/ml for 18 hour in 100 ml RPMI, 10% fetal calf serum, 50 ~lM b-mercaptoethanol, 1% phytohemagglutinin (PHA) and 10 ng/ml phorbol 12-myristate 13-acetate (PMA). RNA was extracted by the guanidinium CsCI method and poly A+ RNA prepared by oligo-dT cellulose chromatography (Maniatis et al., Molecular Cloning: A Laborator,v Manual, Cold Spring Harbor, 1982). cDNA was prepared from pol~ A+ RNA essentially as described by Gubler and Hoffman, Gene 25:263-269 (1983). The cDNA was rendered double-stranded using DNA polymeruse 1, blunt-ended with T4 DNA polymerase, methylated with EcoRI
methylase to protect EcoRI cleavage sites within the cDNA, and ligated to EcoRl linkers.
These constructs were digested with EcoR I to remove all but one copy of the linkers at each end of the cDNA, ligated to EcoRI-cut and dephosphorylated arrns of phage AgtlO (Huynh et 2() ul., DNA Cloning: A Practlcal Approach, Glover, ed., IRL Press, pp. 49-78) and packaged into ~ phage extracts (Stratagene, San Diego, CA, USA) according to the manufacturer's instmctions 50(),0()() recombinants were plated on E. coli strain C600hfl- and screened by standard plaque hybridization techniques using the following probes.
Two oligonucleotides were synthesized, with sequences complementary to selected 5' und 3' sequences of the hulL-3 gene. 1'he 5' probe, complementury to a seqllence encoding part of thr hulL-3 Ieader, hud lhe seqllence :S'-GAG'I~GGAGCAGG~GC~GG~C-3'. The 3' probe, correspondhlt3 to u region ellcoding umino acids 12~- 13() of the muture proteill, had the sequence ~'-GATCGCG~\GCC'I'CAAAG'I`CG'I`-~' The melhod of synttlesis W;IS .I Stllndllrd ulltomllted triester metllod substuntisllly simitur to thut disck~sed by Sood et ul" Nucl Acfds 3() Res 4:2557 tl977) ~uld Hirose et al, ~ I,et~. 28:~449 (197X) Followin~ synthesis, oligollllcleQtides were debk7cked ulld purified by preparutive ~el electrophoresis, For use as screenillg probes, the oligollucleotides were terminully rudiolabeled with 32P-ATP and T4 polynucleotide kinase usillg techlliqlles similur to those disclosed by Maniatis et al. The E. coli straill used for libraly screening wus C6()0hfl- (Hllynh et al, 19X5, supra).
Thirteen positive plaques were purified and re-probed separately with the two hybridization probes Eleven clones hybridized to both oligonucleotides The cDNA inserts from several positive recombinant phage were subcloned into an EcoR1-cut derivative (pGEMBL18) of the standard cloning vector pBR322 containing a polylinker having a unique EcoRl site, a BamH1 site and numerous other unique restriction sites. An exemplary vector of 2~
this type, pCEMBL, is described by Dente et al., Nucl. Acids Res. 11:1645 (1983), in which the p~omoters for SP6 and r7 polymerases flank the multiple cloning sites. The nucleotide sequences of selected clones were detennined by the chain temtination method. Specifically~
partial EcoRt digestion of ~GTtO:IL-3 clones 2, 3, 4 and 5 yielded fragments ranging from 850 bp to 1,000 bp in size which were separately subcloned into the EcoR1 site of pGEMBL18. The inserts of the pGEMBL:rhuIL-3 subclones were sequenced using a universal primer that binds adjacent to the multiple cloning site of pGEMBL18, and synthetic oligonucleotide primers derived from tlte huL-3 sequence.

B. Modi~lcation of N-Glvcosvlation Sites Encoded bv huIL-3 cDNA and Assemblv of Expression Vector for rhuIL-3 (Pro8 ASPIS ASp70). The twO asparagine-linked glycosylation sites present in the natural protein (Asnl5 and Asn70) were altered by changing the codons at these positions to ones that encode aspartic acid. This prevents N-linked glycosylation (often hyperglycosylation) of the secreted protein by the yeasr cells, and a more homogeneous product is obtained. These changes were made as described below upon subcloning the huIL-3 cDNA into the yeast expression vector plXY 120.
The yeast expression vector plXY120 is substantially identical to pBC102-K22, described in EPA 243,153, except that the following synthetic oligonucleotide containing multiple cloning sites was inserted from the Asp7 18 site (amino acid 79) near lhe 3' end of the 2() a-factor signal peptide to the Spel site contained in the 211 sequences:
A~p718 ~col GTACCTTTGGATAAAAGAGACTACAAGGACGACGATGACAAGAGGCCTCCATGGATCCCCCGGGACA
GAAACCTATTTTCTCTGATGTTCCTGCTGCTACTGTTCTCCGGAGGTACCTAGGGGGCCCTGTGATC
~amHl Spel In addition, a 514-bp DNA frugmcllt derivcd from tlle single-strundcd bacteriophage fl containing the origin of replication and intcrgenic region WUS inserted at the Nrul site in the pBR322 DNA sequences. ï`he prescnce of thc fl origin of replication enubles generation of 30 single-stranded copies of the vector whcn trallsfonncd into nppropriate (male) struins of E. c oli alld superinfected witll bacteriophllgc fl. ~I'his cnl)llbility facilit.ltes ONA sc(lucncing of the vector lmd ullows lhc possibility of in vilro m~lt;lgen~sis The yeast expression vector plXY12() W~ls digested with the restriction en~ymes Asp718, which clcuves near tlle 3' elld of thc ~x-factor leuder peptide (nucleotide 237), ulld 3S Balt1HI, whicll cleuves in the polylinker. I~le large vector fragment was purified and lig;lted to tlle following DNA fragmellts: (I) a hull,-3 cDNA frugment derived from plasmid GEMBLlX:hlllL-3 from the Clal site (nucleotide 58 of mature hulL-3) to the BamH1 site (3' to the hulL-3 cDNA in a polylinker); and (2) the following synthe~ic oligonucleotide linker A:

2 ~

GTA CCT TTG GAT AAA AGA GAC TAC AAG GAC GAC GAT GAC AAG GCT CCC ATG ACC CAG
GA AAC CTA TTT TCT GTG ATG TTC CTG CTG CTA CTG TTC CGA GGG TAC TGG GTC
AC~ ACG CCC TTG AAG AC~ AGC TGG GTT ~A~ TGC TCT AAC ATG AT
TGC TGC GGG AAC TTC TGG TCG ACC CAA CTA ACG AGA TTG TAC TAG C

Oligonucleotide A regenerates the sequence encoding the C-terminus of the -factor leader peptide and fusing it in-frame to the octapeptide DYKDDDDK, which is, in turn, fused to the 10 N-terminus of mature rhuIL-3. This fusion to the rhuIL-3 protein allows detection with antibody speci~lc for the octapeptide and was used initially for monitoring the expression and puri~lcation of rhuIL-3. This oligonucleotide also encodes an amino acid change at position 15 (Asnl5 to Aspl5) to alter this N-linked glycosylation site. The underlined nucleotides in oligonucleotide A represent changes from the wild type cDNA sequence. Only the A to G and 15 C to T changes at nucleotides 43 and 45, respectively (counting from the codon corresponding to the N-terrninal alanine of the mature hulL-3 molecule), result in an amino acid change (Asp15). The other base changes introduce convenient restriction sites (AhaII and PvulI) without altering the amino acid sequence. The resulting plasmid was designated plXY139 and contains a rhulL-3 cDNA with one remaining N-linked glycosylation consensus sequence 2() (Asn70).
Plasmid plXY 139 was used to perfonn oligonucleotide-directed mutagenesis to remove the second N-linked glycosylation consensus sequence by changing Asn70 to Asp70. The in vitro mutagenesis was conducted by a method similar to that described by Walder and Walder, Cene 42:133 (1'~86). The yeast vector, plXY139, contuins the origin of r- plication for the 25 single-stranded bucteriophage fl and is capable of generuting single-stranded DNA when present in a suituble (mule) strain of E. coli slnd superillfeeted with helper phnge.
Single-stranded DNA was generated by trunsforming E. c~71i strain JM107 und superitlfecting witll helper pha~e IRI. Single-strlmded DNA wus isolated and ~Innealed to the following mutngellic oli~onllcleotide 13, GTC AAG AGrl~rr~ CAG ~AC ~JCA TCA GCA
3() AAT G, which provides a codon switch subslituling Asp for Asn ut position 7() of matllre hulL-3. Annealing alld yeas~ trlmsformation conditions were done sls described by Walder and Walder, supr~l Yenst lrnllsfonnunts were selected by growth on medi1lm lacking tryptophan, pooled, and I~NA ex(racted as deseribed by llolm et al, Gene 4~:16'~ (19~6). This ONA, containing a mixture of wild type alld IlltltUllt plasmid DNA, WllS used to transforrn E coli RRl 35 to ampicillill resistance. The resultillg colonies were screened by hybridization to radiolabeled oligonucleotide B USillg stand~urd techniques Plasmids comprising DNA encoding hulL-3 Asp7(~ were identified by the hybridization to radiolabeled oligonucleotide B under stringent conditions and veri~led by nucleotide sequencing.
The resulting yeast expression plasmid was designated pIXY138, and contained the40 huIL-3 gene encoding the Aspl5 Asp70 amino acid changes and ~he octapeptideDYKDDDDK
at the N-terminus. The final yeast expression plasrnid is identical to pIXY138 except that it 2 ~ 8 lacks the nucleotide sequences coding for the octapeptide, thus generating mature rhuIL-3 as the product.
The final yeast expression plasmid was constructed as described below. The yeastexpression vector pIXY120 was cleaved with the restriction enzymes Asp718 and BarnH1 as 5 described above. The large vector fragment was ligated together with (1) a huIL-3 cDNA
fragrnent derived from plasmid pIXY138 that extended from the Aha2 site (which cleaves a nucleotide 19 of mature huIL-3) to the BamH1 site 3' to the cDNA, and (2) the following synthetic oligonucleotide C:

GA AAC CTA TTT TCT CGT GGG TAC TGG GTC TGC TGC
Pro Leu A~p Ly~ Arg Ala Pro Met Thr Gln Thr Thr Oligonucleotide C regenerates the 3' end of the a-factor leader peptide from the Asp718 site 15 (the amino acids Pro-Leu-Asp-Lys-Arg) and the N-terrninal seven amino acids of huIL-3 to the AhalI site. The resulting plasmid was designated plXY151. This vector, when present in yeast, allows glucose-regulated expression and secretion of rhulL-3 (pro8 Aspl5 Asp70).

C. Expression Vector for rhuGM-CSF ~Leu23 Asp27~ alu39) Containin~ Modified N-2() Glvcosy~ n Sites. The wild-type gene coding for human GM-CSF, resident on plasmid pHG23, has been deposited with the American Type Culture Collection (ATCC), 12301 Parklawn Drive, Rockville, Maryland 20852, USA, under accession number 39900. The wild-type gene inserted in a yeast expression vector, pYfHuGM, has also been deposited with the ATCC under accession numbcr 53157. In order to provide a non-glycosyla~ed ~malog 25 of human GM-CSF, oligomlcleotide-directed site-spccific mutagenesis procedures were employed to eliminute potential N-glycosylation sites, as dcscribcd in PCT publication WO
89/()3881. A plasmid cncodil1g this analog, hllaM-CSP (Lcu23 Asp27 Glu39), was dcposited with thé ATCC as plaslllid L2()7-3 in t'. c oli str,lin RR Mlllder accession numbcr 67231.

hxamplc 2 oll~tm~tiQn of ~[~x~er~ Vcctor ~ aM-CSr/lL-~ sic)n l'ro~ein In order to crellte a secretion vector for expressing a fusion construct having human GM-CSF alld hlltn;lll IL-3 separated by a linker sequence, a precursor plasmid was first 35 constrllcted by directly fusing DNAs encoding GM-CSF and IL-3 together without regard to reading frame or intervening sequences. A cDNA fragment encoding nonglycosylated human GM-CSF was excised from plasmid L207-3 as a 977bp restriction fragment (Sphl to Sspl).
The IL-3 cDNA was excised from pIXY151 by digestion with Asp718, which was then blunt ended using the T4 polymerase reaction of ~Ianiatas et al. (Molecular Cloning: A Laboratc~ry 40 Manual, Cold Spring Harbor, 1982, p. 118) and further digested with Xhol giving an 803 bp Q ~
fragment. These two fragments were then directly ligated to a pIXY151 vector fragment cut with Sphl and Xhol. This plasmid was called GM/IL-3 direct fusion.
The GM/IL-3 direct fusion plasmid was used as a template in oligonucleotide-directed mutagenesis using methods similar to those described by Walder and Walder, supra. The 5 following oligonucleotide was then synthesized GCCAGTCCAGGAGGGTGGCGGTGGATCCGGCGGTGGTGGATCTGGTGGCGGCGGCTCAGCTCCCATGACCC
ProValGlnGluGlyGlyGlyGlySerGlyGlyGlyGlySerGlyGlyGlyGlySerAlaProMetThr 0 ---GM-CSF---><--------------Linker-----------------------><---IL-3----This oligonucleotide overlaps the 3' end of GM-CSF by 13 bp but does not include the stop codon, contains the Gly Ser linker, and overlaps the 5' end of TL-3 by 13 bp. The linker sequence was a modified version of the linker described by Huston et al. (Proc. Natl. Acad.
Sci. USA 85:5879-5883, 1988) but was optimized for codon usage in yeast as per Bennetzen et al. (J. Biol. Chem. 257:3026, 1982).
Single stranded plasmid DNA was made from the GM/IL-3 direct fusion using R408 helper phage (Stratagene) and the methods of Russel et al. (Cene 45:333-338, 1986).
Oligonucleotide directed mutagenesis was then carried out by annealing the above2() oligonucleotide to the single stranded plasmid DNA and transforming yeast strain XV2181 with annealed DNAas described by Walder and Walder, supra. The yeast vector eontains the origin of replication for the single stranded bacteriophage fl and is capable of sponsoring single stranded DNA production when present in a suitable (male) strain of E. coli and superinfected with helper phage. Yeast transformants were selected by growth on medium lacking tryptophan, pooled, and DNAwas extracted as described by ~lolm et al, (Cene 42:169, 19~6). This DNA,containing a mixture of mutant and wild type plalsmid DNA, was used to transforrn E. coli RRI to nmpicillin resistnnce, Tlle resulting colonies were sereened by tlybridization to radiolabeled oligonllcleotide usin~ stall(lar(l techniques. Plnsmids eomprising DNA encoding GM-CSF/litlker/11,-3 were idenlified by lheir hybridizaltion to radiolabeled 3() oligonucleotide contllinin~ the linker under stringent conditions and vwified by nucleoti(le sequencing.
Duril1g nucleotide sequellcil1g it was discovered that a mutation had occurred within the linker region. The nllcleotide se~luellce TGa'l'GGA'l'CTGG was deleted (see sequence), reslllting in the expression of a protein in which the sequence of amino acids GlyGlySerGly were deleted. Tllis mutation did not change the reading frame or prevent expression of a biologienlly active protein. The resulting plasmid was designated plXY321 and expressed the ~;lsion protein huGM-CSFI~Leu23Asp27Glu39]/Gly4SerGlysSer/hulL-3[Pro8Aspl5Asp7 2~6~8 Example 3 Expression and Purifiçation of GM-CSF/IL-3 Fusion Protein The host strain, XV2181, a diploid S. cerevisiae strain, was formed by mating XV617-1-3B [a, his6, leu2-1, ~rpl-l, ura 3, steS], obtained from the University of Washington, Department of Genetics Yeast Strain Barlk, Seattle, WA, USA, and X2181-lB [a, ~rpl-l, gall, adel, his2], obtained from the Yeast Genetic Stock Center, University of California, Berkeley, CA, USA. The host strain was transforrned with the expression plasmid by the method of Sherrnan et al., Laboratory Course Manual for Methods in Yeast Genetics, Cold Spring 10 Harbor Laboratory, 1986.
Yeast eontaining the expression plasmid pIXY321 (see lD, above) was maintained on YNB-trp agar plates stored at 4C. A preeulture was started by inoculating several isolated reeombinant yeast eolonies into one liter of YNB-trp medium (6.7 g/L Yeast Nitrogen Base, 5 g/L easamino aeids, 40 mg/L adenine, 160 mg/L uraeil, and 200 mg/L tyrosine), and was 15 grown overnight in two 2-liter flasks at 30C with vigorous shaking. By morning the eulture was saturated, in stationary phase, at an OD600 of 2 to 7. The ferrnenters (three machines of 1() liter working volume), previously eleaned and sterilized, were filled to 80% of their working eapacity with SD-2 medium (4.0 g/L ammonium sulfate, 3.2 g/L monobasie potassium phosphate, 3.() g/L yeast extract, l.() g/L eitric aeid, 0.1 g/L sodium ehloride, S
2() ml/L 2% ealeium ehloride, 2.5 ml/L vitamin l01 solution, 0.S ml/L traee elements solution, 0.5 ml/L 20% magnesium sulfate, 2.0 ml/L glucose) and maintained at 3()C with 500-600 rpm agit,ltion and 10-16 Ipm aeration. The inoeulum was added. After two hours of growth a nutrient feed of 5()% glllcose was begun at a rate such that 50 ~IL is added over n period of 10-12 hours. The nutrient feed WllS then shifted to 5()% ethanol added nt 3()-40 ml/hr until 25 harvest.
Total elapsed time of ferment.ltioll was appt~ximsltely 2() hours, nfter whieh optieal density (6()0 nm) ranged from 3() to 45. The fentlenters were then coolecl to 2()C, p~l of the yeast beer wus adjllsted to 8.0 by the additic)n of S M NaOI l, a-ld the reslllting material filtered through u Millipore Pellicon hlt~r system eqllippe(l wit~) a 0.45 ~lm filter cassette, and eolleeted 30 hl a sterile 1() L carboy.
One liter of yeast supermltllllt eontnining GM-CS~/lL-3 fusion protein wus concentrllted to 5() ml 011 an Amicoll YM-I() lneIllbr;lne. Ttle yeust broth coneentrate was then further pulified by preparlltive HPLC by npplyin~ to a I cm X 25 em eolumn packed with 5,u C-18 siliell (Vydae, Sep.lr;~tions Group, Hesperia, CA, USA) that was equilibrated in 0.1%
35 trifluoroacetie acid in water (Solvent A) prior to application of the yeast concentrate.
Alternatively, the emde yeast broth can be pumped direetly on to the C- 18 column. Following npplication of the materinl, the column was flushed with Solvent A until the optical absorbance of the eMuent approached base line values. At this time a gradient of 0.1% trifluoroacetic acid in acetonitrile (Solvent B) was established from 0% B to 100% B at a rate of change of 1-2% B

2 ~ 8 per minute and at a flow rate of 2 mVminute. One minute fractions were collected. Aliquots of the fractions were analyzed for protein content by dot blot with a rabbit polyclonal antisera to IL-3. ~M-CSFtlL-3 eluted in fraction 50 at approximately 50% acetonitlile.
HPLC fractions which were positive for GM-CSF/IL-3 fusion protein by dot blot were S pooled and bound to SP-Sepharose in 20 mM ~-alanine, pH 4. Fusion protein was eluted with 0.5 M NaCl, 100 mM Tris-HCI, pH 8. Frac2ions containing fusion protein wereidentified by SDS-PAGE.
The ion exchange fractions containing protein having a molecular weight of 35,000 were pooled, concentrated to 100 ~1 and further purified by FPLC gel filtration on a Superose 12 column. The column was eluted with PBS. Fractions containing only the purified 35,000 MW fusion protein were identified by SDS-PAGE.
The biological activities (units/mg) and binding affinities of the GM-CSF/~L-3 prepared substantially as described above were detennined as set forth in Examples 4 and 5.
.

Example 4 Biolo~ical Activitv of GM-CSF/IL-3 Fusion Protein in Thvmidine Incolporation Assav In order to determine its level of biological activity, the GM-CSF/IL-3 fusion protein prepared as described in Example 3 was assayed for ability to stimulate proliferation of AML-2() 193 cells in a thymidine incorporiation assay. The AML-193 cell line is a GM-CSF dependent hlltnan monocytic leukemia cell line originally described by Santoli et al. (J. Immunol.
135~:3348, 1987). The cells were grown in Iscove's Modified Dulbecco's Media (IMDM) with 25 mM HEPES, 2(H) nM L-glutamine, S ~,lg/ml insulin, S ~g/ml trnnsferrin, S ng/ml sodium selenite, 2.5% heat inactivated fetal bovine serum, antibiotics, and S ng/ml of purified recombinant hum,ln GM-CSF. The cclls wcrc split twice weekly and were scedcd inîo fresh medium at a density of 3(H),(XN)/ml, A thymidine incorl)oratiol1 s~ssay was employcd ~o cxamitlc the capacity of known growth factors and ullktlown sul~ern;lt;lllts lo stimulate prolifer;ltion of AMt,-193. AMI.^193 cells were wllshecl by centrifll~;ltioll nll(l res1lspcndcd in ussay mcdium col11posed of IMDM as 3() above except that fetnl calf serum alld/or GM-CSF wus not included, Pure GM-CSF, 11,-3 or GM-CSF/IL-3 t;lSiOll protcin was added to the first well of a g6 well flat bottom tissue culture plllte at a fin~ll concelltr~ltioll of 4(X) n~ ll in 5() ml of medium. These samples were then serially diluted by 3-fold through the additionul t I wells of the microtitre plate. Fifty 1ll of medi1lm containil1g 375() AML-193 cells was added to each well and plates were incubated at 37C for 138 hours in a fully humidified atmosphere of 6% CO2 in air. Tritiated thymidine (0.5 mCi/well) was added to each well for an additional 6 hours of incubation and the samples were harvested with an automated sample harvester and counted by liquid scintillation. One unit of activity is defined as the amount of growth factor required to stimulate half-maximal thymidine incorporation.

2B54~08 Simultaneous titration of IL-3, GM-CSF or GM-~SF/IL-3 fusion pro~ein at identical concentrations revealed that the fusion protein was a more potent proliferation stimulus than either factor alone or IL-3 and GM-CSF combined. The specific activity of the IL-3, GM-CSF
and GM-CSF/IL-3 fusion protein is set forth in Table A, below.
s TABLE A

Molecule S~ecific ~ctivi~v 10 IL 3 1.65 x 105 GM-CSF 9.74 x l(P
IL-3 + GM-CSF 1.39 x 105 GM-CSF/IL-3 1.81 x lo6 The specific activity of GM-CSF/IL-3 fusion protein is approximately 10-fold higher than IL-3 or GM-CSF alone or GM-CSF plus IL-3 combined.

Example S
Bir~ ACtivitY of GM-CSF/IL-~ Fusion Protein in Equilibrium Binding Assav Binding affinities of human IL-3, GM-CSF and fusion protein for receptors on human cells lines were determined by inhibition of 1251-labeled IL-3 or GM-CS~; binding.
A. Radiola~oeling of GM-CS~ and IL-3. Recombinant hum.m GM-CSF/IL-3 fusion 25 protein was expressed in yeast cells and purified substantially as dcscribed above.
Recombinant human IL-3 and GM-CSF, engineered to contain the octapeptide DYKDDDDK
were expressed in yeast and purified using a monoclonlll antibody spccirlc to the octapeptide substantially as described in Hopp ct nl. (Bil)17'ec/lnology 6:t2()4, 19XX). l'he purificd GM-CSF and IL-3 proteins were radiolllbeled using a eommcrcially av.liluble enzymobead 30 radioiodination rengcnt (BioRad), sllbstantially as described by Park et sll. (J. Biol. Chcm.
261:4177, 198fi). Briefly, 2-l() llg of recombin:lnt protein in ~ 1().2 M sodium phosphate, pH 7.2, was combilled with 50 111 enzylllobcad reagent, 2 ItlCi of sodillm iodide in 2() ~,11 of 0.05 M sodium phosphslte pH 7 and 1() ,ul of 2.5% ~-D-glucose. After l() min at 25C, sodium azide (10 1ll of 5() mM) and sodium metabislllfite (lO ~l of S mg/ml) were added sequentially 3~ and incubation contimled for 5 minuîes at 25C. The reaction mixture was fractionated by gel filtration on a 2 ml bed volume of Sephadex G-25 (Sigma) equilibrated in Roswell Park Memorial Institute (RPMI) 1640 medium containing 2.5% (w/v) bovine serum albumin (BSA), 0.2% (w/v) sodium azide and 20 mM Hepes, pH 7.4 (binding medium). The final pools of 125I-IL-3 and 125I-GM-CSF were diluted to a working stock solution of 1 x 10-7 M in binding medium and stored for up to one month at 4C without det~ctable loss of receptor binding activity. The specifi~ activity of radiolabeled preparation of GM-CSF is routinely in the range of 1-5 x 1015 cpm/mmole. The specific activity of IL-3 is in the range of 3-6 x 1015 cpm/mmole.
S B. Binding Assays. Binding assays were performed using lM-l, KG-l, HL-60 and AML-193 cells . JM-l, HL-60 attd KG-l cells were obtained and prepared as descnbed by Park et al. (J. Biol. Chem. 264:5420, 1989). AML-193 cells were obtained and prepared as described above in Example 4. As described by Park et al. (supra), 125I-GM-CSF does not bind to JM- 1 cells nor does GM-CSF inhibit binding of 125I-IL-3 to JM-l cells, indicating that these cells possess receptors capable of binding only IL-3. Conversely, 125I-IL-3 does not bind to HL-60 cells nor does IL-3 inhibit binding of 125I-GM-CSF to HL-60 cells, indicating that these cells possess receptors capable of binding only GM-CSF. In contrast, both KG-l and AML-193 cells bind 1251-GM-CSF and 1251-IL-3 and, in addition, both IL-3 and GM-CSF
are able to partially compete specific binding of the heter~logous radiolabeled ligand, with approximately equivalent capacities. This suggests that these cell lines possess receptors that bind only IL-3, receptors that bind only GM-CSF, and receptors that bind both GM-CSF and IL-3, all with high affinity.
In order to determine the affinity of binding (Kl) of IL-3, GM-CSF and GM-CSF/IL-3 fusion protein, inhibition assays were performed in which the ability of varying concentrations 2() of these unlabeled proteins to inhibit binding of 1251-lL-3 to JM- I celis, 1251-GM-CSF to ~IL-60 cells and 1251-lL-3 and 1251-GM-CSF to KG- I and ~MI,- 193 cells were measured. Assays were performed by incubatin~ cells (3.3 x 1()7/ml) with 3 x 1()-1() M 1251-GM-CSF or 1251-lL-3 and varying concentr;ltions of unl,lbeled 11,-3, GM-CSI~ or GM-CSF/IL-3 fusion protein for 3()-6() minutes at 37C, Binding was ~Issnycd USillg the phth~ ltc oil separation method disclosed by Dower et al. (J. Immun(ll. t32:75 1, 1~X4), csscntially as dcscribed by Park ct al.
(J. Biol. Chem 2fi/:4177, 19X6). 1~nta was an;lly7.cd as dcscribc(l hy i'nrk et ~11. (Blood 74:5~, 1989), Binding aflinilics were detenllille-l ~or IL-~, GM-CSI an(l GM-CSf~/ll,-~, as shown in T~lble B, bclow, 3() 'I'AI~l,E B

Labcl~l Unl~ lcd KlV~Ilu~ct~ l!
Liisll d _ C'om~ or ~ IL60 KC-I AML-lg3 1251 1L 3 IL-3 1.8x1()1 -- 2.0x101 4.1x109 CM-CSF/IL-3 6.1x109 -- 2.8xlO1 1 1.5x101 51-GM-CSF GM-CSF - 1.2x101 3.2x101 1.6xlOlG
CM-CSF/IL-3 -- 6.8xlO9 1.9x101 I.Sx101 2~5t~ ~ 8 The expenments used to obtain the dat~ in Table B were conducted using different cell lines in different expenments and accordingly show some variation, making direct comparison dif~lcult. In order to enable direct compaIison of this data, the Kl values of both the controls and the fusion proteins were no~nalized to the KI value for the control on one cell line. IL-3 5 data were normalized to a KI=1.8xlO10 M-1 on JM-l cells, and GM-CSF data were nonnalized to a Kl=1.2xlO10 M-l on HL-60 cells to give the values as set forth below in Table C.

TABLE C
Labeled UnL~b~led Normalized KI Values (M-l!
Lieand ComDetitor JM-I HL60 KG-1 AML-193 1251-lL-3 IL-3 1.8x101 -- 1.8x101 1.8x10l GM-CSF/IL-3 6.1x109 -- 2.5xlO11 6.8x101 51-GM-CSF GM-CSF - 1.2x101 1.2x101 1.2x101 (~M-CSF/IL-3 - 6.8xlO9 7.1xlO9 I.lx101 _. ~

20 Comparison of the normalized data indicates that the GM-CSF/IL-3 fusion protein and GM-CSF bind with approximately the same affinity to receptors for GM-CSF on HL-60, KG-I and AML-193 cells. In contrast, the GM-CSF/IL-3 fusion protein and IL-3 bind with different affinities: GM-CSF/IL-3 fusion protein binds with lower amnity than IL-3 to receptors on JM-1 cells (which have only IL-3 binding receptors); in contrast, the GM-CSF/IL-3 fusion 25 protein binds w;th a signi~lcantly higher affinity than IL-3 to receptors on KG-I and AML-193 cells tboth of which have GM-CS~/IL-3 receptors). Using thc JM- I cell line as a standard for normal binding nffinity of the GM-CSF/IL-3 filsion protein to a receptor (i.e., for binding to a receptor which is capnble of binding only a single ligand), thc GM-CSF/IL-3 fusion protein binds to KG-1 cells wilh a 41.0-fokl hi~her binding affinity, nnd to AML-193 cells with lln 3() 11. I-fold higher bindin~ affinity.
Not wishillg to be bound by any particular theory, it is believed that the higher binding nffinity of GM-CSF/IL-3 fusion proîcin to KG-I nnd AML-193 cclls is related to the presence in both of these cell lines of the GM-CSF/IL-3 receptor, In pilrticular, the higher binding ~lffinity of the GM-CSF/IL-3 filsion protein to the AML-193 cell line may explain the higher 3~ biologicnl nctivity of the GM-CSF/IL-3 fusion protein in the thymidine incorporation assay of Example 4 which utili~ed the AML- 193 cell line.

2 ~ Q 8 Example 6 Effect of GM-CSF/~L-3 on Prolifer~tion of Human Bone Marrow The biological effect of GM-CSF/IL-3 on the proliferation of unfractionated human S bone marrow was compared with that of GM-CSF and IL-3 alone. Non-adherent~ lowdensity, T cell deplated cultures of human bone marrow were plated in methylcellulose (BFU-E, CFU-GEMM, 40,000 cells per plate) or agar (CFU-GM; 40,000 cells per eulture) as deseribed by Lu et al., Blood 61:250 (1983). Methylcellulose eultures eontained 1 unit per plate erythropoietin and accounts for the background of 48+'7 BFU-E in the absence of eytokine. Cultures were incubated in a 5% 2. 5% CO2, 90% N2 atmosphere for 14 days and counted with an inverted microscope. These values represent the mean +1 standard deviation of duplicate or triplicate data points in one of two representative experiments.
TABLE D

Colonies (Mean~S.D.) Cvtokine Dosc (v~/ml! CFU-GEMM CFU-GM BFU-~3 2() Nonc ot 3+ I t 48i2t GM-CSF/IL-3 5000 10.3il 97 t2 107 tS
2500 10.3iO.9 74 tl 119iS
1250 6.8_0.5 59_2 125~8 625 4.3 tO.5 52_6 83 t5 312 2.5_0.7 35.+2 70t4 156 I.OtO.4 24t3 49ifit 78 ()t 16t2 sot3t 3() GM CSF+IL 3 5000+5(X)() 10.0tO.7 54_3 94t6 250X)+25()() N.5~().5 52_3 81:t3 1250+125() 5.0~().4 47:t4 56t3t 625+fi25 2.5~().3 27:t2 44 t3t 312+312 I .Oit) 14:t 1 42i3t ISfi tlSfit ().3i~).3 I(k~l 44:t3t aM-CSI~ 5(XX) fi.8LO.6 42i4 fi3t3 25(X) 3.5~0.7 42t2 61tl 1250 1.5i().3 ~-.t2 47~4t (i25 ().~ (),3t 21 ~;-2 4fi~3 t 312 ()t lfitl t IL-3 5(XX) 4.()t().7 16iO.3 67i2 250X) 4.() ~ ().4 9 t 1 66_2 1250 1.3t().3 4_1t 49-+2t 625 ot ~ 45i3t 312 ()t t t t = value equal to media control * p<0.05 compared to media control 2 ~
TABLE E

Colonies (Meanis.D~!
S Cvtokine . Dose (~/ml) CFU-GEMM CFU-GM BFU-E
None ~+ot ot 44+4t GM-CSF/llL-3 5000 - 43+5 2500 lO.O~l~ 45+5 125ilO~
1250 9.O~O* 23+2 127+7 625 6.0~1~ 15+2 97+2*
312 5.5+0.5 8+1 71~1~
156 2.0~1t 5~1 47~3t 78 2 0~0t 2~0.3 44+1t GM-CSF+IL-3 5000+5000 6.5il.5~ 32~2 93i3 2500+2500 S.Oil~ 21i2 71~8~
1250+1250 I S~o.St 13~0.3 44i2t 625+625 2.0i0t 7+2 45~5t 312+312 t 3~1 t 156+156 t liO.9t t -t = value equal to media control ~ p<0.05 compared to media control Tables D and E indicate that GM-CSF plus IL-3 is approximately two fold more potent than either GM-CSF or IL-3 alone in enhancing proliferation of unfractionated human bone marrow cells. The GM-CSF/IL-3 fusion protein is equivalent in potentcy to a mixture of GM-CSF and 30 IL-3, and that the mixture of GM-CSF and IL-3 showed an approximately two fold enhancement compared to GM-CSF or IL-3 alone.

Example 7 A cDNA encoding u fusion protein comprising an N-terminal IL-3 und a C-te~ninal GM-CSF w~ls constructed as follows. Thc yeast expression vector plXY120 (described in Exumple lB) was digested with the restriction enzymcs Asp7 18, which clcaves ncar the 3' end of the ~-fi~ctor leader peptid~ (nucleotide 237), ~nd Ncol, which cleaves in the polylinker. The 4a largc vector fragment was purified and ligated to an approximately 500bp Asp718-Ncol fragment (encoding GM-CSF(Lell23Asp27Glu39)) from a partial digest of L207-3 (ATCC
67231), to yield plXY273. A 9kb Asp718-Bgl2 fragment of plXY273 (still containing the GM-CSF(Leu23Asp27Glu39) cDNA) was then ligated to an Asp718-Nrul fragment encoding human IL-3 (Pro8AsplSAsp70) from pIXY151 (described in Example lB) and the following 45 double stranded oligonucloetide:

2~6~8 5' CGATCTTTGGTGGCGGTGGATCCGGCGGTGGTGGTGGATCTGGTGGCGGATCTGCTCCAGCTA 3' 3' GCTAGAAACCACCGCCACCTAGGCCGCCACCACCACCTAGACCACCGCCTAGACGAGGTCGATCTAG 5' -IL-3--><~ -----LINKER--------- ------------------><---GM-CSF----5 This oligonucleotide overlaps the 3' end of IL-3 by 8bp but does not include the stop codon, contains the Gly-Ser linker, and overlaps the 5' end of GM-CSF by lObp. The resulting vector was terrned pIXY344 and was used to express an IL-3/GM-CSF fusion proteinessentially as described above in Exarnple 3.

Example 8 Binding Activitv of IL-3/GM-CSF Fusion Protein in Equilibrium Binding Assav Binding affinities of human IL-3, GM-CSF and IL-3/GM-CSF fusion protein (produced as described in Example 8) for receptors on human cells lines were deterrnined by 15 inhibition of 125I-labeled IL-3 or GM-CSF binding as described in Example 5 above.
Binding assays were performed using JM-I, HL-60 and KG-1 cells, which were obtained and prepared as described by Park et al. (J. Biol. Chem. 264:5420, 1989). JM-1 cells possess receptors capable of binding IL-3, but not GM-CSF. Conversely, HL-60 cells possess receptors capable of binding GM-CSF, but not IL-3. KG- 1 cells possess receptors for 20 both GM-CSF and IL,-3.
Binding affinities (Kl)were detennined for IL-3, GM-CSF and GM-CSF/IL-3, as shown in Table F, below.

TABLE F
~
L~bclcd ~)nlabcl~l or ~1~1 Hl.60 KO-l 1251 IL~3 IL-3 f~.()x1()9 5.7xl()') 3() aM-CSr~ 2.5x1()9 - 3.()xl()1() IL~3/GM-CS1~ 1.2x1()9 2.2x101() t251 C;M-CSF GM-CSr - 1.5xl()1() N.D.
GM-CSF/11-3 - 5.4x109 N.D.
IL.-3/GM-CSr - 3.0xlO9 N.D.

.
N.D. = no dalll availablc The above data indicate that the GM-CSF/IL-3 and IL-3/GM-CSF fusion proteins bind to JM-1 cells with an affinity lower than that of IL-3 alone. In contrast, the GM-CSF/IL-3 and IL-40 3/GM-CSF fusion proteins bind to KG- 1 cells with a significantly higher affinity than that of 2~60~
IL-3 alone. The Kl value for both GM-C.SF/IL-3 and IL-3/(~M-CSF fusion proteins on KG-1 cells is 10-20 fold higher than on JM-l cells. Similarly, the Kl values determined for GM-CSF/L-3 and IL-3/GM-CSF on HL-60 cells are similar.
In view of the data shown in Examples 4-6 (which show a colrelation between binding S af~mity and enhanced biological activity), the above binding data suggest that the IL-3/GM-CSF fusion protein will have increased biological activity.

Claims (24)

1. A fusion protein having a formula selected from the group consisting of R1-R2, R2-R1, R1-L-R2 and R2-L-R1 wherein R1 is GM-CSF; R2 is IL-3; and L is a linker peptide sequence.

2. A fusion protein according to claim 1, wherein said linker sequence comprises amino acids selected from the group consisting of Gly, Asn, Ser, Thr and Ala.

3. A fusion protein according to claim 2, wherein the length of said linker sequence is 5 to 15 amino acids.

4. A fusion protein according to claim 3 selected from the group consisting of amino acid residues 1-271 depicted in Figure 1 and amino acid residues 1-275 depicted in Figure 2.

5. A fusion protein according to claim 3, wherein the fusion protein is huGM-CSF[Leu23Asp27Glu39]/Gly4SerGly5Ser/huIL-3[Pro8Asp15Asp70].

6. A fusion protein according to claim 1, wherein the human GM-CSF is selected from the group consisting of huGM-CSF, huGM-CSF[Leu23Asp27Glu39], huGM-CSF[Leu23], huGM-CSF[Leu23Asp27], huGM-CSF[Glu39], huGM-CSF[Asp27Glu39], huGM-CSF[Leu23 Glu39] and huGM-CSF[Asp27] and human IL-3 is selected from the group consisting of huIL-3, huIL-3[Pro8Asp15Asp70], huIL-3[Asp70], huIL-3[Asp15Asp70], huIL-3[Pro8Asp15], huIL-3[Pro8Asp70], and huIL-3[Asp15].

7. A DNA sequence encoding the protein of claim 1.

8. A DNA sequence encoding the protein of claim 4.

9. A DNA sequence encoding the protein of claim 5.

10. A DNA sequence encoding the protein of claim 6.

11. A DNA sequence according to claim 1, which is degenerate as a result of the genetic code to the DNA sequences defined in Figure 1 or Figure 2.

12. A recombinant expression vector comprising a DNA sequence according to claim 7

13. A recombinant expression vector comprising a DNA sequence according to claim 8.

14. A recombinant expression vector comprising a DNA sequence according to claim 9.

15. A recombinant expression vector comprising a DNA sequence according to claim 10.

16. A recombinant expression vector comprising a DNA sequence according to claim 11.

17. A process for preparing a fusion protein comprising GM-CSF and IL-3, comprising the step of culturing a suitable host cell comprising a vector according to claim 12 under conditions promoting expression.

18. A process for preparing a fusion protein comprising GM-CSF and IL-3, comprising the step of culturing a suitable host cell comprising a vector according to claim 13 under conditions promoting expression.

19. A process for preparing a fusion protein comprising GM-CSF and IL-3, comprising the step of culturing a suitable host cell comprising a vector according to claim 14 under conditions promoting expression.

20. A process for preparing a fusion protein comprising GM-CSF and IL-3, comprising the step of culturing a suitable host cell comprising a vector according to claim 15 under conditions promoting expression.

21. A process for preparing u fusion protein comprising GM CSF and IL-3, comprising the step of culturing a suitable host cell comprising a vector according to claim 16 under conditions promoting expression.

22. A composition for regulating immune or inflammatory responses in a mammal, comprising all effective amount of a fusion protein according to claim 1, and a suitable diluent or carrier.

23. A method for regulating immune responses in a mammal, comprising administering an effective amount of a composition according to claim 22.

24. A fusion protein comprising GM-CSF linked to IL-3.