EP0302103A1 - Colony stimulating factors having reduced levels of carbohydrate - Google Patents

Abstract

Proteins characterized by GM-CSF type activity and having the following structure: A-R1-B-R2-C, where A, B and C are sequences of GM-CSF peptides as illustrated in Table 1 , where A comprises the sequence Ala-1 up to Leu-26, B comprises the sequence Arg-30 up to Met-36 and C comprises the sequence Val-40 up to Glu-127. R1 and R2 each represents an independently selected peptide sequence or a peptide bond linking the above-mentioned sequences A, B and C, so that R1 is a peptide bond or a peptide sequence different from Asn-Leu-Ser and / or R2 is a bond peptide or a peptide sequence different from Asn-Glu-Thr.

Description

COLONY STIMULATING FACTORS HAVING REDUCED LEVELS OF CARBOHYDRATE

This invention relates generally to genetically engineered variants of human GM-CSF which have GM-CSF biological activity, a process for obtaining the proteins from genetically engineered cells, and therapeutic compositions containing the proteins for therapeutic uses involving stimulating primate hematopoietic progenitor cells which are responsive to recombinant or natural human GM-CSF.

The hematopoietic system provides a unique opportunity for the development of useful protein pharmaceuticals. In this system, hematopoietic progenitor cells found in the bone marrow continuously divide and differentiate ultimately resulting in the release of mature blood cells into the periphery. A number of different polypeptide growth factors have been identified which are capable, at least in vitro, of regulating the production of mature blood cells (reviewed in Metcalf, 1984, The Hemopoietic Colony Stimulating Factors, Amsterdam, Elsevier). In principle, any of these growth factors could prove valuable in the treatment of cytopenias, both naturally arising as well as those induced by chemotherapy or irradiation therapy for cancer [Gasson et al., 1983, "Lymphokines and Hematopoiesis" in Normal and Neoplastic Hematopoiesis (Golde & Marks, Eds.) Liss, NY]. These proteins are particularly attractive as therapeutic agents because their target cells in the bone marrow are readily accessible to soluble drugs via the circulatory system.

The hematopoietic growth factors which are also known as the colony stimulating factors were originally identified by their ability to stimulate the clonal expansion of bone marrow progenitor cells in semi-solid medium (Metcalf,

1984, supra). Traditionally, the colony stimulating factors are classified based upon the types of blood cells found in the colonies grown in vitro. Thus granulocytemacrophage colony stimulating factor or GM-CSF was identified as a hematopoietic growth factor which is capable of stimulating the formation of colonies containing both granulocytes and macrophages (reviewed in Metcalf,

1985, Science 229; 16). Both the murine and human GM-CSF cDNAs have been cloned and used to produce the respective recombinant growth factors (Gough et al., 1984, Nature 309:763; Wong et al., 1985, Science 228:819). With these developments, it became possible for the first time to perform detailed studies of the physical and biological properties of the these proteins. (Wong et al., 1985, in Cancer Cells 3; Growth Factors and Transformation, p. 235, Cold Spring Harbor Laboratory, NY; Metcalf et al., 1986, Blood 67:34; Tomonage et al., 1986, Blood 67:31). These studies have greatly extended the range of the known activities of GM-CSF. Thus GM-CSF has proved to be a potent activator of mature neutrophils, eosinophils and monocytes. (Weisbart et al., 1985, Nature 341:361; Grabstein et al., 1986, Science 232:506). In addition, Nathan and co-workers have found that the human GM-CSF in the presence of erythropoietin stimulates normal human progenitors to form colonies containing erythroid cells suggesting that this hematopoietin may interact with earlier progenitors than originally thought (Seif et al., 1985, Science 230:1171; Emerson et al., 1985, J. Clin Invest. 76:1287; Donahue, et al., 1985).

The large scale production of recombinant human GM-CSF has permitted us to begin to study the biological properties of this molecule in vivo (Donahue et al., 1986, Nature 321: 6073, 872-875). Previously, we have shown that the human GM-CSF is a potent stimulator of hematopoiesis in a primate model. Other investigators have used a murine model to show that another hematopoietin, murine interleukin 3 (IL- 3) , can stimulate murine blood cell production in vivo (Kindler et al., 1986, Proc. Natl. Acad. USA 83:1001). In preparation for initiating clinical studies with the recombinant human GM-CSF, we have extended our studies of the in vivo properties of this protein and variants thereof.

The variants of this invention are active colony stimulating factors which may be produced in more homogeneous form and which may possess improved pharmacokinetic profiles relative to natural or recombinant GM-CSF.

The polypeptide backbone of natural human GM-CSF includes two consensus Asn-linked glycosylation sites: Asn-Leu-Ser at positions 27-29 and Asn-Glu-Thr at positions 37-39. Natural human GM-CSF and the "recombinant" version thereof may be produced and recovered in somewhat heterogeneous form, in that both, either or neither of the two consensus glycosylation sites may be occupied by carbohydrate moieties. For the purpose of this disclosure, the numbering of amino acids begins with Ala-1 of the mature protein having an N-terminus comprising Ala-Pro-Ala-Arg- Ser-Pro... .

This invention involves novel glycosylation site variants of human GM-CSF which contain no Asn-linked carbohydrate moieties or only one such moiety. Significantly, we have found that the GM-CSF proteins which are characterized structurally by a reduced presence of carbo-hydrate moieties relative to fully glycosylated natural or recombinant GM-CSF are characterized biologically by an improved specific activity (up to 10-fold higher) relative to fully glycosylated natural or recombinant GM-CSF. This reduced presence of carbohydrate moieties results from amino acid substitution at one or both of the consensus N-linked glycosylation recognition sites present in the native human GM-CSF molecule. Asn-linked glycosylation recognition sites are presently believed to comprise tripeptide sequences which are specifically recognized by the appropriate cellular glycosylation enzymes. These tripeptide sequences are either asparagine-X-threonine or asparagine-X-serine, where X is any amino acid, except perhaps proline. A variety of amino acid substitutions, insertions or deletions at one or more of the three positions of a glycosylation recognition site results in non- glycosylation at the modified peptide sequence. By way of example, Asn-27 of GM-CSF has been replaced with Gln in one embodiment, Asn-37 has been replaced with Gln in another embodiment, and both Asn's have been replaced with Gln in a third embodiment. In the case of the double Gln replacement, the resultant glycoprotein (Gln-27, Gln-37) contains no N-linked carbohydrate moiety in contrast to native human GM-CSF which is a mixture of proteins containing zero, one or two such moieties. Similarly, the Gln-27 and Gln-37 variants each contain zero or one carbohydrate moiety. Those skilled in the art will appreciate that analogous glycoproteins having the same degree of glycosylation or non-glycosylation may be prepared by deleting Asn-27 and/or Asn-37 and/or by substituting other amino acids at positions 27 and 37 and/or by substituting for or deleting one or more amino acids at other positions within the respective glycosylation recognitions sites, e.g. at Ser-29 and Thr-39, as mentioned above and/or by substitution, insertion or deletion at one or more of the "X" positions of the tripeptide sites. This invention encompasses such nonglycosylated and monoglycosylated human GM-CSF variants. These variants may be described schematically with reference to the polypeptide of formula (1) below:

A-R¹-B-R²-C (1) wherein A, B and C represent the following domains of human GM-CSF, substantially as depicted in Table 1: A comprises the polypeptide sequence Ala-1 through Leu-26, B comprises Arg-30 through Met-36 and C comprises Val-40 through Glu- 127. R¹ and R² of formula (1) above each represent a peptide bond or peptide sequence linking the aforementioned polypeptide domains A, B and C by peptide bonds. Thus, the compounds of this invention have a peptide sequence substantially the same as human GM-CSF (as shown in Table 1) except at one or both of R¹ and R². In the variants of this invention one or both of R¹ and R² are (i) tripeptide sequences other than consensus glycosylation sites, (ii) dipeptide sequences, (iii) single amino acid residues, or (iv) peptide bonds, as described above. The moieties selected for R¹ and R² may be the same or different from one another, i.e. may be independently selected. Peptide sequences other than consensus glycosylation sites present in the polypeptide sequence of human GM-CSF are tabulated below in Table 2.

Table 2: Alternative Peptide Sequences

R¹ R²

(wt) (Asn Leu Ser) (Asn Glu Thr)

I ( X Leu Ser) ( X Arg Thr)

II (Asn Y Ser) (Asn V Thr)

III (Asn Leu Z) (Asn Arg Z )

X = any amino acid except Asn, or a peptide bond

Y = " " " " Leu, or a peptide bond

Z = " " " " Thr or Ser, or a peptide bond

V = " " " " Glu, or a peptide bond wt= wild type, i.e. prior to mutagenesis Table 1: Peptide and nucleotide sequence of GM-CSF including altenative nucleotides and bases with regions corresponding to R1 and R2 underlined

GCT GGAGG ATC TGG CTG CAG AGC CTG CTG CTC TTG GGC ACT GTG GCC TCC MET Trp Leu Gln Ser Leu Leu Leu Leu Gly The Val Ala Cys

Ser Arg

T GAGC ATC TCT GCA CCC GCC CGC TCG CCC AGC CCC AGC ACG CAG CCC TGG GAG CATSer lle Ser Ala Pro Ala Arg Ser Pro Ser Pro Ser Thr Gln Pro Trp Glu Bis

1

GTG AAT GCC ATC CAG GAG GCC CGG CGT CTC CTG AAC CTG AGT AGA GAC ACT GCT

Val Asn Ala lle Gln Glu Ala Arg Arg Leu Leu Asn Leu Ser Arg Asp Thr Ala 26 30 lle Val

A G

GCT GAG ATG AAT GAA ACA GTA GAA GTC ATC TCA CAA ATG TTT GAC CTC CAG GAG

Ala Glu MET Asn Glu Thr Val Glu Val lle Ser Glu MET Phe Λap Leu Gln Glu 36 40

A T

CCG ACC TGC CTA CAG ACC CGC CTG GAG CTG TAC AAG CAG GGC CTG CGG GGC AGC

Pro Thr Cys Leu Gln Thr Arg Leu Glu Leu Tyr Lys Gln Gly Leu Arg Gly Ser

CTC ACC AAC CTC AAG GGC CCC TTG ACC ATG ATG GCC AGC CAC TAC AAG CAG CAC Leu Thr Lys Leu Lys Gly Pro Leu Thr MET MET Ala Ser His Tyr Lys Gin His

100 Ile A G T

TCC CCT CCA ACC CCG GAA ACT TCC TGT GCA ACC CAG ACT ATC ACC TTT GAA AGT Cys Pro Pro Thr Pro Glu Thr Ser Cys Ala Thr Gln Thr Ile Thr Phe Glu Ser

Thr C TTC AAA GAG AAC CTG AAG GAC TTT CTG CTT GTC ATC CCC TTT GAC TGC TGG GAG Phe Lys Glu Asn Leu Lys Asp Phe Leu Leu Val He Pro Phe Asp Cys Trp Glu

Gly G CCA GTC CAG GAG TGA GACCGGCCAG ATGAGGCTGG CCAAGCCCCG GAGCTGCTCT CTCATGAAAC Pro Val Gln Glu

127

G C T

AACAGCTACA AACTCAGGAT GGTCATCTTG GAGGGACCAA GGGGTGGGCC ACAGCCATGG TGGGAGTGGC

G

CTCGACCTCC CCTGGGCCAC ACTGACCCTG ATACAGGCAT GGCAGAAGAA TGGGAATATT TTATACTGAC AGAAATCAGT AATATTTATA TATTTATATT TTTAAAATAT TTATTTATTT ATTTATTTAA GTTCATATTC CATATTTATT CAAGATGTTT TACCGTAATA ATTATTATTA AAAATATGCT TCTAAAAAAA AAAAAAAAAA By way of example, several glycoproteins of this invention are depicted below in Table 3.

Table 3 : Exemplary Compounds

A-R¹-B-R²-C** Compound R¹

1 Gln Leu Ser Asn Glu Thr 2 Gln Leu Ser Gln Glu Thr

3 Asn Leu Ser Gln Glu Thr

4 Asn Leu Ala Asn Glu Thr 5 Asn Leu Ser Asn Glu He 6 Asn Leu Ala Asn Glu He

7 Gln Leu Ser Asn Glu He

8 -- Leu Ser Asn Glu Thr

9 Asn Leu Ser -- Glu Thr

10 Asn Leu -- Asn Glu Thr

11 Asn Leu -- -- Glu Thr

12 Asn Leu -- Gin Glu Thr

13 Asn -- Ser Asn Glu Thr

** A, B and C are as defined above; mutagenized positions are underlined; non-mutagenized R groups bear N-linked carbohydrate moieties covalently bonded thereto; "--" indicates the site of a deleted amino acid, i.e. represents a peptide bond. This invention further encompasses other variants of GM-CSF, e.g. allelic variants and other variants having one or more amino acids deleted and/or replaced with different amino acids so long as such variants are characterized biologically by possessing GM-CSF-type biological activity and structurally by possessing modifications at one or both consensus N-linked glycosylation sites, as described herein. Such variants are also characterized structurally (i) by being encoded by a DNA sequence capable, or capable but for the degeneracy of the genetic code, of hybridizing to a DNA sequence encoding native human GM-CSF, under stringent hybridization conditions, as is known in the art; or (ii) by having a peptide sequence at least about 90%, and preferably at least about 95%, homologous to the peptide sequence of human GM-CSF, so long as one or both of the consensus Asn-linked glycosylation sites are modified to other than a consensus Asn-linked glycosylation site.

In one subgenus of the invention the proteins contain one so-called "complex carbohydrate" sugar moiety characteristic of mammalian glycoproteins. As exemplified in greater detail below, such "complex carbohydrate" glycoproteins may be produced by expression of a DNA molecule encoding the desired polypeptide sequence in mammalian host cells. Suitable mammalian host cells and methods for transformation, culture, amplification, screening and product production and purification are known in the art. See e.g. Gething and Sambrook, Nature 293:620-625 (1981), or alternatively, Kaufman et al., Molecular and Cellular Biology 5: (7):1750-1759 (185) or Howley et al., U.S. Patent No. 4,419,446.

A further aspect of this invention involves monoglycosylated variants as defined above in which the carbohydrate moiety is a processed form of the initial dolicol-linked oligosaccharide characteristic of insect cell-produced glycoproteins, as opposed to a "complex carbohydrate" substituent characteristic of mammalian glycoproteins, including mammalian derived GM-CSF. Such insect cell-type glycosylation is referred to herein as "high mannose" carbohydrate for the sake of simplicity. For the purpose of this disclosure, complex and high mannose carbohydrates are as defined in Kornfeld et al., Ann. Rev. Biochem. 54: 631-64 (1985) . "High mannose" variants in accordance with this invention are characterized by a variant polypeptide backbone as described above and as exemplified in Table 3. Such variants may be produced by expression of a DNA sequence encoding the variant in insect host cells. Suitable insect host cells as well as methods and materials for transformation/transfection, insect cell culture, screening and product production and purification useful in practicing this aspect of the invention are known in the art. Glycoproteins so produced also differ from natural GM-CSF and from GM-CSF produced heretofore by recombinant engineering techniques in mammalian cells in that the variants of this aspect of the invention do not contain terminal sialic acid or galactose substituents on the carbohydrate moieties or other protein modifications characteristic of mammalian derived glycoproteins.

The variant proteins of this invention which contain no N-linked carbohydrate moieties may also be produced by expressing a DNA molecule encoding the desired variant, e.g. compounds 2,6,7,11 and 12 of Table 3, in mammalian, insect, yeast, fungal or bacterial host cells. As indicated above suitable mammalian and insect host cells, and in addition, suitable yeast, fungal and bacterial host cells , as well as methods and materials for transformation/transfection, cell culture, screening and product production and purification useful in practicing this aspect of the invention are also known in the art.

cDNAs encoding these compounds may be readily prepared and mutagenized at one or both of the codons for R¹ and R² and may be inserted into expression vectors and expressed in host cells by the methods disclosed herein.

As should be evident from the preceding, all variants of this invention are prepared by recombinant techniques using DNA sequences encoding GM-CSF analogs which contain fewer or no potential N-linked glycosylation sites relative to natural human GM-CSF. Such DNA sequences may be produced by conventional site-directed mutagenesis of DNA sequences encoding GM-CSF.

DNA sequences encoding human GM-CSF have been cloned and characterized. One such clone is available from the American Type Culture Collection, 12301 Parklawn Drive, Rockville, Maryland 20852 (USA) where it has been deposited in plasmid pCSF-1 (in E. coli MC1061) under accession No. 39754. If desired, the human GM-CSF clone may be readily excised from pCSF-1 as a ~780 basepair (bp) EcoRI fragment. As mentioned above, DNA sequences encoding individual variants of this invention may be produced by conventional site-directed mutagenesis of a DNA sequence encoding human GM-CSF or analogs or variants thereof. Such methods of mutagenesis include the Ml3 system of Zoller and Smith, Nucleic Acids Res. 10:6487-6500 (1982); Methods Enzymol. 100:468-500 (1983); and DNA 3:479-488 (1984), using single stranded DNA and the method of Morinaga et al., Bio/technology, 636-639 (July 1984), using heteroduplexed DNA. Several exemplary oligonucleotides used in accordance with such methods to convert an asparagine residue to threonine or glutamine, for example, are shown in Table 4. It should be understood, of course, that DNA encoding each of the glycoproteins of this invention may be analogously produced by one skilled in the art through site-directed mutagenesis using (an) appropriately chosen oligonucleotide(s). Expression of the DNA by conventional means in a mammalian, yeast, fungal, bacterial, or insect host cell system yields the desired variant. Mammalian expression systems and the variants obtained thereby are presently preferred.

The mammalian cell expression vectors described herein may be synthesized by techniques well known to those skilled in this art. The components of the vectors such as the bacterial replicons, selection genes, enhancers, promoters, and the like may be obtained from natural sources or synthesized by known procedures. See Kaufman et al., J. Mol. Bio., 159:51-521 (1982); Kaufman, Proc. Natl. Acad. Sci. USA 82:689-693 (1985).

Table 4: Exemplary Olicfonucleotides*

No. Sequence Mutation

1. TCTACTCAGCTGCAGGAGACG N-27 > Gin

2. TACTGTTTCCTGCATCTCAGC N-37 > Gin

3. TCTACTCAGTTGTCAGGAGACG N-27 > Thr

4. TACTGTTTTGTCATCTCAGC N-37 > Thr

*Codons for replacement amino acids are underlined. As those skilled in this art will appreciate, oligonucleotides can be readily constructed for use in deleting one or more amino acids or for inserting a different (replacement) amino acids at a desired site by deleting one or more codons or substituting the codon for the desired amino acid in the oligonucleotide, respectively. Other mutagenesis oligonucleotides can be designed based on an approximately 20-50 nucleotide sequence spanning the desired site, with replacement or deletion of the original codon(s) one wishes to change. Established cell lines, including transformed cell lines, are suitable as hosts. Normal diploid cells, cell strains derived from in vitro culture of primary tissue, as well as primary explants (including relatively undifferentiated cells such as hematopoietic stem cells) are also suitable. Candidate cells need not be genotypically deficient in the selection gene so long as the selection gene is dominantly acting.

The host cells preferably will be established mammalian cell lines. For suitable integration of the vector DNA into chromosomal DNA, and for subsequent amplification of the integrated vector DNA, both by conventional methods, CHO (Chinese hamster ovary) cells are presently preferred. Alternatively, the vector DNA may include all or part of the bovine papilloma virus genome (Lusky et al., Cell, 36: 391-401 (1984) and be carried in cell lines such as C127 mouse cells as a stable episomal element. Other usable mammalian cell lines include HeLa, COS-1 monkey cells, mouse L-929 cells, 3T3 lines derived from Swiss, Balb-c or NIH mice, BHK or HaK hamster cell lines and the like.

Stable transformants then are screened for expression of the product by standard immunological or activity assays. The presence of the DNA encoding the variant proteins may be detected by standard procedures such as Southern blotting. Transient expression of the DNA encoding the variants during the several days after introduction of the expression vector DNA into suitable host cells such as COS-1 monkey cells is measured without selection by activity or immunological assay of the proteins in the culture medium.

In the case of bacterial expression, the DNA encoding the variant may be further modified to contain different codons for bacterial expression as known in the art and preferably is operatively linked in-frame to a nucleotide sequence encoding a secretory leader polypeptide permitting bacterial expression, secretion and processing of the mature variant protein, also as is known in the art. The compounds expressed in mammalian, insect, yeast, fungal or bacterial host cells may then be recovered, purified, and/or characterized with respect to physiochemical, biochemical and/or clinical parameters, all by known methods.

These compounds have been found to bind to monoclonal antibodies directed to human GM-CSF and may thus be recovered and/or purified by immunoaffinity chromatography using such antibodies or by other conventional methods or methods described hereinafter. Futhermore, these compounds possess human GM-CSF-type activity, e.g. , compounds of this invention effectively stimulate the proliferation of granulocytes and macrophages, as measured in conventional assays.

This invention also encompasses compositions for hematopoietic therapy which comprise a therapeutically effective amount of a variant described above in admixture with one or more pharmaceutically acceptable parenteral carriers and/or conventional excipients. Such composition can be used in the same manner as that described for human GM-CSF and should be useful in humans or other primates. The exact dosage and method of administration will be determined by the attending physician depending on potency and pharmacokinetic profile of the particular compound as well as on various factors which modify the actions of drugs, for example, body weight, sex, diet, time of administration, drug combination, reaction sensitivities and severity of the particular case. The following examples are given to illustrate embodiments of the invention. It will be understood that these examples are illustrative, and the invention is not to be considered as restricted thereto except as indicated in the appended claims.

Those of ordinary skill in the art to which this invention pertains will appreciate that other host cells, promoters and vectors containing the relevant cDNA, etc. as discussed above, may also be used in the practice of each embodiment of this invention.

The DNA manipulations employed are, unless specifically set forth herein, in accordance with Maniatis et al., Molecular Cloning: A Laboratory Manual (Cold Spring Harbor, NY 1982).

Experimental Examples

Methods: Cell Culture

The growth of the Mo T-cell line has been described previously (Wong et al., 1985, Science, supra). The Cl0-MJ2 T-cell line was grown as described by Arya et al., 1984, Science 223:1086). These cells were induced to make GM-CSF by incubation at 5×10⁵ cells/ml for 24 hrs in the presence of 0.3% phytohemagglutin (PHA) and 5 mg/ml phorbolmyrystic acetate (PMA) in the presence of 10% fetal calf serum. Peripheral blood lymphocyte (PBL) conditioned medium was prepared similarly using Ficoll separated PBL's at a final density of 1×10⁶ cells/ml. Monkey COS-1 cell transfections were performed using the DEAE-dextran protocol with chloroquin treatment as described previously (Wong et al., 1985, Science, supra). The cells were pulse-labeled by incubation with 0.5 mC of ³⁵S-methionine in 0.5ml (per 10cm dish) of Dulbecco's Modified Eagle's Medium (DMEM) for 4 hours, 48 hours after the chloroquin treatment. In one transfection, 10 g/ml tuniσamycin (Sigma) was added 30 minutes prior to labeling to inhibit the addition of N-linked carbohydrate. The plasmid pCSF-1 which was isolated by expression cloning was introduced into Chinese Hamster Ovary cells (CHO) using standard methods (Kaufman and Sharp, 1982, J. Mol. Biol. 159:601) and the GM-CSF sequences in the resulting cells were amplified to approximately 200 copies per cell by step-wise selection with methotrexate to yield a cell line (CHO-D2) which expresses high levels of human GM-CSF. Confluent dishes of CHO-D2 cells were pulse-labeled for four hours with 0.5 mC of ³⁵S-methionine in 0.5 ml DMEM. Analysis of Natural GM-CSF's

Ten ml samples of conditioned medium from Mo cells, C10-MJ2 cells, lectin-stimulated PBL's, and from primary blasts from a patient with acute myeloblastic leukemia (kindly provided by J. Griffin, Dana Farber Cancer Institute, Boston, Massachusetts.) were passed over a three ml antibody column prepared from an antiserum raised in sheep by immunization with recombinant human GM-CSF. The bound GM-CSF was eluted with 0.1M glycine, pH 2.8 and concentrated by centrifugation through a Centricon 10 concentrator (Amicon). Aliqouts of these samples and an aliquot of CHO-D2 conditioned medium were fractionated by electrophoresis through a 12% polyacrylamide gel and the positions of the GM-CSF species in each sample visualized by transferring the protein to nitrocellulose (BioRad Transblot System) and incubating the resulting filters with a rabbit anti-GM-CSF antibody. The antibody bound to the filters was visualized by treatment with ¹25I-staph-A protein (NEN) followed by autoradiography.

Site Specific Mutacrenesis of GM-CSF

Specific mutations in the GM-CSF sequence which eliminated either the first (Site 1) or second (Site 2) potential sites of N-linked carbohydrate addition were generated using the gapped heteroduplexed site-directed mutagenesis as described by Morinaga, et al., 1984, Biotechnology 2: 636. This was accomplished using two synthetic 21mers (#1585 having the sequence d(TCTACTCAGCTGCAGGAGACG) and #1590 having the sequence d(TACTGTTTCCTGCATCTCAGC)) which spanned the DNA sequence for either Site 1 or Site 2 such that the codons for either asparagine 27 (#1585) or asparagine 37 (#1590) were positioned in the center of the oligonucleotide and were converted to glutamine codons. pCSF-1 DNA was linearized by treatment with Sall which cleaves this plasmid at a unique site in the tetracycline resistance gene and p91023 (b) was linearized at its unique cloning site by cleavage with EcoRI. The linearized plasmid DNA's were mixed in equimolar amounts, denatured by base treatment and allowed to re-anneal to form heteroduplexes. These heteroduplexes were annealled with either oligonucleotide # 1585 or # 1590 then repaired by treatment with the large fragment of DNA polymerase I (Klenow fragment) in the presence of all four deoxynucleotide triphosphates, ATP and T4 DNA ligase. The products of these reactions were used to transform E. coli MC 1061 and the desired mutants identified by colony hybridization with the appropriate oligonucleotide. The double mutant in which both sites were eliminated was made beginning with the DNA from the Site 1 single mutant by an identical mutant strategy. All of the mutations were confirmed by DNA sequence analysis.

pCSF-1 containing the mutagenized cDNA encoding a GM-CSF variant of this invention, or, if desired, an alternative conventional mammalian, bacterial, or insect expression vector containing the mutagenized cDNA, may then be used to transfect or transform desired corresponding host cells. Culture of the transfected of transformed host cells by conventional means yields the desired variant, which may then be recovered, and if desired, further purified, by conventional methods such as immunoaffinity chromatography using antibodies, e.g. directed to native or recombinant GM-CSF, or by the purification method described below.

Alternative mammalian expression vectors are well known in the art and include vectors such as pMT2. The preparation of pMT2 from pMT2-VWF (ATCC No. 67122) and use thereof is described in published International application WO 87/07144 (see, page 23). Suitable bacterial, fungal and alternative mammalian expression vectors and methods for their use are des cribed in published International application WO 88/00206. Suitable yeast, and alternative bacterial and mammalian, expression vectors and methods for their use are described in published International application WO 86/00639.

Purification of ³⁵S-Met-labeled GM-CSF

Conditioned medium from metabolically labeled CHO-D2 cells was diluted 1:1 with 0.02M Tris HCl, pH 7.4, 1.0mM EDTA, 0.01% Tween 80, and passed over a 1 ml column of DEAE-Trisacryl M (LKB). The column was rinsed in the same buffer containing 50mM NaCl and the GM-CSF eluted with 250mM NaCl in the same buffer. To isolate the GM-CSF having either one (1-N) or both (2-N) N-linked carbohydrate sites occupied, the GM-CSF eluted from the DEAE column was passed over a Vydac C4 analytical reverse phase column as described previously (Wong et al., 1985, Cancer Cells, supra). The 2-N GM-CSF, which is less hydrophobic, eluted from this column first (about 41% acetonitrile). To isolate GM-CSF lacking N-linked carbohydrate, the DEAE purified protein was passed over a 1 ml column of hydroxyl apatite (BioRad, DNA grade). The column was washed with 10mM sodium phosphate, pH 7.0 to elute the 1-N and 2-N protein and the 0-N GM-CSF (lacking N-linked carbohydrate) was eluted with 25 mM sodium phosphate, pH 7.0. To isolate all of the forms of GM-CSF at once, the labeled CHO-D2 medium was passed over the anti-GM-CSF antibody column (3ml) (as described for the conditioned medium samples above). The extent of sialylation of the labeled GM-CSF was assessed by binding to the lectin Ricinus Communis Agglutinin I (RCA 120, agarose-bound, from Vector Laboratories) (Debray et al., 1983, in Lectins: Biology, Biochemistry, Clinical Biochemistry, 3:335 (Bog-Hausen & Spengler, eds) Walter de Gruyter, Berlin, NY). To do this, the GM-CSF was diluted 1:1 with 0.01M Tris-Ci, pH 7.4, 0.1M NaCl, 0.01% Tween 80 and passed over a 0.2 ml column of RCA 120. The column was washed with 10 volumes of the same buffer and the bound protein eluted with the same buffer containing 0.5 M galactose. In general, none of the labeled GM-CSF could be bound to the RCA 120. Essentially all of the 1-N and 2-N GM-CSF specifically bound to the RCA-120 after neuriminidase treatment indicating that at least one sialic acid residue had been removed from each molecule of GM-CSF having N-linked carbohydrate. Little if any of the O-N GM-CSF was bound to RCA 120 suggesting either that the sialic acid on the 0- linked carbohydrate is resistant to neuriminiadase treatment or that the desialylated O-N GM-CSF does not bind to ricin 120.

Rat Clearance Studies

Each sample of ³⁵S-labeled GM-CSF was concentrated by centrifugation in a Centricon 10 microcentrator (Amicon) to 0.2 ml in PBS. These samples (10⁶ cpm) were injected rapidly into the tail vein of a 200-300 gram male Sprague-

Dawley rat that had previously been anesthetized with sodium pentobarbital (15 mg/250g). At various times after the injection, 0.2-0.3 ml aliqouts of blood were collected from the end of the tail into tubes containing 40 units of heparin. After the final sample, the rat was exsanguinated and the major body organs (lung, kidney, and liver) were collected and weighed. 0.05 ml aliqouts of plasma were spotted onto the glass fiber filters and washed gently with cold 5% trichlorσacetic acid then cold methanol and were finally dried and counted in anhydrous scintillation fluid (NEN). Radioactivity in each organ was determined by homogenizing 200-300 mg samples in 1ml of 1% SDS, heating to 100 C, and counting 0.1 ml in 5 ml of Aquasol (NEN). Administration of GM-CSF to Monkeys

GM-CSF was administered to three different monkeys (M. fasicularis) by continuous infusion through catheters surgically implanted in the jugular veins as described previously (Donahue, et al. 1986). The GM-CSF was purified by the Genetics Institute Pilot Development Laboratory using lentil-lectin affinity chromatography and reverse phase HPLC essentially as described by Gasson et al., 1984, Science 226:1339. This GM-CSF had a specific activity of 1-2 ×10⁶ units per ml in our CML proliferation assay as discussed elsewhere (Donahue et al., 1986, Nature, supra). The protein was found to be a mixture of about 30% 1-N and 70% 2-N GM-CSF.

Results

Carbohydrate Structure of GM-CSF

From the earliest attempts to purify human GM-CSF, it was recognized that this hematopoietin is heterogeneously glycosylated (Lusis et al., 1981, Blood 57: 13; Wong et al., 1985, Science, supra and Cancer Cells, supra). However, the extent of this heterogeneity, proved to be even greater than was originally thought. The GM-CSF derived from different sources, including lectin-stimulated peripheral blood lymphocytes (PBL's), several T-cells lines and primary leukemic blast cells from a patient with acute mylogenous leukemia ranges in apparent molecular mass between 14.5 kD and about 32-34 kD. As the molecular mass predicted from the primary sequence is 14.5 kD, some forms of GM-CSF comprise more than 50% carbohydrate.

Although the size range of the GM-CSF from different sources was found to be similar, the distribution of sizes within that range can be extremely variable. For example, GM-CSF produced by normal PBL's consisted largely of molecules at the high end of the range (22 to about 32-34 kD) while the AML blast cell derived protein on the average was much less glycosylated. The distribution of species of recombinant GM-CSF produced by our engineered Chinese Hamster Ovary cell line (CHO-D2), was similar to the size distribution observed for the GM-CSF from normal PBL's.

The distribution of the different species of GM-CSF from all of the different sources was observed to fall into three general size classes, most clearly evident in the case of the AML-derived protein. Because the primary sequence of GM-CSF contains two consensus sequences (Asn-X- Thr/Ser) (Winzler, 1973, in Hormonal Proteins and Peptides 1: 1 (CH. Li, ed)., Academic Press, NY) for asparagine linked carbohydrate (Site 1 is Asn 271 Site 2 is Asn 37) we predicted that these size classes are generated by the state of occupancy of these two sites. We expected that molecules with both Site 1 and Site 2 modified would fall in the largest size class (22 to about 32-34 kD); molecules

with either Site 1 or Site 2 occupied would be in the intermediate size class (18-22 kD) and molecules in which neither site is occupied would fall in the smallest size class (14-18 kD). To test this prediction, we used oligonucleotide primed site-directed mutagenesis to generate mutations in the GM-CSF sequence which eliminated either one or both of the consensus N-linked carbohydrate addition sequences by converting the asparagine codons to codons for glutamine. These mutants DNAs, each constructed in the expression vector p91023 (b) (Wong et al., Science, supra), were used to transfect monkey COS-1 cells. The expressed proteins were visualized by SDS PAGE analysis of conditioned medium from the transfected cells which had been pulse-labeled with ³⁵S-met. Alteration of the carbohydrate addition sites resulted in the predicted shift in the size distribution of the GM-CSF. Elimination of either Site 1 or Site 2 prevented the synthesis of any molecules falling in the largest size class while the simultaneous alteration of both Site 1 and Site 2 prevented the expres s i on o f GM-CSF in either the large or intermediate size classes. The double mutant directed the expression of GM-CSF with a size distribution similar in range to that observed when COS-1 cells transfected with the wild-typ e GM- CS F s equence were treated with tunicamycin , a drug which inhibits the addition of asparagine linked carbohydrate (Duskin and Mahoney, 1982 , J . Biol . Chem 257 : 3105) . These results demonstrate that the maj or siz e heterogeneity of GM-CSF results from different degrees of occupancy of the two sites of N-linked carbohydrate. We contemplate that the variants of this invention possess pharmacokinetic profiles and biological activities which are similar, generally, to those of the natural or recombinant GM-CSF proteins possessing corresponding levels of N-linked glycosylation.

Role of Carbohydrate in Clearance of GM-CSF

The function of the carbohydrate modification of GM-CSF is poorly understood. The activity of GM-CSF measured in vitro has been found to decrease with increasing carbohydrate content yet the natural protein from PBL's is highly glycosylated. To begin to study the effects of the carbohydrate modifications on the clearance of the protein from the bloodstream, we have used a rat model system to follow the fate of metabolically labeled GM-CSF after intravenous injection. In this study, we used conventional protein fractionation to resolve the three distinct sizes of GM-CSF expressed by our CHO-D2 cell line. As has been observed with erythropoietin (Steinberg et al., 1986, Blood 67:646), the clearance of GM-CSF from the bloodstream of a rat follows biphasic kinetics. The initial phase (called the a phase) of the clearance largely reflects distribution of the protein into all extracellular fluid of the animal while true plasma clearance is represented by the second (or β phase) portion of the decay curve (Shargel, L. and A.B.C. Yu, 1985, Applied Biopharmaceutical and Pharmacokinetics, Appleton-Century-Crofts, Norwalk, Conn.). The extent of modification of the GM-CSF via N-linked carbohydrate had a significant effect on the a but not the β portion of the curve. In the case of GM-CSF having no N- linked carbohydrate, more than 60% of the GM-CSF was lost from the blood stream with an apparent half life of about 2.5 minutes. The remaining 40% was cleared with an apparent half life of 16 minutes. Approximately 60% of the GM-CSF having one N-linked carbohydrate residue was lost from circulation with an apparent half life of 4 minutes while the remainder was cleared during the β phase having a 15 minute half life. In contrast, the GM-CSF in which both N-linked glycosylation sites were occupied, only 10% of the GM-CSF was lost from circulation during the β phase with a half life of 4.5 minutes and 90% of the injected protein was cleared during the β phase with an apparent half life of 15 minutes. These differences may reflect differences in the rate of distribution of the GM-CSF in the extra cellular body fluids or a mechanism of clearance also contributing to the observed altered kinetics.

In analyzing the distribution of the labeled GM-CSF in the various organs of the rat thirty minutes post injection, it was clear that the predominant site of clearance for this protein is the kidney. Because the major organs of the rat are highly vascular, their blood content would be expected to contribute significantly to the labeled GM-CSF found in each organ. If we assume for example that there is no significant clearance of GM-CSF in the lungs then it is apparent that the total level of radioactivity in the liver (which has 5-7 times the mass of the lungs) largely resulted from blood contamination and not specific clearance of the glycoprotein in this organ. The high specific activity of GM-CSF found in the kidney, however, clearly demonstrated that this organ is an important site of clearance for this protein.

In a separate study, we used the enzyme neuraminidase to remove the terminal sialic acid residues of metabolically labeled GM-CSF. The desialylated GM-CSF was isolated by binding to a column to which the lectin Ricin 120 had been covalently attached . This lectin has been shown to specifically bind terminal galactose residues which are exposed upon enzymatic removal of sialic acid from glycoproteins (Debray et al . , 1983 , supra) . Prior to the treatment with neuraminidase, none of the labeled GM-CSF was bound by the lectin column. Post treatment, all of the labeled GM-CSF having N-linked carbohydrate was bound to the column and specifically eluted by competition with 0.5 M galactose . A comparison of the clearance of the sialylated and desialylated GM-CSFs was made in our rat model . In this experiment, the rats were sacrificed 15 minutes after injection and the distribution of CSF in the various organs determined. From this experiment, it was clear that the exposure of the terminal galactose residues by removal of sialic acid dramatically altered the organ clearance of GM-CSF: desialylated GM-CSF was cleared in the liver while the sialylated protein was cleared by the kidney. Because both of the samples represented mixtures of different s iz e classes of GM-CSF the kinetics of clearance were difficult to analyze. However, the levels o f c ircul at ing GM- CS F 15 minutes after inj ection demonstrated that the desialylated GM-CSF was cleared more rapidly than the fully sialylated protein. Dose Responsiveness of GM-CSF in Primates

The effect of the carbohydrate on the clearance of GM-CSF in the rat might be expected to influence the efficacy of the protein in vivo. Previously, we have shown that the CHO-D2 derived GM-CSF is a potent stimulator of hematopoiesis in M. fasicularis (macaque) and M. mulatta (rhesus) (Donahue et al., 1986, Nature, supra). In preparation for human clinical trials, we have undertaken a study of the dose response relationships of GM-CSF administration in monkeys. In early experiments, we found that desialylated GM-CSF (as assessed by 100% binding to a ricin 120 column) was ineffective at stimulating primate hematopoiesis (data not shown). In contrast, GM-CSF with a much higher sialic acid content (10 moles per mole of GM- CSF) has been found to elicit a rapid leukocytosis when administered at a rate of 10μg/kg/day. The observed response in the levels of circulating blood cells is dose dependent; when GM-CSF was administered at a rate of 5 μg/kg/d, a minimal response was observed. The dose dependency is also evident in the rate of increase in the white cell count, which increased with increasing rates of administration of the protein.

Discussion

Several of the natural hematopoietins including GM-CSF, erythropoietin (Goldwasser et al., 1974, J. Biol. Chem. 249:4202) and CSF-1 (reviewed in Stanley, 1985, Methods in Enzymology, 116: 564, Academic Press, NY) are extensively modified by the addition of asparagine linked carbohydrate. The modification of GM-CSF in this fashion is particularly heterogeneous and varies greatly between the different sources of the protein. The generation of mutant forms of GM-CSF plus our detailed structural analysis of the native sequence have revealed that this heterogeneity results in large part from different states of occupancy of the two sites for additions of N-linked sugar. The recombinant GM-CSF made either in CHO cells or monkey COS-1 cells is also further modified by the addition of O-linked sugar at several different sites in the molecule, generating further heterogeneity in the structure of the protein.

The function of the carbohydrate modifications of the hematopoietins is unclear. The specific activity of GM-CSF measured in vitro is significantly depressed in the largest most fully glycosylated forms of the protein relative to the smaller, less heavily glycosylated molecules. For this reason, current efforts have focused on the effects of the carbohydrate structures on the clearance of the hormone from the blood stream. Here we have shown that the effective half life of GM-CSF in the blood stream of a rat following a single intravenous bolus injection is significantly increased by the addition of N-linked carbohydrate. The clearance of GM-CSF in the rat follows biphasic kinetics and it is the first or a phase that is prolonged by the carbohydrate modification. In this study, we have not distinguished between simple distribution of the hormone throughout the extra-cellular fluid of the animal and specific organ-mediated clearance. It will be of interest to determine if the effect is on the retention of the molecule within the circulatory system of the animal or if there is a site of rapid clearance of GM-CSF that is blocked by the carbohydrate modification of the protein. Because GM-CSF is largely cleared in the kidney, this second site of clearance most likely would also reside in this organ.

The clearance of asialoglycoproteins from the blood stream is well documented (Neufled et al., 1980, The Biochemistry of Glycoproteins and Proteoglycans, p. 241 (Lennarz, ed) Plenum Press, NY). The receptor for the penultimate galactose residues of glycoproteins which are exposed when complexed carbohydrate is desialylated is found in the liver and has been extensively studied (Schwarz et al., 1981, J. Biol. Chem 256: 8878). As expected, desialylated GM-CSF is rapidly cleared from the circulation in the liver. This rapid clearance appears to reduce the effectiveness of the molecule in elevating the white count in monkeys. The stimulation in the levels of circulating blood cells achieved with extensively sialylated GM-CSF was much more dramatic than observed using the desialylated protein. The stimulation with the sialated GM-CSF was dose dependent as the higher doses resulted in a more rapid increase in the numbers of circulating blood cells.

We have used a mammalian cell expression system to produce large quantities of glycosylated GM-CSF that structurally closely resembles that natural molecule. This material is a potent stimulator of hematopoiesis in monkeys and has no deleterious effects on the animals. These studies support the hypothesis that recombinant GM-CSF will be effective in treating cytopenias in humans that frequently result from chemical or radiation therapy for cancer. We further hope that this molecule will prove effective in speeding the recovery of patients undergoing bone marrow transplantation. If this proves to be case, the high mortality from this treatment might be greatly reduced.

Claims

What is claimed is:

1. Proteins characterized by possessing GM-CSF-type biological activity and having a peptide sequence substantially as shown in Table 1 except that 1 - 6 amino acids are replaced with different amino acids and/or deleted within the regions Asn-27 through Ser-29 and Asn-37 through Thr-39, such that one of both of said regions are completely deleted or replaced by a single amino acid residue, a dipeptide sequence, or a tripeptide sequence other than Asn-X-Ser or Asn-X-Thr' where X is any amino acid except Pro.

2. A protein of claim 1 characterized by possessing zero or one N-linked carbohydrate moieties.

3. A cDNA encoding a protein of claim 1.

4. A host cell containing a cDNA of claim 3 which is capable of expressing the cDNA.

5. A method for producing a protein of claim 1 which comprises culturing a host cell containing and capable of expressing a cDNA encoding the protein.

6. A pharmaceutical composition for hematopoietic therapy comprising a therapeutically effective amount of a protein of claim 1 in admixture with one or more pharmaceutically acceptable parenteral carriers and/or conventional excipients.