WO1991002058A1

WO1991002058A1 - Production of vascular endothelial cell growth factor

Info

Publication number: WO1991002058A1
Application number: PCT/US1990/004227
Authority: WO
Inventors: Edmund G. Tischer; Judith A. Abraham; John C. Fiddes; Richard L. Mitchell
Original assignee: California Biotechnology Inc.
Priority date: 1989-07-27
Filing date: 1990-07-27
Publication date: 1991-02-21
Also published as: JP3523645B2; JPH10309191A; AU6079890A; EP0484401B1; EP0484401A1; ES2094159T3; ATE142693T1; DK0484401T3; CA2063810C; DE69028535T2; US5194596A; DE69028535D1; CA2063810A1; JP2004000255A; JP2007068545A; EP0484401A4; JPH05501350A

Abstract

Isolated DNA sequences, expression vectors and transformant cells are provided which allow for the large scale production of vascular endothelial cell growth factor. The vascular endothelial cell growth factor is useful in the treatment of wounds in which neovascularization or reendothelialization is required for healing.

Description

PRODUCTION OF VASCULAR ENDOTHELIAL CELL GROWTH FACTOR

Background of the Invention

The invention relates to the field of wound healing. In particular, the invention relates to the production of a wound healing agent that is mitogenic for vascular endothelial cells and consequently iε useful in promoting neovascularization (angiogenesiε) and re- endothelialization of inner vascular surfaces. The invention provides methods and means for producing vascular endothelial cell growth factor by means of recombinant DNA technology. Angiogenesis, i.e. the growth of new capillary blood vessels, is a process which is crucial to the proper healing of many types of wounds. Consequently, factors that are capable of promoting angiogenesis are useful as wound healing agents. Angiogenesis is a multi-step process involving capillary endothelial cell proliferation, migration and tissue penetration. A number of known growth factors, including basic and acidic fibroblast growth factor, transforming growth factor alpha and epidermal growth factor, are broadly mitogenic for a variety of cell types as well as being angiogenic and are, therefore, potentially useful in promoting tissue repair. Broad spectrum itogenicity is desirable in many types of wound healing applications. There are, however, specific types of wound healing applications in which it would be desirable to have a more cell-specific mitogenic activity. For example, following vascular graft surgery, balloon angioplasty or to promote collateral circulation in post-myocardial infarction patients, it would be desirable to employ a wound healing agent incorporating a mitogenic factor having mitogenic activity that is highly specific for vascular endothelial cells since proliferation of other cell types along with endothelial cells could cause blockage and/or reduced blood flow. At present, no highly suitable mitogenic factor is widely available for this type of application.

Recently, a mitogen apparently specific for vascular endothelial cells was isolated from media conditioned by bovine folliculo stellate cells and its partial amino acid sequence determined (Gospodarowicz et al., PNAS (1989) 6_(19):7311-7315; Ferrara and Henzel, BBRC (1989) JL6_1(2) :851-858) . This factor appears to be a homodimer of two approximately 23 kD subunits. A partially homologous factor that is the mouse homolog of the bovine protein described by Gospodarowicz et al. and Ferarra et al . has also been isolated from the conditioned media of murine AtT20 cellε (ATCC CCL 89) and its N-terminal amino acid sequence determined (Plouet et al., EMBO J. (1989) (12) :3801-3806) . Both factors have been demonstrated to have mitogenic activity for vascular endothelial cells and for none of the other cell types tested and are therefore useful in a number of types of wound healing applications. Unfortunately, it is not practical and economical to obtain commercial quantities of these factorε by purification from their native sources.

Summary of the Invention

The present invention provides methods and means for obtaining commercial scale quantities of vascular endothelial cell growth factor for use as a wound healing agent. In particular, the present invention provides DNA sequences that encode amino acid sequences of mammalian vascular endothelial cell growth factor. These DNA ^* sequences are inserted into expression vectors under the control of regulatory elements that direct the expression of the encoded amino acid sequences in a suitable expression host. The vascular endothelial cell growth factor expressed in this manner can be recovered and formulated into pharmaceutical compositions that are useful in a variety of wound healing applications in which angiogenesis and/or re- endothelialization play an important role.

In the course of providing the DNA sequences herein that encode vascular endothelial cell growth factor, we discovered that two different forms of the coding region are produced i^n vivo in the mRNA for the factor. These two forms, which apparently arise through alternative message splicing, differ in the length of the open reading frame due to the presence or absence of a 44-codon insert in the mature protein coding region. The predicted higher molecular weight factor, comprising a 164-amino acid sequence in the bovine case and a 165-amino acid sequence in the human case, is believed to correspond to the approximately 23 kD subunit isolated by Gospodarowicz et al . ( supra) and by Ferrara and Henzel ( supra) . The novel, lower molecular weight factor predicted from the coding region lacking the insert comprises a 120-amino acid εequence in the bovine case and 121-amino acid sequence in the human case. The lower molecular weight form differs not only in the length of the amino acid sequence, but also in the presence of a Lys residue at position 114 in the bovine protein (position 115 in the human protein) that iε not present in the higher molecular weight form because of the differential mesεage εplicing which occurε within the corresponding codon at this position. For convenience, these two forms of vascular endothelial cell growth factor will be referred to, respectively, as bVEGF* 5 and bVEGF-^O for the bovine factor (hVEGF*Lg5 and hVEGF-^i for the human factor).

The 121-amino acid form of hVEGF and the corresponding 120-amino acid bovine protein differ in their properties from hVEGF;j_g5 and bVEGF-^64 in a manner which provides therapeutic advantages in the clinical use of the protein. In particular, hVEGF-|_2i and

do not bind heparin, whereas the longer forms are characterized by strong heparin binding. The absence of heparin binding affinity leaves more of the protein free to bind to vascular endothelial cell growth factor receptor and increases the half-life and distribution of the protein in circulation.

Surprisingly, the amino acid sequence deduced from N-terminal sequence analysis and an isolated DNA sequence (shown in Fig. 3a indicates a significant level of sequence homology between bovine vascular endothelial cell growth factor and corresponding sequences from each of the A-chain and B-chain subunits of human platelet-derived growth factor (PDGF), with complete conservation of eight cysteine residues among the mature forms of all three sequences. Accordingly, hybrid dimeric proteins can be prepared comprising a first polypeptide chain and a second polypeptide chain, wherein one of the chains comprises at least a portion of the amino acid sequence of the A-chain or the B-chain subunit of platelet-derived growth factor and the other chain comprises at least a portion of the amino acid sequence of vascular endothelial cell growth factor. Preparation of the hybrid proteins allows one to "tailor" the properties of the molecule such that the hybrid exhibits a profile of mitogenic activity between that of vascular endothelial cell growth factor and platelet-derived growth factor. The PDGF B-B homodimer is mitogenic for vascular smooth muscle cells but not for vascular endothelial cells. Conversely, the vascular endothelial cell growth factor of the present invention has the opposite specificity. A hybrid factor may stimulate both cell types and therefore be useful as a broader-spectrum mitogen in wound healing therapies.

Brief Description of the Drawings

Fig. 1 is a representation of five DNA sequences generated by a modification of the polymerase chain reaction process. One of the five (pET-19A; clone no. 5) encodes amino acidε no. 15 to 38 of the mature, εequenced form of bovine vascular endothelial cell growth factor. The figure also includes a consensus DNA sequence derived from the five DNA sequences, as well as a translation of pET-19A. Each sequence includes DNA linkers at either end which represent an EcoRI restriction site and a Hindlll restriction site.

Fig. 2 is a schematic representation of the method by which the five DNAs of Fig. 1 were generated and amplified from bovine folliculo stellate cell mRNA.

Fig. 3a is a representation of a DNA sequence, as well as its deduced amino acid sequence, derived from a clone, deεignated 11B'. The illustrated sequence encodes amino acids no. 15 to 120 of bovine vascular endothelial cell growth factor Fig. 3b is a representation of a synthetic DNA sequence, based on preferred codon usage in human cells, which encodes amino acids no. 1 to 19 of bVEGF*L20 ^and which overlaps the 5' end of the DNA sequence of Fig. 3a. This synthetic DNA can be enzymatically joined to the isolated DNA sequence of Fig. 3a, after the DNA sequence in Fig. 3a has been digested with the restriction enzyme Accl, to produce a DNA sequence encoding the full length, mature

protein.

Fig. 4 is a representation of isolated DNA sequences encoding the A-chain and B-chain subunits of human platelet-derived growth factor, and the amino acid sequences of the precursorε of these two proteins as deduced from the DNA sequences.

Fig. 5 is a photograph of an ethidium bromide stained polyacrylamide gel containing DNA produced by amplification of a portion of the mRNA encoding bovine vascular endothelial cell growth factor.

Fig. 6 is a representation of the isolated cDNA sequences encoding bVEGF-^O and bVEGF*_|_54. The boxed DNA sequence beginning at base 342 represents the insert sequence that is present in the alternatively spliced cDNA which encodes bVEGF_]^. The amino acid sequence given immediately below the nucleotide sequence represents the deduced sequence for bVEGF*-^. The deduced amino acid sequence for bVEGF^O is identical to that of bVEGFi64 through position 113 (Glu). The carboxyl-terminal sequence of bVEGFi20' beginning at position 111 (Arg) is given in italics below the bVEGFi sequence in Fig. 6. Fig. 7 is a representation of the native human DNA sequences encoding the mature forms of human vascular endothelial cell growth factor, hVEGF;*_21 ^and hVEGFigs- The DNA sequences shown represent composites of sequences obtained from human genomic and human cDNA clones. The bracketed amino acid (Gly) encoded by codon 7 represents an inserted amino acid relative to the sequences of

and bVEGF*L54,

Fig. 8 is a representation of portions of the DNA sequences of the overlapping genomic inserts in two bacteriophages, _^which together contain eight exons encoding the various forms of hVEGF, along with contiguous splice junctions

Fig. 9 is a representation of two oligonucleotide primers used to amplify a full length cDNA sequence for hVEGF₁₂l ^{froπι u937 cel1 mRNA}-

Fig. 10a is a schematic representation of pLEN, an expression vector used to express hVEGF^i in Chinese hamster ovary cell culture. Fig. 10b is a schematic representation of pMTN, another expreεsion vector used to express hVEGF-^i in Chinese hamster ovary cell culture.

Detailed Description of the Invention

As used herein, the term "vascular endothelial cell growth factor" refers to a mammalian protein that has mitogenic activity for vascular endothelial cells and that:

(a) has an amino acid sequence which either is encoded by a DNA sequence that is capable of hybridizing, under standard hybridization conditions, to the DNA sequence shown in Fig. 3a; or

(b) is substantially homologous to the amino acid sequence of bVEGF]_20' VEGF]^, hVEGF^i ^{or hVEGF}165' ^sh°^wn in Fig. 6 and Fig. 7.

An amirϊo acid sequence is considered to be "substantially homologous" herein if the level of amino acid sequence homology is at leaεt 50% and, preferably at least 80%, compared with the protein in question. "Standard hybridization conditions", as used herein means the use of -1-

40% Formamide Buffer (described below) as the prehybridization/hybridization buffer and washing in 1 x SSC, 0.1% SDS at 50°C.

The amino acid sequence numbering system used herein for vascular endothelial cell growth factor is based on the mature forms of the protein, i.e. the post- translationally procesεed forms. Accordingly, the residue numbered one in the bovine or human proteins is alanine, which is the first reεidue of the iεolated, mature formε of theεe proteins.

Mitogenic activity for vascular endothelial cellε can be determined by an assay which uses, as target cells, adrenal cortex-derived capillary endothelial cells (ACE cells). This assay is carried out essentially as described in Gospodarowicz et al. , J. Cell Physiol. (1986) 127;121- 136), the disclosure of which is incorporated herein by reference. Generally, stock cultures of ACE cells are maintained in the presence of Dulbecco's modified Eagle's medium (DMEM-21) supplemented with 10% calf serum. The antibiotics penicillin (50 IU/ml), streptomycin (50 yg/ml), gentamycin (50 _/yg/ml), and Fungizone (0.25 yg/ml) and 2mM L- glutamine can also be added to the medium. Cells are passaged weekly on tisεue culture diεheε at a εplit ratio of between 1:40 and 1:200 (the preferred εplit ratio is that which gives 2.5 x 10*^ cells in 15 ml of medium in T75 flasks). For the mitogenic assay, cells are seeded in 12 well cluster plates at a density of 5 x IO"-" cells per well in 1 ml Dulbecco's modified Eagle's medium supplemented with 10% calf serum and antibiotics as described in Gospodarowicz et al., Europ. J. Cell. Biol. (1988) 4_6:144-151.

Alternatively, the ACE cells are plated in 35 mm dishes or 6 well cluster plates at a density of 5 - 10 x 10^ cells per dish or well in 2 ml of medium as described in Gospodarowicz, et al. , J. Cell Physiol. (1986) 127:121-136. Ten-microliter aliquots of appropriate dilutions of each sample are then added to duplicate or triplicate wells in the dishes on days 0 and 2. After 4 or 5 days in culture, the plates are trypsinized and cell densities determined in a Coulter counter. For purposes of deεcription herein, a factor iε considered to have mitogenic activity for vascular endothelial cells if the cell denεity at the end of thiε assay is at least 1.5 times and preferably at least 3 times the cell density of control wells receiving no factor additions.

Although the DNA sequence illustrated in Fig. 3a was obtained from a bovine cell cDNA library, and therefore represents a sequence which encodeε a bovine protein, the DNA sequence provided allows for the retrieval of sequences encoding homologous proteins from other mammalian specieε. Accordingly, we have employed the illustrated bovine sequence as a probe to retrieve DNA sequenceε encoding the correεponding human proteinε. Also included within the εcope of "vascular endothelial cell growth factor" herein are biologically active fragments thereof, as well as N-terminally or C- terminally extended versions thereof or analogε thereof substituting and/or deleting or inserting one or more amino acid reεidues which retain qualitatively the biological activities of the protein described herein. Preferred analogε include those in which one or more cysteine residues not required for biological activity are substituted by a different amino acid residue, preferably εerine. Subεtitution. of one or more cyεteine residues reduces the opportunity for intermolecular and intramolecular disulfide bond formation, thereby rendering the molecule more εtable. There are nine cyεteine reεidueε that are preεent in hVEGF]_2i, t» EGF-]_20 VEGF^g5 and bVEGF-]^. Of theεe, eight are conserved with PDGF. Accordingly, the most preferred analog is one in which the ninth cysteine residue is substituted by serine. This cysteine residue is present at position 160 of

and position 116 of hVEGF*L2l and the corresponding positions in the bovine forms. Amino acid substitutions can be accomplished by site specific mutagenesis of the DNA sequences described herein using well known techniques (see, e.g., Zoller, M.J. and Smith, M., Nucleic Acids Research (1982) 10:6487-6500). While the native form of the bovine vascular endothelial cell growth factor described herein is apparently glycosylated, there is currently no evidence that glycosylation iε essential for biological activity. Accordingly, biologically active non-glycosylated or partially glycosylated forms, which will be produced by prokaryotic or eukaryotic hosts using the expression sequences provided herein, are included within the scope of "vascular endothelial cell growth factor". Expression of DNA sequences encoding vascular endothelial cell growth factor in Chinese hamster ovary cell culture under the conditions deεcribed herein resulted in approximately 50% of the expreεεed VEGF being modified by N- linked glycoεylation. There iε a εingle site for N-linked glycoεylation at the Aεn reεidue at poεition 75 of hVEGF^l (correεponding to poεition 75 of hVEGFigs and poεition 74 of bVEGF^O and bVEGF-*^). Furthermore, following expreεεion of hVEGF*L2i ^anc- secretion into cell culture media, we have iεolated dimeric protein εpecies which correspond in molecular weight to dimers of vaεcular endothelial cell growth factor in which both subunits are either glycosylated or unglycosylated and dimers in which one of the subunitε iε glycoεylated and the other iε not glycoεylated.

Vaεcular endothelial cell growth factor—aε isolated by Gospodarowicz et al. ( supra) and by Ferrara and Henzel ( supra)—iε a dimeric protein of approximately 45-46 kD, as determined by SDS polyacrylamide gel electrophoresis. Bovine vascular endothelial cell growth factor was obtained in homogeneous form from cell culture media conditioned by folliculo stellate cellε, by a proceεε which involved the steps of ammonium εulfate precipitation; heparin-Sepharose affinity chromatography; εize-excluεion gel chromatography; cation exchange chromatography; and reverεe phase high performance liquid chromatography. Similar procedures may be employed to purify a corresponding protein from conditioned media of cultured cells from other mammalian species which are known to produce vascular endothelial cell growth factor, for example, murine AtT20 cells. We have also determined, by Northern blot analyses, that human fetal vascular smooth muscle cells are a good source of human vascular endothelial cell growth factor and mRNA encoding the factor. Isolated bovine vascular endothelial cell growth factor obtained as described above was sequenced using the Edman degradation technique on an automated gas-phase protein sequenator. A single major 24-amino acid N-terminal sequence was obtained, indicating that the protein is homodimeric. Following tryptic digestion of the protein and amino acid sequencing of various peptide fragments, it was determined, according to overlapping amino acid sequences, that the bovine protein has the following 41-amino acid N- terminal sequence*:

A P M A E G G Q K P H E V V K F M D V Y Q R S F C R P I E T L V D I F Q E Y P D E

(*Using the standard single letter abbreviation code for amino acids) <

Using the N-terminal amino acid sequence for bovine vascular endothelial cell growth factor described above, a number of unsucceεεful efforts were made to retrieve a full or partial length cDN encoding the protein by probing a folliculo stellate cell cDNA library using degenerate oligonucleotide probe mixtures encoding portions of the amino acid sequence. The DNA segment of Fig. 3a was ultimately retrieved from the cDNA library using a probe generated by amplifying that portion of the nucleotide sequence encoding amino acids 15 to 38 (and two-thirds of the codon for amino acid 39) by a modification of the polymerase chain reaction method. The polymerase chain reaction method for amplifying a desired DNA εequence iε deεcribed in detail in U.S. Patentε No. 4,683,202 and 4,683,195, the diεcloεures of which are incorporated herein by reference. The procedure allows the amplification of a desired nucleotide εequence, even if the bulk of the sequence is not known, provided one is able to provide oligonucleotide primers that will hybridize to either end of the sequence that it is desired to amplify. The polymerase chain reaction procesε has been employed to amplify a desired segment of cDNA using degenerate oligonucleotides as primers (Lee et al., Science (1988) 1288-1291).

The DNA probe used to retrieve the cDNA of Fig. 3a was selected from the five homologous sequences shown in Fig. 1. These sequences were obtained by a procedure which iε illustrated schematically in Fig. 2 and which is described in greater detail in the examples which follow. In accordance with the illustrated procedure, poly(A)⁺ RNA from bovine folliculo stellate cells was precipitated with an anti-sense primer consiεting of a 16-fold degenerate synthetic oligonucleotide mixture based on the amino acid sequence of amino acids no. 35 to 39 of bovine vascular endothelial cell growth factor. The 24-base oligonucleotide primer consisted of 14 bases reflecting the amino acid sequence, with a 10-base EcoRI linker on the 5' end. The oligonucleotide primer sequences in the mixture which hybridized to the poly(A)⁺ RNA served to prime the synthesiε of a DNA strand complementary to a section of the desired mRNA in the presence of deoxynucleotide triphosphateε (dNTPε) and reverεe tranεcriptaεe. A εecond DNA strand, complementary to the first εyntheεized εtrand, waε then prepared by hybridizing the first synthesized strand to a sense-strand primer consisting of an 8-fold degenerate synthetic 24-base oligonucleotide mixture based on the amino acid sequence of amino acids no. 15 to 19 of bovine vascular endothelial cell growth factor. The εecond εtrand oligonucleotide primer contained a 14-baεe region reflecting the amino acid εequence, joined to a 10-baεe Hindlll linker on the 5' end. The oligonucleotide primer εequences in the mixture which hybridized to the first synthesized DNA strand served to prime the syntheεiε of a εecond DNA strand in the presence of dNTPs and DNA polymerase I, Klenow fragment. Since the 10 base linker εequence in the primer could not hybridize to the first strand DNA, second strand synthesiε was carried out at a temperature, i.e. 28°C, at which the remaining 14 nucleotides could be expected to remain hybridized to the first strand DNA. DNA polymerase I, Klenow fragment, was used for the second εtrand εynthesis, since Thermus aquaticus (Tag) DNA polymerase, which is normally used in the polymerase chain reaction, would not be effective to catalyze DNA synthesis at this temperature. Second εtrand synthesis produced a senεe εtrand coding for that portion of bovine vaεcular endothelial cell growth factor extending from amino acid no. 15 to amino acid no. 38 (and including two-thirds of the codon for amino acid 39).

The two synthesized DNA strandε were then separated and the desired εequence waε amplified by a repeated sequence of reactions in which the single stranded DNAs were uεed aε templateε for the syntheεiε of complementary εtrandε in the presence of both the senεe- and anti-εense oligonucleotide primer mixtures and Thermuε aquaticuε (Tag) DNA polymerase. After each synthesis of complementary strandε, the reaction mixture was heated to separate the strandε and the reaction waε repeated.

After 30 cycleε of amplification, the DNA from the polymerase chain reaction mixture was subjected to electrophoresis on a 6% polyacrylamide gel. The DNA in the gel was stained with ethidiu bromide and the band having the appropriate size for the coding εequence for amino acids 15 to 38 (and two-thirds of the codon for amino acid 39) together with the Hindlll and EcoRI linkerε from the priming oligonucleotideε waε cut from the gel. The ethidium bromide εtained gel iε shown in the photograph of Fig. 5, where the dominant band repreεenting the deεired amplified sequence can be clearly visualized. DNA waε electroeluted from the exciεed gel fragment containing the dominant band, digeεted with Hindlll and EcoRI and ligated into Hindlll- and EcoRI- cut Ml3mpl9 and Ml3mpl8 phage vectors. DNA sequence analysis of white plaques isolated after transformation of the ligation mixtures into E^ coli JM103 host cells demonεtrated that a cDNA εequence encoding amino acidε 15 to 38 (and containing two-thirdε of the codon for amino acid 39) of vaεcular endothelial cell growth factor indeed had been obtained. The amplified DNA sequence contained in one of these recombinant phage (pET-19A) waε employed aε a probe to retrieve a cloned cDNA εequence from a bovine folliculo stellate cell cDNA library. The isolated cloned sequence consiεted of an 797-baεe pair insert coding for all but the 14 N-terminal amino acids of one of the mature forms of bovine vascular endothelial cell growth factor. The isolated cloned insert was ligated into the EcoRI site of pUC8 to create the plasmid designated pST800. The insert contained the nucleotide sequence shown in Fig. 3a (in Fig. 3a, the EcoRI linkers on each end of the insert are not shown; hence the sequence is numbered beginning with nucleotide 7 of the insert). The coding region of amino acids no. 15 to 120 of bVEGF-^O is represented by nucleotides no. 9 to 326 of Fig. 3a.

The amino acid εequence predicted from the iεolated DNA εequence εhown in Fig. 3a containε one potential εite for N-linked glycoεylation at the aεparagine reεidue at amino acid no. 74 (correεponding to amino acid no. 75 in the human form) . Since an N-linked glycosylation of approximately 3 kD at this site would predict a total molecular weight of about 17 kD for the encoded protein, which is considerably smaller than the apparent molecular weight of 23 kD observed for the vascular endothelial cell growth factor subunitε iεolated by Goεpodarowicz et al . and by Ferrara and Henzel, it iε apparent that the iεolated cDNA encodeε a different form of vascular endothelial cell growth factor than that previously observed. A polymerase chain reaction experiment indicated that alternative forms of the vascular endothelial cell growth factor coding region exist.

A polymerase chain reaction waε primed from folliculo stellate poly(A)⁺ RNA using a senεe oligonucleotide corresponding to bases 70-126 in Fig. 6, and an antisense oligonucleotide corresponding to baseε 513-572. Polyacrylamide gel analysis of the products after digestion with BstNI (which cutε within each of the pri erε) revealed two major species of approximately 300 and 450 bp. Both of theεe products were εubcloned into M13 vectorε and sequenced. The DNA εequence of the smaller product (311 bp) corresponded to that predicted if the PCR amplified the cDNA sequence carried in pST800. The sequence of the larger 5 fragment waε identical to that of the 311 bp product, except for an inεert of 132 bp (boxed sequence in Fig. 6). Analysis of human vascular endothelial cell growth factor genomic clones, obtained as described below, has indicated that this inεert occurs at an exon-intron junction,

10 εuggeεting that the two forms of the coding region arise through alternative exon splicing.

The DNA sequenceε εhown in Fig. 1 or Fig. 3a are useful in the retrieval of DNA coding for full length bovine vascular endothelial cell growth factor, or for the

15 corresponding vascular endothelial cell growth factor of other species, including the corresponding human protein, or for related proteins in the same gene family as vascular endothelial cell growth factor.

To obtain bovine vascular endothelial cell growth

20 factor cDNA clones containing sequence information upstream from that present in pSTδOO (Fig. 3a), we employed a modification of the "RACE" polymerase chain reaction technique described by Froh an, M.A. et al., PNAS (USA) (1988) 8_5;8998-9002. A linker was ligated onto the 5' end

25 of the duplex resulting from primer extension of the vascular endothelial cell growth factor mRNA, after which polymerase chain reaction was carried out uεing aε primerε the original primer-extension oligonucleotide and an oligonucleotide complementary to the linker. Sequence

30 analysiε of the reεulting polymeraεe chain reaction products, after digestion of the primers with Hindlll and subcloning into M13 vectors, gave sequenceε encoding the mature amino terminuε of vaεcular endothelial cell growth factor (Fig. 6). The longeεt cDNA clone obtained extended

35 14 bp 5' to the beginning of the mature protein coding region (AGTGGTCCCAGGCTGCACCC... ) , revealing four additional amino acids of the vaεcular endothelial cell growth factor precursor (WSQAAPMA... ) . In order to retrieve DNA εequenceε for the human forms of vascular endothelial cell growth factor, the sequenceε in Fig. 1, Fig. 3a, or Fig. 6, preferably the εequenceε of Fig. 3a or Fig. 6 or εegments thereof, are used as probes to retrieve the desired sequences from cDNA or genomic libraries. Genomic libraries can be prepared by known techniques and are now widely commercially available. A suitable genomic DNA library from which genomic DNA sequences encoding human vascular endothelial cell growth factor can be isolated is a human fibroblast genomic library (Stratagene Inc., La Jolla, CA) . This library, obtained from the W138 cell line, harbors >15 kb DNA inεertε in the Lambda FIX™ vector. Alternatively, a genomic library can be prepared by the technique diεcloεed by Frischauf, A.M. in Methods in Enzymology, eds. Berger, S.L. and Kimmel, A.R., Vol. 152, pp. 190-199 (1987) Academic Preεε, N.Y. Methodε for preparing cDNA librarieε are alεo well known to those skilled in the art (see, e.g., Kimmel, A.R. and Berger, S.L., ibid. , pp. 307-316). Preferably, the cDNA library is prepared from a cell line or tissue source which actively produces vascular endothelial cell growth factor. For the iεolation of a cDNA εequence encoding the human protein, it iε preferred to employ a cDNA library prepared from fetal human vaεcular εmooth muεcle cellε. A DNA sequence encoding vascular endothelial cell growth factor, which is obtained as described above, is inserted into a suitable expression vector under the control of regulatory sequenceε capable of directing expreεεion of the DNA εequence in a deεired hoεt. If the DNA εequence retrieved iε a genomic εequence containing intronε, then it is desirable to insert the sequence into an expression vector that is compatible with a eukaryotic host. Expresεion of the genomic DNA encoding vascular endothelial cell growth factor in a eukaryotic host is accompanied by correct splicing of the encoded RNA to remove intron εequenceε, thereby producing an mRNA template encoding the deεired protein. Alternatively, a εynthetic DNA εequence can be constructed from εynthetic oligonucleotideε that repreεents the coding sequence obtained after the intron sequenceε in the genomic clone have been removed. Expression vectors containing this synthetic sequence or the cDNA sequence encoding vascular endothelial cell growth factor can be uεed to expreεs the protein in prokaryotic or eukaryotic hosts. Exemplary control sequence DNAs and hosts are described below under Standard Procedures.

Biologically active vascular endothelial cell growth factor iε produced in accordance with the teachings of this invention, as a homodimeric molecule. In this context, the term "homodimeric" refers to a dimer in which the two subunits have the same primary amino acid structure. As previouεly indicated, one or both of the εubunitε may be modified by N-linked glycoεylation or neither of the εubunits modified. A fully active protein iε produced by expresεion and/or recovery of the polypeptide sequence encoded by the DNA εequence of the invention under conditions which allow the formation of disulfide bonds in order to form a dimer. The present invention also provides for the production of chimeric, dimeric proteins in which a portion of the primary amino acid structure corresponds to a portion of either the A-chain subunit or the B-chain εubunit of platelet-derived growth factor and a portion of the primary amino acid structure corresponds to a portion of vascular endothelial cell growth factor. In particular, there iε provided a chimeric growth factor comprising a first polypeptide chain and a second polypeptide chain, said chains being disulfide linked, wherein the first polypeptide chain comprises at least a portion of the amino acid sequence of either the A-chain subunit or the B-chain subunit of platelet-derived growth factor and the second chain comprises at least a portion of the amino acid sequence of vascular endothelial cell growth factor. Platelet-derived growth factor is an approximately 30 kD dimer which has been isolated in both homodimeric and heterodimeric forms. Platelet-derived growth factor heterodimer contains two polypeptide chains, designated the A- and B-chainε. The mature A- and B-chains exhibit approximately 40% amino acid sequence ho ology, with complete conservation of eight cysteine residues. Platelet- derived growth factor exiεtε i_ji vivo as either an A-A or a B-B homodimer or as an A-B heterodimer.

DNA sequences encoding the A-chain and B-chain subunitε of platelet-derived growth factor have been iεolated and εequenced (A-chain cDNA (human) is disclosed in Betsholtz, C, et al. , Nature (1986) 320; A-chain genomic DNA (human) in Bonthron, D.T., et al., PNAS (1988) 85:1496; B-chain cDNA (human) in Collins, T., et al . , Nature (1985) 316 :748; and B-chain genomic DNA (human) in Chin, I.-M., et al., Cell (1984) _37:123). Fig. 4 illustrateε the iεolated cDNA sequences and deduced precursor amino acid sequences for the A- and B-chain subunits of human platelet- derived growth factor with a BamHI linker joined to the 5' end and an EcoRV linker joined to the 3' end of the cDNAs in each caεe. Each of theεe εequenceε can be inεerted into a suitable expression vector under the control of appropriate regulatory elements and expressed in a suitable host, such as for example, E . coli or a eukaryotic host such as Chinese hamster ovary (CHO) cellε or yeast.

Examples of the chimeric growth factor proteins contemplated by the present invention include: a dimeric protein conεisting of the full length A-chain εubunit of platelet-derived growth factor linked by diεulfide bondε to a full length vaεcular endothelial cell growth factor polypeptide chain; and a dimeric protein consiεting of the full length B-chain εubunit of platelet-derived growth factor linked by diεulfide bondε to a full length vaεcular endothelial cell growth factor polypeptide chain. In other, less preferred embodiments, the dimeric protein can conεist of two disulfide-linked polypeptide chains in which one or both of the chains conεists of an N-terminal segment having an amino acid sequence corresponding to an N-terminal portion of either the A- or B-chain subunit of platelet- derived growth factor or vaεcular endothelial cell growth factor and a C-terminal εegment correεponding to a C- terminal sequence selected from one of the other two chains. For example, one can prepare a dimer in which one polypeptide chain consistε of the N-terminal one-half of the A-chain of platelet-derived growth factor linked through a peptide bond to the C-terminal one-half of vascular endothelial cell growth factor; and the other polypeptide chain conεists of the entire amino acid εequence of vascular endothelial cell growth factor. Conversely, one of the polypeptide chains can be composed of an N-terminal portion of vascular endothelial cell growth factor linked to a C- terminal εegment of the A- or B-chain εubunit of platelet- derived growth factor and the other polypeptide chain can have the amino acid εequence of vaεcular endothelial cell growth factor. Numerous different hybrid combinations can be prepared, as will be readily apparent.

In order to prepare the chimeric growth factors of the invention, a DNA sequence encoding each deεired chain iε inserted into a suitable expression vector, e.g. a plasmid, under the control of regulatory sequences capable of directing its expresεion in a host cell. Host cells are then transformed ^"with the expression vectors. If desired, a single host may be cotransformed with expression vectors for each of the two chains. Alternatively, separate host cells can be transformed with the vectors encoding the two polypeptide chainε. The polypeptide chainε are then expreεsed and recovered in a conventional manner. If the polypeptide chains are expresεed with secretion signal sequences such that they are secreted from host cellε, they may naturally form the correct dimer structure during synthesis and secretion. The dimers may then be purified using the techniques described in Gospodarowicz et al., PNAS (1989) 86(19) :7311-7316 and Ferrara and Henzel, BBRC (1989) 161(2) :851-858. If the correct dimer structure is not obtained by this route, or if the two chainε of the chimera are synthesized in different hosts, then an example of one means of refolding and dimerizing the chains would be to treat the partially-purified or purified chainε with guanidine-HCl, a2Sθ3 and a2S4θg, as described in more detail in the exampleε below. The reεulting S-εulfonated, denatured proteinε are then refolded and dimerized together in the preεence of 5 mM glutathione, 0.5 mM glutathione disulfide and urea before final purification.

Compositions and Uses

Vascular endothelial cell growth factor (bVEGF-^O' bVEGF;L64' hVEGF-L21 ^or hVEGF-j^) provided by the invention is useful as a wound healing agent, particularly in applications where it is desired to re-endothelialize vascular tissue, or where the growth of a new capillary bed (angiogenesiε) iε important.

Vaεcular endothelial cell growth factor can, therefore, be used in the treatment of full-thickness wounds such as dermal ulcerε, including the categorieε of preεεure sores, venous ulcers and diabetic ulcers. In addition, vascular endothelial cell growth factor can be used in the treatment of full-thicknesε burns and injuries where angiogenesis is required to prepare the burn or injured site for a skin graft or flap. In this case, the vascular endothelial cell growth factor is either applied directly to the site or it is used to soak the skin or flap that is being transplanted prior to grafting. In a similar fashion, vaεcular endothelial cell growth factor can be uεed in plastic surgery when reconstruction is required following a burn, other trauma or for cosmetic purposeε.

Angiogeneεis is also important in keeping woundε clean and non-infected. Vaεcular endothelial cell growth factor can, therefore, be used in association with general surgery and following the repair of cuts and lacerations. It is particularly useful in the treatment of abdominal wounds with a high risk of infection. Neovascularization iε alεo key to fracture repair εince blood veεεelε develop at the site of bone injury. Administration of vascular endothelial cell growth factor to the site of a fracture iε, therefore, another utility.

In cases where vascular endothelial cell growth factor iε being uεed for topical wound healing, aε deεcribed above, it may be administered by any of the routes deεcribed below for the re-endothelialization of vascular tissue, or more preferably by topical means. In these cases, it will be adminiεtered as either a solution, spray, gel, cream, 5 ointment or as a dry powder directly to the site of injury. Slow release devices directing vascular endothelial cell growth factor to the injured site will also be used. In topical applications, vascular endothelial cell growth factor will be applied at a concentration ranging from 50 to

10 1,000 yg/ml either in a single application, or in dosing regimens that are daily or every few days for a period of one to several weeks. Generally, the amount of topical formulation administered iε that which iε εufficient to apply from about 0.1 to 100

of vaεcular endothelial

15 cell growth factor, based on the surface area of the wound. Vaεcular endothelial cell growth factor can be used as a post-operative wound healing agent in balloon angioplasty, a procedure in which vascular endothelial cells are removed or damaged, together with compresεion of

20 atheroεclerotic plaqueε. Vaεcular endothelial cell growth factor can be applied to inner vaεcular εurfaceε by systemic or local intravenous application either as intravenous bolus injection or infusions. If desired, the vascular endothelial cell growth factor can be administered over time

25 using a micrometering pump. Suitable compositionε for intravenouε administration comprise vascular endothelial cell growth factor in an amount effective to promote endothelial cell growth and a parenteral carrier material. The vascular endothelial cell growth factor can be present

30.in the compoεition over a wide range of concentration, for example, from about 50 _/yg/ml to about 1,000 /g/ml using injections of 3 to 10 ml per patient, administered once or in dosing regimens that allow for multiple applicationε. Any of the known parenteral carrier vehicles can be used,

35 such as normal saline or 5-10% dextroεe.

Vascular endothelial cell growth factor can also be used to promote endothelialization in vascular graft εurgery. In the case of vascular grafts using either transplanted vesselε or synthetic material, for example, vascular endothelial cell growth factor can be applied to the surfaces of the graft and/or at the junctions of the graft and the existing vasculature in order to promote the growth of vaεcular endothelial cellε. For εuch applications, the vascular endothelial cell growth factor can be applied intravenously aε deεcribed above for balloon angioplasty or it can be applied directly to the surfaces of the graft and/or the existing vaεculature either before or during surgery. In such cases, it may be desired to apply the vascular endothelial cell growth factor in a thickened carrier material so that it will adhere to the affected surface. Suitable carrier materials include, for example, 1-5% carbopol. The vascular endothelial cell growth factor can be present in the carrier over a wide range of concentrations, for example, from about 50 _/yg/mg to about 1,000 _/yg/mg. Alternatively, the vascular endothelial cell growth factor can be delivered to the site by a micrometering pump as a parenteral solution. Vascular endothelial cell growth factor can alεo be employed to repair vascular damage following myocardial infarction and to circumvent the need for coronary bypass surgery by stimulating the growth of a collateral circulation. The vascular endothelial cell growth factor is administered intravenously for this purpose, either in individual injections or by micrometering pump over a period of time as described above or by direct infusion or injection to the site of damaged cardiac muscle.

Vascular endothelial cell growth factor can also be used aε a growth factor for the _in vitro culturing of endothelial cellε. For εuch uεes, vascular endothelial cell growth factor can be added to the cell culture medium at a concentration from about 10 pg/ml to about 10 ng/ml.

The hybrid growth factor moleculeε of the invention will be expected to exhibit mitogenic profiles falling between those of platelet-derived growth factor and vascular endothelial cell growth factor or, in some cases, may be employed as inhibitors of angiogenesiε. The moεt pronounced distinction between the activities of the two factors is that platelet-derived growth factor exhibits subεtantial mitogenic activity on εmooth uεcle cells and fibroblasts, but not on endothelial cells, whereas vascular endothelial cell growth factor exhibits the opposite specificity. The mitogenic activity of PDGF A- and/or B-chain on smooth muscleε cells and fibroblasts tends to impart tensile strength to a healing wound. Therefore, the growth factor which is a hybrid between platelet-derived growth factor and vascular endothelial cell growth factor can be applied to a wound in order to induce neovascularization and impart tensile strength to the wound area during and after healing. The hybrid growth factors are applied in essentially the same manner and at the same dosageε aε described above for vascula-r endothelial cell growth factor.

Elucidation of the DNA εequences encoding the various forms of vascular endothelial cell growth factor and their deduced amino acid sequences also provides the meanε for producing inhibitors of vascular endothelial cell growth factor activity. Inhibition of the angiogenic activity of vascular endothelial cell growth factor is uεeful, for example, in retarding or preventing the growth of tumors, since neovascularization is required to provide the necessary blood supply to a growing tumor. One may administer antibodies to vascular endothelial cell growth factor or one may administer fragments of vascular endothelial cell growth factor which are capable of binding vascular endothelial cell growth factor receptor but which do not exhibit the angiogenic activity of full-length vascular endothelial cell growth factor.

Antibodies to the expresεed and iεolated vaεcular endothelial cell growth factor proteinε can be produced by known techniqueε. The therapeutic antibodies may be polyclonal or monoclonal. Antibodies are prepared using standard immunization protocols in rabbits, mice or other suitable animal and recovering the antisera. in addition, antibody-secreting cellε from the immunized ani alε can be immortalized using fusion techniques to produce hybridomas which can be screened for antibodies immunoreactive with the vaεcular endothelial cell growth factor (εee e.g. "Antibodieε: A Laboratory Manual", E. Harlo and D. Lane, Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. ) . The determination of an appropriate treatment regimen (i.e., doεage, frequency of administration, εyεtemic vε. local, etc.) is within the skill of the art. For administration, the antibodies will be formulated in a unit dosage injectable form (solution, suεpenεion, emulεion, etc.) in association with a pharmaceutically acceptable parenteral vehicle. Such vehicles are usually nontoxic and nontherapeutic. Examples of such vehicles are water, saline, Ringer's solution, dextrose solution, and Hank's solution. Nonaqueous vehicleε εuch aε fixed oils and ethyl oleate may also be uεed. A preferred vehicle iε 5% (w/w) human albumin in εaline. The vehicle may contain minor amounts of additives, such as substances that enhance isotonicity and chemical stability, e.g., buffers and preservativeε. The antibody iε typically formulated in such vehicles at a concentration of about 20 yg/ml to 20 mg/ml. There iε alεo provided herein a method to inhibit angiogeneεiε, e.g. to retard or prevent the growth of a tumor, which involveε adminiεtration of a heterodimeric protein having two different εubunitε, each of the subunits being selected from the mature amino acid εequence of hVEGFi2l hVEGF-L55 and hVPFi89- The term hVPF^βg referε to a 189-amino acid protein which haε now been found to arise from differential meεsage splicing of the transcribed genomic DNA sequence shown in Fig. 8. The homodimeric form of hVPF-|_89 is referred to aε vaεcular permeability factor. The amino acid sequence of hVPF iε homologouε with both hVEGF]_2i and hVEGF-[65 ^{for the} N-terminal 114 amino acids of the mature protein and the C-terminal 6 amino acids. However, hVPF-j_89 contains a εequence of 68 amino acidε, beginning at position 115, which are not present in hVEGF*L2i. The complete cDNA coding sequence and deduced amino acid sequence of hVPF^gg are discloεed in Keck, P. et al., Science (1989) 246:1309-1312, the diεclosures of which are incorporated herein by reference. While not wishing to be bound by any particular theory or mechanism of action, it is believed that dimeric PDGF-like growth factors exert their biological activity by a mechanism wherein each of the subunits of the dimer binds a separate receptor subunit on the cell surface, with binding to the two adjacent receptor subunits being necessary to trigger receptor activity. By administering a heterodimer composed of different subunits of the vascular endothelial cell growth factor family, each of the subunits may bind a εpecific receptor εubtype, thereby blocking the bound receptor from interacting effectively with endogenouε, homodimeric vaεcular endothelial cell growth factor. However, εince the other subunit of the adminiεtered homodimer is a different vascular endothelial cell growth factor subtype, the dimer will be incapable of binding a second receptor subunit of the same subtype to trigger the biological activity of the receptor.

The heterodimeric protein which is employed in this manner to inhibit angiogenesiε can be produced by expreεεing the deεired amino acid sequences for hVEGF-^l' hVEGFτ_g5 or hVPF]_89 in separate hostε or by co-expreεεion in the same host. Dimerization is then carried out by procedureε deεcribed elεewhere herein. The desired heterodimer can be separated from homodimeric forms by any convenient method for size separation of proteins. The heterodimer is administered to a host in need of anti-angiogenic treatment, e.g. an individual suffering from a tumor, in a manner εimilar to that in which vaεcular endothelial cell growth factor is administered aε an angiogenic agent. The precise dosing regimen and dosage is within the skill of those in the art to determine.

An alternative means of inhibiting angiogenesiε is the administration of a heterodimer in which one subunit is selected from hVEGF*^*_2i, hVEGF^g5 and hVPF^89' ^anα- ^tne other subunit is a biologically inactive fragment or an analog of hVEGF-j_2i, hVEGF*L65 or hVPF-j^g in which one or more amino acids are substituted by different amino acids which render the subunits inactive.

Since vascular endothelial cell growth factor is produced at elevated levels in tumors, one can employ anti- vascular endothelial cell growth factor antibodies in a conventional immunoaεεay to detect the preεence of a tumor in an individual. Antibodies to vascular endothelial cell growth factor, preferably to human vascular endothelial cell growth factor, e.g. hVEGFi21 ^or hVEGFι 5, can be produced by known techniqueε. The antibodies can be monoclonal or polyclonal. Vaεcular endothelial cell growth factor levelε are determined in a fluid sample, preferably a serum sample, of an individual suspected of having a tumor, using any of the conventional immunoasεay techniqueε which employ antibodies to the subεtance being meaεured. Preferred aεsays are the sandwich type immunoasεayε such as a sandwich type enzymeimmunoassay or radioimmunoassay. Circulating levels of vascular endothelial cell growth factor can reliably be measured by these techniques to the 1-1000 pg/ml level. The meaεured level of vascular endothelial cell growth factor is compared with control levels taken from normal individuals, i.e. individuals who do not have tumors. Elevated levels of circulating vascular endothelial cell growth factor are considered diagnoεtic for tumorε.

Standard Procedures

Most of the procedures which are used to transform cells, conεtruct vectorε, extract messenger RNA, prepare cDNA libraries, and the like are widely practiced in the art and most practitioners are familiar with the εtandard resource materials which describe specific conditions and procedures. However, for convenience, the following paragraphs may serve as a guideline.

Hosts and Control Sequences

Both prokaryotic and eukaryotic systems may be used to express the vascular endothelial cell growth factor encoding sequences: prokaryotic hosts are, of course, the moεt convenient for cloning procedures. Prokaryotes most frequently are represented by various strainε of E^ coli; however, other microbial εtrainε may alεo be uεed. Plasmid vectorε which contain replication sites, selectable markers and control sequences derived from a species compatible with the microbial host are used; for example, E_^ coli is typically transformed uεing derivativeε of pBR322, a plaεmid conεtructed from partε of three naturally-occurring plasmids, two obtained from species of Salmonella, and one isolated from E^ coli by Bolivar, et al., Gene (1977) 2_:95. pBR322 contains geneε for ampicillin and tetracycline reεiεtance, and thuε provideε multiple εelectable markers which can be either retained or destroyed in constructing the desired vector. Commonly used prokaryotic control sequences (also referred to herein as "regulatory elements") which are defined herein to include promoters for transcription initiation, optionally with an operator, along with ribosome binding site εequences, include such commonly used promoters as the beta-lactamase (penicillinase) and lactose (lac) promoter εystems (Chang, et al., Nature (1977) 198:1056 ) and the tryptophan (trp) promoter system (Goeddel, et al., Nucleic Acids Res. (1980) 8_:4057 and the lambda derived P_L promoter and N-gene ribosome binding site (Shimatake, et al., Nature (1981) 292:128) . In addition to bacteria, eucaryotic microbes, such aε yeaεt may alεo be used as hostε. Laboratory strains of Saccharomyces cerevisiae, Baker'ε yeaεt, are moεt uεed although a number of other εtainε or εpecieε are commonly available. Vectorε employing, for example, the 2 μ origin of replication of Broach, J.R., Meth. Enz. (1983) 101:307 , or other yeast compatible origins of replication (see, for example, Stinchcomb, et al., Nature (1979) 282:39, Tschumper, G., et al., Gene (1980) 1^0:157 and Clarke, L. et al., Meth. Enz. (1983) 101:300) may be used. Control εequenceε for yeast vectors include promoters for the synthesis of glycolytic enzymes (Hess, et al., J. Adv. Enzyme Reg. (1968) 1_: 149 ; Holland, et al., Biochemistry (1978) 17:4900). Additional promoters known in the art include the promoter for 3-phoεphoglycerate kinase (Hitzeman, et al. , J. Biol. Chem. (1980) 255:2073) . Other promoterε, which have the additional advantage of transcription controlled by growth conditions and/or genetic background are the promoter regions for alcohol dehydrogenase 2, isocytochrome C, acid phosphatase, degradative enzymes asεociated with nitrogen metaboliεm, the alpha factor system and enzymes responεible for maltose and galactose utilization. It is also believed terminator sequences are desirable at the 3' end of the coding sequences. Such terminators are found in the 3' untranεlated region following the coding εequenceε in yeast- derived genes.

It is also, of course, possible to express genes encoding polypeptides in eukaryotic host cell cultures derived from multicellular organismε. See, for example, Axel, et al. 4,399,216. These systems have the additional advantage of the ability to splice out introns and thus can be used directly to express genomic fragments. These syεtemε can also provide post-translational modification mimicing those occurring in some natural proteins. Useful host cell lines include VERO and HeLa cellε, and Chinese hamster ovary (CHO) cells. Expresεion vectors for such cells ordinarily include promoterε and control εequenceε compatible with mammalian cellε εuch aε, for example, the commonly uεed early and late promoterε from Simian Viruε 40 (SV40) (Fierε, et al., Nature (1978) 2_7_3:113), or other viral promoters such as those derived from polyoma, Adenovirus 2, bovine papilloma virus, or avian sarcoma viruses. The controllable promoter, hMT-II_A (Karin, M., et al. Nature (1982) 299:797-802) may also be uεed. General aspects of mammalian cell host system transformations have been described by Axel (supra). It is apparent that "enhancer" regions are also important in optimizing expression; these are, generally, sequenceε found upstream or downstream of the promoter region in non-coding DNA regions. Origins of replication may be obtained, if needed, from viral sources. However, integration into the chromoεome is a common mechanism for DNA replication in eucaryotes.

Transformations Depending on the host cell used, transformation is done using standard techniques appropriate to such cells. The calcium treatment employing calcium chloride, as deεcribed by Cohen, S.N. PNAS (1972) 9_:2110, or the RbCl₂ method described in Maniatis, et al. , Molecular Cloning: A Laboratory Manual (1982) Cold Spring Harbor Press, p. 254 and Hanahan, D. , J. Mol. Biol. (1983) 166:557-580 may be used for procaryotes or other cells which contain substantial cell wall barriers. For mammalian cellε without such cell walls, the calcium phosphate precipitation method of Graham and van der Eb, Virology (1978) £2:546, optionally as modified by Wigler, M., et al. , Cell (1979) 16:777-785 may be used. Tranεformationε into yeaεt may be carried out according to the method of Beggs, J.D., Nature (1978) 275:104-109 or of Hinnen, A., et al. , PNAS (1978) 75:1929.

Vector Construction

Construction of suitable vectorε containing the deεired coding and regulatory elementε for expreεεion of the DNA sequenceε provided herein employε εtandard ligation and restriction techniques which are well understood in the art. Isolated plasmids, DNA sequences, or synthesized oligonucleotides are cleaved, tailored, and religated in the form desired.

The DNA sequences which form the vectors are available from a number of sources. Backbone vectors and control systemε are generally found on available "hoεt" vectorε which are uεed for the bulk of the εequenceε in the constructions. Typical sequences have been set forth in Hosts and Control Sequenceε above. For the pertinent coding εequence, initial construction may be, and usually is, a matter of retrieving the appropriate sequences from cDNA or genomic DNA libraries. However, once the sequence is disclosed it is possible to synthesize the entire gene εequence mi vitro εtarting from the individual nucleotide derivativeε. The entire gene sequence for genes of sizeable length, e.g., 500-1000 bp may be prepared by synthesizing individual overlapping complementary oligonucleotides and filling in εingle stranded nonoverlapping portions using DNA polymerase in the presence of the deoxyribonucleotide triphosphates. Thiε approach haε been used succeεsfully in the construction of several genes of known sequence. See for example, Edge, M.D., Nature (1981) 292:756; Nambair, K.P., et al., Science (1984) 223:1299; Jay, Ernest, J. Biol. Chem. (1984) 259:6311.

Synthetic oligonucleotides are prepared by either the phosphotriester method as described by Edge, et al. , Nature (εupra) and Duckworth, et al., Nucleic Acids Res. (1981) j):1691 or the phoεphoramidite method aε deεcribed by Beaucage, S.L., and Caruthers, M.H. Tet. Letts. (1981) 2_2:1859 and Matteucci , M.D., and Caruthers, M.H., J. Am. Chem. Soc. (1981) 103:3185 and can be prepared using commercially available automated oligonucleotide synthesizerε. Kinaεing of εingle εtrandε prior to annealing or for labeling iε achieved using an excesε, e.g., approximately 10 unitε of polynucleotide kinaεe to 1 nmole substrate in the presence of 50 mM Tris, pH 7.6, 10 mM MgCl2, 5 mM dithiothreitol, 1-2 mM ATP, 1.7 pmoles γ-³²P-ATP (2.9 mCi/mmole), 0.1 mM spermidine, 0.1 M EDTA.

Once the components of the deεired vectorε are thuε available, they can be exciεed and ligated uεing standard restriction and ligation procedures.

Site specific DNA cleavage is performed by treating with the suitable reεtriction enzyme (or enzymeε) under conditionε which are generally underεtood in the art, and the particulars of which are specified by the manufacturer of these commercially available restriction enzymes. See, e.g., New England Biolabs, Product Catalog. In general, about 1 /yg of plasmid or DNA sequence is cleaved by one unit of enzyme in about 20 μl of buffer solution: in the examples herein, typically, an excesε of reεtriction enzyme is used to insure complete digestion of the DNA substrate. Incubation times of about one hour to two hours at about 37°C are workable, although variations can be tolerated. After each incubation, protein is removed by extraction with phenol/chloroform, and may be followed by ether extraction, 5 and the nucleic acid recovered from aqueous fractionε by precipitation with ethanol. If deεired, εize separation of the cleaved fragments may be performed by polyacrylamide gel or agarose gel electrophoresis using standard techniques. A general description of size separationε iε found in Methods

10 in Enzymology (1980) 6_5:499-560.

Restriction cleaved fragments may be blunt ended by treating with the large fragment of E^ coli DNA polymerase I (Klenow) in the presence of the four deoxynucleotide triphosphates (dNTPs) using incubation timeε of about 15 to

15 25 minutes at 20 to 25°C in 50 mM Tris pH 7.6, 50 mM NaCl, 6 mM MgCl2, 6 mM DTT and 0.1-1.0 mM dNTPε. The Klenow fragment fillε in at 5' εingle-stranded overhangs but chews back protruding 3' εingle strands, even though the four dNTPs are present. If desired, selective repair can be

20 performed by supplying only one of the, or selected, dNTPs within the limitations dictated by the nature of the overhang. After treatment with Klenow, the mixture is extracted with phenol/chloroform and ethanol precipitated. Treatment under appropriate conditionε with SI nuclease or

25 BAL-31 resultε in hydrolysiε of any εingle-εtranded portion. Ligations are performed in 15-50 μl volumes under the following εtandard conditionε and temperatures: for example, 60 mM Triε-Cl pH 7.5, 16 mM MgCl₂, 10 mM DTT, 33 _/yg/ml BSA, and either 40 yM ATP, 0.01-0.02 (Weiss) units

30 T4 DNA ligaεe at 0°C (for "εticky end" ligation) or 1 mM

ATP, 0.3-0.6 (Weiss) units T4 DNA ligaεe at 14°C (for "blunt end" ligation). Intermolecular "εticky end" ligationε are usually performed at 33-100 /yg/ml total DNA concentrationε (5-100 nM total end concentration). Intermolecular blunt

35 end ligations are performed at 1 /yM total ends concentration.

In vector construction employing "vector fragments", the vector fragment is commonly treated with bacterial alkaline phoεphatase (BAP) or calf intestinal alkaline phosphatase (CIP) in order to remove the 5' phosphate and prevent self-ligation of the vector. Digestions are conducted at pH 8 in approximately 10 mM Tris-HCl, 1 mM EDTA using about 1 unit of BAP per _/yg of vector at 60°C for about one hour, or in 50 mM Tris-HCl (pH 9.0), 1 mM MgCl2, 0.1 M ZnCl2, 1 mM Spermidine, 1 unit CIP at 37°C for 60 minuteε (for protruding 5' ends) or 15 minutes at 37°C and then 15 minutes at 56°C (for blunt ends or recesεed 5' endε). In order to recover the nucleic acid fragmentε, the preparation iε extracted with phenol/chloroform and ethanol precipitated. Alternatively, religation can be prevented in vectorε which have been double digested by additional restriction enzyme digestion and separation of the unwanted fragments.

For portions of vectors derived from cDNA or genomic DNA which require sequence modificationε, εite εpecific primer directed mutagenesis may be used (Zoller, M.J., and Smith, M. Nucleic Acids Res. (1982) 10:6487-6500 and Adelman, J.P., et al., DNA (1983) 2_:183-193). This is conducted using a primer synthetic oligonucleotide, complementary to a single stranded phage DNA to be mutagenized, except for limited miεmatching which represents the desired mutation. Briefly, the synthetic oligonucleotide iε used as a primer to direct εyntheεiε of a εtrand complementary to the phage, and the reεulting partially or fully double-εtranded DNA iε tranεformed into a phage-εupporting hoεt bacterium. Cultureε of the tranεformed bacteria are plated in top agar, permitting plaque formation from single cells which harbor the phage. Theoretically, 50% of the new plaques will contain the phage having, aε a εingle strand, the mutated form; 50% will have the original sequence. Plaque liftε of the reεulting plaqueε onto nitrocelluloεe are waεhed after hybridization with kinased synthetic primer at a wash temperature which permits binding of an exact match, but at which the mismatches with the original strand are sufficient to prevent binding. Plaqueε which hybridize with the probe are then picked, cultured, and the DNA recovered.

Verification of Conεtruction In the constructions set forth below, correct ligations for plasmid construction are confirmed by first transforming E^ coli strain MC1061 obtained from Dr. M. Casadaban (Caεadaban, M., et al., J. Mol. Biol. (1980) 138:179-207) or other εuitable host with the ligation mixture. Successful transformants are selected by ampicillin, tetracycline or other antibiotic resiεtance or using other markers depending on the mode of plasmid construction, as is understood in the art. Plas idε from the transformants are then prepared according to the method of Clewell, D.B., et al., PNAS (1969) 6_2:1159, optionally following chloramphenicol amplification (Clewell, D.B. J. Bacterial. (1972) 110:667) . Several mini DNA preps are commonly used, e.g., Holmes, D.S. et al. , Anal. Biochem. (1981) 114:193-197 and Birnboim, H.C., et al. , Nucleic Acids Res. (1979) ]_-.1513-1523. The isolated DNA is analyzed by restriction and/or sequenced by the dideoxy nucleotide method of Sanger, F. et al., PNAS (1977) 7_4:5463 as further described by Mesεing, et al., Nucleic Acidε Reε. (1981) :309, or by the method of Maxa , et al., Methodε in Enzymology (1980) 6J5:499.

Hoεtε Exemplified

Hoεt εtrainε uεed in cloning and prokaryotic expression herein are as followε: For cloning and εequencing, and for expression of constructions under the control of most bacterial promoters, E. coli strainε εuch as B, MC1061, DHl, RRl, C600hfl^_, K803, HB101, JA221, JM101, and JM103 were used.

Illuεtrative Procedures

The following examples are intended to illustrate the invention as a means of better understanding it. The examples are not, however, intended to limit the scope of the invention in any way. The DNA encoding vaεcular endothelial cell growth factor is obtained initially by firεt obtaining a pivotal probe by meanε of amplification of the deεired εequence in a preparation of folliculo εtellate poly(A)⁺ RNA. However, it would not be neceεεary to repeat the procedure, εince the εequence of the pivotal probe iε now known and could thuε be conεtructed chemically jin vitro. In addition, a plasmid containing the sequence illustrated in Fig. 3a as been deposited at the American Type Culture Collection, Rockville, MD.

In the following exampleε the bufferε deεcribed below have the indicated compositions:

Buffer Composition

40% Formamide 50 M HEPES pH 7.0

40% formamide

5 x Denhardt's (50x = 1% Ficoll;

1% polyvinylpyrollidone;

1% bovine serum albumin) 5 x SSC (20 x SSC = 3M NaCl;

0.3 M sodium citrate) 50 /yg/ml sheared DNA

50% Formamide Same aε above, but εubεtitute 50% formamide for 40% formamide

Short Oligo Same aε above, but with no formadide Prehybridization added

Long Oligo 6 x SSC Prehybridization 50 mM εodium phoεphate (pH 6.8)

5 x Denhardt's

100 _/yg/ml sheared DNA

20% formamide Example 1 Amplification of Probe from Folliculo Stellate Cell mRNA Referring to Fig. 2, 5 /yg of poly(A)⁺ RNA from bovine folliculo stellate cellε (cells are isolated as described in Ferrara, N. et al., (1986) In: Methods in Enzymology, Conn, P.M. ed., Vol. 124, pp 245-253, Academic Press, N.Y.; Ferrara, N. et al., (1987) PNAS, 84:5773-5777) was ethanol precipitated with 1 /yg of an anti-senεe priming oligonucleotide baεed on the known amino acid εequence of amino acids no. 35 to 39 of bovine vaεcular endothelial cell growth factor. The oligonucleotide, designated #4296, was a 24-mer having a 16-fold degeneracy. The degeneracy was confined to a region of 14 baseε corresponding to the anti- sense strand of the coding region for amino acids 35 to 39. At the 5' end of the 14 bases there waε added a 10-baεe linker containing an EcoRI restriction site. The sequence of the oligonucleotide primer was as follows:

5'-GCC GAA TTC GGG ^TA ξϊC !jTG AA-3'

The mRNA and oligonucleotide were dissolved in 55 μl of 36 mM KCl, 9 mM MgCl2, 45 mM Tris pH 7.5, 12 units RNasin and 0.5 mM each of the four dNTPε. The sample was heated to 70°C for 2 minutes and then waε brought to room temperature. For εynthesis of a DNA anti-sense strand complementary to a portion of the mRNA adjacent to the site of hybridization of the primer, 60 units of avian myeloblastoεiε viruε reverεe transcriptase was added, and the reaction was allowed to stay at room temperature for 2 minutes and then brought to 42°C for 45 minutes. The sample was then extracted with phenol and chloroform, precipitated with ethanol and dried. A second DNA strand waε then εyntheεized using all of the firεt εyntheεized εtrand aε template. For εecond εtrand εyntheεis, the dried pellet waε diεsolved in 50 μl of 50 mM NaCl, 7 mM MgCl2, 7 mM Tris pH 7.5 and 1 M each of the four dNTPs. There was alεo added 1 /yg sense strand oligonucleotide primer, which was based on the known amino acid sequence of bovine vascular endothelial cell growth factor at amino acid positions no. 15 to 19. The oligonucleotide, designated #4295, was a 24-mer with an 8- fold degeneracy. The degeneracy was confined to the 14-base region corresponding to the sense strand of the coding region for amino acids 15 to 19. At the 5' end of these 14 bases there was added a 10-base linker containing a Hindlll restriction site. The sequence of the oligonucleotide primer was as follows:

5'-GCC AAG CTT GAA ^TT ξ.AT GGA ξiGT-3'

The sample was heated to 100°C for 2 minutes and then brought to 28°C. Second strand synthesis was carried out at 28°C for 10 minutes with the addition of 10 units of DNA polymerase I, Klenow fragment. Afterwards, the polymerase enzyme was inactivated by heating at 100°C for 2 minutes. The DNA sequence extending between the two primer hybridization sites was amplified by a repetitive series of enzymatically catalyzed polymerization reactions using an automated thermal cycler (Perkin Elmer Cetus DNA Thermal Cycler). For the chain reaction 5 μl of the above reaction was brought to 100 μl in 1 x reaction mix by the addition of 10yl lOx Tag buffer mix (supplied in a polymerase chain reaction kit from Cetus Corp.), 52 μl dH2θ, and 16 μl containing all 4 dNTPs at 1.25 mM each. In addition, 10% DMSO (final concentration) and 1 μg of each of the sense and anti-sense oligonucleotides described above were added, along with 2 μl of Tag polymerase supplied in the Cetuε kit. The reaction mix waε covered with 200 μl of mineral oil and placed in the thermal cycler. The cycler waε programmed to repeat the following cycle: 1. Denature at 94°C, 1 minute

2. Anneal at 55°C, 2 minuteε

3. DNA synthesis at 72°C, 3.5 minutes

The amplification reaction waε carried out for 30 cycleε. A portion of the DNA from the amplification reaction (20 μl ) was loaded onto a 6% polyacrylamide gel and subjected to electrophoresis using Haelll-digested pUC8 DNA to provide εize markerε. The gel waε stained with ethidium bromide. The stained gel is shown in Fig. 5. A major band (marked with an arrow in Fig. 5) running between 80 and 100 baεe pairε correεponded to the appropriate length DNA (94 base pairε) to encode the two oligonucleotide primerε aε well as the amino acid coding region segment bracketed by the two primers. This band was cut from the gel and the DNA was electroeluted from the gel slice at 30 volts in 0.5 x Tris borate EDTA buffer (0.045 M Triε baεe, 0.045 M boric acid, 0.001 M EDTA). The DNA obtained was precipitated with ethanol.

Example 2 Subcloning and Sequencing of Amplified Probe The DNA that was electroeluted from the gel as described in Example 1 was subcloned in bacteriophage Ml3mpl8 and Ml3mpl9. One-half of the DNA obtained from the gel was dissolved in 20 μl of water and digested with Hindlll in standard Hindlll digestion buffer for 90 minuteε at 37°C. The concentration of Triε-HCl (pH 7.5) in the reaction waε raiεed to 85 mM, EcoRI waε added and the reaction waε incubated a further 90 minuteε at 37°C. In εeparate reactionε, approximately one-tenth of the digeεted preparation per reaction waε then ligated, in the preεence of T4 DNA ligaεe, with Ml3mpl8 phage and Ml3mpl9 phage double-stranded DNA (Yanisch-Perron, et al., Gene (1985)3_3:103) that had been cut with Hindlll and EcoRI. Each ligation mix was then tranεfected into E^ coli JM103 uεing standard techniques and plated onto L plateε containing 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal) and isopropyl-β-D-thiogalactopyranoεide (IPTG). In the caεe of the Ml3mpl9 reaction, after plaques formed on the plates, portions of the plaques were lifted onto nitrocellulose filter paper, lysed by treatment with NaOH according to standard techniqueε, and baked for 2 hourε at 80°C in a vacuum oven. In order to screen for the presence of the desired insert εequence, the plaque lift was probed with a radiolabelled sample of the oligonucleotide primer #4296. Probe #4296 hybridized to numerous plaques on the plaque lift and four were chosen for further analysiε. In the caεe of the Ml3mpl8 reaction, four plaqueε for further analysis were picked based on the fact that they were white, indicating that an insert fragment had been ligated between the EcoRI and Hindlll siteε of the phage vector.

The four plaqueε from each of the Ml3mpl8 and Ml3mpl9 infected plateε that were picked were used to infect JM103 and replicative form (RF) DNA was prepared from the infected cultures using standard techniques. The RF DNA from each infection was then cut with Haelll and loaded onto a polyacrylamide gel. Electrophoresis waε conducted uεing Haelll-cut pUC8 and Haelll-cut M13mpl8 RF aε εize markers. Upon visualization with ethidium bromide, the DNA in all eight sample lanes waε εhown to contain an inεert of the correct εize to encode the amino acid sequence lying in the region between and including the amino acids used to design the primers used in the polymerase chain reaction.

One of the Ml3mpl9 plaques and the four Ml3mpl8 plaques shown to have the correct insert sequence length were picked for further analysis. Single stranded DNA was then prepared for sequencing according to εtandard procedureε (εee Messing, J., Methods in Enzymology (1983) 101 :20-78 ) . The εequenceε of the inεertε in the five isolated clones are given in Fig. 1. One region of the sequencing gel (nucleotides 32-35 in Fig. 1) was not unambiguously readable for the four Ml3mpl8 clones. Excluding the unreadable region of the Ml3mpl8 sequences, four of the five sequenced cloneε encoded the εame amino acid sequence corresponding to that portion of the bovine vascular endothelial cell growth factor encoded by the mRNA region extending between and including the hybridization sites of the two primer sequences used in the polymerase chain reaction. The fifth sequenced clone contained a single encoded amino acid difference, encoding His (CAC), rather than Pro (CCC), at the position corresponding to amino acid no. 27 of bovine vascular endothelial cell growth factor. Fig. 1 gives the DNA sequenceε of the inserts in the five clones, as well aε a consensuε DNA sequence and the deduced amino acid sequence for the insert in the Ml3mpl9 clone. It can be seen that most of the non-homologouε nucleotideε between the five iεolated sequences occurred at positions located in the primer sequences used in the polymeraεe chain reaction, indicating that in εome inεtances degenerate oligonucleotides in the primers having single- nucleotide mismatches may have hybridized with the protein- encoding sequence in the vascular endothelial cell mRNA and subsequently been amplified. Other sequence differences presumably were the result of either polymerase errors or polymorphisms in the vascular endothelial cell growth factor mRNA. While these sequences may not correspond precisely to the native DNA sequence, four of the five nonethelesε encode the correct amino acid sequence (excluding the unreadable region in the Ml3mpl8 clones) for vascular endothelial cell growth factor. Moreover, they can be used as probes to isolate full-length DNA sequenceε for bovine vascular endothelial cell growth factor or for isolating DNAs encoding the corresponding protein in non-bovine specieε. The picked phage in which the amplified DNA had been ligated into Ml3mpl9 waε renamed pET-19A; the inεerted fragment in thiε phage waε uεed aε the probe to screen a folliculo stellate cell cDNA library, as described in Example 3 below.

Example 3

Retrieval of cDNA Encoding Bovine Vascular

Endothelial Cell Growth Factor (120-Amino Acid Form)

A bovine folliculo stellate cell cDNA library was prepared in λgtlO bacteriophage according to a modification of the procedure of Huynh, et al., in DNA Cloning, D.M. Glover ed. , Vol. I, p.. 49, IRL Press, Washington, D.C. (1985). Therpoly(A)⁺ RNA used to make the cDNA for the library waε obtained from folliculo stellate cells iεolated and expanded from bovine pituitary by Dr. Deniε

Goεpodarowicz according to publiεhed procedureε (Ferrara, et al., PNAS (1987) 4:5773-5777). The cDNA library (approximately 1.5 x 10^ phage) in λgtlO was plated on C600hfl^~ cells (30 plates, 5 x 10^*^ phage/plate) . Two plaque lifts from each plate were made onto nitrocellulose filter papers. The filters were immerεed in a denaturing εolution (0.2 M NaOH, 1.5 M NaCl) for 3 minuteε, followed by a neutralization εolution (2 x SSC, 0.4 M Triε pH 7.5) for 3 minutes and a wash solution (2 x SSC) for 3 minutes. The filters were then air-dried and baked at 80°C for 2 hours in a vacuum oven. One set of the filters was prehybridized in 200 ml of 40% Formamide Buffer at 42°C. To prepare a probe for screening the filterε a single-stranded preparation was made of the Ml3mpl9-derived phage pET-19A, iεolated aε described in Example 2 above, using standard methods (Mesεing, J., Methodε Enzymol. (1983) 101:20-78) . Thiε preparation waε annealed with the "univerεal" primer (Meεεing, J., Methodε Enzymol. (1983) 101:20-78) , and a complementary εtrand for pET-19A was syntheεized by extending the primer using Klenow-fragment polymerase and α-^-^P-dNTPs.

One εet of plaque liftε was screened with this radiolabelled probe. The probe was heated to 100°C for 2 minutes to melt the double-stranded DNA and then set on ice. The probe (1 ml; 5 x 10¹-' cpm) was then added to the 200 ml of 40% Formamide Buffer used for prehybridization and mixed thoroughly. The prehybridized filters were added and incubated overnight at 42°C in a rocking water bath. The filters were then washed in 1 x SSC (20 x SSC equals 3 M NaCl, 0.3 M Na citrate) containing 0.1% SDS for several hours at 50°C. After washing, the filterε were expoεed to X-ray film at -70 to -80°C overnight. Approximately 32 putative poεitive cloneε were identified in the initial εcreen with the primer-extended probe derived from pET-19A. Ten cloneε, identified lc-lOc were selected for further screening.

The second set of plaque lifts was screened with a radiolabelled synthetic oligonucleotide probe designed on the basis of the εequence obtained for the amplified DNA insert in pET-19A, εhown in Fig. 1. The oligonucleotide, identified as probe #4340, was a 39-mer oligonucleotide corresponding to the anti-sense strand and having the nucleotide sequence:

5'-CAC CAG GGT CTC GAT GGG ACG GCA GAA GCT GCG CTG GTA-3'

The filters were pre-hybridized in 100 ml of Long Oligo Prehybridization Buffer at 43°C for approximately 6 hourε. The filterε in the pre-hybridization buffer were then heated for 10 minuteε in a 65°C water bath. The probe (5 x 10° cpm) , radiolabelled uεing γ-³²P-ATP and polynucleotide kinase, was added to the 65°C buffer and the temperature was slowly cooled to room temperature by shutting off the heat to the water bath. The following day, the filters were removed from the hybridization buffer and waεhed for 2 hourε [6~in 3 x SSC, 0.1% SDS at 45°C with a change of wash buffer at 1 hr. The filters were dried and exposed to X-ray film overnight at -70 to -80°C.

Of the cloneε that hybridized with probe #4340, only one clone appeared to correεpond to one of the 32 cloneε from the firεt set of filters that had hybridized with pET-19A. This clone was not one of the original 10 picked for further analysiε; therefore, the clone waε picked and designated clone lie. The filters probed with #4340 were rewashed under more εtringent conditionε (1 x SSC, 0.1% SDS, 65°C), whereupon the number of putative poεitive cloneε waε reduced to approximately 6, including the clone lie which had hybridized with pET-19A.

The picked cloneε 2c-llc were plated out for a εecond round of εcreening. To "pick" the cloneε, plugε of each clone were removed from the appropriate areas of the agar plates by placing the open end of a sterile Falcon 12 x 75 mm tube over the desired area of the plate corresponding to the poεitive signal on the filter lifted from the plate, pushing the tube down through the agar, and picking up the plug with a sterile spatula. Each plug was then placed into 1 ml SM. buffer (100 mM NaCl; 8 mM MgS0₄; 50 mM Tris pH 7.5; 0.01% gelatin), vortexed, and allowed to sit approximately 20 minutes at room temperature to allow the phage to diffuse out of the agar. One μl of the reεulting εuεpenεion for each picked clone was suspended in 1 ml of SM buffer (1:1000 dilution). 10 μl of each 1:1000 phage dilution was then transferred to a Falcon 75 x 100 mm tube containing 590-600 μl of plating cells (CδOOhfl^***) and then serially diluted out by transferring 10 μl from this tube to a second tube containing 590-600 μl of plating cells, and from that tube to a third tube. Phage were absorbed for 20 minutes at 37°C and the phage/C600hf1^~ mix in each tube was plated out with approximately 10 ml of top agarose in 150 mm agar plates. The plates were incubated overnight at 37°C. Plates having approximately 5,000 phage per plate were used to make plaque liftε onto nitrocelluloεe filter paperε. The DNA on the filterε waε then denatured by treating the filterε with NaOH in the same manner described above for the plaque lifts from the primary εcreen of the cDNA library. After baking, the filterε were prehybridized by immersing them in plastic sealable bagε (3 filterε/bag) each containing 10 ml of 50% Formamide Buffer, in a rocking water bath for 2 hourε at 42°C.

To prepare a probe for εcreening the filterε a double-stranded (replicative form) preparation was made of the Ml3mpl9-derived phage pET-19A. Thiε preparation waε digested with EcoRI and Hindlll and the 82-base pair inεert fragment repreεenting the amplified DNA εegment was isolated by gel electrophoresiε and then labelled by filling in εingle-εtranded endε with Klenow-fragment polymeraεe and _α32p__cj_NTp_ε# The p_ro e was boiled for 2 minutes to melt the double-stranded DNA, cooled on ice and then added to the prehybridization buffer in which the filters were immersed. The filters were incubated overnight in a rocking water bath at 42°C and washed in 0.1 x SSC, 0.1% SDS wash buffer at 50°C for 1-1/2 hours with two buffer changes. The filterε were exposed to X-ray film overnight at -70 to -80°C. Three positive clones on the plate representing the re-plating of clone lie hybridized to the 82 base pair insert fragment from pET-19A. In addition, there appeared to be one questionable poεitive clone on the plate repreεenting the re-plating of clone 7c. Theεe four positive clones were excised from the agar plates using the wide end of a Pasteur pipet, diluted in plating cells and re-plated for a third round of screening in a manner similar to that previouεly deεcribed. The three plateε produced for the third round screening of the three positive clones on the plate representing clone lie in the second round screening were designated 11A, 11B and lie. Two plaque liftε were prepared from each plate on nitrocelluloεe filter paper, as previously described. The DNA on the filters was denatured and baked using procedures εimilar to thoεe described above. The first set of plaque lifts was screened with radiolabelled probe #4340 (previously described) and the second set of plaque lifts waε screened with the previouεly deεcribed probe prepared from the 82 baεe pair inεert of pET-19A.

The firεt set of filters waε prehybridized by immerεion in plaεtic bagε containing 7 ml of Short Oligo Prehybridization Buffer at room temperature. Radiolabelled probe #4340 waε added to the prehybridization buffer containing theεe filterε. The temperature waε brought to 65°C by placing the bagε for a few minuteε in a εhaking water bath set at 65°C. The heat for the water bath waε then shut off, allowing the temperature to return slowly to room temperature. Incubation was allowed to proceed for approximately 2-1/2 days at room temperature.

The second set of filters was prehybridized by immersion in plastic bags containing 10 ml of 50% Formamide Buffer, and incubation at 42°C. The probe prepared from the 82 base pair insert of pET-19A was boiled and added directly to the prehybridization buffer containing the second set of filters. The hybridization reaction waε incubated at 42°C in a rocking water bath for approximately 2-1/2 dayε.

The- set of filters hybridized with probe #4340 waε washed in 1 x SSC, 0.1% SDS at 55°C. The set of filters hybridized with the probe derived from pET-19A were washed at 55°C in 0.1 x SSC, 0.1% SDS. Both setε of filterε were expoεed to X-ray film for 3-1/2 hourε. Several plaques on plateε 11A, 11B and 11C hybridized εtrongly to both probe #4340 and the 82 base pair insert from pET-19A. No plaques on the plate representing a dilution of the pick from plate 7c of the εecond round screening hybridized to either probe. Two strongly hybridizing plaques from plate 11A were picked using the thin end of a Pasteur pipet. These clones, designated 11A' and 11B', were diluted in plating cells and replated, as previouεly deεcribed, for a fourth round of εcreening. Two plaque liftε were prepared on nitrocellulose filter paper from each of the plates prepared from positive cloneε llA' and 11B' . The DNA on the filterε waε denatured and baked uεing procedureε εimilar to those described above. Each set of filterε waε prehybridized by immerεion in a plastic bag containing 10 ml of Short Oligo Prehybridization Buffer and incubating at room temperature. The first plaque lift from each of plates 11A' and 11B' waε εcreened with radiolabelled probe #4340 (previouεly deεcribed). The proh^p waε added to the plaεtic bag containing the filters and prehybridization buffer and incubated first at 65°C and then with slow cooling as described above. The second plaque lift from each of plateε llA' and 11B' was screened with a radiolabelled 48-fold degenerate mixed oligonucleotide probe, identified as probe #4255, which was based on the amino acid sequence derived from an internal tryptic fragment of bovine vascular endothelial cell growth factor and which has the sequence

5'-CA£ Aτ GGX GA§ ATG-3' T

The hybridization conditionε were the εame aε juεt descrjl" -! for probe #4340. The plaque lifts from plates llA' and 11B' were washed with 1 x SSC, 0.1% SDS at 55°C (#4340) or 3 x

SSC, 0.1% SDS at 30°C and then with 3 M tetramethylammonium chloride (TMACl), 0.05 M Tris-HCl, pH 8.0, 0.1% SDS, 0.002 M

EDTA at 45°C (#4255) and exposed to film. The plaques from both plates were found to hybridize to both probe #4340 and probe #4255, with all plaques on each filter hybridizing. It was concluded that cloneε llA' and 11B^* constitute single, pure cloneε.

The DNA inεert in clone 11B' waε εequenced by firεt digeεting the phage DNA with EcoRI, fractionating the digeεt on a 6% polyacrylamide gel, electroeluting the approximately 800 baεe pair insert fragment, and ligating the fragment into EcoRI-cut Ml3mpl8. The insert was then sequenced by the dideoxynucleotide procedure, using standard methods. The nucleotide sequence, as well as the encoded amino acid sequence, is shown in Fig. 3a (only nucleotideε 7 through 795 of the 797-nucleotide inεert sequence are shown in the figure; the EcoRI linker εequences on each end have been omitted) . The 797-nucleotide insert sequence encodes the known portion of the amino acid sequence of bovine vasculai. endothelial cell growth factor, beginning at amino acid no. 15 of the known protein sequence. The open reading frame extends to an in-frame translation stop codon at nucleotide 327. The insert sequence of clone 11B' was ligated into the EcoRI site of plasmid pUC8. The resulting plasmid, designated pST800, has been deposited in an E^ coli JM83 host at the American Type Culture Collection, Rockville, MD with Accession No. 68060.

A full length coding εequence for a mature form of bovine vascular endothelial cell growth factor is represented by the sequence of Fig. 3a, taken together with the sequence in Fig. 3b. The double-stranded DNA εequence shown in Fig. 3b, with the translation initiation codon, ATG, near its 5' end, representε a sequence of nucleotideε, εelected on the baεiε of preferred codon choice for gene expreεεion in human cellε, which encodes the indicated N- terminal portion of the bovine protein (preceded by an initiating methionine reεidue) and which overlaps the coding sequence shown in Fig. 3a. The DNA sequence of Fig. 3b can be synthesized using known methods of oligonucleotide synthesiε and enzymatically joined to a portion of the sequence shown in Fig. 3a, which can be conveniently obtained from the plasmid deposited at ATCC, in order to produce a full-length coding sequence for a mature form of bovine vaεcular endothelial cell growth factor. Before the εequenceε in Figs. 3a and 3b are joined, the sequence in Fig. 3a is exciεed from the plaεmid pST800 uεing EcoRI, and the iεolated insert iε digeεted with NlalV which cutε the inεert five timeε, all within the 3' untranslated region. A linker encoding a convenient restriction site, e.g. Hindlll, is then joined at the 5'-most NlalV site via blunt-end ligation. The resulting ligation mix is then digested with Accl and the linker enzyme (e.g. Hindlll), to releaεe a 325 base pair fragment (AccI-NlalV) of the pST800 insert with a digested linker ligated at the 3' end (at the NlalV site). Thiε fragment iε purified and ligated to the synthetic fragment shown in Fig. 3b. After digestion of the ligation mix with Ncol and the restriction enzyme that cleaves the linker (e.g. Hindlll, if a Hindlll linker is used), a fragment is produced with the desired coding sequence for a mature form of bovine vascular endothelial cell growth factor, flanked on the 5' εide with a digeεted Ncol site, and on the 3' side by a digested restriction εite uεeful for inεertion of the fragment into an expreεεion vector. The compoεite sequence is inserted into an appropriate expresεion vector under the control of regulatory elementε capable of directing its expresεion in a prokaryotic or eukaryotic host. For expreεsion in E^ coli, a convenient vector would be pKK233-2 (Amman and Brosiuε, Gene (1965) 4_(J:183-190) , which iε commercially available from Pharmacia, Inc. Inεertion of the compoεite εequence between the Ncol and Hindlll sites of this vector would place the coding sequence under the control of the trc promoter. The expression vector is then used to transform a suitable host, such as ______ coli and the transformants are cultured under conditions in which the encoded DNA is expresεed. The expressed protein iε then recovered by means which are conventional in the art. Of course, other sequences could be joined to the sequence in Fig. 3a. For example, the sequence in Fig. 3b could be altered so that the 5' end repreεentε an Ndel εite, rather than Ncol. For expreεεion in mammalian cellε the coding sequence in Fig. 3b could be extended in the 5' direction such that it encodes the amino-terminal sequence of bovine vascular endothelial cell growth factor operably joined to a secretion signal εequence, e.g. the εignal εequence for human growth hormone.

Alternatively, the coding sequence represented by Fig. 3a can be used as a probe, under standard conditions for DNA hybridization, to retrieve native, full-length DNA sequences encoding bovine vascular endothelial cell growth factor or the. corresponding protein in other mammalian specieε, including man. The εequence of Fig. 3a can be used as a probe to retrieve the desired sequences from either cDNA or genomic DNA libraries.

Clones extended toward the 5' end of the bovine vascular endothelial cell growth factor mRNA were generated by priming first-εtrand cDNA εyntheεiε as described above in Example 1 uεing aε a primer the antisense oligonucleotide 4338 (5'-GCCAAGCTTGCACCAGGGTCTCGATGGGACGGCAGAA-3' ) and then ligating onto the 5' end of the resulting duplex a partially double-stranded linker molecule consisting of the oligonucleotides no. 4537 and 4514 (5'-GATCGCGG-3' and 5'- CCGCGATCAAGCTTCCCGGGAATTCGGC-3' , reεpectively). Finally, the products were amplified by polymerase chain reaction using as primers the oligonucleotides 4338 and 4315 (5'- GCCGAATTCCCGGGAAGCTTGATCGCGG; complementary to 4514). Upon sequencing by the dideoxynucleotide method, the reεulting cloneε gave the 5' εequenceε εhown in Fig. 6.

Example 4 Retrieval of cDNA Encoding Bovine Vaεcular Endothelial Cell Growth Factor (bVEGF-*g^)

To isolate VEGF formε other than

(Fig- 3a), firεt-εtrand cDNA εyntheεiε waε carried out uεing folliculo εtellate poly(A)⁺ RNA aε a template and using aε a primer the antiεenεe oligonucleotide 4456 (5'- GTAGTTCTGTGTCAGTCTTTCCTGGTGAGACGTCTGGTTCCCGAAACCCTGAGGGAGGCT -3'). The resulting productε were then amplified by 30 rounds of polymerase chain reaction, using as primers the antisense oligonucleotide 4456 and the sense oligonucleotide 4414 (5'-

TTCTGCCGTCCCATCGAGACCCTGGTGGACATCTTCCAGGAGTACCCAGATGAGATT- 3'). Polyacrylamide gel analysiε and DNA εequencing of the productε revealed two εpecieε of cDNA encoding vascular endothelial cell growth factor, as shown in Fig. 6. The open reading frame which includes the 132 bp insert εhown in the box in Fig. 6 encodeε bVEGF-^ .

Example 5 Retrieval of Genomic DNA Encoding Human Vaεcular Endothelial Cell Growth Factor Human genomic cloneε containing DNA encoding amino acid εequences of vascular endothelial cell growth factor were isolated from a commercially available human lung fibroblast genomic library (Stratagene Inc., La Jolla, CA) . One μl of stock phage (approximately 3 x 10---0 phage/ml) waε diluted into 1 ml SM Buffer and 20yl of CHCI3 were added. LE 392 cells (hsdR514 (r^*-,m^_), supE44, supF5δ, lacYl or

Δ(lacIZY)6, galK2, galT22, metBl, trpR55, λ-) were grown in NZYM medium to an O.D.g_QO of 0.5 and the cells were spun out and resuεpended in 10 mM MgSθ4. In each of 30 tubeε, 2 μl of diluted phage εtock were mixed with 0.6 ml of cellε and incubated at 37°C for 15 minuteε. Ten ml of NZYM top agar were added to each tube, and the contentε of each tube were plated out on a 150 mm NZYM plate and incubated at 37°C for 16 hourε before reducing the temperature to 4°C. Two plaque lifts onto nitrocellulose filter paper were taken from each of the 30 plates. The DNA on the filters was denatured and baked as described above in Example 3. The filters were then prehybridized in 40% Formamide Buffer at 37°C for 6 hours.

The filters were probed with a radiolabelled probe which had been prepared by nick translation of gel-purified EcoRI insert fragment from the plasmid pST800, a plasmid made as described in Example 3 above. The 797 base pair EcoRI insert of pST800 contains a cDNA fragment encoding a portion of bovine vascular endothelial cell growth factor. The probe was boiled at 100°C for 2 minutes to melt the double-stranded DNA and then cooled on ice. The probe was added directly to the prehybridization buffer containing the filterε at 10^ cpm/mr and the filterε were hybridized overnight at 37°C. The filterε were waεhed in 1 x SSC, 0.1% SDS at 50°C with 3 changeε of waεh buffer, then blotted dry and expoεed to X-ray film overnight at -70 to -80°C. The exposed filmε indicated approximately 200 poεitives per plate with 19 clones being characterized as strong positives.

Of the 19 strong positiveε, 12 were picked and diluted as described in Example 3 above, and replated for a second round of screening. Two sets of plaque lifts were prepared, as previously described, from the replated phage. The filters were prehybridized in 40% Formamide Buffer at 37°C for 6 hours. One set of filters was hybridized overnight at 37°C with the same probe used in the first round screen. The other set of filters waε hybridized with nick-tranεlated radiolabelled pUCδ in order to enεure that the picked positives from the first round were not hybridizing with sequences derived from the vector used to subclone the probe εequence. The filterε were waεhed in 1 x SSC, 0.1% SDS at 50°C, and exposed to film. On this second round screening,^' six out of the twelve replated clones were εtill poεitive with the probe derived from pSTδOO, and not with the pUC8 probe.

Positive plaqueε were picked repreεenting each of the εix positive clones in the second round screen. The six picked plaques were diluted aε before and replated for a third round of εcreening. Additionally, the 7 strong positive cloneε from the firεt round which had not been rescreened were also picked and replated for a second round of εcreening. Two plaque liftε were prepared, as previously described, from each of the replated cloneε. The filters were prehybridized in 40% Formamide Buffer at 37°C for 5 hours.

.** To one εet of plaque lifts of the 6 poεitive cloneε from the εecond round of εcreening there were added 10'-' cpm/ml of the nick-tranεlated 797-baεe pair insert probe derived from pST800 as previously described. The filterε and probe were hybridized overnight at 37°C in the prehybridization buffer. To the second set of plaque lifts of these 6 positive cloneε there waε added a probe which waε prepared by nick-tranεlation of an EcoRI-Hpall fragment of the aforementioned 797-baεe pair inεert of pSTδOO. The EcoRI-Hpall fragment conεisted of 331 base pairε at the 5' end of the 797-baεe pair inεert, thereby eliminating the 3' end of the inεert which was rich in A and T nucleotides and may have accounted for falεe poεitive hybridizationε in earlier screening rounds. The probe waε hybridized in the prehybridization buffer overnight at 37°C.

To both εetε of plaque liftε of the 7 replated firεt round poεitive cloneε there were added IO*' cpm/ml of the nick-tranεlated 797-baεe pair probe. The probe waε hybridized in the prehybridization buffer overnight at 37°C. All of the filterε were waεhed in 1 x SSC, 0.1% SDS at 50°C for 2 hourε with 2 changeε of buffer. The filterε were dried and expoεed to X-ray film at -70 to -δ0°C for 3 hourε. Of the 7 firεt round poεitive cloneε, 4 gave poεitive signals with the 797-baεe pair probe, on the second-round screening. Of the 6 positive clones from the second round screening, 4 gave positive signals with both the 797-base pair and 331-base pair probes on the third round screening.

The 4 second-round poεitiveε were picked and εubjected to a third round of εcreening, with one εet of plaque lifts being screened with the radiolabeled 797-base pair probe, and the other set being hybridized with the radiolabeled 331-base pair probe, as described above. All 4 were found to hybridize with both probes. All eight clones that hybridized on the third round screening with both the 797-base pair and 331-baεe pair probeε were picked aε εingle plaqueε. Phage DNA prepε were prepared according to εtandard methodε. Fragmentε of the genomic DNA inserts in the phage were then transferred to Ml3mpl8 and M13mpl9 phage for sequencing, to confirm that they encoded human vascular endothelial cell growth factor. One of the bacteriophage containing a genomic clone encoding human vascular endothelial cell growth factor has been deposited at -the American Type Culture Collection, with accession number ATCC 40636.

Sequence analysis of this clone gave the coding εequence for mature human vascular endothelial cell growth factor, as well_, as the sequence encoding the four amino acids of the signal sequence immediately upstream of the firεt amino acid of the mature protein. A second genomic clone which overlapped this clone at the 5' end gave the upstream sequence encoding the remainder of the 26-amino acid signal sequence of human vascular endothelial cell growth factor, in order to obtain the second genomic clone which contained the signal sequence and the 5' untranslated region for human vascular endothelial cell growth factor, a genomic library (the same one aε before) waε εcreened uεing oligonucleotide 5'-

CTCTCTTGGGTACATTGGAGCCTTGCCTTGCTGCTCTACCTTCACCATGCCAAG. The oligonucleotide εequence waε derived from a bovine cDNA which waε generated via PCR. Approximately 1.2 x 10*-* phage were εcreened. Duplicate nitrocelluloεe filter plaque liftε were treated with NaOH, neutralizing, and 6 x SSC bufferε, for 4 min each. After air drying, they were baked in a vacuum oven for 2 hourε at 80°C. Prehybridization was carried out using short oligo prehybridization buffer at room temperature for 6 hours. 2 x 10*-- cpm/ml of the [³²P]- labeled oligonucleotide waε added to the filterε and hybridized in the same buffer overnight at room temperature. Filters were _^washed in 1 x SSC, 0.1% SDS at 50°C for 2 hours with two changes of buffer. Filters were exposed to X-ray film overnight. First round positiveε were subject to plaque purification and rescreening. One positive clone resulted.

The composite genomic sequence obtained from these two cloneε is represented in Fig. δ. The figure repreεentε eight exons (indicated by roman numerals), which encode the native signal sequence and all of the forms of human vascular endothelial cell growth factor which can arise from differential message splicing. The complete intron sequences are not represented, but rather only the

"junction" sequences contiguous with each of the exons are presented. As the exons are drawn in Fig. 8, mature h^VE^GFl.21 is encoded by exons II-V and VIII; hVEGF¹⁶⁵ is encoded by exons I-V, VII and VIII; and mature h PF-^g is encoded by exons II-VIII.

Example 6 Retrieval of cDNA Encoding Human Vascular Endothelial Cell Growth Factor

Cell Source of Vascular Endothelial Cell Growth Factor mRNA

Vascular smooth muscle cells produce high levels of mRNA encoding the vascular endothelial cell growth factor protein and are therefore a good source of mRNA for the preparation of a cDNA library enriched in vascular endothelial cell growth factor sequences. Fetal human vascular smooth muscle (fhVSM) cells are cultured in low glucose Dulbecco's Modified Eagle's Medium (DMEM-16, GIBCO) supplemented with 10% (v/v) Fetal Bovine Serum (HYCLONE), 2 mM L-glutamine, 100 U each of penicillin and streptomycin per ml, and recombinant human basic fibroblast growth factor (added at a concentration of 1 ng/ml every 48 hours). Cellε are εubcultivated at confluence, and are typically εeeded at 25% confluence.

Poly(A)⁺ RNA Iεolation

For the iεolation of vaεcular εmooth muεcle cell mRNA, the cellε are typically grown to confluence and then treated or not with phorbol myriεtate acetate (PMA) to additionally stimulate synthesis of vascular endothelial cell growth factor mRNA. The cell monolayer is rinsed twice with 5 to 20 ml of Dulbecco's Phosphate Buffered Saline (D- PBS) to remove residual media before isolating RNA according to the methods of Chirgwin et al., Biochemistry (1979) 1_8:5294-5299. With this method, cells in the monolayer are lysed by the direct addition of a lysis buffer (4,0 M guanidine thiocyanate, 0.1% Antifoam A, 25 mM εodium citrate, 0.5% N-lauroyl εarcoεine, and 100 mM β- mercaptoethanol) . After εhearing DNA by paεεage through a syringe needle, RNA iε directly precipitated by the addition of acetic acid and ethanol. The precipitated RNA iε then resuspended in diethylpyrocarbonate (DEP)-treated deionized water (D-H2O, typically about 400 μl ) and 2.6 ml of guanidine-HCl buffer (7.5 M guanidine hydrochloride, 25 mM sodium citrate, 5 mM dithiothreitol) is added and the RNA precipitated by the addition of acetic acid and ethanol. The precipitated RNA is again resuspended in about 400 μl of DEP-treated d-H2θ and precipitated by the addition of sodium acetate and ethanol.

Total cellular RNA isolated by the guanidine- thiocyanate procedure (above) is further fractionated by oligo d(T)-cellulose chromatography to iεolate poly(A)⁺ RNA following eεtabliεhed procedureε (Edmonds, M., et al., PNAS (1971) 68.1336; Aviv, H. and Leder, P., PNAS (1972) 69_:140δ) .

cDNA Syntheεiε and Cloning of Vaεcular Endothelial Cell Growth Factor cDNA in λZAPII cDNA synthesis is performed according to the methods of Gubler and Hoffmann (Gene, 2_5_:263-269) using a cDNA syntheεiε kit purchaεed from Boehringer-Mannhei Biochemicalε. The method is briefly described as follows: first εtrand cDNA synthesis iε primed uεing oligo d(T)]_5 aε a primer to begin εyntheεis by reverse transcriptaεe from the 3'-endε in 5-20 μg of fhVSM poly (A)⁺ RNA. Limited digestion of the reεulting RNA-DNA hybrid with RNase H provides 3'-OH primers for syntheεis of the second DNA strand using IS^ coli DNA polymerase I. DNA polymerase is then used to remove any remaining overhanging 3'-endε, yielding a blunt-ended cDNA product. Before insertion into a lambda cloning vector such as λZAPll (Stratagene Inc., La Jolla, CA) , the blunt-ended cDNA is methylated (e.g. with EcoRI methylaεe) according to εtandard procedureε to block cleavage of a particular εubεet 5 of the restriction sites present in the cDNA (i.e., methylation with EcoRI methylaεe will block cleavage of the cDNA by EcoRI). The cDNA is then ligated to oligonucleotide linkers (e.g. EcoRI linkers, GGAATTCC), the linkers are cleaved with the appropriate restriction endonucleaεe (e.g.

10 EcoRI), and the cDNA is finally ligated into a suitable cloning site in the lambda vector (e.g. the EcoRI site of λZAPII) after removal of excesε linkerε. Subεequent to ligating the cDNA to vector armε, the cloned cDNA iε "packaged" with a lambda packaging extract εuch aε Gigapack

15 II Gold (Stratagene, Inc., La Jolla, CA) .

After packaging, the lambda phage are titered on the appropriate hoεt εtrain (e.g. XLl-Blue, Stratagene, Inc., La Jolla, CA, for the λZAPII vector), then plated on 150 mm plates of NZYM agar at a titer of between 10,000 and

20 50,000 pfu/plate. Following growth for 6-δ hourε at 37°C, the plateε are chilled to 4°C and plaque liftε onto nitrocellulose (BA65, SCHLEICHER AND SCHUELL) or Hybond-N (AMERSHAM) membranes are prepared according to standard procedures (Benton, W.D. and Davis, R.W. , Science (1977)

25 196:160) . Clones containing sequences homologous or partially homologous to vaεcular endothelial cell growth factor εequences are detected by hybridization to ³2p- labeled bovine vascular endothelial cell growth factor probes derived from the cDNA inεert in pSTδOO (Example 3

30 above) or vaεcular endothelial cell growth factor εequence- specific oligonucleotideε baεed on the εequence given in Fig. 3a. The hybridizationε are carried out in εtandard hybridization bufferε containing between 20% and 50% formamide and between 0 and 10% dextran εulfate, and are

35 performed at between 37° and 42°C. Cloneε hybridizing to the vaεcular endothelial cell growth factor probes are subsequently single-plaque purified and the related sequences subcloned into bacteriophage M13 vectors such as M13mplδ and Ml3mpl9 for DNA sequence analysis.

The human cDNA sequence for vascular endothelial cell growth factor can be used to predict the εpecific amino acid sequence of the human vascular endothelial cell growth factor gene productε. The cDNA can also be joined to tranεcriptional control elements in constructs designed to express the human vascular endothelial cell growth factor protein product in bacteria such aε E^ coli, or in yeaεt or mammalian cells.

Example 7

DNA' and Amino Acid Sequences of Human Vaεcular

Endothelial Cell Growth Factor (hVEGF-*2i and hVEGF-j g _$ )_

Following the procedureε set forth in Example 6, cDNA clones encoding human vascular endothelial cell growth factor were prepared and isolated. Sequence analyεiε of several cloneε confirmed that alternative meεsage splicing occurs analogously to the bovine case. Accordingly, there are expressed forms of the human protein which correspond to the bVEGF-L20 ^anc-- bVEGFιg4 proteins (however, since the human proteins contain an additional amino acid at poεition 7 not found in the bovine formε of vascular endothelial cell growth factor, the human formε of vaεcular endothelial cell growth factof¹ contain 121 and 165 reεidueε, reεpectively). A cDNA clone containing a portion of the coding region for hVEGF*L2i, designated λH3, haε been depoεited with the American Type Culture Collection with accession number 40728. :A cDNA clone containing a portion of the coding region for hVEGF*_j_ , designated λH2, haε alεo been depoεited with acceεεion number 40727. Further cloneε can be obtained in an analogouε fashion encoding the entire primary translation products for hVEGF^i and hVEGFi^. Based on composite εequence information obtained from the depoεited human genomic clone of Example 5 and εeveral cDNA cloneε obtained by the procedure deεcribed in Example 6, the native DNA coding εequences for

and hVEGFχg5 were determined. The DNA sequences are shown in Fig. 7. The boxed sequence of 132 nucleotides compriεes the DNA sequence corresponding to the alternatively spliced portion of the message. When this sequence is present in the translated message, the encoded protein is hVEGF- 5, the amino acid sequence of which is given directly above the nucleotide sequence of Fig. 7. When thiε εequence iε not preεent in the tranεlated meεεage, the encoded protein is hVEGF"L21- This form of the protein has the same amino acid sequence as hVEGFigs through position 114. The carboxyl- terminal εequence of hVEGFi21» beginning at position 112, is shown in italics below the nucleotide sequence in Fig. 7. Contiguous cDNA sequenceε encoding hVEGF^l and hVEGF-^g5 can be generated from εynthetic oligonucleotides, or through the use of polymeraεe chain reactions from human fetal vascular smooth muscle poly(A)⁺ RNA, uεing methodε analogous to those described in Example 4.

Example 8 Expresεion of Polypeptideε Having Amino Acid Sequenceε of Formε of Human Vaεcular Endothelial Cell Growth Factor A cDNA clone containing the entire coding region for the primary tranεlation product of human vascular endothelial cell growth factor (hVEGF^l or hVEGF-|_g5) is most conveniently used in complete or truncated (modified) form to produce the recombinant protein in a variety of hostε as set forth in Standard Procedures above. However, expresεion in mammalian εyεtemε iε favored aε the host is capable of post translational processing analogous to that experienced by the natively produced protein, and either cDNA or genomic sequenceε may be uεed, aε the hoεt iε alεo capable of processing introns.

Thuε, a cDNA or genomic clone containing the entire coding region for either form of human vaεcular endothelial cell growth factor is prepared for insertion into a host vector, illustrated by, but not limited to, those deεcribed below. To construct the vectors, the cloned cDNA or genomic insert is excised from the cloning vector in which it was isolated. The insert is provided with Ncol, BamHI, EcoRI or other appropriate linkerε if neceεεary, and then inserted into an appropriate host vector such as pHSl or its derivatives as described below. Alternatively, I vitro mutagenesiε may be uεed to introduce convenient reεtriction sites or additional coding sequences into the cloned insert, before excision of the insert and insertion into an appropriate vector.

Construction of Host Vectors pHSl The plasmid pHSl is suitable for expression of inserted DNA in mammalian hosts. It contains approximately 840 base pair of the human metallothionein-II_A (hMT-II_A) sequence from pδ4H (Karin, M. , et al. , Nature (1962) 2_99_:797-802) which spans from the Hindlll site at position -765 of the hMT-II_A gene to the BamHI cleavage site at base +70. To construct pHSl, plasmid p84H was digested to completion with BamHI, treated with exonuclease BAL-31 to remove terminal nucleotides, and then digeεted with Hindlll. The desired approximately 840 base pair fragment was ligated into pUC8 (Vieira, J. , et al., Gene (1982) l :259-268) which had been opened with Hindlll and Hindu digestion. The ligation mixture was used to transform E^_ coli HB101 to Amp^R, and one candidate plasmid, designated pHSl, was isolated and sequenced by dideoxy sequencing. pHSl containε the hMT-II_A control εequenceε upεtream of a polylinker containing convenient reεtriction εiteε (BamHI, Smal, and EcoRI) .

The workable host plasmid pHSl can be further modified to contain additional control elements beεideε the metallothionein promoter. In particular, the enhancer elementε of viral systems, such aε SV40, can be included, aε well as termination signalε aεsociated with the 3' untranslated regions of other proteins such aε human growth hormone (hGH) . Viral Enhancer

A pair of host expreεεion vectorε containing the SV40 enhancer in operable linkage to the MT-II_A promoter waε conεtructed by inserting an 1118 base pair SV40 DNA fragment into the Hindlll site preceding the MT-II_A promoter sequenceε of pHSl. The SV40 DNA fragment spans the SV40 origin of replication and includes nucleotide 5172 through nucleotide 5243 (at the origin), the duplicated 72 base pair repeat from nucleotide 107-250, and continues through nucleotide 1046 on the side of the origin containing the 5' end of late viral geneε. Thiε Hindlll 1118 baεe pair fragment iε obtained from a Hindlll digest of SV40 DNA (Buchman, A.R., et al., DNA Tumor Viruseε, 2nd ed (J. Tooze, ed.), Cold Spring Harbor Laboratory, New York (1981), pp. 799-841), and cloned into pBR322 for amplification. The pBR322 vector containing the SV40 fragment waε cut with Hindlll, and the 1118 baεe pair SV40 DNA fragment was isolated by gel electrophoresiε and ligated into Hindlll- digeεted, CIP-treated, pHSl. The reεulting vectorε, deεignated pHSl-SV(9) and pHSl-SV(lO), contain the SV40 fragment in oppoεite orientation preceding the MT-II_A promoter. In pHSl-SV(9), the enhancer iε about 1600 base pair from the 5' mRNA start site of the MT-II_A promoter, in the opposite orientation it is approximately 960 base pair from the 5' mRNA start site. Both orientations are operable, but the orientation wherein the enhancer sequenceε are proximal to the start εite provideε higher levelε of expreεεion. It iε believed that deletionε which place the enhancer 250-400 baεe pairε upεtrea of the tranεcription start are optimal.

Additional vectorε were conεtructed which place the SV40 enhancer 3' terminus 190 base pairε, 250 base pairs, and 360 base pairε respectively upstream from the 5' end of the MT-II_A promoter TATA box. The constructions were based on the mapping of the upstream regulatory regions of the human MT-II_A promoter described by Karin, M., et al., Nature (1984) 308:513-519. All constructionε retain the εequenceε containing the duplicated εiteε for regulation by heavy metalε, but the constructionε with the 190 baεe pair and 250 baεe pair separations do not retain the εequence for glucocorticoid regulation which is further upstream from these heavy metal regulatory εiteε. These vectorε, deεignated pHS'-SVl90, pHS'-SV250, and pHS'-SV360 are prepared aε outlined below. All conεtructions are identical except for the length of sequence containing the metallothionein promoter and upstream region which is supplied aε a fragment exciεed from pHSl.

For pHS'-SVl90, pHSl iε digeεted with SacII, blunted, and ligated to Kpnl linkerε. The DNA iε then digeεted with EcoRI and Kpnl to liberate the appropriate portion of the MT-II_A control εequenceε. Similarly, for pHS'-SV250, pHSl iε digeεted with Hgal, blunted, ligated to Kpnl linkerε and digeεted with EcoRI and Kpnl; for pHS'- SV360, Ddel iε used in the initial digestion.

An intermediate vector containing the SV40 enhancer is prepared by inserting the Hindlll/Kpnl fragment of SV40 (which extends from position 5172 to position 298 and which contains the enhancer element 50 base pairs from the Kpnl site) into Kpnl/Hindlll digested pUCl9 to obtain pUC-SV. (pUCl9 containε three convenient reεtriction εiteε in the polylinker region, in order, Hindlll, Kpnl, and EcoRI.) The finished vectors are obtained by inserting the Kpnl/EcoRl fragments prepared as described above into Kpnl/EcoRI digested pUC-SV.

All of the foregoing modified vectors, thuε, take advantage of the SV40 enhancer element. Other viral enhancerε could, of courεe, be uεed in an analogouε manner.

Tranεcription Termination Sequences To provide tranεcription termination control sequences, DNA representing the coding εequence and 3' untranslated sequence of human growth hormone was ligated into pHSl. The intermediate vector can provide the hGH 3' untranslated sequence to coding sequenceε subsequently ligated into the vector in place of the hGH coding sequence. The genomic εequenceε encoding hGH were iεolated from p2.6-3 (DeNoto, et al. , Nucleic Acids Res. (1981) 1):3719) by digestion with BamHI, which cuts at the 5' end of the first exon, and EcoRI, which cuts 3' of the functional gene, followed by polyacrylamide gel purification. The isolated fragment was ligated into BamHI/EcoRI digested pHSl and the ligation mixture transformed into E^ coli MC1061 to Amp^R. Successful transformants were screened by restriction analysis, and a strain containing the desired plasmid, pMT-hGHg, was further propagated to prepare quantities of plaεmid DNA.

In a manner εimilar to that deεcribed above for construction pHSl-SV(9) or pHSl-SV(lO), but subεtituting for pHSl, pMT-hGHg, a pair of vectorε containing the hGH gene under the control of the MT-II_A promoter, and operably linked to the SV40 enhancer, and deεignated, reεpectively, phGHg-SV(9) and phGHg-SV(10 ) , were obtained. The ligation mixtures were uεed to tranεform E___ coli MC1061 to Amp^R, and the correct constructions verified.

Construction of Expreεεion Vectors phGHg-SV(lO) is used as a host vector to accommodate human vascular endothelial cell growth factor. phGHg-SV(lO) is digested with BamHI and Smal, blunted with Klenow, and treated with CIP to exciεe the hGH coding εequence. This opened vector is ligated to the insert fragment derived from a cDNA or genomic clone encoding full- length vascular endothelial cell growth factor to obtain expression vector pVEGF-SV(10 ) . As shown in Fig. 8, the full primary tranεlation product of the vaεcular endothelial cell growth factor gene containε a 26 amino acid secretion signal εequence which iε capable of effecting εecretion of mature vaεcular endothelial cell growth factor into mammalian cell culture media. If deεired, εynthetic oligonucleotides can be added to the coding sequence of mature vascular endothelial cell growth factor to operably join a heterologouε εecretion εignal εequence (e.g. from hGH) to the vaεcular endothelial cell growth factor produced in the primary tanslation product. In either case, secretion of the product should result.

In addition, other host vectors may be used to obtain expression of the vascular endothelial cell growth factor gene or cDNA sequenceε, including pHSl and pHSl modified to contain the various configurations of SV40 enhancer as above described. Finally, the host vectors may be further modified such that they encode not only vaεcular endothelial cell growth factor, but the neomycin resistance gene (obtained from pSV2:NEO) and/or the human metallothionein-II_A protein as well (called pMT-VEGF-NEO or pMT-VEGF-NEO-MT) .

These vectors are generically deεignated pMT-VEGF for the purpoεes of the discuεεion below.

Production of Vascular Endothelial Cell Growth Factor by Mammalian Recombinants

Chinese hamεter ovary (CHO)-Kl cellε are grown in medium composed of a 1:1 mixture of F12 medium and DME medium with 12% fetal calf serum. The competent cells are co-transformed with pMT-VEGF and pSV2:NE0 (Southern, P., et al., J. Hoi. Appl. Genet. (1982) l.:327-341), pSV2:NEO contains a functional gene conferring resistance to the neomycin analog G418. In the transformation, 1 μg of pSV2:NEO and 10 μg of pMT-VEGF are applied to cells in a calcium phosphate-DNA co-precipitate according to the protocol of Wigler, M., et al., Cell (1979) l_6:777-765, with the inclusion of a two minute "shock" with 15% glycerol after four hours of exposure to the DNA. Alternatively, the cells can be tranεformed with 10 μg pMT-VEGF-NEO or pMT- VEGF-NEO-MT, uεing the εame calcium phosphate protocol. A day later, the cells are subjected to 1 mg/ml G41δ to provide a pool of G41δ-reεiεtant colonies. After εufficient growth of the pool of resistant colonies, the pool is assayed for vascular endothelial cell growth factor production in either cell-asεociated or εecreted form. Succeεεful G418-reεiεtant transfor ants, also having a stable inheritance of pMT-VEGF, or other vascular endothelial cell growth factor expression plasmid, are plated at low density for purification of clonal isolateε. Small amountε of theεe iεolateε are grown in multi-well plates after exposure to 2 x IO^-4 M zinc chloride for convenient assay of vascular endothelial cell growth factor production. Vascular endothelial cell growth factor determinations are made by mitogenic asεays testing endothelial cell mitogenic activity present in the cells and/or conditioned medium, or by standard ELISA or radio- immunoassays against the antiεera prepared againεt the appropriate vaεcular endothelial cell growth factor protein or peptideε uεing standard methods. Clonal iεolateε which produce large amountε of the deεired vaεcular endothelial cell growth factor, preferably in εecreted form, are εelected.

The cellε are seeded at 1/10 confluency in basal medium (1:1 mix of F12 medium and DME medium) supplemented with 10% fetal calf serum, incubated overnight, and then induced for vascular endothelial cell growth factor production by addition of zinc chloride in the concentration range of 1 x IO^-4 M to 3 x IO^-4 M.

In another method for establiεhing vaεcular endothelial cell growth factor cellε, CHO cellε are cotransformed with pMT-VEGF, pSV:NEO, and pHSl containing a human MT-II_A inεert (pHSl-MT) or with pMT-VEGF-NEO-MT. After G418 εelection, pooled resistant colonies are selected for cadmium resiεtance (due to expression of MT-II_A protein) by growing them in the presence of 10 /yM CdCl2 with 100 /yM ZnCl2 as inducer. Pools of resistant clones are then assayed, as described above, to measure vascular endothelial cell growth factor production levels.

By including in the expresεion vector conεtruction an operable secretion signal sequence for vascular endothelial cell growth factor, such as the native vaεcular endothelial cell growth factor εignal or the εignal derived from hGH, εecretion using the normal constitutive pathwayε could be effected using CHO or other mammalian cell hosts. Effecting secretion has some advantages, of course, since the protein purification taεk becomeε much εi pler, and folding of the protein may be closer to the native configuration, eliminating the need for refolding εteps. Purification of the εecreted vascular endothelial cell growth factor can then be carried out according to the procedures set forth in Example 9, or by other standard methods known in the art, which may include such steps as ion-exchange chromatography, size-exclusion chromatography, separation by hydrophobicity, e.g. reverεe-phaεe HPLC, and antibody-affinity chromatography using anti-vaεcular endothelial cell growth factor antibodies produced according to known techniques.

Example 9 Recovery of Polypeptide and Formation of Dimeric Vascular Endothelial Cell Growth Factor When expresεed in a mammalian expreεεion εyεtem in εuch a way aε to obtain εecretion of the produced growth factor, hVEG ι 5 can be purified by the procedure of Goεpodarowicz, et al. , PNAS (1969) 6(19) :7311-7315. The medium conditioned by the hoεt cells expresεing the growth factor iε centrifuged at 10000 g for 15 to 30 minutes. The supernatant solution is adjusted to pH 5 to 6 with 1 N HCl, and at least 500 g of ammonium εulfate iε added per liter. The εolution is stirred for 2 to 6 hours at 4°C and then centrifuged for 30 to 60 minutes at 10000 g. The supernatant is discarded, and the pellet is retained for further purification.

The pellet is rediεεolved in 5 to 25 mM Triε, pH 6.5 to 6.0, containing 25 to 100 mM NaCl. The εolution iε dialyεed overnight against the εame buffer. Any precipitated material is centrifuged out of solution and discarded (10000 g, 30 to 60 minutes). The solution is loaded onto a column of heparin-Sepharoεe, which iε equilibrated with the εame buffer uεed for dialyεiε. After all of the protein solution iε loaded, the column iε washed with the equilibration buffer until the eluant absorbance returns to baseline levels. The protein is step-eluted from the column with equilibration buffer containing between 0.1 M and 2.5 M NaCl. Active fractionε are combined and concentrated in an Amicon εtirred cell with a 10,000 MW cutoff membrane.

The concentrated biologically active material which iε collected from the heparin-Sepharoεe column iε applied to a column of Bio-Gel P-60 equilibrated in phoεphate buffered saline. The column is eluted in the same buffer and the biologically active fractions are combined. This material is diluted three fold with 20 mM HEPES pH 8.3 and loaded onto a Mono-S column. The column is eluted with a gradient of 0.0 M to 1.0 M NaCl in the same buffer. For structural studies, the final purification is accomplished on a Vydac C4 reverse phase column (RP-HPLC) with a gradient of 10 to 60% acetonitrile in water containing 0.1% trifluoroacetic acid.

When a bacterial expreεεion εyεtem iε uεed to produce the protein in incluεion bodies, the product is purified in a manner analogous to the method of Hoppe, et al., Biochemistry (1989) 2£:2956-2960. The cells are suspended in 5 to 25 mM Tris pH 6.5 to 8.0, 1 mM EDTA and are ruptured by pasεage through a microfluidizer. The solution iε centrifuged at 10000 g for 15 to 30 minuteε and the pellet iε waεhed with 5 to 25 mM Triε pH 6.5 to 8.0, 1 mM EDTA and 1 to 2% Triton X-100.

The pellet iε reεuεpended in 20 M Tris pH 7.5, 1 mM EDTA, 6 M guanidine-HCl, 0.1 M Na2S0₃, and 0.01 mM Na2S4θ , and the solution is left at room temperature for 4 to 12 hours. This step convertε the molecule to the monomeric form. Inεoluble material iε removed by centrifugation.

The resulting S-sulfonated protein is chromatographed on Sephacryl S-200 equilibrated in 10 to 50 mM Triε pH 6.5 to 8.0, 1 mM EDTA, 3 to 6 M guanidine-HCl. Those fractions containing the protein are pooled and are dialysed againεt water. Final purification of the S- εulfonated protein is accomplished on RP-HPLC C4 chromatography. The protein is eluted with a linear gradient of 0% to 100% acetonitrile (chamber A for making the gradient iε 0.1% trifluoroacetic acid in water and chamber B is 0.1% trifluoroacetic acid in acetonitrile).

The protein is dissolved to a final concentration of 0.1 to 0.5 mg/ml in 50 mM Tris pH 8.0, 1 mM EDTA, 5 mM glutathione and 0.5 mM glutathione disulfide with enough urea to maintain solubility of the protein. At this step the protein folds to form the native homodimeric εtructure of vaεcular endothelial cell growth factor. After two dayε the protein iε repurified on same RP-HPLC εyεte as above, or by affinity chromatography steps such as heparin- Sepharose or Mono-S. Monomers are separated from dimers by chromatography on S-Sepharose in 20 mM Triε-HCl, pH 7.5; the dimerε are eluted from the column with 20 mM Tris, pH 7.5 containing 0.7 M NaCl. Since we have discovered that hVEGF^21 (corresponding to bVEGF^Q) lacks the heparin binding characteristics of hVEGFi s and hVPF^gg, purification of hVEGF-^i and bVEGF^O iε carried out using stepε other than heparin-Sepharoεe chromatography.

Example 10 Wound Healing Formulations Containing Vascular Endothelial Cell Growth Factor

A parenteral solution suitable for adminiεtration intravascularly via catheter to a wound εite can be prepared by dissolving vascular endothelial cell growth factor (e.g. bVEGF^*L20* bVEGF-"^, hVEGF-^i or hVEGF-^*^) in water for injection, together with a suitable amount of buffer to maintain stable pH in the range of 5.0 to 7.0 and a suitable amount of sodium chloride to attain iεotonicity. A typical compoεition iε aε followε: mg/ml Vascular endothelial cell growth factor 0^05-1.0 Citric acid 0.2

Sodium chloride 8.5 0.01 N sodium hydroxide to adjuεt pH to 6.0 Water for injection sufficient to make 1.0 ml

The solution described above can also be applied topically to a wound εite with the assistance of a mechanical spray pump.

An aqueous gel, suitable for topical application to a wound site, can be prepared by dispersing the thickening agent hydroxyethylcelluloεe (250H grade) in an aqueous solution containing buffer, preservative and tonicity modifier. When the thickening agent is completely dissolved, a concentrated aqueous solution of vascular endothelial cell growth factor is added and mixed until the product iε uniform. The following pharmaceutical compoεition iε typical of εuch a gel: mg/ml

Vaεcular endothelial cell growth factor 0.05-1.0 Hydroxyethylcelluloεe (250H) 20

Chlorhexidine gluconate 2.5

Citric acid 0.5 Glycerin 20

0.01 N εodium hydroxide to adjuεt pH to 6.0 Purified water sufficient to make 1 ml

A dry powder, suitable for dusting onto a wound site, can be prepared by lyophilizing vascular endothelial cell growth factor with a water soluble carrier and comminuting the lyophilized product to yield a powder of uniform particle size. The powder can be applied to the wound εite directly or with the aid of an aeroεol propellant. A typical powder compoεition is prepared as follows: mg/ml Vascular endothelial cell growth factor 0.05-1.0 Dextran (Mol. wt. 1000) 100

Purified water sufficient to make 1 ml

The solution is freeze dried and the reεulting dried substance is ground in a ball mill to a medium particle size of about 75 μ . The powder can be applied by a shaker. In case a large surface area needs to be covered, the powder can be delivered by an aerosol-driven canister containing fluorocarbon (Freon^R), hydrocarbon (iεobutane) or compressed gaε (carbon dioxide) aε the propellant.

Example 11 Preparation of Chimeric Growth Factor

A chimeric molecule containing one chain of vascular endothelial cell growth factor and one A chain of platelet-derived growth factor is prepared by firεt constructing vectors for the recombinant expression of theεe two molecules. Expresεion vectorε to direct the synthesiε of human vascular endothelial cell growth factor in mammalian cells are described in Example 8. The preferred vector is one that is conεtructed εuch that the εynthesized vascular endothelial cell growth factor iε εecreted from the hoεt cell (e.g., phGHg-SV(10) , altered εuch that the coding region for the full primary tranεlation product of vaεcular endothelial cell growth factor, including the native vaεcular endothelial cell growth factor εecretion εignal, is operably inserted between the BamHI and Smal siteε of the parental vector). A εimilar vector is constructed for the expression of the A chain of platelet-derived growth factor by taking a synthetic or partially-synthetic DNA fragment having the εequence εhown in Fig. 4a (which encodes the full primary tranεlation product of the A chain) , digesting with BamHI and EcoRV, and inserting the resulting coding region fragment between the BamHI and Smal sites of phGHg-SV(lO) .

For production of the growth factor chains, the expression plasmids are introduced into mammalian hoεtε cellε, εuch aε CHO cellε, by the calcium phoεphate precipitation method deεcribed in Example 8. Two different tranεformationε of the CHO cells are carried out: in one transformation, the DNA introduced into the cellε repreεents a co-transformation of the vascular endothelial cell growth factor expression plasmid and pSV2:NE0 in a 10:1 weight ratio; in the other transformation, the co-transformation iε with the platelet-derived growth factor A-chain expreεεion vector and pSV2:NEO in a 10:1 weight ratio. G418-resistant pools of transformants are selected from each transformation, and individual growth factor-producing clones are screened for high-level growth factor production by asεays of the conditioned medium for mitogenic activity on endothelial cells (in the caεe of vaεcular endothelial cell growth factor-producing cells) or on ouεe NIH 3T3 cellε obtainable from the ATCC (#ATCC CRL 1658) (in the caεe of platelet-derived growth factor A-chain-producing cellε), or by ELISA or radio-immune aεεayε for the two growth factor chainε developed by methods standard in the art. in an alternative approach, a transformation is carried out in which three plasmids are co-precipitated with calcium phosphate onto the cellε: the expreεεion vector for vascular endothelial cell growth factor, the expresεion vector for platelet-derived growth factor A-chain, and pSV2:NEO (in a weight ratio of 10:10:1). A G418-reεiεtant pool of cloneε iε εelected from the transformation, and individual clones are then screened by ELISA or other antibody-based asεayε for the εimultaneouε εecretion of both of the growth factor chainε. Purification of the vaεcular endothelial cell growth factor chainε from conditioned medium iε carried out aε described in Example 9. Purification of the platelet- derived growth factor A-chains from conditioned medium iε carried out following the protocolε known in the art, e.g., the protocol of Heldin, et al., Nature (1986) 319:511-514. In thiε latter protocol, the conditioned medium containing the secreted growth factor is fractionated by adsorption to Sulphadex beads, followed by elution of the platelet-derived growth factor A-chain material with 1.5 M NaCl in 0.01 M phosphate buffer, pH 7.4. After ammonium sulphate precipitation to concentrate the eluted protein, the sample is resuεpended in 1 M NaCl, 0.01 M phoεphate buffer, pH 7.4, and dialyzed againεt this buffer. The sample is then fractionated over a Sephacryl S-200 column (elution with 1 M NaCl, 0.01 M phosphate buffer, pH 7.4), dialyzed against 1 M acetic acid, lyophilized, dissolved in 1 M acetic acid, and applied to a BioGel P-150 column (elution with 1 M acetic acid). After lyophilization of the fractions containing the A-chain material, the εample iε re-diεεolved in 1 M acetic acid and fractionated by reverεe-phaεe HPLC (elution with a gradient of ft to 50% propanol in 1 M acetic acid, 2 M guanidine-HCl) . A chimeric dimer containing one chain of vaεcular endothelial cell growth factor and one chain of platelet- derived growth factor A-chain iε produced by mixing the two purified samples of these proteinε prepared aε deεcribed above. The mixture iε then denatured and refolded aε described in Example 9. Briefly, the mixture iε first denatured and S-sulfonated by treatment with 20 mM Triε, pH 7.5, 1 mM EDTA, 6 M guanidine-HCl, 0.1 M Na2S0₃, 0.01 mM Na2S4θ for 4 to 12 hourε at room temperature. After fractionatiori over Sephacryl S-200 and purification by reverεe-phaεe HPLC (see Example 9), the S-sulfonated chainε are lyophilized and then diεεolved at a final concentration of 0.1 to 0.5 mg/ml in 50 mM Triε-HCl, pH 8.0, 1 mM EDTA, 5 mM glutathione, 0.5 mM glutathione diεulfide with enough urea to maintain solubility of the protein chains. Monomers are separated from dimerε by chromatography on S-Sepharoεe in 20 mM Triε-HCl, pH 7.5, uεing εtepε of increasing concentrations of NaCl in the Triε buffer for elution. The chimeric dimerε are then εeparated from the homodimers by a combination of the steps used to purify the individual homodimerε before the denaturation and refolding εtepε were carried out. Additionally, the chimeric molecule can be purified by paεεage over an anti-vaεcular endothelial cell growth factor antibody column, followed by passage over an anti-platelet-derived growth factor A-chain antibody column, following procedures known in the art.

Example 12 Retrieval of Full-length cDNA Encoding Human Vascular Endothelial Cell Growth Factor (hVEGF-|_?ι)

Cell Source of Vascular Endothelial Cell Growth Factor mRNA Human U937 promonocytic leukemia cells produce high levels of mRNA encoding the vascular endothelial cell growth factor protein and are therefore a good source of mRNA for the preparation of cDNA encoding human vaεcular endothelial cell growth factor. U937 cellε obtained from the American Type Culture Collection were maintained in RPMI-1640 (GIBCO) εupplemented with 10% fetal bovine serum (FBS), 1% L- glutamine and 100 U each of penicillin and streptomycin per ml. Cells were typically subcultivated every 3 to 4 days and were typically seeded at a density of 5 X 10--' cells per ml of medium.

Poly(A)^*1* RNA Isolation

For the isolation of U937 cell mRNA, the cells were grown to a density of about 5 X IO-* cellε per ml. Cellε were pelleted from the medium by gentle centrifugation at 500 X g for 5 minuteε. The cell pellet waε washed once with 50 ml of ice-cold Dulbecco's Phoεphate Buffered Saline (D- PBS) to remove reεidual media before iεolating mRNA according to the methodε of Chirgwin et al., Biochemiεtry (1979) 118:5294-5299. With thiε method, cellε in the washed cell pellet were lysed by the direct addition of a lyεiε buffer (4.0 M guanidine thiocyanate, 0.1% Antifoam A, 25 mM εodium citrate, 0.5% N-lauroyl εarcoεine, and 100 mM β- mercaptoethanol) . After εhearing DNA by paεεage through a syringe needle, RNA was directly precipitated by the addition of acetic acid and ethanol. The precipitated RNA was then resuspended in diethylpyrocarbonate (DEP)-treated deionized water (D-H2O, typically about 400 μl ) and 2.6 ml of guanidine-HCl buffer (7.5 M guanidine hydrochloride, 25 mM εodium citrate, 5 mM dithiothreitol) waε added and the RNA precipitated by the addition of acetic acid and ethanol. The precipitated RNA waε resuspended again in DEP-H2O and precipitated by the addition of sodium acetate and ethanol. 5 Total cellular RNA isolated by the guanidine- thiocyanate procedure (above) was further fractionated by oligo d(T)-cellulose chromatography to iεolate poly(A)⁺ RNA following eεtabliεhed procedureε (Edmondε, M. , et al., PNAS (1971) 6_8:1336; Aviv, H. and Leder, P., PNAS (1972) 10 69_:1408) .

Synthesis and amplification of vascular endothelial cell growth factor cDNA by reverse tranεcription and polymerase chain reaction (PCR) amplification.

15 To generate a hVEGF-^l expresεion caεεette, oligonucleotide primers (Fig. 9) were used to prime a PCR from U937 poly(A)⁺ RNA (see above), yielding a vascular endothelial cell growth factor fragment containing BamHI cloning sites at each end and termination codons in all

20 reading frames at the 3' end. First-εtrand cDNA εyntheεiε waε performed by annealing 1 μg of the antiεenεe oligonucleotide (primer 4738, Fig. 9) to 5 μg of U937 poly(A)⁺ RNA followed by polymerization with AMV reverεe transcriptase for 2 hours at 42°C using a cDNA syntheεiε kit

25 (Boehringer-Mannheim) . Following firεt-εtrand cDNA εyntheεis, the reactions were extracted with phenol and chloroform and precipitated with ethanol. One-tenth of the cDNA waε then amplified by 30 roundε of PCR (Saiki, R.K. et al., Science (198δ) 239:487-491) in a Perkin Elmer Cetuε

30 DNA Thermal Cycler uεing the anti-εenεe oligonucleotide 4738 and a εense strand oligonucleotide (primer 4741, Fig. 9) as primerε. The products of the PCR reaction were then fractionated on a 5% polyacrylamide gel, and the band corresponding to the hVEGF-^21 expresεion caεsette was

35 eluted, digeεted with BamHI, and ligated into Ml3mplδ. The sequence was confirmed by the method of Sanger et al., __ _ Mol. Biol. (I960) 143:161-176. Example 13 Expreεεion and Secretion of hVEGF-i j - in CHO Cellε Following εequence confirmation, the BamHI fragment containing the vaεcular endothelial cell growth factor expreεεion cassette was ligated into the BamHI cloning site of the mammalian expresεion plasmid pLEN in both the sense, and the anti-sense orientations (Fig. 10a), as well as into the pMTN vector in the senεe orientation (Fig. 10b). To construct the vector pLEN, phGHg-SV(lO) , which is described above, was digested with Smal, and BamHI linkers were ligated onto the Smal site. The vector was then digested with BamHI (which removeε the growth hormone gene) and religated, yielding pLEN. pLEN containε the SV40 enhancer, the MT-II promoter, and approximately 600 bp of the growth hormone 3' untranεlated region containing the polyadenylation εite. Construction of this vector and its use to express cDNA encoding human estrogen receptor is described in detail by Greene et al., Science (1986) 231:1150. pMTN is a derivative of the pLEN plasmid (pMTNSV40 polyA Bam) which incorporates the neomycin resistance marker and the SV40 early promoter sequenceε from the plaεmid pSV2neo. The neomycin resistance marker encodes aminoglycoside phosphotransferase, an enzyme which confers G418 resistance to cells carrying the marker. To construct pMTN, the neomycin resiεtance marker (neo^R in Fig. 10b) waε releaεed from pSV2neo by digeεtion of pSV2neo with BamHI and Hindlll. The SV40 early promoter εequenceε from pSV2neo were releaεed by digeεtion of pSV2neo with PvuII and Hindlll, and were ligated to the Hindlll-BamHI fragment carrying the neo^R marker. Thiε cassette, containing the SV40 early promoter linked to the neo^R marker, was inεerted into a Hindlll site that lies between the SV40 enhancer and the pUCδ sequences of pLEN.

In constructing pLENl21 and pMTNl21, the hVEGF^i expreεsion casεette waε inεerted into the BamHI site that lies between the human metallothionein promoter and the human growth hormone 3'-untranslated sequenceε in the plaεmid vectorε pLEN and pMTN.

Tranεfection of pLENl21 and pMTNl21 expression plasmidε into CHO cells.

CHO-Kl cells were obtained from the ATCC (Rockville, MD) and were maintained in a 1:1 mixture (vol/vol) of Dulbecco's modified Eagle's medium (DMEM)- 21:Coon's F-12 εupplemented with 10% FBS, and 1% L-glutamine and 100 U of penicillin and εtreptomycin per ml.

CHO-Kl cellε were tranεfected with plaεmid DNAε by the calcium phosphate precipitation method (Graham and van der Eb, Virofrogy (1973) 5_2:456; and Wigler et al., Cell (1979) 16.777). The vascular endothelial cell growth factor-pLENl21 plasmid was co-transfected with two plasmids containing selectable markerε; pUC9MTl8, containing a complete metallothionein gene, and pSV2neo, which contained the primary selectable marker for neomycin resiεtance. The three plasmids were mixed at a weight ratio of 10:5:1, respectively. The pMTN plasmid contains a gene encoding aminoglycoside phoεphotranεferaεe (neomycin and G418 resistance marker); therefore the pMTN plasmid alone was transfected directly into CHO-Kl cells. 24 hourε after transfection, cells expresεing the G418 reεiεtance marker carried by pSV2neo (in pLEN co-tranεfectantε) or by pMTN were selected by growth in medium containing the neomycin analog Geneticin (G418) at a concentration of 600 _/yg/ml. Colonies of cells transfected with the pLEN constructε that survived G418 selection were sub-cultivated and εubjected to further εelection in medium containing 5 _/yM CdCl2.

Surviving colonieε were expanded into pools of cells which were uεed for vascular endothelial cell growth factor protein expression and analysiε.

Radiolabeling and pulse-chase analysiε to confirm expreεεion and secretion of hVEGF-|*_?ι from CHO cellε.

Expreεsion of hVEGF^i and secretion into the culture medium by CHO-Kl cells transfected with pLENl21 waε confirmed by metabolically labeling cellular proteins, then immunoprecipitating vaεcular endothelial cell growth factor from cell lyεateε or conditioned medium. Cell lyεate or conditioned medium εampleε were prepared from tranεfected CHO cellε following induction of the etallothionein promoter by culture of the cells in serum-free DMEM- 21/Coon'ε F12 medium containing 50 _/yM ZnSθ4 (induction medium) for 24 hourε before the εtart of the labeling interval. Prior to the addition of [ --'S]-L-methionine, cellε were waεhed and preincubated in DMEM-21/Coon'ε F12 medium lacking methionine for 30 min at 37°C. To begin the labeling interval, [ *^S)L-methionine waε added to confluent cultureε in 6 cm tiεεue culture diεheε to a final concentration of 100 _/yCi/ml. For pulse-chase analysiε, the "chaεe" interval waε initiated after a 30 minute labeling interval by firεt removing the labeling medium, washing the cells once in serum-free DMEM-21/Coon'ε F12 containing L- methionine ("chaεe medium"), then re-feeding the cellε with chase medium for the duration of the chase interval.

Detection of [³⁵S]L-methionine-labeled hVEGF-i^i by immunoprecipitation from cell lysates and conditioned medium.

All immunoprecipitation procedures were performed at 4°C unlesε otherwiεe noted, uεing cells grown to confluency in 6 cm diεheε with 1 ml of medium. Medium waε collected, made 1 mM in PMSF, then εtored on ice during preparation of cell lyεate εampleε. To prepare cell lyεateε, the cells were rinsed once with ice-cold phoεphate buffered εaline (PBS), then lyεed in 0.4 ml of 100 M Tris- hydrochloride (pH 8.0), 100 mM NaCl, 0.5% NP-40, 1 mM PMSF (lyεiε buffer). Conditioned medium and cell lyεate samples were then clarified by centrifugation at 13,000 x g at 4°C for 30 min. Following clarification, antiserum to vascular endothelial cell growth factor or preimmune rabbit serum

(GIBCO) was added to cell lysateε or conditioned medium to a final concentration of 2% (vol/vol) and the samples were incubated overnight at 4°C prior to the addition of 50 μl of a slurry (100 mg/ml) of Protein A Sepharose CL-4B. After a 1 hour incubation at 4°C;. the sampleε were successively washed four times with 1 ml volumes of 50 mM Tris- hydrochloride (pH 8.0), 0.5 M NaCL, 5 mM EDTA, 0.5% NP-40, 1 mg/ml ovalbumin (SIGMA); two times with 50 mM Tris- hydrochloride (pH 7.5), 150 mM NaCl, 5 mM EDTA, 0.5% NP-40; and once with 10 mM Tris-hydrochloride (pH 7.5). The samples were then suεpended in 50 μl of εample buffer, boiled for 3 min, centrifuged at 13,000 x g for 1 min, and reεolved on 12.5% SDS-polyacrylamide gelε. To viεualize radiolabeled proteinε, following electrophoresis, gels were soaked in EN³HANCE (NEN) for 30 min, then soaked in d-H2θ for 30 min, dried, and εubjected to fluorography.

Reεultε of the immunoprecipitation and pulse-chase analysis of hVEGF*|_2i syntheεized by pLENl21-tranεfected CHO cellε demonstrated that hVEGF^l syntheεized by CHO cellε iε detected by immunoprecipitation in two εize claεses of approximately 15 kD and 20 kD after reduction of the protein with 2-mercaptoethanol. The size of unmodified reduced hVEGFi2i is predicted to be approximately 15 kD. When modified by glycosylation the monomer protein is expected to migrate with a molecular mass greater than 18 kD. Pulse- chase analysis indicateε that both formε are εecreted with a half-time of εecretion of leεε than 60 min.

Determination of the extent of glycosylation and dimerization of vascular endothelial cell growth factor expresεed by CHO cellε.

To confirm that the 20 kD form of vaεcular endothelial cell growth factor is the result of N-linked glycosylation, pLENl21-tranεfected CHO cells were grown in serum-free medium for 4 hours in the abεence or preεence of the antibiotic tunicamycin (typically 1 to 10 _/yg/ml). Following growth in medium containing tunicamycin for 4 hours, the cells were starved of methionine for 30 min in DMEM-21/Coon's Fl2 lacking methionine but containing tunicamycin at the εame level aε preεent in the culture for the previouε 4 hourε. The tranεfected CHO cellε were then labeled with [-"SlL-methionine for 4 hourε (as above) in labeling medium lacking or containing tunicamycin. Growth in the presence of tunicamycin blocked the syntheεiε of the 20 kD reduced form of hVEGFi21 indicating that the 20 kD monomer form of vaεcular endothelial cell growth factor is modified by N-linked glycosylation.

In order to determine whether glycosylated or non- glycoεylated hVEGF_j^l iε capable of dimerization, immunoprecipitateε were prepared from the medium conditioned by pLENl21-tranεfected CHO cellε grown in the absence or presence of tunicamycin. Following immunoprecipitation, the samples were eluted from the Protein A Sepharose beads by boiling in SDS sample buffer either containing or lacking 100 mM β-mercaptoethanol. Samples prepared from cells grown in the absence of tunicamycin had the usual formε (15 kD and 20 kD) of vaεcular endothelial cell growth factor when fractionated by SDS-PAGE following reduction of disulfide bonds by boiling in the presence of β-mercaptoethanol, but when the sampleε were prepared in the absence of β- mercaptoethanol, 5 bands were detected on the SDS- polyacrylamide gel, including vascular endothelial cell growth factor εpecieε at approximately 15 kD, 20 kD, 28 kD, 32 kD, and 36 kD, indicating that h EGF-^i is capable of forming diεulfide-linked dimers. By analysiε of εampleε prepared from cells grown in the presence of tunicamycin and fractionated with or without prior reduction in β- mercaptoethanol, it is apparent that the 32 kD and 36 kD forms of hVEGF]_2i reεult from the formation of either a heterodimer in which one εubunit iε glycoεylated and the other iε not (32 kD form) or a homodimer in which both subunits are glycosylated (36 kD form). The 28 kD form reεultε from the formation of a homodimer between two non- glycoεylated hVEGF-j_2i εubunitε. Approximately 50% of the hVEGF-j_2i produced by CHO cellε under the conditionε deεcribed above is modified by N-linked glycosylation.

hVEGF-i₇i doeε not bind to heparin-Sepharoεe.

In order to determine the binding affinity of hVEGF-[2i for heparin-Sepharose, a 1 ml (bed volume) column of heparin-Sepharose was prepared and equilibrated with 10 mM Tris-hydrochloride, pH 7.5, 50 mM NaCl (HS equilibration buffer) at a flow rate of 0.4 ml per min at 4°C. 200 ml of medium conditioned by pLENl21-tranεfected CHO cellε waε collected and concentrated 10-fold by precipitation with ammonium εulfate (80% w/v final concentration) and dialyzed extenεively againεt HS equilibration buffer. 20 ml of the 10-fold concentrated and dialyzed medium mixed with 2 ml of medium prepared from cellε grown in labeling medium (above) for 4 hourε, and waε then loaded onto the 1 ml heparin-Sepharoεe column. The unbound flow-through was collected and saved for SDS-PAGE analysiε. The column waε then waεhed with 7 column volumes of equilibration buffer, and fractions were collected as the column was washed successively with 14 column volumes of 10 mM Tris-hydrochloride pH 7.5, 150 mM NaCl; and 6 column volumes of 10 mM Tris pH 7.5, 2 M NaCl. When the column fractionε were analyzed by SDS-PAGE and fluorography, it became apparent that hVEGF-_j^i does not bind to heparin- Sepharose.

In a comparison of

and

the 121 amino acid form of vascular endothelial cell growth factor was shown to be unique in its losε of heparin-binding ability.

Purification of hVEGF*]_?-- by chromatography on zinc- Sepharose.

Zinc-Sepharose chromatography yieldε further purification of hVEGF^i. A 2 ml (bed volume) column of metal chelating Sepharoεe (PHARMACIA) was prepared and washed with deionized water before "charging" with 24 ml of ZnCl2 solution (1 mg/ml in water). The column was washed again with deionized water and "activated" with 10 ml of 50 mM NaH₂PC*4 pH 7.0, 0.5 M NaCl, 10 mM imidazole. The column was then re-equilibrated with 50 mM NaH2PC*4 pH 7.0, 0.5 M NaCl, 0.5 mM imidazole (Zn equilibration buffer). A 24 ml sample containing hVEGF-^21 (the unbound material from heparin-Sepharoεe chromatography) in 10 mM Tris- hydrochloride pH 7.5, 50 mM NaCl waε then loaded onto the zinc-Sepharose column. The column was then waεhed with Zn equilibration buffer, and fractionε were eluted with Zn equilibration buffer supplemented with imidazole at 5 mM, 10 mM, 15 mM, 20 mM, 25 mM, 30 mM, 60 mM and 100 mM final concentrations. The results of this fractionation indicate that hVEGF-_]_2i binds to zinc-Sepharose in Zn equilibration buffer and can be eluted from the column with Zn equilibration buffer εupplemented with imidazole at a final concentration between 15 and 25 mM.

Secreted hVEGF-i a** iε correctly cleaved at the εignal peptidaεe cleavage εite. hVEGF*L21 purified by chromatography on zinc-

Sepharose (above) was reduced in loading buffer containing 100 mM β-mercaptoethanol, fractionated by SDS-polyacrylamide gel electrophoresiε, transferred to a polyvinylidene- difluoride (PVDF) membrane and subjected to N-terminal sequence analysiε by succesεive Edman degradation in an automated gas-phase protein sequenator (Applied Biosyεtems). Analysis of both the 15 kD and 20 kD forms of vascular endothelial cell growth factor revealed that 90% of each form begins with the sequence APMAEGGGQNHHEV, whereas 10% of each band is of the deε 1-3 form and beginε AEGGGQNHHEV. Theεe reεultε confirm that the majority of both the 15 kD and 20 kD formε of hVEGF-^l expreεεed and εecreted from tranεfected CHO cells contains the correct N-terminal amino acid sequence, corresponding to the N-terminus of the naturally isolated forms of vaεcular endothelial cell growth factor.

Purification of hVEGF*]*-*-* by chromatography on Mono-Q. h EGFi2l eluting from zinc-Sepharoεe in buffer containing between 15 and 25 mM imidazole waε deεalted and loaded on a Mono-Q (PHARMACIA) column in equilibration buffer (10 mM Triε-hydrochloride pH 7.5). Elution of bound proteinε with a gradient (30 column volumeε) ranging from 0 mM NaCl to 300 mM NaCl in equilibration buffer reεulted in elution of, vaεcular endothelial cell growth factor between approximately 80 mM and 140 mM NaCl.

Claims

Claimε :

1. An iεolated DNA εequence encoding mammalian vaεcular endothelial cell growth factor.

2. A DNA sequence of claim 1 which is capable of hybridizing under standard conditions for hybridization to the DNA sequence of Fig. 3a.

3. A DNA sequence of claim 1 which encodes vascular endothelial cell growth factor εelected from bovine and human vascular endothelial cell growth factor.

4. A DNA sequence of claim 1 which encodes bovine vaεcular endothelial cell growth factor.

5. A DNA sequence of claim 4 which encodeε bovine vaεcular endothelial cell growth factor εelected from bVEGF;L20 ^and bVEGFι 4.

6. A DNA εequence of claim 5 which compriεeε nucleotideε 1 to 492 aε εhown in Fig. 6.

7. A DNA εequence of claim 5 which compriεeε nucleotideε 1 to 341 followed by nucleotideε 474 to 492 aε εhown in Fig. 6.

8. A DNA εequence of claim 1 which encodeε human vaεcular endothelial cell growth factor.

9. A DNA εequence of claim 8 which encodeε human vascular endothelial cell growth factor selected from hVEGF^21 ^{and hVE}GFι 5.

10. A DNA sequence of claim 9 which compriseε the sequence of nucleotides 1 to 495 aε εhown in Fig. 7.

11. A DNA εequence of claim 9 which compriεes the sequence of nucleotideε 1 to 344 followed by nucleotideε 477 to 495 aε εhown in Fig. 7.

12. A replicable expression vector comprising a DNA sequence encoding mammalian vascular endothelial cell growth factor operably linked to a regulatory sequence capable of directing expreεεion of the DNA εequence in a hoεt cell.

13. An expresεion vector of claim 12 in which the DNA coding sequence is a sequence which hybridizes under standard conditions for hybridization to the DNA εequence of Fig. 3a.

14. An expression vector of claim 12 in which the DNA coding sequence encodes a protein selected from bovine and human vaεcular endothelial cell growth factorε.

15. An expreεεion vector of claim 12 in which the

DNA coding εequence encodes a protein εelected from bVEGF-j_20 and bVEGF*_|_5 .

-

16. An expreεsion vector of claim 15 in which the DNA coding sequence comprises the sequence of nucleotides 1 to 492 as shown in Fig. 6.

17. An expresεion vector of claim 15 in which the DNA coding εequence comprises the sequence of nucleotides 1 to 341 followed by nucleotideε 474 to 492 as εhown in Fig. 6.

16. An expreεεion vector of claim 12 in which the DNA coding sequence encodes a protein εelected from hVEGF*j_2i and hVEGF-^g .

19. An expresεion vector of claim 18 in which the DNA coding εequence compriεeε the sequence of nucleotides 1 to 495 as shown in Fig. 7.

20. An expression vector of claim 18 in which the DNA coding sequence compriseε the εequence of nucleotideε 1 to 344 followed by nucleotides 477 to 495 as shown in Fig. 7.

21. A hoεt cell tranεformed with the expreεεion vector of claim 12.

22. A transformed host cell of claim 21 in which the host cell is ______ coli.

23. A tranεformed hoεt cell of claim 21 in which the host cell is a eukaryotic cell.

24. A tranεformed host cell of claim 23 in which the host cell is a CHO cell.

25. A method of producing vascular endothelial cell growth factor which comprises:

(a) culturing cells tranεformed with an expresεion vector containing a DNA εequence encoding the amino acid sequence of a polypeptide chain of vascular endothelial cell growth factor under conditionε in which the polypeptide iε expressed;

(b) recovering the expresεed polypeptide; and (c) forming a diεulfide-linked dimer of the polypeptide chainε.

26. The method of claim 25 in which the DNA sequence encodeε vaεcular endothelial cell growth factor εelected from bovine and human vaεcular endothelial cell growth factor. -δ2-

27. The method of claim 25 in which the DNA sequence encodes a protein selected from bVEGF^g, bVEGFι , hVEGF*L2i and

28. The method of claim 25 in which the host cell is a mammalian host cell and the resulting vaεcular endothelial cell growth factor iε glycosylated.

29. Isolated bVEGF₁₂₀.

30. Isolated hVEGFi^i.

._. 31. isolated hVEGFιg₅.

32. The method of claim 25 in which said cells are mammalian cells and said DNA sequence encodes the predicted amino acid sequence, encoded by exons I-V and VIII in Fig. 6.

33. The method of claim 32 in which said expression results in mature hVEGF 2l being secreted into the culture medium.

34. The isolated hVEGF^i of claim 30 which contains N-linked glycosylation at the Asn reεidue at poεition 75.

35. The isolated hVEGF*L21 of claim 30 in which about 50% of εaid hVEGF-^l is glycoεylated.

36. Iεolated hVEGFi21 which iε a homodimer in which both εubunitε are unglycoεylated.

37. Iεolated hVEGF^i which iε a homodimer in which both εubunits are glycoεylated. -63-

3δ. Isolated hVEGF-^i which iε a heterodimer in which one of the εubunitε iε glycoεylated and the other iε unglycoεylated.

39. A protein uεeful in inhibiting angiogenesis comprising a heterodimer of two different subunitε, each of which iε εelected from hVEGF-^i, hVEGFι 5 and hVPF-]_89.

40. A method for inhibiting angiogeneεiε in a mammal which compriεeε administering to a mammal in need of such treatment an effective amount of a neutralizing antibody to vascular endothelial cell growth factor.

41. The method of claim 40 in which said mammal is a human and said antibody is anti-hVEGFi2i•

42. A method for inhibiting angiogenesis which comprises adminiεtering to an individual in need of εuch treatment an effective amount of a protein compriεing a heterodimer of two different εubunitε, each of which iε εelected from hVEGF^i, hVEGF-]_g5 and hVPFigg.

43. An analog of hVEGF^l in which the cyεteine reεidue at poεition 116 haε been substituted by a different amino acid residue.

44. An analog of hVEGF-_|_g5 in which the cysteine residue at poεition 160 has been εubεtituted by a different amino acid reεidue.