AU2004206993B2 - High affinity oligonucleotide ligands to growth factors - Google Patents

Description

AUSTRALIA

PATENTS ACT 1990 DIVISIONAL and ADDITIONAL APPLICATION NAME OF APPLICANT: Gilead Sciences, Inc.

ADDRESS FOR SERVICE: DAVIES COLLISON CAVE Patent Attorneys 1 Nicholson Street Melbourne, 3000 INVENTION TITLE: "High affinity oligonucleotide ligands to growth factors" The following statement is a full description of this invention, including the best method of performing it known to us: Q:\OPER\JEH\2004\Sept\l2494700 245.doc 1/9/04 HIGH AFFINITY OLIGONUCLEOTIDE LIGANDS TO GROWTH FACTORS FIELD OF THE INVENTION Described herein are methods for identifying and preparing high-affinity nucleic acid ligands to TGFB, PDGF, and hKGF. The method utilized herein for identifying such nucleic acid ligands is called SELEX, an acronym for Systematic Evolution of Ligands by EXponential enrichment. This invention includes high affinity nucleic acid ligands ofTGFI, PDGF, and hKGF. Further disclosed are RNA and DNA ligands to TGF31 and PDGF and RNA ligands to hKGF. Also included are oligonucleotides containing nucleotide derivatives chemically modified at the positions ofpyrimidines. Additionally disclosed are RNA ligands to TGFB1 and hKGF containing 2'-NH 2 -modifications or 2-F-modifications and RNA ligands to PDGF containing 2'-F modifications. This invention also includes high affinity nucleic acid inhibitors of TGF 1, PDGF, and hKGF. The oligonucleotides of the present invention are useful as pharmaceuticals or diagnostic agents.

BACKGROUND OF THE INVENTION

TGFO

The transforming growth factor -B (TGFB) polypeptides influence growth, differentiation, and gene expression in many cell types. The first polypeptide of this family that was characterized, TGFB1 has two identical 112 amino acid subunits which are covalently linked. TGFB1 is a highly conserved protein with only a single amino acid difference distinguishing human from mice forms.

There are two other members of the TGFB gene family that are expressed in mammals. TGFB2 is 71% homologous to TGFBI(de Martin et al., (1987) EMBQ 1. f:3673-3677), whereas TGFB3 is 80% homologous to TGFBI(Derynck et al., (1988) EMBO J 2:37 3 7 37 4 3 The structural characteristics of TGFB1 as determined by nuclear magnetic resonance (Archer et al., (1993) Biochemistry 32:1164-1171) agree with the crystal structure of TGFB2 (Daopin et al., (1992) Science 257:369-374; Schlunegger and Grutter (1992) Nature 358;430-434).

Even though the TGFB's have similar three dimensional structures, they are by no means physiologically equivalent. There are at least three different extracellular receptors, type I, II and III, involved in transmembrane signaling of TGFB to cells carrying the receptors. For reviews, see Derynck (1994) TIBS 19.548-553 and Massague (1990) Annu. Rev. Cell Biol 6:597-641. In order for TGFB2 to effectively interact with the type II TGFB receptor, the type III receptor must also be present (Derynck (1994) TIBS 19i548-553). Vascular endothelial cells lack the type III receptor. Instead endothelial cells express a structurally related protein called endoglin (Cheifetz et al., (1992) J. Biol. Chem.

262:19027-19030), which only binds TGFB1 and TGFB3 with high affinity. Thus, the relative potency of the TGFB's reflect the type of receptors expressed in a cell and organ system.

In addition to the regulation of the components in the multifactorial signaling pathway, the distribution of the synthesis ofTGFB polypeptides also affects physiological function. The distribution of TGFB2 and TGFB3 is more limited (Derynck et al., (1988) EMBO J 2:3737-3743) than TGFB1, e.g.,TGFB3 is limited to tissues of mesenchymal origin, whereas TGFB1 is present in both mesenchymal and epithelial cells.

TGFB1 is a multifunctional cytokine critical for tissue repair. High concentrations of TGF31 are delivered to the site of injury by platelet granules (Assoian and Sporn, (1986) J Cell Biol. 102 1217-1223.). TGFB1 initiates a series of events that promote healing including chemotaxis of cells such as leukocytes, monocytes and fibroblasts, and regulation of growth factors and cytokines involved in angiogenesis, cell division associated with tissue repair and inflammatory responses. TGF1 also stimulates the synthesis of extracellular matrix components (Roberts et al., (1986) Proc. Natl. Acad Sci USA 83:4167-4171; Spom et al., (1983) Science 219:1329-1330; Massague, (1987) Cell 49:437-438) and most importantly for understanding the pathophysiology of TGFB 1, TGFBI autoregulates its own synthesis (Kim et al., (1989) J Biol Chem 264:7041-7045).

A number of diseases have been associated with TGF11 overproduction.

Fibrotic diseases associated with TGFB1 overproduction can be divided into chronic conditions such as fibrosis of kidney, lung and liver and more acute conditions such as dermal scarring and restenosis. Synthesis and secretion of TGFB1 by tumor cells can also lead to immune suppression such as seen in patients with aggressive brain or breast tumors (Arteaga et al., (1993) J Cln Invest 22: 2569-2576). The course of Leischmanial infection in mice is drastically altered by TGFB1 (Barral-Netto et al., (1992) Science 252:545-5 4 7 TGFI1 exacerbated the disease, whereas TGFB1 antibodies halted the progression of the disease in genetically susceptible mice. Genetically resistant mice became susceptible to Leshmanial infection upon administration of TGFB1.

The profound effects of TGFB1 on extracellular matrix deposition have been reviewed (Rocco and Ziyadeh, (1991) in Contemporary Issues in Nephrology v23, Hormones, autocoids and the kidney. ed. Jay Stein, Churchill Livingston, NY pp391-410; Roberts et al., (1988) Rec. Prog. Hormone Res. 44:157-197) and include the stimulation of the synthesis and the inhibition of degradation of extracellular matrix components. Since the structure and filtration properties of the glomerulus are largely determined by the extracellular matrix composition of the mesangium and glomerular membrane, it is not surprising that TGFB1 has profound effects on the kidney. The accumulation of mesangial matrix in proliferative glomerulonephritus (Border et al., (1990) Kidney nt. 32:689-695) and diabetic nephropathy (Mauer et al., (1984) J. Clin Inest.24: 1143-1155) are clear and dominant pathological features of the diseases. TGFI1 levels are elevated in human diabetic glomerulosclerosis (advanced neuropathy) (Yamamoto et al., (1993) Proc. Natl. Acad. Sci. 9Q:1814-1818). TGFBI is an important mediator in the genesis of renal fibrosis in a number of animal models (Phan et al., (1990) Kidney Int. 32:426; Okuda et al., (1990) J. linInvest.

86:453). Suppression of experimentally induced glomerulonephritus in rats has been demonstrated by antiserum against TGFB1 (Border et al., (1990) Nature 346:371) and by an extracellular matrix protein, decorin, which can bind TGFB1 (Border et al., (1992) Nature 160:361-363).

Too much TGFI31 leads to dermal scar-tissue formation. Neutralizing TGFB1 antibodies injected into the margins of healing wounds in rats have been shown to inhibit scarring without interfering with the rate of wound healing or the tensile strength of the wound (Shah et al., (1992) Lancet 332:213-214). At the same time there was reduced angiogenesis, reduced number of macrophages and monocytes in the wound, and a reduced amount of disorganized collagen fiber deposition in the scar tissue.

TGFB1 may be a factor in the progressive thickening of the arterial wall which results from the proliferation of smooth muscle cells and deposition of extracellular matrix in the artery after balloon angioplasty. The diameter of the restenosed artery may be reduced 90% by this thickening, and since most of the reduction in diameter is due to extracellular matrix rather than smooth muscle cell bodies, it may be possible to open these vessels to 50% simply by reducing extensive extracellular matrix deposition. In uninjured pig arteries transfected in vivo with a TGF81 gene, TGFB1 gene expression was associated with both extracellular matrix synthesis and hyperplasia (Nabel et al., (1993) Proc. Natl.

Acad. Sci USA 9010759-10763). The TGFB 1 induced hyperplasia was not as extensive as that induced with PDGF-BB, but the extracellular matrix was more extensive with TGFB1 transfectants. No extracellular matrix deposition was associated with FGF-1 (a secreted form of FGF) induced hyperplasia in this gene transfer pig model (Nabel (1993) Nature 362:844-846).

There are several types of cancer where TGFI1 produced by the turnor may be deleterious. MATLyLu rat cancer cells (Steiner and Barrack, (1992) Mol.

Endocrinol. 6:15-25) and MCF-7 human breast cancer cells (Arteaga et al., (1993) Cell Growth and Differ. 4:193-201) became more tumorigenic and metastatic after transfection with a vector expressing the mouse TGFI1. In breast cancer, poor prognosis is associated with elevated TGFB (Dickson et al., (1987) Po. Natl.

Acad Sci. USA 84:837-841; Kasid etal., (1987) Cancer Res. 47:5733-5738; Daly e al., (1990) J Cel Biochem 41:199-211; Barrett-Lee et (1990) Br. J Cancer 61:612-617; King el al., (1989) Steroid Biochem 34:133-138; Welch et al., (1990) Proc Natl. Acad Sci. 28:7678-7682; Walker et al., (1992)urJ Cancer 238: 641-644) and induction of TGFB1 by tamoxifen treatment (Butta et al., (1992) Canr Res 52:4261-4264) has been associated with failure of tamoxifen treatment for breast cancer (Thompson et al., (1991) Br J Cancer :60 9 6 1 4 Anti TGFB 1 antibodies inhibit the growth of MDA-231 human breast cancer cells in athymic mice (Arteaga et al., (1993) JClin Invest 22 2569-2576), a treatment which is correlated with an increase in spleen natural killer cell activity. CHO cells transfected with latent TGF81 also showed decreased NK activity and increased tumor growth in nude mice (Wallick et al., (1990) J Exp Med 122:1777-1784). Thus, TGFB1 secreted by breast tumors may cause an endocrine immune suppression.

High plasma concentrations of TGFB 1 have been shown to indicate poor prognosis for advanced breast cancer patients (Anscher et al.. (1993) INEn Med 328:1592-8). Patients with high circulating TGFB before high dose chemotherapy and autologous bone marrow transplantation are at high risk for hepatic veno-occlusive disease (15-50% of all patients with a mortality rate up to and idiopathic interstitial pneumonitis (40-60% of all patients). The implication of these findings is 1) that elevated plasma levels of TGFBI can be used to identify at risk patients and 2) that reduction of TGFB1 could decrease the morbidity and mortality of these common treatments for breast cancer patients.

PDGF

Platelet-derived growth factor (PDGF) was originally isolated from platelet lysates and identified as the major growth-promoting activity present in serum but not in plasma. Two homologous PDGF isoforms have been identified, PDGF A and B, which are encoded by separate genes (on chromosomes 7 and 22).

The most abundant species from platelets is the AB heterodimer, although all three possible dimers (AA, AB and BB) occur naturally. Following translation, PDGF dimers are processed into =30 kDa secreted proteins. Two cell surface proteins that bind PDGF with high affinity have been identified, a and B (Heldin et al., Proc. Natl. Acad. Sci.. 2: 3664 (1981); Williams et al., Proc. Natl. Acad. Sci., 22: 5867 (1981)). Both species contain five immunoglobulin-like extracellular domains, a single transmembrane domain and an intracellular tyrosine kinase domain separated by a kinase insert domain. The functional high affinity receptor is a dimer and engagement of the extracellular domain of the receptor by PDGF results in cross-phosphorylation (one receptor tyrosine kinase phosphorylates the other in the dimer) of several tyrosine residues. Receptor phosphorylation leads to a cascade of events that results in the transduction of the mitogenic or chemotactic signal to the nucleus. For example, in the intracellular domain of the PDGF B receptor, nine tyrosine residues have been identified that when phosphorylated interact with different src-homology 2 (SH2) domain-containing proteins including phospholipase C-g, phosphatidylinositol 3'-kinase, GTPase-activating protein and several adapter molecules like She, Grb2 and Nck (Heldin, Cell, S0: 213 (1995)). In the last several years, the specificities of the three PDGF isoforms for the three receptor dimers (aa, aB, and BB) has been elucidated. The a-receptor homodimer binds all three PDGF isoforms with high affinity, the B-receptor homodimer binds only PDGF BB with high affinity and PDGF AB with approximately 10-fold lower affinity, and the aB-receptor heterodimer binds PDGF BB and PDGF AB with high affinity (Westermark Heldin, Acta Oncologica, 22: 101 (1993)). The specificity pattern results from the ability of the A-chain to bind only to the a-receptor and of the B-chain to bind to both a and B-receptor subunits with high affinity.

The earliest indication that PDGF expression is linked to malignant transformation came with the finding that the amino acid sequence of the PDGF-B chain is virtually identical to that of p28s', the transforming protein of the simian sarcoma virus (SSV) (Waterfield et al. Nature, 34: 35 (1983); Johnsson et al., EMBO 3: 921 (1984)). The transforming potential of the PDGF-B chain gene and, to a lesser extent, the PDGF-A gene was demonstrated soon thereafter (Clarke et al., Nature 308: 464 (1984); Gazit et al., Cll, 12: 89 (1984); Beckmann et al., Science, 241: 1346; Bywater et al., Mol Cell. Biol 2753 (1988)). Many tumor cell lines have since been shown to produce and secrete PDGF, some of which also express PDGF receptors (Raines et al., Peptide Growth Factors and Their Receptors, Springer-Verlag, Part I, p 173 (1990)). Paracrine and, in some cell lines, autocrine growth stimulation by PDGF is therefore possible. For example, analysis of biopsies from human gliomas has revealed the existence of two autocrine loops: PDGF-B/8-receptor in tumor-associated endothelial cells and PDGF-A/a-receptor in tumor cells (Hermansson et al., Proc.

Natl. Acad. Sci., 7748 (1988); Hermansson et al., Cancer Res., 2: 3213 (1992)). The progression to high grade glioma was accompanied by the increase in expression of PDGF-B and the B-receptor in tumor-associated endothelial cells and PDGF-A in glioma cells. Increased expression of PDGF and/or PDGF receptors has also been observed in other malignancies including fibrosarcoma (Smits et al., Am. J. Pathol., 140: 639 (1992)) and thyroid carcinoma (Heldin et al., Endocrinology, 122: 2187 (1991)).

In view of its importance in proliferative disease states, antagonists of PDGF may find useful clinical applications. Currently, antibodies to PDGF (Johnsson et al., (1985) Proc. Natl. Acad. Sci.. U. S. A. 82: 1721-1725; Ferns et al., (1991) Science 153: 1129-1132; Herrenet al., (1993) Biochimica et Biophysica Acta 1173, 194-302) and the soluble PDGF receptors (Herrenet al., (1993) Biochimica et Biophysica Acta 1173: 194-302; Duanet al., (1991) J. Biol.

Chem. 266: 413-418; Teisman et al., (1993) J. Biol. Chem. 268: 9621-9628) are the most potent and specific antagonists of PDGF. Neutralizing antibodies to PDGF have been shown to revert the SSV-transformed phenotype (Johnsson et al., (1985) Proc. Natl. Acad. Sci.. U. S. A. 82: 1721-1725) and to inhibit the .development of neointimal lesions following arterial injury (Ferns et al., (1991) Science 153: 1129-1132). Other inhibitors of PDGF such as suramin (Williams et al., (1984) J. Biol. Chem. 25: 5287-5294; Betsholtz et al., (1984) CilL 39 447-457), neomycin (Vassbotn et al., (1992) J. Biol. Chem. 262 15635-15641) and peptides derived from the PDGF amino acid sequence (Engstr6m et al., 1992) J. Biol. Chem. 267: 16581-16587) have been reported, however, they are either too toxic or lack sufficient specificity or potency to be good drug candidates.

Other types of antagonists of possible clinical utility are molecules that selectively inhibit the PDGF receptor tyrosine kinase (Buchdunger et al., (1995) Proc. Natl.

Acad. Sci.. U. S. A. 22: 2558-2562; Kovalenk et (1994) Cancer Res- 4: 6106-6114).

hKGF a) Biochemical properties of IZKGF Human Keratinocyte Growth Factor (hKGF) is a small (26-28KD) basic heparin-binding growth factor and a member of the FGF family. hKGF is a relatively newly identified molecule, which is also known as FGF-7 (Finch et al., (1989) Scec 244:752-755). It is a growth factor specific for epithelial cells (Rubin et al., (1989) Proc Natl Acad Sci USA 1U:802-806), and its main function is in developmentlmorphogenesis (Werner et al., (1994) Sience 2.6&:8l9-822) and in wound healing (Werner et (1992) Proc NatI Acad Sci USA U:6896-6900). The major in vivo source of hKGF is stromal fibroblasts (Finch et al., (1989) Scec 244:752-755). Microvascular endothelial cells (Smola et al., (1993) J Ce1ll iQ M12:417-429) and very recently, activated intraepithelial gd T cells (Boismenu et (1994) Science 2kj:1253-1255) have also been shown to synthesize hKGF. hKGF expression is stimul ated in wounds (Werner et al., (1992) Proc Natl Acad Sci USA B6896-6900). Several cytokines are shown to be hKGF inducers (Brauchle et al., (1994) QncQogene 2:3199-3204), with IL- I the ,most potent one (Brauchie et al., (1994) Onogn 2:3199-3204; Chedid et al., (1994)J LBiolChem2M:10753-10757). Unlike bFGF, hKGF has a signal peptide and thus is secreted by producing cells (Finch et al., (1989) Science244:752-755).

hKGF can be overexpressed in E coli and the recombinant protein (-19-21 KD) is biologically active (Ron et al., (1993) J BiolChem 2M1:2984-2988). The E. coi derived recombinant protein is 10 times more mitogenic than the native protein (Ron et al., (1993) JLBiolChem 2612984-2988). This difference may be due to glycosylation. The native protein has a potential Asn glycosylation site (Ron et al., (1993) J iotLChem 2U&:2984-2988).

The hKGF bioactivity is mediated through a specific cell surface receptor (Mild et al., (1991) Scec 2iL:72-75). The hKGF receptor is a modified FGF receptor resulting from alternative splicing of the C-terminal extracellular region of the FGF-R2 (Mild et al., (1992) Proc Natl Acad Sci USA B2:246-250).

NIH/3T3 cells transfected with the hKGF receptor express high affinity (-200 pM) binding sites for hKGF (Miki et al., (1992) Proc Natl Acad Sci USA 89:246-250). The approximate number of specific binding sites per NIH/3T3 cell is about 500,000 Bottaro and S. Aaronson, personal communication). The hKGF receptor binds hKGF and aFGF with similar affinities, and bFGF with about 20 fold less affinity (Miki el al., (1991) Science 251:72-75; Miki et al., (1992) Proc Natl Acad Sci USA 89:246-250). A variant of the hKGF receptor has been found to be an amplified gene one gene, multiple copies), designated K-SAM, in a human stomach carcinoma cell line (Hattori et al., (1990) Proc natl Acad Sci USA 87: 5983-5987).

Heparin has been reported to be an inhibitor of hKGF bioactivity (Ron et al., (1993) J Biol Chem 2: 2 9 84 2 9 8 8 This is in contrast to the agonistic effect of heparin for aFGF (Spivak-Kroixman et al., (1994) Cell 7:1015-1024).

b) Role of hKGF in human disease The recombinant hKGF molecule has been available only since 1993.

Therefore, there is limited information on the role of hKGF in human disease.

The published literature, however, contains evidence that strongly suggests a role for hKGF in at least two human diseases, namely psoriasis and cancer. hKGF has also been implicated in inflammatory bowel disease Finch, personal communication).

Psoriasis Psoriasis is a skin disorder which can be debilitating (Greaves et al., (1995) N Eng J Medicine 332: 581-588), characterized by hyperproliferation of the epidermis and incomplete differentiation of keratinocytes, together with dermal inflammation (Abel et al., (1994) Scientific American Medicine III-1 to III-18; Greaves et al., (1995) N Eng J Medicine 332:581-588). There is not yet an effective treatment for psoriasis (Anonymous, (1993) Drug Market Development 4:89-101; Abel et al., (1994) Scientific American Medicine III-1 to III-18; Greaves et al., (1995) N Eng J Medicine 332:581-588). Psoriasis occurs in 0.5 to 2.8 percent of the population with the highest incidence in Scandinavia. In the US in 1992, it was estimated that 4-8 million people affected with psoriasis spent about $600 million for various drugs and related therapies, none of which is very effective. Most of the expenditure was made by about 400,000 patients with severe psoriasis spending $1,000-1,500 annually on treatment. There are about 200,000 new cases of psoriasis every year.

The basic cause of the disorder is not known, but it results from a primary or secondary defect in the mechanisms that regulate epidermal keratinocyte cell division (Abel et al., (1994) Scientific American Medicine III-1 to III-18).

Psoriasis responds to steroids and cyclosporine and in that sense is characterized as an immune disease (Abel et al., (1994) Scientific American Medicine III-I to 111-18). Since hKGF is the primary specific growth factor for keratinocytes, its overexpression and deregulation are primary candidates as the cause of keratinocyte hyperproliferation in psoriasis. The demonstration that the immune system is a prime regulator of hKGF release (Boismenu et al., (1994) Science 266: 1253-1255; Brauchle et al., (1994) Oncogene 2: 3199-3204; Chedid et al., (1994) J Biol Chem 26: 10753-10757) strengthens the notion that hKGF deregulation is the cause of psoriasis. Furthermore, application of hKGF in porcine wounds creates a histological appearance resembling psoriasis (Staiano-Coico et al., (1993) J Ex Med J.8:865-878); keratinocyte derived hKGF in transgenic mice causes pathology reminiscent to psoriasis (Guo et al., (1993) EMBOJ 12: 973-986); in situ hybridization experiments demonstrated a moderate and a strong upregulation of hKGF and hKGF receptors respectively in psoriasis Finch, personal communication). In situ hybridization experiments also demonstrated involvement of hKGF in another immune disease namely, inflammatory bowel disease Finch, personal communication).

Cancer It is well established in the literature that deregulation of the expression of growth factors and growth factor hKGF and/or its receptor is expected to be the transformation event in some human cancers. The transforming ability of the hKGF system has been demonstrated in vitro (Miki et al., (1991) Science 251:72-75). In another study, carcinoma cell-lines have been found to express the hKGF receptor and to respond to hKGF but not to aFGF, while sarcoma cell-lines WO 96/38579 11 do not express hKGF receptors and respond to aFGF but not to hKGF (Ishii et al., (1994) Cancer Res 54:518-522).

Gastrointestinal Cancer Several poorly differentiated stomach cancers have an amplified gene, designated K-sam, which is an isoform of the hKGF-receptor (Katoh et al., (1992) Proc Natl Acad Sci USA 89:2960-2964). In vivo administration of hKGF to rats causes proliferation of pancreatic ductal epithelial cell (Yi et al., (1994) Ami Pathol 145:80-85), hepatocytes, and epithelial cells throughout the gastrointestinal tract (Housley Et al., (1994) J Clin Invest 94:1764-1777).

Lung Cancer Administration of hKGF to rats causes type II pneumocyte hyperplasia similar to the bronchoalveolar cell variant of lung carcinoma (Ulich et al., (1994) JClin Invest 91:1298-1306).

Breast Cancer In vivo, hKGF causes mammary duct dilation and rampant epithelial hyperplasia, both of which are common features of breast cancers (Ulich et al., (1994) Am J Pathol 144:862-868; Yi et al., (1994) Am J Pathol 145:1015-102 2 However, the ductal epithelium of breastfeeding rats is resistant to the growth promoting effects of hKGF and this is of interest in regard to epidemiological observations that pregnancy in women decreases susceptibility to breast cancer and that dairy cows almost never develop breast cancer (Kuzma, 1977, Breast in Pathology, Mosby There is additional supporting evidence implicating hKGF in breast cancer. hKGF mRNA has been detected recently in normal human breast tissue and in 12 of 15 breast tumor samples tested (Koos et al., (1993) J Steroid Biochem Molec Biol 45:217-225). The presence of hKGF mRNA in breast tumors considered in conjunction with the observation that hKGF is present in nonneoplastic mammary glands and that hKGF causes rampant proliferation of mammary epithelium suggests that hKGF may be an autocrine or paracrine growth factor important in the regulation of the growth of normal and neoplastic mammary epithelium (Ulich et al., (1994) Am J Pathol 144:862-868).

Infiltrating ductal mammary adenocarcinoma is characteristically enveloped by a desmoplastic stroma that has been postulated to represent a defensive host response to the carcinoma (Ulich et al., (1994) Am J Pathol 144:862-868). Since hKGF is stroma derived it is possible that the desmoplasmic stroma contributes rather than inhibits the growth of the tumor.

Prostate Cancer The growth promoting effect of androgens on prostate tumors appears to be mediated through hKGF (Yan et al., (1992) Mol Endo 6:2123-2128), as androgens induce the expression of hKGF in prostate stroma cells. Prostate tumors that are androgen dependent in vivo, are androgen independent in vitro, but hKGF dependent (Yan et al., (1992) Mol Endo 6:2123-2128). In agreement with the role of hKGF as andromedin is the observation that hKGF functions in epithelial induction during seminal vesicle development, a process that is directed by androgen (Alarid et al., (1994) Proc Natl Acad Sci USA 91:1074-1078).

Furthermore, hKGF causes aberrant activation of the androgen receptor, thus probably contributing to the failure of androgen ablation therapy in prostate cancer (Culig et al., (1994) Cancer Res 54:5474-5478). Based on this information, it is possible that genetic alterations cause hKGF to escape androgen regulation and thus convert the androgen dependent tumor into an androgen independent, highly malignant tumor. Such tumors would still be able to express the androgen regulated marker PSA, as hKGF also causes the aberrant activation of the androgen receptor. It is also likely that hKGF might be responsible for Benign Prostate Hypertrophy (BPH), a common health problem in older men Bottaro, personal comunication).

d) hKGF Competitors To date, a monoclonal antibody and a short hKGF-receptor derived peptide have been described as hKGF competitors (Bottaro et al., (1993) JBiol Chem 268:9180-9183). The monoclonal antibody, designated 1G4, has a Kd of 200pM for hKGF. The short peptide inhibits hKGF binding to the cell surface of NIH/3T3 cells expressing the human receptor with a Ki of about 1-5 AM. Bottaro et al. (WO 94/25057) provide hKGF-receptor peptides which inhibit binding between hKGF and its receptor. Also provided is a method of assaying test compounds for the ability to inhibit hKGF receptor-mediated cell proliferation.

e) Assaying for receptor-growth factor interaction Blocking the interaction of growth factors and lymphokines with their cell surface receptor using antagonists has been an approach for disease treatment.

The discovery of such antagonists requires the availability of biochemical assays for the receptor-growth factor or lymphokine interaction. A classic assay has been the competitive inhibition of radiolabeled growth factor or lymphokine (tracer) to its cell surface receptor. These types of assays utilize cell lines that express the relevant receptor on their surface and determines the amount of cell bound tracer in the presence of various concentrations of potential antagonists. Additionally, other assays utilize membrane extracts from cell lines that express the relevant receptor, and tracer binding is followed by filter binding (see Nenquest Drug Discovery System: Human Tumor Necrosis Factor-Alpha, NEN Research Products, E. I. DePont de Nemours Co. Boston, MA) or by immobilizing the membrane extracts onto solid supports (Urdal et al., (1988) J Biol Chem 263:2870-2877; Smith et al., (1991) Bioch Bioph Res Comm 176:335-342).

Receptor induced electrophoretic mobility shift of tracer has been applied to identify the presence and size of cell surface receptors by crosslinking the receptor to the tracer and then analyzing on denaturing gels (for example see Kull et al., (1985) Proc natl Acad Sci USA 82:5756-5760; Hohmann et al., (1989) LBiol Chem 264:14927-14934; Stauber et al., (1989) J Biol Chem 264:3573-3576).

The use of native gels and non-crosslinked complexes has not been described for growth factors or lymphokines and their receptors, but has been widely applied to study nucleic acid protein interactions (Sambrook et al., (1989) Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY).

Screening of various cancer cell lines for the presence of hKGF receptors by PCR, revealed that all carcinoma cell lines express hKGF receptor mRNA while sarcoma cell lines do not. The presence of mRNA does not necessarily mean that hKGF receptor will be present on the surface of these cells. For hKGF, only cell based assays have been described using Balb/MK keratinocytes (Weissman, (1983) Cell 32: 599-606) or NIH/3T3 cells transfected with the hKGF receptor (Miki, (1992) Proc. Natl. Acad. Sci. USA 89:246-250).

SELEX

A method for the in vitro evolution of nucleic acid molecules with highly specific binding to target molecules has been developed. This method, Systematic Evolution of Ligands by EXponential enrichment, termed SELEX, is described in United States Patent Application Serial No. 07/536,428, entitled "Systematic Evolution of Ligands by Exponential Enrichment," now abandoned, United States Patent Application Serial No. 07/714,131, filed June 10, 1991, entitled "Nucleic Acid Ligands," now U.S. Patent No. 5,475,096, United States Patent Application Serial No. 07/931,473, filed August 17, 1992, entitled "Nucleic Acid Ligands," now United States Patent No. 5,270,163 (see also PCT/US91/04078), each of which is herein specifically incorporated by reference. Each of these applications, collectively referred to herein as the SELEX Patent Applications, describes a fundamentally novel method for making a nucleic acid ligand to any desired target molecule.

The SELEX method involves selection from a mixture of candidate oligonucleotides and step-wise iterations of binding, partitioning and amplification, using the same general selection scheme, to achieve virtually any desired criterion of binding affinity and selectivity. Starting from a mixture of nucleic acids, preferably comprising a segment of randomized sequence, the SELEX method includes steps of contacting the mixture with the target under conditions favorable for binding, partitioning unbound nucleic acids from those nucleic acids which have bound specifically to target molecules, dissociating the nucleic acid-target complexes, amplifying the nucleic acids dissociated from the nucleic acid-target complexes to yield a ligand-enriched mixture of nucleic acids, then reiterating the steps of binding, partitioning, dissociating and amplifying through as many cycles as desired to yield highly specific, high affinity nucleic acid ligands to the target molecule.

The basic SELEX method has been modified to achieve a number of specific objectives. For example, United States Patent Application Serial No.

07/960,093, filed October 14, 1992, entitled "Method for Selecting Nucleic Acids on the Basis of Structure," describes the use of SELEX in conjunction with gel electrophoresis to select nucleic acid molecules with specific structural characteristics, such as bent DNA. United States Patent Application Serial No.

08/123,935, filed September 17, 1993, entitled "Photoselection of Nucleic Acid Ligands" describes a SELEX based method for selecting nucleic acid ligands containing photoreactive groups capable of binding and/or photocrosslinking to and/or photoinactivating a target molecule. United States Patent Application Serial No. 08/134,028, filed October 7, 1993, entitled "High-Affinity Nucleic Acid Ligands That Discriminate Between Theophylline and Caffeine," describes a method for identifying highly specific nucleic acid ligands able to discriminate between closely related molecules, termed "Counter-SELEX." United States Patent Application Serial No. 08/143,564, filed October 25, 1993, entitled "Systematic Evolution of Ligands by EXponential Enrichment: Solution SELEX," describes a SELEX-based method which achieves highly efficient partitioning between oligonucleotides having high and low affinity for a target molecule. United States Patent Application Serial No. 07/964,624, filed October 21, 1992, entitled "Methods of Producing Nucleic Acid Ligands" describes methods for obtaining improved nucleic acid ligands after SELEX has been performed. United States Patent Application Serial No. 08/400,440, filed March 8, 1995, entitled "Systematic Evolution of Ligands by EXponential Enrichment: Chemi-SELEX," describes methods for covalently linking a ligand to its target.

The SELEX method encompasses the identification of high-affinity nucleic acid ligands containing modified nucleotides conferring improved characteristics on the ligand, such as improved in vivo stability or improved delivery characteristics. Examples of such modifications include chemical substitutions at the ribose and/or phosphate and/or base positions.

SELEX-identified nucleic acid ligands containing modified nucleotides are described in United States Patent Application Serial No. 08/117,991, filed 16 September 8, 1993, entitled "High Affinity Nucleic Acid Ligands Containing Modified Nucleotides," that describes oligonucleotides containing nucleotide derivatives chemically modified at the 5- and 2'-positions ofpyrimidines. United States Patent Application Serial No. 08/134,028, supra, describes highly specific nucleic acid ligands containing one or more nucleotides modified with 2'-amino 2'-fluoro and/or 2'-O-methyl United States Patent Application Serial No. 08/264,029, filed June 22, 1994, entitled "Novel Method of Preparation of 2' Modified Pyrimidine Intramolecular Nucleophilic Displacement," describes oligonucleotides containing various 2'-modified pyrimidines.

The SELEX method encompasses combining selected oligonucleotides with other selected oligonucleotides and non-oligonucleotide functional units as described in United States Patent Application Serial No. 08/284,063, filed August 2, 1994, entitled "Systematic Evolution of Ligands by Exponential Enrichment: Chimeric SELEX" and United States Patent Application Serial No. 08/234,997, filed April 28, 1994, entitled "Systematic Evolution of Ligands by Exponential Enrichment: Blended SELEX," respectively. These applications allow the combination of the broad array of shapes and other properties, and the efficient amplification and replication properties, of oligonucleotides with the desirable properties of other molecules. Each of the above described patent applications which describe modifications of the basic SELEX procedure are specifically incorporated by reference herein in their entirety.

BRIEF SUMMARY OF THE INVENTION The present invention includes methods of identifying and producing nucleic acid ligands to transforming growth factor beta (TGFB), platelet-derived growth factor (PDGF), and human keratinocyte growth factor (hKGF), and homologous proteins, and the nucleic acid ligands so identified and produced.

For the purpose of this application, TGFB includes human TGFB1, TGFB2, and TGFB3 and TGFB's that are substantially homologous thereto. By substantially homologous it is meant a degree of amino acid sequence identity of 70% or more.

For the purposes of this application, PDGF refers to PDGF AA, AB, and BB isoforms and homologous proteins. Specifically included in the definition are human PDGF AA, AB and BB isoforms. In particular, RNA sequences are provided that are capable of binding specifically to TGFB1, PDGF, and hKGF.

Also provided are ssDNA sequences that are capable of binding specifically to TGFp and PDGF. Specifically included in the invention are the RNA ligand sequences shown in Tables 3, 13, 16, and 23 (SEQ ID NOS:12-42, 128-170, 189- 262, 272-304). The RNA ligand sequences of TGFp shown in Table 3 include both pre and post SELEX modifications. Also included in the invention are ssDNA ligands of TGFp and PDGF shown in Tables 6, 8, 9, and Figures 3, 4, and 9 (SEQ ID NOS:55-89, 93-124, 171-176). Also included in this invention are RNA ligands of TGFB1 and hKGF that inhibit the function of TGFB1 and hKGF, presumably by inhibition of the interaction of TGFp and hKGF with their receptors. Also included in this invention are ssDNA ligands of PDGF that inhibit the function of PDGF, presumably by inhibition of the interaction of PDGF with its receptor.

Further included in this invention is a method of identifying nucleic acid ligands and nucleic acid ligand sequences to a target selected from the group consisting of TGFB, PDGF, and hKGF comprising the steps of contacting a candidate mixture of nucleic acids with the target partitioning between members of said candidate mixture on the basis of affinity to the target and (c) amplifying the selected molecules to yield a mixture of nucleic acids enriched for nucleic acid sequences with a relatively higher affinity for binding to the target.

More specifically, the present invention includes the RNA and ssDNA ligands to TGFB identified according to the above-described method, including those ligands shown in Tables 3 and 6 (SEQ ID NOS:12-42, 55-89). Also included are nucleic acid ligands to TGFB that are substantially homologous to any of the given ligands and that have substantially the same ability to bind TGFB and inhibit the function of TGFB. Further included in this invention are nucleic acid ligands to TGFB that have substantially the same structural form as the ligands presented herein and that have substantially the same ability to bind TGFB and inhibit the function of TGFB.

Additionally, the present invention includes the ssDNA and RNA ligands to PDGF identified according to the above-described method, including those ligands shown in Tables 8 and 13, and Figures 3, 4, and 9 (SEQ ID NOS:93-124, 128-176). Also included are DNA and RNA ligands to PDGF that are substantially homologous to any of the given ligands and that have substantially the same ability to bind PDGF. Further included in this invention are nucleic acid ligands to PDGF that have substantially the same structural form as the ligands presented herein and that have substantially the same ability to bind

PDGF.

In addition, the present invention includes the RNA ligands to hKGF identified according to the above-described method, including those ligands shown in Tables 16 and 23 (SEQ ID NOS:189-262, 272-304). Also included are RNA ligands to hKGF that are substantially homologous to any of the given ligands and that have substantially the same ability to bind hKGF and inhibit the interaction of hKGF with its receptor. Further included in this invention are nucleic acid ligands to hKGF that have substantially the same structural form as the ligands presented herein and that have substantially the same ability to bind hKGF and inhibit the interaction of hKGF with its receptor.

The present invention also includes other modified nucleotide sequences based on the RNA ligands identified herein and mixtures of the same.

Further included in this invention is a method of assaying a test compound for the ability to inhibit hKGF receptor-mediated cell proliferation comprising the steps of contacting the test compound with a hKGF nucleic acid ligand and a keratinocyte growth factor; and detecting the ability of the test compound to inhibit binding between the hKGF nucleic acid ligand and the keratinocyte growth factor.

Also included in this invention is a method of assaying a test compound for the ability to inhibit the interaction of a growth factor with its plasma membrane bound receptor comprising the steps of solubilizing cells containing the plasma membrane bound receptor; creating a plasma membrane extract of the cells; reacting the extract with labeled growth factor alone and in the presence of the test compound thereby creating complexes; analyzing the complexes by electrophoresis under native conditions; visualizing the complexes by imaging; and comparing the image of the extract with labeled growth factor alone to the image of the extract in the presence of the test compound to determine whether the test compound inhibited the interaction between the growth factor and its plasma membrane bound receptor.

Further included in this invention is a method for assaying cells to determine whether they express a growth factor plasma membrane bound receptor comprising the steps of solubilizing the cells; creating a plasma membrane extract of the cells; reacting the plasma membrane extract with a labeled growth factor; analyzing the reaction between the plasma membrane extract with the labeled growth factor by electrophoresis under native conditions; (e) comparing the electrophoresis of step with electrophoresis of labeled growth factor; and visualizing the results of the electrophoresis to determine whether a complex is formed with altered mobility relative to the mobility of a labeled growth factor alone.

BRIEF DESCRIPTION OF THE FIGURES Figure 1 shows the binding analysis of the 40D7 DNA library for TGFB1.

Binding data obtained from Round 19 (triangles) and Round 0 (circles) are shown.

Figure 2 shows the results of the PAI-luciferase assay of TGFB1 incubated with oligonucleotides (0.1 iM) or anti-TGFB (60 pg/ml).

Figure 3 shows the consensus secondary structure for the sequence set shown in Table 9. R A or G, Y C or T, K G or T, N and N' indicate any base pair.

Figure 4 shows the minimal ligands 20t, 36t and 41t folded according to the consensus secondary structure motif. [3T] represents a linked thymidine nucleotide added to reduce 3'-exonuclease degradation.

Figures 5A, 5B and 5C show the binding of minimal high affinity DNA ligands to PDGF-AA, AB and BB, respectively. The fraction of 32 P 5' end-labeled DNA ligands bound to varying concentrations of PDGF was determined by the nitrocellulose filter binding method. Minimal ligands tested were 20t 36t and 41t Oligonucleotide concentrations in these experiments were 10 pM (PDGF-AB and PDGF-BB) and =50 pM (PDGF AA). Data points were fitted to eq. 1 (for binding of the DNA ligands to PDGF-AA) or to eq. 2 (for binding to PDGF AB and BB) using the non-linear least squares method. Binding reactions were done at 37C in binding buffer (PBSM with 0.01% HSA).

Figure 6 shows the dissociation rate determination for the high affinity interaction between the minimal DNA ligands and PDGF AB. The fraction of 2 P end-labeled ligands 20t 36t and 41t all at 0.17 nM, bound to PDGF AB (1 nM) was measured by nitrocellulose filter binding at the indicated time points following the addition of a 500-fold excess of the unlabeled competitor. The dissociation rate constant (kff) values were determined by fitting the data points to eq 3. The experiments were performed at 37 0 C in binding buffer.

Figure 7 shows the effect of DNA ligands on the binding of 2

I-PDGF-BB

and 'SI-PDGF-AA to PDGF a-receptors expressed in PAE cells.

Figure 8 shows the effect of DNA ligands on the mitogenic effect of PDGF-BB on PAE cells expressing the PDGF B-receptors.

Figure 9 shows the 2'-O-methyl-2'-deoxy- and 2'-fluoro-2'-deoxyribonucleotide-substitution pattern compatible with high affinity binding to PDGF-AB. Underlined symbols indicate 2'-O-methyl-2'-deoxynucleotides; italicized symbols indicate 2'-fluoro-2'-deoxynucleotides; normal font indicates 2'-deoxyribonucleotides; indicates inverted orientation thymidine nucleotide (Glen Research, Sterling, VA); PEG in the loops of helices II and III indicates pentaethylene glycol spacer phosphoramidite (Glen Research, Sterling, VA).

Figure 10A shows the saturation binding of radiolabeled hKGF on the surface of the PC-3 cells. TB (total binding) is the binding observed in the 21 absence of competing unlabeled hKGF, whereas NSB (nonspecific binding) is the binding observed in the presence of 100 fold molar excess of unlabeled hKGF.

SB (specific binding) demonstrates the specific binding, and this curve was derived by subtracting the NSB curve from the TB curve. Figure 10B is the Scatchard analysis of the data points shown in 10A for the SB curve.

Figure 11 shows the shift of the electrophoretic mobility due to plasma membrane extracts from PC-3 cells. In lanes 1-8, the membrane extracts were reacted with various concentrations of radiolabeled hKGF as shown under each lane. In addition to the radiolabeled hKGF (as shown under each lane) for lanes 9-12 a 100 fold molar excess was included of unlabeled hKGF. Cl and C2 represent two observed complexes due to the presence of hKGF binding moieties in the PC-3 plasma membrane extracts.

Figures 12A D show the proposed alignment of 2'F and 2'NH 2 ligands.

Lower case, italicized sequence residues indicate the constant region of the template. In the consensus sequences, capital and lower case letters are used for residues found in greater than or equal to 80% and 60% of the members of each family respectively. Kd and K, values are also shown next to the designation of each ligand. The Ki values shown here were calculated using the formula where IC50 is the measured half maximal inhibitory concentration of each ligand in the PC-3 cell assay as described in Example 16; C is the concentration of 'SI-KGF; and Kd is the equilibrium dissociation constant of KGF for its receptor, (about 150 pM). The ligands marked with stars show biphasic binding curves.

Figures 12A and 12B show the proposed alignment of2'F ligands. The majority of 2'F ligands can be folded into pseudoknot structures. Two classes are proposed as shown. The summary structure for each class is also shown. Bases participating in stem I (S1) are underlined with single lines while bases of stem 2 (S2) are underlined with double lines. Spaces were introduced for alignment of the various elements of the pseudoknots.

Figures 12C and 12D show the proposed folding of2'NH 2 ligands. These ligands are assigned into two classes. As shown in the summary structures, class 1 and class 2 ligands can form a stem-loop and dumbbell structure, respectively.

Spaces were introduced to allow sequence alignment. Residues participating in stems are underlined. In the summary structures, periods indicate a variable number of residues. Ligands 2N and 54N are circular permutations of the same dumbbell structure. For alignment of the corresponding loops these ligands are wrapped around two lines.

Figure 13 shows the minimal sequence requirement for binding of ligand 6F and 14F to hKGF. The predicted folding of each ligand is shown. Constant regions of the ligands are shown in lower case. Conserved sequences are underlined. Circles and triangles mark the 3' ends of active and inactive truncates respectively.

DETAILED DESCRIPTION OF THE INVENTION This application describes high-affinity nucleic acid ligands to TGFp, PDGF, and hKGF identified through the method known as SELEX. SELEX is described in U.S. Patent Application Serial No. 07/536,428, entitled Systematic Evolution of Ligands by EXponential Enrichment, now abandoned, U.S. Patent Application Serial No. 07/714,131, filed June 10, 1991, entitled Nucleic Acid Ligands, United States Patent Application Serial No. 07/931,473, filed August 17, 1992, entitled Nucleic Acid Ligands, now United States Patent No. 5,270,163, (see also PCT/US91/04078). These applications, each specifically incorporated herein by reference, are collectively called the SELEX Patent Applications.

In its most basic form, the SELEX process may be defined by the following series of steps: 1) A candidate mixture of nucleic acids of differing sequence is prepared.

The candidate mixture generally includes regions of fixed sequences each of the members of the candidate mixture contains the same sequences in the same location) and regions of randomized sequences. The fixed sequence regions are selected either: to assist in the amplification steps described below, to mimic a sequence known to bind to the target, or to enhance the concentration of a given structural arrangement of the nucleic acids in the candidate mixture.

I

The randomized sequences can be totally randomized the probability of finding a base at any position being one in four) or only partially randomized the probability of finding a base at any location can be selected at any level between 0 and 100 percent).

2) The candidate mixture is contacted with the selected target under conditions favorable for binding between the target and members of the candidate mixture. Under these circumstances, the interaction between the target and the nucleic acids of the candidate mixture can be considered as forming nucleic acid-target pairs between the target and those nucleic acids having the strongest affinity for the target.

3) The nucleic acids with the highest affinity for the target are partitioned from those nucleic acids with lesser affinity to the target. Because only an extremely small number of sequences (and possibly only one molecule of nucleic acid) corresponding to the highest affinity nucleic acids exist in the candidate mixture, it is generally desirable to set the partitioning criteria so that a significant amount of the nucleic acids in the candidate mixture (approximately 5-50%) are retained during partitioning.

4) Those nucleic acids selected during partitioning as having the relatively higher affinity to the target are then amplified to create a new candidate mixture that is enriched in nucleic acids having a relatively higher affinity for the target.

By repeating the partitioning and amplifying steps above, the newly formed candidate mixture contains fewer and fewer weakly binding sequences, and the average degree of affinity of the nucleic acids to the target will generally increase. Taken to its extreme, the SELEX process will yield a candidate mixture containing one or a small number of unique nucleic acids representing those nucleic acids from the original candidate mixture having the highest affinity to the target molecule.

The SELEX Patent Applications describe and elaborate on this process in great detail. Included are targets that can be used in the process; methods for partitioning nucleic acids within a candidate mixture; and methods for amplifying partitioned nucleic acids to generate enriched candidate mixture. The SELEX 24 Patent Applications also describe ligands obtained to a number of target species, including both protein targets where the protein is and is not a nucleic acid binding protein.

The nucleic acid ligands described herein can be complexed with a lipophilic compound cholesterol) or attached to or encapsulated in a complex comprised of lipophilic components a liposome). The complexed nucleic acid ligands can enhance the cellular uptake of the nucleic acid ligands by a cell for delivery of the nucleic acid ligands to an intracellular target. U.S. Patent Application No. 08/434,465, filed May 4, 1995, entitled "Nucleic Acid Ligand Complexes," which is incorporated in its entirety herein, describes a method for preparing a therapeutic or diagnostic complex comprised of a nucleic acid ligand and a lipophilic compound or a non-immunogenic, high molecular weight compound.

The methods described herein and the nucleic acid ligands identified by such methods are useful for both therapeutic and diagnostic purposes. Therapeutic uses include the treatment or prevention of diseases or medical conditions in human patients. Diagnostic utilization may include both in vivo or in vitro diagnostic applications. The SELEX method generally, and the specific adaptations of the SELEX method taught and claimed herein specifically, are particularly suited for diagnostic applications. SELEX identifies nucleic acid ligands that are able to bind targets with high affinity and with surprising specificity. These characteristics are, of course, the desired properties one skilled in the art would seek in a diagnostic ligand.

The nucleic acid ligands of the present invention may be routinely adapted for diagnostic purposes according to any number of techniques employed by those skilled in the art. Diagnostic agents need only be able to allow the user to identify the presence of a given target at a particular locale or concentration. Simply the ability to form binding pairs with the target may be sufficient to trigger a positive signal for diagnostic purposes. Those skilled in the art would also be able to adapt any nucleic acid ligand by procedures known in the art to incorporate a labeling tag in order to track the presence of such ligand. Such a tag could be used in a number of diagnostic procedures. The nucleic acid ligands described herein may specifically be used for identification of the TGFp, PDGF, and hKGF proteins.

SELEX provides high affinity ligands of a target molecule. This represents a singular achievement that is unprecedented in the field of nucleic acids research. The present invention applies the SELEX procedure to the specific targets of TGFp, PDGF, and hKGF. In the Example section below, the experimental parameters used to isolate and identify the nucleic acid ligands to TGFp, PDGF, and hKGF are described.

In order to produce nucleic acids desirable for use as a pharmaceutical, it is preferred that the nucleic acid ligand binds to the target in a manner capable of achieving the desired effect on the target; be as small as possible to obtain the desired effect; be as stable as possible; and be a specific ligand to the chosen target. In most situations, it is preferred that the nucleic acid ligand have the highest possible affinity to the target.

In co-pending and commonly assigned U.S. Patent Application Serial No.

07/964,624, filed October 21, 1992 now U.S. Patent No. 5,496,938, methods are described for obtaining improved nucleic acid ligands after SELEX has been performed. The '624 application, entitled Methods of Producing Nucleic Acid Ligands, is specifically incorporated herein by reference. Further included in this patent are methods for determining the three-dimensional structures of nucleic acid ligands. Such methods include mathematical modeling and structure modifications of the SELEX-derived ligands, such as chemical modification and nucleotide substitution.

In the present invention, SELEX experiments were performed in order to identify RNA and DNA ligands with specific high affinity for TGFp from degenerate libraries containing 40 or 60 random positions (40N or 60N) (Tables 1 and This invention includes the specific RNA ligands to TGF p shown in Table 3 (SEQ ID NOS:12-42), identified by the methods described in Examples 1 and 2. This invention further includes RNA ligands to TGFp which inhibit TGF p function, presumably by inhibiting the interaction of TGF p with its receptor. This invention includes the specific ssDNA ligands to TGF p1 shown in 26 Table 6 (SEQ ID NOS:55-89) identified by the methods described in Examples and 6.

In the present invention, two SELEX experiments were also performed in order to identify ssDNA and RNA with specific high affinity for PDGF from degenerate libraries containing 40 and 50 random positions (40N and respectively (Tables 7 and 12). This invention includes the specific ssDNA and RNA ligands to PDGF shown in Tables 8, 9 and 13 and Figures 3, 4, and 9 (SEQ ID NOS:93-124, 128-176), identified by the methods described in Examples 7 and In the present invention, a SELEX experiment was also performed in search of RNA ligands with specific high affinity for hKGF from degenerate libraries containing 40 random positions (40N) (Table 14). This invention includes the specific RNA ligands to hKGF shown in Tables 16 and 23 (SEQ ID NOS:189-262, 272-304), identified by the methods described in Examples 16 and 17. This invention further includes RNA ligands to hKGF which inhibit the interaction of hKGF with its receptor.

The scope of the ligands covered by this invention extends to all nucleic acid ligands of TGFp, PDGF, and hKGF, modified and unmodified, preferably those identified according to the SELEX procedure. More specifically, this invention includes nucleic acid sequences that are substantially homologous to the ligands shown in Tables 3, 6, 8, 9, 13, 16, and 23 and Figures 3, 4, and 9 (SEQ ID NOS:12-42, 55-89, 93-124, 128-176, 189-262, 272-304). By substantially homologous it is meant a degree of primary sequence homology in excess of most preferably in excess of 80%. A review of the sequence homologies of the nucleic acid ligands shown in Tables 3 and 6 (SEQ ID NOS.:12-42, 55-89) for TGFp, Tables 8 and 13 (SEQ ID NOS:93-124, 128-170) for PDGF, and Tables 16 and 23 (SEQ ID NOS:189-262, 272-304) for hKGF shows that sequences with little or no primary homology may have substantially the same ability to bind a given target. For these reasons, this invention also includes nucleic acid ligands that have substantially the same structure and ability to bind TGFp, PDGF, and hKGF as the nucleic acid ligands shown in Tables 3, 6, 8, 9, 13, 16, and 23 and 27 Figures 3, 4, and 9 (SEQ ID NOS:12-42, 55-89, 93-124, 128-176, 189-262, 272- 304). Substantially the same structure for PDGF includes all nucleic acid ligands having the common structural elements shown in Figure 3 that lead to the affinity to PDGF. Substantially the same ability to bind TGFp, PDGF, or hKGF means that the affinity is within one or two orders of magnitude of the affinity of the ligands described herein. It is well within the skill of those of ordinary skill in the art to determine whether a given sequence substantially homologous to those specifically described herein has substantially the same ability to bind TGFp, PDGF, or hKGF.

This invention also includes the ligands as described above, wherein certain chemical modifications are made in order to increase the in vivo stability of the ligand or to enhance or mediate the delivery of the ligand. Examples of such modifications include chemical substitutions at the sugar and/ or phosphate and/or base positions of a given nucleic acid sequence. See, U.S. Patent Application Serial No. 08/117,991, filed September 9, 1993, entitled High Affinity Nucleic Acid Ligands Containing Modified Nucleotides which is specifically incorporated herein by reference. Other modifications are known to one of ordinary skill in the art. Such modifications may be made post-SELEX (modification of previously identified modified or unmodified ligands) or by incorporation into the SELEX process.

Example 20 describes post-SELEX procedure modification of a nucleic acid ligand to basic fibroblast growth factor (bFGF). The nucleic acid ligand was modified by the addition of phosphorothioate caps and substitution of several ribopurines with 2'-deoxy-2'-O-methylpurines.

As described above, because of their ability to selectively bind TGFP, PDGF, and hKGF, the nucleic acid ligands to TGFp, PDGF, and hKGF described herein are useful as pharmaceuticals. This invention, therefore, also includes a method for treating TGF-p-mediated pathological conditions by administration of a nucleic acid ligand capable of binding to TGF p, a method for treating PDGF-mediated pathological conditions by administration of a nucleic acid ligand capable of binding to PDGF, and a method for treating hKGF-mediated pathological conditions by administration of a nucleic acid ligand capable of binding to hKGF.

Therapeutic compositions of the nucleic acid ligands may be administered parenterally by injection, although other effective administration forms, such as intraarticular injection, inhalant mists, orally active formulations, transdermal iontophoresis or suppositories, are also envisioned. One preferred carrier is physiological saline solution, but it is contemplated that other pharmaceutically acceptable carriers may also be used. In one preferred embodiment, it is envisioned that the carrier and the ligand constitute a physiologically-compatible, slow release formulation. The primary solvent in such a carrier may be either aqueous or non-aqueous in nature. In addition, the carrier may contain other pharmacologically-acceptable excipients for modifying or maintaining the pH, osmolarity, viscosity, clarity, color, sterility, stability, rate of dissolution, or odor of the formulation. Similarly, the carrier may contain still other pharmacologically-acceptable excipients for modifying or maintaining the stability, rate of dissolution, release, or absorption of the ligand. Such excipients are those substances usually and customarily employed to formulate dosages for parental administration in either unit dose or multi-dose form.

Once the therapeutic composition has been formulated, it may be stored in sterile vials as a solution, suspension, gel, emulsion, solid, or dehydrated or lyophilized powder. Such formulations may be stored either in a ready to use form or requiring reconstitution immediately prior to administration. The manner of administering formulations containing nucleic acid ligands for systemic delivery may be via subcutaneous, intramuscular, intravenous, intranasal or vaginal or rectal suppository.

The following examples are provided to explain and illustrate the present invention and are not intended to be limiting of the invention. Examples 1-4 describe initial experiments to identify RNA with specific high affinity for TGFp1. Example 1 describes the various materials and experimental procedures used in Examples 2-4. Example 2 describes a representative method for identifying RNA ligands by the SELEX method which bind TGFp 1. Example 3 describes the affinities the ligands have for TGFp 1 and demonstrates that the ligands are capable of inhibiting the function of TGFp 1, presumably by inhibiting the interaction of TGFp 1 with its receptor. Example 4 describes which regions of the ligands are believed to be necessary for TGFp 1 binding and inhibition of TGFp 1 receptor binding. Example 5 describes another representative method for identifying RNA and DNA ligands by the SELEX method which bind TGFp 1.

Example 6 reports on the binding analysis, bioassay, and sequences of a ssDNA SELEX library. Example 7 describes the various materials and experimental procedures used in evolving ssDNA ligands to PDGF described in Examples 8-13.

Example 8 describes the ssDNA ligands to PDGF and the predicted secondary structure of selected nucleic acid ligands. Example 9 describes the minimal sequence necessary for high affinity binding. Example 10 describes the kinetic stability of PDGF-Nucleic Acid Ligand complexes. Example 11 describes the thermal melting properties for selected ligands. Example 12 describes photocrosslinking of nucleic acid ligands and PDGF. Example 13 describes the inhibition by DNA ligands of PDGF isoforms on cultured cells and inhibition of the mitogenic effects of PDGF in cells by DNA ligands. Example 14 describes the modification of nucleic acid ligands to PDGF with modified nucleotides.

Example 15 describes the experimental procedures used in evolving RNA ligands to PDGF and shows the ligand sequences. Example 16 describes the various materials and experimental procedures used in evolving nucleic acid ligands to hKGF described in Examples 17-19. Example 17 describes the RNA ligands to hKGF, the affinities the ligands have for hKGF, and the specificity of the RNA ligands to hKGF. Example 18 describes inhibition of hKGF binding to cell surface receptors. Example 19 reports on the inhibition of mitogenic activity of hKGF by a selected ligand. Example 20 describes the modification of nucleic acid ligands to bFGF with 2'-deoxy-2'-O-methylpurines.

EXAMPLES

EXAMPLE 1. EXPERIMENTAL PROCEDURES This example provides the general procedures followed and incorporated in Examples 2-4.

A. Materials.

Human recombinant TGFB1 used in this SELEX procedure was acquired from Genentech. Human recombinant TGFB1 can also be purchased from R&D systems, Minneapolis, MN, USA.

Biotinylated TGFB1 was prepared by reacting TGFB1 at 3.6 UM with an 11 fold molar excess of sulfo-NHS-biotin (Pierce, Rockford, IL, USA) in NaHCO3 for 3 hr. in an ice bath. The reaction was acidified with 0.036 volumes of 10% acetic acid and applied to a 40 mg. Vydac (The Separations Group, Hesperia, CA, USA) reverse phase column made in a siliconized pipet tip to separate unreacted biotin from biotinylated TGFB 1. The column was prewashed with 200 pl ethanol followed by 200 pl 1% acetic acid, the biotinylation reaction was applied, free biotin was washed through with 200 pl of 50 mM sodium acetate pH 5.5, followed by 200 gl of 20% acetonitrile and finally eluted with 200 ,l of acetonitrile. The sample was lyophilized and resuspended in 50 mM sodium acetate pH 5.0 at 40 ,M and stored at 40 C. The TGFB1 was spiked with 100,000 cpm iodinated TGFB 1 in order to follow recovery and to assess the success of the biotinylation reaction by measuring the fraction of the radioactivity that would bind to streptavidin coated agarose beads (Pierce) before and after biotinylation.

An aliquot of the TGFB1 before and after biotinylation was subjected to analytical reverse phase chromatography. The biotinylated TGFB1 substantially ran as a single peak which was retarded with respect to the unbiotinylated TGFB. A small amount (5 ofunreacted TGFB1 could be detected. The efficiency of binding of the iodinated, biotinylated TGFB1 to streptavidin (SA) agarose beads (30,l) was under the binding conditions used for SELEX partitioning.

Iodinated TGFB1 was prepared by the lactoperoxidase method sodium phosphate pH 7.3, 0.16% glucose) with BioRad Enzymo beads (BioRad, 31 Richmond, CA, USA) and the bound iodine separated from the free iodine by gel filtration on G25 Sephadex in 50 mM sodium acetate 0.01% Tween.

The mink lung cell line expressing the luciferase reporter gene under the control of PAl 1 promoter (Abe et al. (1994) Anal. Biochem. 216:276-284) was a gift from Dr. Dan Rifkin (Department of Cell Biology, New York Medical Center, New York, New York 10016). Luciferase was assayed by reagents purchased from Analytical Luminescence Laboratory, San Diego, CA, USA.

NH modified CTP and UTP were prepared according to the method of Pieken et al. (1991) Science 25:314-317. DNA oligonucleotides were synthesized using standard procedures either at NeXstar Pharmaceuticals, Inc.

(Boulder, CO, USA) or by Operon Technologies (Alameda, CA, USA). All other reagents and chemicals were purchased from standard commercial sources and sources have been indicated.

B. SELEX procedure SELEX ligands that bind to TGFB1 were derived essentially as described in U.S. Patent No. 5,270,163 (see also,Tuerk and Gold (1990) Science 249:505-510). To generate the starting pool of PCR template, PCR product from twenty separate PCR reactions each containing 16.1 pmol ofunpurified, single stranded DNA (at least a total of 2 x 101'to 2 x 10" different molecules) were pooled before the first transcription. PCR conditions were 50 mM KCI, 10 mM Tris-HCL, pH 9, 0.1% Triton-X100, 1.5 mM MgCl 2 0.2 mM of each dATP, dCTP, dGTP, and dTTP, 2MM each primer and 0.075 units/l ofTaq DNA polymerase, 100 Il per reaction in a siliconized microfuge tube. All PCR cycles took advantage of hot start using Ampliwax (Perk and Elmer, Norwalk, CN, USA). Duration of the initial PCR was 10 cycles; a PCR cycle was 94 520 720 An initial denaturation was 940 C for 4' and the final extension at 72° C for PCR reactions were combined, phenol/ chloroform extracted, and isopropanol precipitated (2.0 M ammonium acetate, 50% isopropanol) to remove primers.

Transcription reactions contained 200 nM DNA, 0.9 mM GTP, 0.9 mM 2'-NH,-UTP, 0.9 mM 2'-NH 2 -CTP, 0.5mM ATP, 87 mM Tris-HCI pH 8.0, 17 mM MgC1 2 4.4 mM spermidine, 22 mM DTT, 100 mg/ml acetylated BSA (Promega, Madison, WI, USA) and 4 units/ml T7 RNA polymerase. (2'-F-UTP and 2'-F-CTP (United States Biochemical, Cleveland, OH, USA) were used at mM, whereas UTP and CTP were used at 0.9mM each). Transcription reactions were incubated overnight at 280 C (at least 10 hours). After transcription the template was digested by addition of 2,l RQ1 Dnase (Promega) for 15' at 280 C, and then extracted with phenol/CHCl 3 followed by three ethanol precipitations from ammonium acetate (3.9M ammonium acetate, 72% ethanol).

The RNA molecules were incubated with TGFBI bound to SA agarose beads as described below in Krebs-Ringer solution (KR) (120 mM NaCI, 4.8 mM KCI, 10 mM Na phosphate buffer pH 7.4, 1.2 mM MgSO 4 2.6 mM CaC1 2 modified to include 20mM Na-Hepes pH 7.5 and 0.2% Triton X100 (Pierce).

This buffer is referred to as KRHT.

TGFB I-RNA complexes were separated from unbound RNA by washing the beads. Recovery of the selected 2'-NH 2 or F pyrimidine modified RNA from the agarose beads required guanidine thiocyanate extraction (SM GnSCN, 10 mM Tris-HCI, 0.1 mM EDTA, pH 7.0, 0.1M beta mercaptoethanol) or from Seradyne SA coated beads by 2% SDS (0.1 M Tris-HCI pH 7.5, 50 mM NaCI, 1 mM Na 2 EDTA, 2% SDS, 1.5mM DTT). Regular 2'-OH RNA was easily recovered under less harsh conditions with the same buffer used for the Seradyne beads containing only 0.2% SDS. After extraction and precipitation to purify and concentrate the RNA, the sample was reverse transcribed with a cloned MMLV RT with the RNase H sequence deleted. The reaction contained less than or equal to 16 nM RNA, 10 ,M 3' primer, 50 mM Tris-HCL pH 8.3, 75 mM KC1, 5 mM MgCl 2 10 mM DTT, 0.5 mM dNTP's. Prior to addition of buffer the RNA and the primer were boiled together. After addition of buffer and salts the reaction was annealed for 10 min at 28* C before addition of 600 units of Superscript reverse transcriptase (Bethesda Research Labs, Gaithersburg, MD, USA) and synthesis at C for 1 hour.

PCR amplification of this cDNA pmol) resulted in approximately 250 pmol double-stranded DNA, of this, 40pmols was transcribed and used to initiate the next round of SELEX.

C. Partitioning Method for SELEX.

2.5 pmols biotinylated TGFB1 were bound to 30 Ml SA agarose beads (Pierce) in 200 p, KRHT. The mixture was incubated on a rotator at 37 0 C for to 30 minutes. The beads were washed three times by centrifugation and resuspension in 200 ,1 cold KRHT to remove unbound TGFB 1, and resuspended in a final volume of 500 gl KRHT. RNA containing 2'-NH 2 pyrimidines was heated at 70° C for three minutes (RNAs containing 2'-OH or 2'-F pyrimidines were heated at 950 C) and diluted into KRHT containing TGFB1 bound to SA beads. The final concentration of RNA is 1 IM and the TGFB1 was 5nM. Binding occurs with rotation at 370 C for 30 minutes. Beads were washed by centrifugation and resuspension three times with 200 cl binding buffer to remove unbound RNA. RNA was eluted from the beads as described above.

D. Binding assays.

Two binding assays for ligands to TGFB1 gave equivalent results wherever tested. In the SA bead assay the biotinylated TGFB1 was serially diluted in KRHT in polypropylene tubes (Linbro, ICN, Irvine, CA, USA) and bound to the beads as described above. After unbound TGFBI was washed away, trace quantities of 3

P

-labeled RNA(<0.1 nM) were added to each tube and vortexed to mix. After static incubation at 37 0 C for 30 minutes, the unbound RNA was removed by washing three times with 200 ,e of KRHT.

In the nitrocellulose filter binding assay, TGFB1 was serially diluted in KRH containing 0.1% defatted BSA (Fluka radioimmunoassay grade, Fluka, Hauppauge, NY, USA) as carrier instead of Triton X-100. Incubation with RNA tracer was as above. Samples were pipetted with a multiwell pipettor onto a multiwell manifold holding a sheet of wet BioRad 0.45 micron nitrocellulose, aspirated, and washed three times with 200 1l KRH (containing no BSA). The filters were air dried and counted in a liquid scintillation counter (Beckmann Instruments, Palo Alto, CA) The equation used to fit the binding ofligands to TGFB1 describes the binding of a ligand to a receptor (in this case TGFB 1) that follows the laws of mass action and for which there is a single binding site: Y=Bmax*X/(Kd+X): where Y is the fraction of the ligand bound, is the maximum fraction of the ligand bound, X is the concentration of TGFBI and Kd is the dissociation constant of TGFBI1 and the ligand. Data points were fit by nonlinear regression using the computer program Graphpad Prism (Graphpad Software, San Diego, CA). The algorithm minimized the sum of the squares of the actual distance of the points from the curve. Convergence was reached when two consecutive iterations changed the sum-of-squares by less than 0.01%.

E. Cloning and Sequencing.

SELEX experiments are described in Table 2. Primers for SELEX experiments 1 and 2 shown in Table 1 contain recognition sites for the restriction endonucleases SacI primer T7SacBam; SEQ ID NO:7) and XbaI primer 3XH; SEQ ID NO:9). PCR products from SELEX experiments I and 2 were cloned directionally into SacI/XbaI digested pGem 9zf(Promega). 5' primer T7SB2N (SEQ ID NO:8) and 3' primer 3XH (SEQ ID NO:9) (Tablel) were used for SELEX experiments 3-9. PCR products from SELEX experiments 3-9 were cloned directionally into the BamHI/XbaI site of a modified pGem9zf:BamH cloning vector. The BamH1 site was engineered into pGem9zf in the following way. A clone isolated from library 2 (lib2-6-2) that did not bind to TGF 1 (sequence not shown) was digested with BamHl and Xbal. The sequence flanking the cloning site of the modified pGem9zf vector is shown in Table 1 (SEQ ID NOS:10-11).

After digestion of the plasmid with restriction endonuclease and dephosphorylation with CIP (calf intestinal phosphatase), vectors were gel purified. Inserts were ligated and recombinant plasmids were transformed into E.

coli strain DH1OB (Bethesda Research Labs). Plasmid DNA was prepared by alkaline lysis, mini prep procedure. Twenty-two clones representing 9 unique sequences were sequenced at random from libraries I and 2. 50 clones were sequenced from libraries 3-9 using a single dideoxy G reaction (called G track).

The sequencing ladders were compared and organized for similarities. Selected clones from each family were chosen for complete sequence analysis. TGFB1 binding assays were performed on transcripts representing different G sequences in each library. Out of a total of 140 binding assays, 27 ligands bound with a Kd less than 10 nM, and 21 of these were sequenced. Clones were sequenced with the Sequenase sequencing kit (United States Biochemical Corporation, Cleveland,

OH).

F. Ligand Truncation.

Truncation experiments were carried out to determine the minimal sequence necessary for high affinity binding of the RNA ligands to TGFB For 3' boundary determination, RNA ligands were 5' end-labeled with y- 2 P-ATP using T4 polynucleotide kinase. 5' boundaries were established with 3' end-labeled ligands using a- 2 P-pCp and T4 RNA ligase. After partial alkaline hydrolysis, radiolabeled RNA ligands were incubated with TGFBI at concentrations ranging from 1 nM to 50 nM and protein-bound RNA was separated by nitrocellulose partitioning. RNA truncates were analyzed on a high-resolution denaturing polyacrylamide gel. A ladder of radioactively labeled ligands terminating with G-residues was generated by partial RNase TI digestion and was used as markers.

G. Inhibition of TGF01 function.

TGFB1 signal transduction begins with binding to a cell surface receptor and results in the induction of transcription of a variety of genes. One of these genes is Pail. The TGFB 1 assay utilizes the mink lung epithelial cell (MLEC) carrying the luciferase reporter gene fused to the Pai 1 promoter. The MLEC has TGFB1 receptors on its cell surface. Thus one can measure the response of the cells to TGFB1 and the effective concentration of TGFB1 in the culture media by measuring the luciferase enzyme activity after a period of induction.

Mink lung epithelial cells (MLEC) carrying the Pail/luc construct were maintained in DME containing 10% fetal bovine serum and 400 ig/ml G418.

MLEC-Pail/luc cells were plated at 3.2 xl0 4 cells per well in a 96 well Falcon plate, in 100 ,l of DME 10% fetal bovine serum overnight. Media was made from autoclaved water. The cells were washed three times (100 lA) in serum free 36 DME plus Solution A Solution A is 30 mM Hepes pH 7.6, 10 mM glucose, mM KCI, 131 mM NaCI, 1.0 mM disodium phosphate. Samples (100pl) were added in DME containing 20 mM Hepes pH 7.5, and 0.1% BSA (Fluka, radioimmunoassay grade). All samples were in triplicate. After six hours at 370 C in a 5% CO, incubator the media was removed and cells were washed three times (1001 each) in cold PBS. Lysis buffer (75 (Analytical Luminescence Laboratory) was added and the plates incubated on ice for 20 min. The plates were sealed and frozen at -80° C until assayed. Samples (25,l) were assayed for luciferase activity with the Enhanced Luciferase Assay Kit from Analytical Luminescence Laboratory (San Diego, CA, USA) according to the manufacturers instructions using the Berthold Microlumat LB96P luminometer. Luminescence is reproducibly a function of TGFB1 concentration added to the media.

Ligands tested for inhibition of TGF 1 activity were tested at a minimum of five concentrations. The ligands were serially diluted in DME, 20 mM Hepes pH 7.5, 0.1% Fluka BSA in polypropylene tubes and an equal volume of media containing 12 pM TGF 1 was added to each tube, vortexed and .transferred to the cells without further incubation. From the TGF31 standard curve included on every plate the effective concentration of TGFB1 in the presence of the inhibitory ligands was determined by the reduction in luminescence measured. Some ligands were tested at both 3 pM and 6 pM TGFi31 with the same results. To determine the ICso (the concentration of SELEX ligand necessary to reduce the TGFB1 activity the five values obtained for each ligand were plotted and the value at 50% inhibition was determined graphically using Graphpad Prism assuming a hyperbolic fit of the data and using non-linear regression.

EXAMPLE 2. RNA LIGANDS TO TGFO1 A. SELEX experiments In order to generate RNA ligands to TGFBI, nine SELEX experiments, as summarized in Table 2, were performed using the methods described in Example 1. As shown in Table 1, the RNA pools differ in the number of random bases present in the central portion of the molecules: 40 nucleotides in the 40N6 (SEQ ID NO:2) SELEX and 64 nucleotides in the 64N6 and lib2-6-1RN6 (SEQ ID NOS:1, 3) SELEX experiment. Since the goal was to select RNA ligands that not only bound to TGFB 1 but also blocked receptor binding, the large random region (64N) was chosen. In two experiments, a shorter random region (40N) was also included. Ligands to TGF31 were very rare with 40N and were qualitatively the same as the 64N6 ligands selected.

The sequences of clones from the SELEX experiments are shown in Table 3 (SEQ ID NOS:12-42).The pyrimidines used in the various SELEX experiments differed at the 2' position of the sugar (Table In the first two SELEX experiments, ligands were evolved as 2'-OH pyrimidines. Ligands were post-SELEX modified with 2'-NH, or substituted pyrimidines to see if they retained TGFB1 binding. Since the 2' substitutions rendered the ligands resistant to RNase A they were also tested in the cell culture assay for inhibition of TGFI31 activity. One such ligand lib2-6-1 (Group D, Table 3; SEQ ID NO:35) when substituted with 2'-NH 2 -UTP and 2'-F-CTP was shown to inhibit TGF 1 receptor mediated activity. To select more ligands, six more independent SELEX experiments (lib3-7 and lib9) were performed using the 2'-F and 2'-NH 2 -substituted pyrimidines during the evolution process. In experiment lib8 the biologically active clone lib2-6-1 (SEQ ID NO:35) was randomized and subjected to re-selection to see if the binding and inhibition behavior of the clone could be improved. Lib8 was evolved as a 2'-OH pyrimidine RNA. In some cases, post-SELEX modification ofTGFB1 ligands derived from experiments 3-9 were performed, to determine if a ligand evolved with 2'-F pyrimidine Ssubstitutions would also bind with 2'-NH, substitutions.

Each starting pool for a SELEX experiment contained 3 x 101' RNA molecules (500 pmol). The affinity of the starting RNA for TGFB1 was estimated to be greater than 50 mM. After 4 rounds of SELEX, the affinities of the evolving pools had improved to approximately 10 nM and did not shift significantly in subsequent rounds. RNA was bulk sequenced and found to be non-random and cloned.

38 Libl took 20 rounds to evolve since optimum concentrations of TGF31 were not used until round 15 and libraries 5, 6 and 7 took longer to evolve because optimum conditions for recovery of bound ligands during the partitioning step in SELEX were not introduced until round 8. Optimum TGFBI concentrations and partitioning conditions are described in Example 1.

B. RNA Sequenes Many clones in a SELEX library are identical or similar in sequence. The libraries were screened by G track and only representatives of each G track type were tested in a binding assay. The binding assay was five points (16.5nM, 5.5nM, 1.8nM, 0.6nM, and 0.2nM) and could detect only those SELEX clones with a Kd less than or equal to 10 nM. RNA ligands that bound well in the binding assay were sequenced. The sequences were inspected by eye and analyzed using computer programs which perform alignments and fold nucleic acid sequences to predict regions of secondary structure. Ligands were classified into five groups B, C, D, and orphans) by sequence homology. Each group has characteristic allowable 2' substitutions.

58 clones were identified by G track from 7 separate SELEX experiments to belong to group A ligands (Table 3; SEQ ID NOS:12-42). 15 clones were sequenced; 13 were similar but not identical, whereas 3 clones, lib3-13 (SEQ ID NO:12), lib5-6 and lib5-13, were identical. Group A ligands were recovered from seven of the eight SELEX libraries which included libraries evolved as 2'-NH,, 2'-OH or 2'-F -substituted pyrimidines as well as a library evolved as 2'-F-UTP, 2'-NH 2 -CTP. Post SELEX modification indicates that 2'-NH 2 -UTP, 2'-F-CTP does not disrupt binding of lib8-9 to TGFB1, thus the structure of Group A ligands appears to not require a specific 2' moiety on the pyrimidine sugar in order to maintain binding.

Group B ligands bind both as 2'-NH, and 2'-F pyrimidine substituted RNA. 28 Group B clones were detected by G track analysis from 3 libraries. Two of the libraries were evolved as 2'-NH, and one as 2'-F library. Four clones were sequenced, two were identical (lib5-47 and lib4-12; SEQ ID NO:28). An internal deletion can occur in group B, as in lib 3-44. The 40N orphan, lib3-42 was placed in Group B on the basis of secondary structure. The internal deletion in lib3-44, the binding affinity, the bioactivity and boundary experiments all support the placement of lib3-42 in this group.

Group C ligands bind as 2'-OH or 2'-F ligands as expected, since members of this group were evolved as 2'-OH ligands in lib 1 and as 2'-F pyrimidine substituted ligands in lib 6. Libl-20-3 (SEQ ID NO:32) was post SELEX modified and as 2'-F derivative. Libl-20-3 did not bind with 2'-NH 2 pyrimidines incorporated.

Group D ligand, lib2-6-1 (SEQ ID NO:35), was isolated after 2'-OH SELEX but was post SELEX modified and binds well as a 2'-NH 2 -UTP and 2'-F-CTP pyrimidine derivative. Lib2-6-1 does not bind well to TGF8I with 2'-NH 2 2'-F or 2'F-UTP, 2'-NH 2 -CTP-substituted pyrimidines. Variants of Group D were only reselected in two other SELEX experiments, lib8, a 2'-OH library, and lib 9, a 2'-NH 2 -UTP, 2'-F CTP library, supporting the observation that there is specificity for the 2' pyrimidine position in this ligand.

The group labeled orphans are not homologous to each other and no variant sequences for these ligands have been determined. G track indicates that eight 40N clones similar to lib3-45 were isolated from two libraries. Two of the eight were sequenced and are identical. Lib3-45 (SEQ ID NO:39) binds whether it contains 2'-NH, or 2'-F substituted pyrimidines or the 2'-F-UTP, 2'-NH 2

-CTP

combination. Libl-20-5 (SEQ ID NO:40) isolated as a 2'-OH ligand binds as a whereas libl-20-12 (SEQ ID NO:41) and lib2-6-8 (SEQ ID NO:42) bind well only as 2'-OH pyrimidines and will not tolerate 2'-NH 2 or 2'-F post SELEX modifications.

As it was unusual that similar sequences were obtained from different SELEX experiments containing different modifications, another set of SELEX experiments was performed in search of RNA and ssDNA ligands to TGFB1 as described in examples 5 and 6 infra.

EXAMPLE 3. INHIBITION OF TGFli RECEPTOR BINDING The Kds and Bm, values reported in Table 4 for Group A ligands are for the 2'-NH, substituted version of the ligand unless otherwise noted. Bmu for the Group A ligands was 0.38±0.12 (n=14) which is in agreement with the measured retention of TGF81 on the nitrocellulose filters. The Kd's for Group A ligands were all similar, 2.2±1.1 nM Where measured nitrocellulose and SA agarose bead binding assays gave equivalent results.

The IC 5 o's in Table 4 for Group A ligands were all tested with the 2'-NH 2 pyrimidine substituted ligands except where indicated. 2'-NH 2 ligands were used in the tissue culture bio-assay since they exhibited the greatest stability under the conditions of the bio-assay. Five out of ten Group A ligands tested inhibited TGFB1 receptor activity. ICo values for the inhibitors were typically 25 fo!d above the Kd for TGFB1. The data are reproducible; the Kd for ligand lib3-13 was 0.83±0.11 nM and the IC,, for lib3-13 (SEQ ID NO:12) was 25±14 nM RNA concentrations in the bioassays are all estimates based on an assumed extinction coefficient and 100% purity of the ligand. The RNA concentrations may, therefore, be overestimated during the bio-assay which in turn would overestimate the IC 5 0 Another five Group A ligands did not inhibit TGFB receptor binding activity. One obvious difference between the non-bioactive ligands, lib2-6-4 (SEQ ID NO:20), lib5-48 (SEQ ID NO:19), and lib6-23 (SEQ ID NO:21), and the bioactive ligands is the substitution at nucleotide 72. Lib7-21 (SEQ ID NO:23) and lib7-43 (SEQ ID NO:24) were tested as 2'-F-UTP, 2'-NH 2

-CTP

ligands for bio-activity. These ligands were not bio-active despite their high affinity to TGFB. In conclusion, binding and bioactivity are separable functions of the TGFB Group A ligands.

Group B ligands have different binding properties than Group A ligands (Table Both the Kd (0.63±0.5 nM, n=4) and (0.14±0.04, n=4) are lower for Group B ligands. One Group B inhibitor, lib4-12 (SEQ ID NO:28), actually appears to stimulate TGFB1 activity in the tissue culture bio-assay at low concentrations. The basis of this mixed agonist/antagonist behavior has not been determined. The best inhibitor in this group, lib3-42 (SEQ ID NO:30) has an IC 50 of 22 nM and had no agonist behavior over the concentration ranges tested.

Group C ligands were tested as 2'-F derivatives and were not bio-active.

Neither was the 2'-F orphan libl-20-5 (SEQ ID NO:40). The orphan, lib3-45 is an example of another ligand with high affinity for TGFB1 and no ability to inhibit TGFB1 receptor binding.

Group D ligands were tested in the bio-assay as 2'-NH 2 -UTP, 2'-F-CTP derivatives. Both lib2-6-1 (SEQ ID NO:35) and the truncated version lib2-6-1-81 (SEQ ID NO:36) can inhibit TGFI1 receptor binding; however, a single mutation from a C to a G at position 53 decreases bio-activity in clone lib8-23. Similarly a 2 base pair deletion in clone lib6-30 (SEQ ID NO:34) at positions corresponding to nucleotides 67 and 68 in lib2-6-1 (SEQ ID NO:35) increases binding by 10 fold but eliminates bio-activity.

Lib2-6-1 (SEQ ID NO:35) was shown to be fully effective only against TGFI1 and not TGFB2 and TGFB3. Lib2-6-1 (SEQ ID NO:35) was biologically active in the presence of 10% horse serum in the cell culture medium in addition to the 0.1% BSA. Thus the ligand demonstrates specificity towards TGFOI which is not interferred with by the presence of the horse serum in this assay. The biggest indication that the inhibition of TGFB1 receptor binding is a specific phenomenon is the fact that not all TGFBI ligands block receptor binding, but the ones that do, do so reproducibly. There are no examples of ligands that do not bind to TGFB1 blocking TGFB1 receptor binding activity.

In summary, RNA ligands that can block TGFB1 receptor binding are a subset of ligands. Binding is necessary but not sufficient for bio-activity.

Roughly 50% of the high affinity ligands tested were inhibitors. Of the inhibitors, were good inhibitors (ICo 25 nM).

EXAMPLE 4. BOUNDARY ANALYSIS Truncation experiments were done on a number of TGFB1 ligands to determine the nucleotides essential for binding. Group A ligands, lib3-13 (SEQ ID NO:12) and lib8-9 (SEQ ID NO:16), were truncated with consistent results.

The fragment lib3-13-79 binds to TGFB1, thus none of the nucleotides 3' to nucleotide 79 in lib3-13 are essential for binding. Similarly when all nucleotides to nucleotide 38 are deleted the remaining fragment, lib3-13-(38-123) can still bind to TGFBI. The 5' boundary is in agreement with the sequence lib6-23 (SEQ ID NO:21), which has a deletion corresponding to nucleotides 19-36 of lib3-13 (SEQ ID NO:25), and still binds to TGFI31. Thus, all high affinity binding determinants for Group A clones may lie wholly within the random region and may correspond to a 42 nucleotide fragment, lib3-13-(38-79). Many Group A ligands contain deletions or substitutions within the predicted essential binding domain, in the region corresponding to lib3-13-(72-81). The deletion and substitution in lib4-32 have no effect on its 3' boundary which corresponds io lib3-13 nucleotide 80. Thus, the 3' boundary is probably correct and the alterations in nucleotide sequence 72-81 are ones that do not significantly alter the nucleic acid structure required for binding. Mutations in this region, most notably nucleotide 72 may, however, modify the ability of the ligand to block TGFB1 receptor binding as noted earlier.

Boundary analysis of the 3' end of Group B ligand, lib4-12 (SEQ ID NO:28), predicts that nothing beyond nucleotide 72 is required for TGFBI binding. When the 5' boundary of lib4-12 was determined, all but the first three nucleotides were required for binding, indicating that the 5' constant region is an essential part of the ligand at least when the boundary of the full length ligand was determined. Assuming that ligand lib3-44 (SEQ ID NO:29) has a similar binding determinant as lib4-12 (SEQ ID NO:28), we can also conclude that nucleotides 37-46 of lib4-12 are not required for binding since these are deleted in lib3-44 and lib3-42 (SEQ ID The 3' constant region is not necessary for binding in Group C and D ligands. Both ligand types bind without the 3' nucleotides in the constant region.

Libl-20-3-82, an 82 nucleotide truncated version of libl-20-3 (SEQ ID NO:32), binds as well as the full length libl-20-3. Likewise binding and bioactivity of lib2-6-1 is unaffected by the 3' truncation found in lib2-6-1-81 (SEQ ID NO:36).

EXAMPLE 5. EXPERIMENTAL PROCEDURES In the preferred embodiment, a second set of SELEX experiments was performed in search of RNA and DNA ligands with specific high affinity for TGFB1 from degenerate libraries containing 40 random positions (40N). This Example provides the general procedures followed and incorporated in Example 6.

A. Materials.

M-MLV superscript reverse transcriptase was purchased from Gibco BRL (Gaithersburg, MD). T7 RNA polymerase was purified according to standard procedures at NeXstar Pharmaceuticals, Inc. (Boulder, CO). Taq DNA polymerase (Amplitaq) was from Perkin Elmer/Cetus (Richmond, CA). T4 polynucleotide kinase, DNA polymerase (Klenow fragment), and alkaline phosphatase were purchased from New England Biolabs, Inc. (Beverly, MA). The 2'-amino substituted nucleotide triphosphates amino-UTP and amino-CTP were synthesized according to standard procedures at NeXstar Pharmaceuticals, Inc.

(Boulder, CO). Other reagents used in this work were of the highest quality obtainable.

B. Nucleic Acids.

RNAs were synthesized by in vitro transcription using double-stranded DNA oligonucleotides and T7 RNA polymerase. DNA oligonucleotides (Table were purchased from Operon, Inc. (Alameda, CA) and purified by 6% preparative polyacrylamide gel electrophoresis. PCR amplification was performed in 50 mM KC1, 10 mM Tris-HC1 (pH 2.5 mM MgCI 2 170 mg/mL BSA, and dNTPs (present at 1 mM each). Taq DNA polymerase was used at 100 units per 0.1 mL reaction, and the and 3'-primers were present at 1 mM. Transcription was performed in 40 mM NaCI, 10 mM dithiothreitol, 50 mM Tris-acetate (pH 8 mM magnesium acetate, 2 mM spermidine, and 2 mM NTP. T7 RNA polymerase was present at 1 unit/mL. The reaction was incubated at 28 degrees for 16 hours and then treated with 20 units of DNAse I for an additional 10 min at 37 degrees.

The reaction was stopped by the addition of one half volume of loading buffer r (93% formamide, 10 mM EDTA, pH 8.0) and heated to 95 degrees for 3 min prior to electrophoresis on a 6 polyacrylamide/8 M urea denaturing gel. The RNA transcript was visualized by UV shadowing and was excised from the gel and eluted into TE buffer (10 mM Tris-acetate pH 8.0, 2 mM EDTA). The RNA transcript was ethanol precipitated, dried under vacuum, and redissolved in distilled HO0. The concentration of RNA as well as single-stranded DNA was quantified by measuring the A 2 6 0 and assuming that 1 A 2 6 0 unit equaled 40 mg/mL and 33 mg/mL, respectively.

C. Evolution of High-Affinity Ligands.

SELEX ligands that bind to TGF1I were derived essentially as described in U.S. Patent No. 5,270,163 (see also Tuerk and Gold (1990) Science 249:505-510) using the oligonucleotides illustrated in Table 5 (SEQ ID NOS:43- 54). The DNA templates contained a 40-nucleotide (40N) variable sequence generated by mixed-nucleotide DNA synthesis, as well as and 3'-fixed sequences, necessary for PCR amplification of the template. The sequence of oligonucleotides 40N7 (SEQ ID NO:43) and 40N8 (SEQ ID NO:49) also contained a T7 RNA polymerase promoter. RNA for the first round of RNA SELEX was transcribed from double-stranded DNA templates generated by primer extension on single-stranded DNA templates 40N7 and 40N8 with the Klenow fragment of DNA polymerase I. RNA SELEX consisted of up to rounds of RNA synthesis, binding to target, partitioning of bound and unbound RNA by nitrocellulose filtration, cDNA synthesis, and PCR amplification to regenerate the double-stranded DNA template. Binding to the target by the RNA pool was performed in binding buffer A (120 mM NaC1, 2.5 mM KCI, 0.12 mM MgSO4, 40 mM HEPES, 20 mM NaH2PO4/Na2HPO4 pH 7.4, 0.01% HSA) at 37 degrees for at least 10 min prior to filtration. In contrast, the first round of single-stranded DNA SELEX was performed by using the synthetically synthesized oligonucleotides 40D7 and 40D8 directly. SELEX consisted of rounds of binding to target, partitioning of bound and unbound single-stranded DNA by nitrocellulose filtration, PCR amplification to generate a double-stranded DNA population, and preparative polyacrylamide gel electrophoresis to purify single-stranded DNA for the next round of SELEX. Binding of the target to the single-stranded DNA pool was performed in binding buffer B (150 mM NaCI, mM Tris-acetate pH 7.5, 0.001% BSA) at 37 degrees for at least 15 min prior to filtration. Radiolabeling of RNA as well as DNA repertoires was performed by incubation of 5 picomoles nucleic acid, 2 units of T4 polynucleotide kinase, and 6 mL ATP (800 Ci/mmol) in a volume of 10 mL at 37 degrees for 30 min.

The concentration of nucleic acid at each round of the SELEX experiment varied between 1500 nM and 1 nM while the concentration of the target TGF-BI varied between 150 nM and 0.03 nM.

D. Cloning and Sequencing of Ligands.

Cloning of the nucleic acid repertoire was performed as described by Tuerk and Gold (1990) Science 249:505-510 using double-stranded DNA that was generated from the RNA repertoire by PCR amplification. PCR-amplified DNA was digested with the restriction enzymes SphI and HindIII and ligated into compatible sites within pGEM. Ligated plasmids were transformed into E. coli and plated onto LB agar containing 5-bromo-4-chloro-3-indolyl B-D-galactoside, isopropyl thiogalactoside, and 100 mg/mL ampicillin. Colonies not expressing B-galactosidase were analyzed. Sequencing of DNA was performed as described by Tuerk and Gold (1990) using the dideoxynucleotide procedure of Sanger et al.

(1977) Proc. Natl. Acad. Sci. USA 74:5463-5467. Plasmids were isolated from E.

coli by the alkaline lysis miniprep procedure (Manitatis et al. (1982) in Molecular Cloning: A laboratory Manual, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY). DNA was incubated in 50 mM Tris-HC1 (pH 60 mM NaCl, 6 mM magnesium acetate, and 1 mM DTT with 0.4 mM dNTP and 0.2 mM dideoxy-NTP for 20 min at 48 degrees. DNA polymerase was present at 4 units per reaction. The reactions were stopped by the addition of 10 mL of loading buffer and heated to 95 degrees for 3 min prior to gel electrophoresis on a 6% polyacrylamine/8 M urea denaturing gel. G-track sequencing was performed as described and provided a convenient method to quickly screen the cloned library for ligands of different sequence. Briefly, the G-track sequencing reaction contained 50 mM Tris-HCl (pH 60 mM NaCI, 6 mM magnesium acetate, and 1 mM DTT with 0.4 mM dNTP, 0.2 mM dideoxy-GTP, and 4 units of DNA polymerase. The reaction was performed at 48 degrees for 20 min and was stopped by the addition of 10 uL of loading buffer and heated to 95 degrees for 3 min prior to gel electrophoresis on a 6% polyacrylamide/8 M urea denaturing gel.

EXAMPLE 6. BINDING ANALYSIS. BIOASSAY RESULTS. AND SEOUENCES OF A ssDNA LIBRARY.

Binding analysis of the 40D7 DNA library for TGF-B1 is shown in Figure 1. Binding data obtained from round 19 (triangles) and round 0 (circles) are shown. The experiment was performed by incubating nucleic acid (less than 1 nM) and the indicated concentration of TGF-bl in Binding Buffer (150 mM NaCI, mM Tris-acetate pH 8.2, 0.001% BSA) for 15 minutes at 37 degrees in a volume of 0.1 mL. Samples were filtered through nitrocellulose and were immediately followed by 3 mL of TE Buffer (10 mM Tris-acetate pH 8.0, 0.1 mM EDTA). The percentage of radiolabel bound was calculated from the amount of radiolabel retained on the nitrocellulose filter and the total radiolabel added to the binding reaction. The results show that the apparent Kd of the 40D7 library is InM, whereas the starting pool has an apparent Kd of 30 nM. Thus, the 40D7 library shows an increase of about three fold in binding.

A PAI-luciferase assay to detect TGF-bl activity in the presence of the nucleic acid libraries generated in Example 5 was performed as described in Abe et al. (1994) Analytical Biochem. 216:276-284. Mink lung epithelial cells containing the PAI-luciferase reporter gene were incubated with TGF-bl (10 pM) and oligonucleotides from the DNA libraries or anti -TGF-B antibody (60 gg/mL).

The mink lung epithelial cells were incubated for 18 hours and oligonucleotides were pre-incubated with TGF-bl before the assay and readded after 8 hours.

Addition of oligonucleotides alone (100 nM) to the cell culture did not affect the assay (data not shown). The identity of the oligonucleotide libraries as well as their effect on luciferase activity is indicated in Figure 2. The ssDNA library 40N7 completely inhibited the activity of TGF-B while the control (an equal concentration of randomized nucleic acid) showed a small stimulation of TGF-B I activity.

Based on the results of the binding analysis and PAI-luciferase assay, DNA ligands from the 40N7 library were sequenced as described in Example The sequences are shown in Table 6 (SEQ ID NOS:55-89). As the DNA 40N7 library showed inhibition in the PAI-luciferase bioassay, it is reasonable to suggest that the individual clones from the library are TGFp1 binders.

EXAMPLE 7. EXPERIMENTAL PROCEDURES This Example provides the general procedures followed and incorporated in Examples 8-15 for the evolution of nucleic acid ligands to PDGF.

A. Materials.

Recombinant human PDGF-AA (Mr-29,000), PDGF-AB (Mr=27,000) and PDGF-BB (Mr=25,000) were purchased from R&D Systems (Minneapolis, MN) in lyophilized form, free from carrier protein. All three isoforms were produced in E. coli from synthetic genes based on the sequences for the long form of the mature human PDGF A-chain (Betsholtz et al.,(1986) Nature 320: 695-699) and the naturally occurring mature form of human PDGF B-chain (Johnsson et al.,(1984) EMBO J. 921-928). Randomized DNA libraries, PCR primers and DNA ligands and 5'-iodo-2'-deoxyuridine-substituted DNA ligands were synthesized by NeXstar Pharmaceuticals, Inc. (Boulder, CO) or by Operon Technologies (Alameda, CA) using the standard solid phase phosphoramidite method (Sinha et al., (1984) Nucleic Acids Res. 12: 4539-4557).

B. Single Stranded DNA (ssDNA) Selex Essential features of the SELEX procedure have been described in detail in the SELEX Patent Applications (see also Tuerk and Gold, Science. 24: 505 (1990); Jellinek et al., Biochemistry. 31: 10450 (1994); Jellinek et al., Proc. Natl.

Acad. Sci.. 9: 11227 (1993)), which are incorporated by reference herein. The initial ssDNA library containing a contiguous randomized region of forty nucleotides, flanked by primer annealing regions (Table 7; SEQ ID NO:90) of invariant sequence, was synthesized by the solid phase phosphoramidite method using equal molar mixture of the four phosphoramidites to generate the randomized positions. The ssDNA library was purified by electrophoresis on an 8% polyacrylamide/7 M urea gel. The band that corresponds to the full-length DNA was visualized under UV light, excised from the gel, eluted by the crush and soak method, ethanol precipitated and pelleted by centrifugation. The pellet was dried under vacuum and resuspended in phosphate buffered saline supplemented with 1 mM MgC12 (PBSM 10.1 mM Na,HPO 1.8 mM KHPO 4 137 mM NaCI and 2.7 mM KCI, 1 mM MgCl,, pH 7.4) buffer. Prior to incubation with the protein, the ssDNA was heated at 90 0 C for 2 minutes in PBSM and cooled on ice.

The first selection was initiated by incubating approximately 500 pmol (3 x molecules) of 5' 2P end-labeled random ssDNA with PDGF-AB in binding buffer (PBSM containing 0.01% human serum albumin The mixture was incubated at 4 0 C overnight, followed by a brief (15 min) incubation at 37 0 C. The DNA bound to PDGF-AB was separated from unbound DNA by electrophoresis on an 8% polyacrylamide gel (1:30 bis-acrylamide:acrylamide) at 4°C and at V/cm with 89 mM Tris-borate (pH 8.3) containing 2 mM EDTA as the running buffer. The band that corresponds to the PDGF-ssDNA complex, which runs with about half the electrophoretic mobility of the free ssDNA, was visualized by autoradiography, excised from the gel and eluted by the crush and soak method.

In subsequent affinity selections, the ssDNA was incubated with PDGF-AB for minutes at 37 0 C in binding buffer and the PDGF-bound ssDNA was separated from the unbound DNA by nitrocellulose filtration, as previously described (Green,et al., (1995) Chemistry and Biology 2, 683-695). All affinity-selected ssDNA pools were amplified by PCR in which the DNA was subjected to 12-20 rounds of thermal cycling (30 s at 93 0 C, 10 s at 52*C, 60 s at 72*C) in 10 mM Tris-Cl (pH 8.4) containing 50 mM KC1, 7.5 mM MgCl 2 0.05 mg/ml bovine serum albumin, 1 mM deoxynucleoside triphosphates, 5 pM primers (Table 7) and 0.1 units/Il Taq polymerase. The 5' PCR primer was 5' end-labeled with polynucleotide kinase and [y-"P]ATP and the 3' PCR primer was biotinylated at the 5' end using biotin phosphoramidite (Glen Research, Sterling, VA). Following PCR amplification, streptavidin (Pierce, Rockford, IL) was added to the 49 unpurified PCR reaction mixture at a 10-fold molar excess over the biotinylated primer and incubated for 15 min at room temperature. The dsDNA was denatured by adding an equal volume of stop solution (90% formamide, 1% sodium dodecyl sulfate, 0.025% bromophenol blue and xylene cyanol) and incubating for 20 min at room temperature. The radiolabeled strand was separated from the streptavidin-bound biotinylated strand by electrophoresis on 12% polyacrylamide/ 7 M urea gels. The faster migrating radiolabeled (non-biotinylated) ssDNA strand was cut out of the gel and recovered as described above. The amount of ssDNA was estimated from the absorbance at 260 nm using the extinction coefficient of 33 zg/ml/absorbance unit (Sambrook et al., (1989) Molecular Cloning: A Laboratory Manual. 2 Ed. 3 vols., Cold Spring Harbor Laboratory Press, Cold Spring Harbor).

C. Cloning and Sequencing.

The amplified affinity-enriched pool from SELEX round 12 was purified on a 12% polyacrylamide gel and cloned between HindlII and PstI sites in JM109 strain of E. coli (Sambrook, et al., (1989) Molecular Cloning: A Laboratory Manual. 2 Ed. 3 vols., Cold Spring Harbor Laboratory Press, Cold Spring Harbor).

Individual clones were used to prepare plasmids by alkaline lysis. Plasmids were sequenced at the insert region using the forward sequencing primer and Sequenase 2.0 (Amersham, Arlington Heights, IL) according to the manufacturer's protocol.

D. Determination of the apparent equilibrium dissociation constants and the dissociation rate constants.

The binding of ssDNA ligands at low concentrations to varying concentrations of PDGF was determined by the nitrocellulose filter binding method as described (Green et al., (1995) Chemistry and Biology 2: 683-695).

The concentrations of PDGF stock solutions (in PBS) were determined from the absorbance readings at 280 nm using the following e 2 so values calculated from the amino acid sequences (Gill, S. and von Hippel, P. H. (1989) Anal. Biochm.

182: 319-326): 19,500 M'cm-' for PDGF-AA, 15,700 M-'cm- 1 for PDGF-AB and 11,800 for PDGF-BB. ssDNA for all binding experiments were purified by electrophoresis on 8% (>80 nucleotides) or 12% (<40 nucleotides) polyacrylamide/7 M urea gels. All ssDNA ligands were heated at 90 0 C in binding buffer at high dilution 1 nM) for 2 min and cooled on ice prior to further dilution into the protein solution. The binding mixtures were typically incubated for 15 min at 37°C before partitioning on nitrocellulose filters.

The binding of DNA ligands to PDGF-AA is adequately described with the bimolecular binding model for which the fraction of bound DNA at equilibrium is given by eq. 1, q=(f/2 2 (1) where and are total protein and total DNA concentrations, Kd is the equilibrium dissociation constant and f is the efficiency of retention of protein-DNA complexes on nitrocellulose "lters (Irvine et al., (1991) J. Mol. Biol.

222: 739-761; Jellinek et al., (1993) Proc. Nat'l. Acad. Sci. USA e0: 11227-11231).

The binding of DNA ligands to PDGF-AB and PDGF-BB is biphasic and can be described by a model in which the DNA ligand is composed of two non-interconverting components and L 2 that bind to the protein with different affinities, described by corresponding dissociation constants, Kdi and Ke (Jellinek et al., 1993) Proc. Nat'l. Acad. Sci. USA 90: 11227-11231). In this case, the explicit solution for the fraction of bound DNA is given by eq. 2, X/ IKd X2 2 [P] q=f (2) S 1l+Kd 1 l+Kd2 [P] with [P]t XIKdi [L]t X2 Kd2 [L]t 1+ l+Kdl l+Kd2[P] where x, and X2(=l-XI) are the mole fractions of L, and The Kd values for the binding of DNA ligands to PDGF were calculated by fitting the data points to eq.

1 (for PDGF-AA) or eq. 2 (for PDGF-AB and PDGF-BB) using the non-linear least squares method.

The dissociation rate constants were determined by measuring the amount of "P 5'-end labeled minimal ligands (0.17 nM) bound to PDGF-AB (1 nM) as a function of time following the addition of 500-fold excess of unlabeled ligands, using nitrocellulose filter binding as the partitioning method. The k,, values were determined by fitting the data points to the first-order rate equation (eq. 3) exp(-kt) (3) where q, q, and q. represent the fractions of DNA bound to PDGF-AB at any time t=0 and respectively.

E. Minimal ligand determinations.

To generate a population of 5' end-labeled DNA ligands serially truncated from the 3' end, a primer complementary to the 3' invariant sequence region of a DNA ligand template (truncated primer 5N2, Table 7; SEQ ID NO:92) was radiolabeled at the 5' end with [y-"P]-ATP and T4 polynucleotide kinase, annealed to the template and extended with Sequenase (Amersham, Arlington Heights, IL) and a mixture of all four dNTPs and ddNTPs. Following incubation in binding buffer for 15 min at 37 0 C, the fragments from this population that retain high affinity binding to PDGF-AB were separated from those with weaker affinity by nitrocellulose filter partitioning. Electrophoretic resolution of the fragments on 8% polyacrylamide/7 M urea gels, before and after affinity selection, allows determination of the 3' boundary. To generate a population of 3' end-labeled DNA ligands serially truncated from the 5' end, the DNA ligands were radiolabeled at the 3' end with [a-"P]-cordycepin-5'-triphosphate (New England Nuclear, Boston, MA) and T4 RNA ligase (Promega, Madison, WI), phosphorylated at the 5' end with ATP and T4 polynucleotide kinase, and partially digested with lambda exonuclease (Gibco BRL, Gaithersburg, MD). Partial digestion of 10 pmols of 3'-labeled ligand was done in 100 ML volume with 7 mM glycine-KOH (pH 2.5mM MgCI 2 1 ug/ml BSA, 15 mg tRNA, and 4 units of lambda exonuclease for 15 min at 37°. The 5' boundary was determined in an analogous manner to that described for the 3' boundary.

F. Melting temperature measurements.

Melting profiles for the minimal DNA ligands were obtained on a Cary Model IE spectrophotometer. Oligonucleotides (320-400 nM) were heated to in PBS, PBSM or PBS with 1 mM EDTA and cooled to room temperature prior to the melting profile determination. Melting profiles were generated by heating the samples at the rate of I OC/min from 15-95°C and recording the absorbance every 0.1 OC. The first derivative of the data points was calculated using the plotting program KaleidaGraph (Synergy Software, Reading, PA). The first derivative values were smoothed using a 55 point smoothing function by averaging each point with 27 data points on each side. The peak of the smoothed first derivative curves was used to estimate the T, values.

G. Crosslinking of 5-iodo-2'-deoxvuridine-substituted DNA ligands to

PDGF-AB.

DNA ligands containing single or multiple substitutions of 5'-iodo-2'deoxyuridine for thymidine were synthesized using the solid phase phosphoramidite method. To test for the ability to crosslink, trace amounts of 3 2 P end-labeled ligands were incubated with PDGF-AB (100 nM) in binding buffer at 370 for 15 min prior to irradiation. The binding mixture was transferred to a 1 cm path length cuvette thermostated at 37° and irradiated at 308 nm for 25-400 s at 20 Hz using a XeCl charged Lumonics Model EX748 excimer laser.

The cuvette was positioned 24 cm beyond the focal point of a convergent lens, with the energy at the focal point measuring 175 mjoules/pulse. Following irradiation, aliquots were mixed with an equal volume of formamide loading buffer containing 0.1% SDS and incubated at 95" for 5 min prior to resolution of the crosslinked PDGF/ligand complex from the free ligand on 8% polyacrylamide/7 M urea gels.

To identify the protein site of crosslinking for ligand 20t-I4, binding and irradiation were done on a larger scale. PDGF-AB and 5' 3 P end-labeled ligand, each at 1 uM in PBSM, were incubated and irradiated (300 s) as described above in two 1 ml reaction vessels. The reaction mixtures were combined, ethanol precipitated and resuspended in 0.3 ml of Tris-HCl buffer (100mM, pH The PDGF-AB/ligand crosslinked complex was digested with 0.17 ,g/Ul of modified trypsin (Boehringer Mannheim) for 20 hours at 37°. The digest mixture was extracted with phenol/chloroform, chloroform and then ethanol precipitated. The pellet was resuspended in water and an equal volume of formamide loading buffer with 5% 1-mercaptoethanol (no SDS), incubated at 950 for 5 min, and resolved on a 40 cm 8% polyacrylamide/7 M urea gel. The crosslinked tryptic-peptide/ligand that migrated as two closely spaced bands about 1.5 cm above the free ligand band was excised from the gel and eluted by the crush and soak method and ethanol precipitated. The dried crosslinked peptide (about 160 pmoles based on the specific activity) was sequenced by Edman degradation (Midwest Analytical, Inc., St. Louis, MO).

H. Receptor Binding Assay.

The binding of 2 I-PDGF-AA and 5 I-PDGF-BB to porcine aortic endothelial (PAE) cells transfected with PDGF a- or B-receptors were performed as described (Heldin et al., (1988) EMBO J. 1, 1387-1394). Different concentrations of DNA ligands were added to the cell culture (1.5 cm 2 in 0.2 ml of phosphate buffered saline supplemented with 1 mg bovine serum albumin per ml together with 2 I-PDGF-AA (2 ng, 100,000 cpm) or 1 25 -PDGF-BB (2 ng, 100,000 cpm). After incubation at 4"C for 90 min, the cell cultures were washed and cell associated radioactivity determined in a y-counter (Heldin et al., (1988) EMBO J. 2, 1387-1394).

L Hlthymidine Incorporation Assay.

The incorporation of ['H]thymidine into PAE cells expressing PDGF B-receptor in response to 20 ng/ml of PDGF-BB or 10% fetal calf serum and in the presence of different concentrations of DNA ligands was performed as described 54 (Mori et al., (1991) J. Biol. Chem. 266, 21158-21164). After incubation for 24 h at 37 0 C, 3 H-radioactivity incorporated into DNA was determined using a B-counter.

EXAMPLE 8. ssDNA LIGANDS OF PDGF High affinity DNA ligands to PDGF AB were identified by the SELEX process from a library of =3 x 10" 4 molecules (500 pmol) of single stranded DNA randomized at forty contiguous positions (Table 7; SEQ ID NO:90). The PDGF-bound DNA was separated from unbound DNA by polyacrylamide gel electrophoresis in the first round and by nitrocellulose filter binding in the subsequent rounds. After 12 rounds of SELEX, the affinity-enriched pool bound to PDGF-AB with an apparent dissociation constant (Kd) of =50 pM (data not shown). This represented an improvement in affinity of =700-fold compared to the initial randomized DNA library. This affinity-enriched pool was used to generate a cloning library from which 39 isolates were sequenced. Thirty-two of these ligands were found to have unique sequences (Table 8; SEQ ID NOS:93- 124). Ligands that were subjected to the minimal sequence determination are marked with an asterisk next to the clone number. The clone numbers that were found to retain high affinity binding as minimal ligands are italicized. All ligands shown in Table 8 were screened for their ability to bind to PDGF AB using the nitrocellulose filter binding method. To identify the best ligands from this group, we determined their relative affinities for PDGF-AB by measuring the fraction of 5' 3 "P end-labeled ligands bound to PDGF-AB over a range of protein concentrations. For the ligands that bound to PDGF-AB with high affinity, the affinity toward PDGF-BB and PDGF-AA was also examined: in all cases, the affinity of ligands for PDGF-AB and PDGF-BB was comparable while the affinity for PDGF-AA was considerably lower (data not shown).

Twenty-one of the thirty-two unique ligands can be grouped into a sequence family shown in Table 9. The sequences of the initially randomized region (uppercase letters) are aligned according to the consensus three-way helix junction motif. Nucleotides in the sequence-invariant region (lowercase letters) are only shown where they participate in the predicted secondary structure.

Several ligands were "disconnected" (equality symbol) in order to show their relatedness to the consensus motif through circular permutation. The nucleotides predicted to participate in base pairing are indicated with underline inverted arrows, with the arrow heads pointing toward the helix junction. The sequences are divided into two groups, A and B, based on the first single stranded nucleotide (from the 5' end) at the helix junction (A or G, between helices II and III).

Mismatches in the helical regions are shown with dots under the corresponding letters (G-T and T-G base pairs were allowed). In places where single nucleotide bulges occur, the mismatched nucleotide is shown above the rest of the sequence between its neighbors.

This classification is based in part on sequence homology among these ligands, but in greater part on the basis of a shared secondary structure motif: a three-way helix junction with a three nucleotide loop at the branch point (Figure These ligands were subdivided into two groups; for ligands in group A, the loop at the branch point has an invariant sequence AGC and in group B, that sequence is The proposed consensus secondary structure motif is supported by base-pairing covariation at non-conserved nucleotides in the helices (Table 10). Since the three-way junctions are encoded in continuous DNA strands, two of the helices end in loops at the distal end from the junction. These loops are highly variable, both in length and in sequence. Furthermore, through circular permutation of the consensus motif, the loops occur in all three helices, although they are most frequent in helices II and III. Together these observations suggest that the regions distal from the helix junction are not important for high affinity binding to PDGF-AB. The highly conserved nucleotides are indeed found near the helix junction (Table 9, Figure 3).

EXAMPLE 9. BOUNDARY ANALYSIS The minimal sequence necessary for high affinity binding was determined for the six best ligands to PDGF-AB. In general, the information about the 3' and 56 minimal sequence boundaries can be obtained by partially fragmenting the nucleic acid ligand and then selecting for the fragments that retain high affinity for the target. With RNA ligands, the fragments can be conveniently generated by mild alkaline hydrolysis (Tuerk et al., (1990) J. Mol. Biol. 2a: 749-761; Jellinek et al., (1994) Biochemistry 10450-10456; Jellinek et al., (1995) Biochemistry 34: 11363-11372; Green etal., (1995J. Mol. Biol. 247: 60-68).

Since DNA is more resistant to base, an alternative method of generating fragments is needed for DNA. To determine the 3' boundary, a population of ligand fragments serially truncated at the 3' end was generated by extending the end-labeled primer annealed to the 3' invariant sequence of a DNA ligand using the dideoxy sequencing method. This population was affinity-selected by nitrocellulose filtration and the shortest fragments (truncated from the 3' end) that retain high affinity binding for PDGF-AB were identified by polyacrylamide gel electrophoresis. The 5' boundary was determined in an analogous manner except that a population of 3' end-labeled ligand fragments serially truncated at the 5' end was generated by limited digestion with lambda exonuclease. The minimal ligand is then defined as the sequence between the two boundaries. It is important to keep in mind that, while the information derived from these experiments is useful, the suggested boundaries are by no means absolute since the boundaries are examined one terminus at a time. The untruncated (radiolabeled) termini can augment, reduce or have no effect on binding (Jellinek et al., (1994) Biochemistry 33: 10450-10456).

Of the six minimal ligands for which the boundaries were determined experimentally, two (20t (SEQ ID NO:172) and 41t (SEQ ID NO:174); truncated versions of ligands 20 and 41) bound with affinities comparable (within a factor of 2) to their full-length analogs and four had considerably lower affinities. The two minimal ligands that retained high affinity binding to PDGF, 20t and 41t, contain the predicted three-way helix junction secondary structure motif (Figure The sequence of the third minimal ligand that binds to PDGF-AB with high affinity, 36t (SEQ ID NO:173), was deduced from the knowledge of the consensus motif (Figure In subsequent experiments, we found that the single-stranded region at the 5' end ofligand 20t is not important for high affinity binding. Furthermore, the trinucleotide loops on helices II and III in ligand 36t (GCA and CCA) can be replaced with pentaethylene glycol spacers (infra). These experiments provide further support for the importance of the helix junction region in high affinity binding to PDGF-AB.

The binding of minimal ligands 20t, 36t, and 41t to varying concentrations of PDGF-AA, PDGF-AB and PDGF-BB is shown in Figures 5A, 5B and 5C. In agreement with the binding properties of their full length analogs, the minimal ligands bind to PDGF-AB and PDGF-BB with substantially higher affinity than to PDGF AA (Figures 5A, 5B, and 5C, Table 11). In fact, their affinity for PDGF-AA is comparable to that of random DNA (data not shown). The binding to PDGF-AA is adequately described with a monophasic binding equation while the binding to PDGF-AB and PDGF-BB is notably biphasic. In previous SELEX experiments, biphasic binding has been found to be a consequence of the existence of separable nucleic acid species that bind to their target protein with different affinities (Jellinek et al., (1995) Biochemistry 34: 11363-11372) and unpublished results). The identity of the high and the low affinity fractions is at present not known. Since these DNA ligands described here were synthesized chemically, it is possible that the fraction that binds to PDGF-AB and PDGF-BB with lower affinity represents chemically imperfect DNA. Alternatively, the high and the low affinity species may represent stable conformational isomers that bind to the PDGF B-chain with different affinities. In any event, the higher affinity binding component is the most populated ligand species in all cases (Figures 5B and For comparison, a 39-mer DNA ligand that binds to human thrombin with a Kd of 0.5 nM (ligand T39 (SEQ ID NO.:177): 5'-CAGTCCGTGGTAGGGCAGGTTGGGGTGACTTCGTGGAA[3T], where [3T] represents a linked thymidine nucleotide added to reduce 3'-exonuclease degradation) and has a predicted stem-loop structure, binds to PDGF-AB with a Kd of 0.23 pM (data not shown).

EXAMPLE 10. KINETIC STABILITY OF PDGF-NUCLEIC

ACID

LIGAND COMPLEXES In order to evaluate the kinetic stability of the PDGF-AB/DNA complexes, the dissociation rates were determined at 37 0 C for the complexes of minimal ligands 20t, 36t and 41t (SEQ ID NOS:172-174) with PDGF-AB by measuring the amount of radiolabeled ligands (0.17 nM) bound to PDGF-AB (1 nM) as a function of time following the addition of a large excess of unlabeled ligands (Figure At these protein and DNA ligand concentrations, only the high affinity fraction of the DNA ligands binds to PDGF-AB. The following values for the dissociation rate constants were obtained by fitting the data points shown in Figure 6 to the first-order rate equation: 4.5 0.2 x 10-' s' (tia 2.6 min) for ligand 20t, 3.0 0.2 x 10- s 3.8 min) for ligand 36t, and 1.7 0.1 x 10 3 s-' (tia 6.7 min) for ligand 41t. The association rates calculated for the dissociation constants and dissociation rate constants (kon=kon/Kd) are 3.1 x 107 M's' l for 3.1 x 107 for 36t and 1.2 x 10' for 41t.

EXAMPLE 11. THERMAL MELTING PROPERTIES In order to examine the ability of minimal ligands 20t, 36t and 41t to assume folded structures, their melting temperatures were determined from the UV absorbance vs. temperature profiles in PBSM or PBSE buffers. At the oligonucleotide concentrations used in these experiments (320-440 nM), only the monomeric species were observed as single bands on non-denaturing polyacrylamide gels (data not shown). Ligands 20t and 41t underwent thermal melting that is well described by a two-state (folded and unfolded) model with linearly sloping baselines (Petersheim and Turner (1983) Biochem. 22:256-263) with Tm values in PBSM buffer of 43.8 0.4 *C and 49.2 0.5 respectively.

In PBSE buffer, similar Tm values were obtained: 44.8 0.5 *C for ligand 20t and 48.0 0.50C for ligand 41t. Ligand 36t exhibited a more complex thermal melting profile in which two distinct transitions were observed. In this case, the data were well described by a three-state model in which the fully folded and the unfolded states are connected through a partially unfolded intermediate results.

Using this model, we obtained two Tm values for ligand 36t: 47.0 0.9 *C and 67.1 3.8 "C in PBSM buffer and 44.2 1.7 "C and 64.3 4.1 *C in PBSE buffer.

EXAMPLE 12. PHOTO-CROSSLINKING OF NUCLEIC ACID LIGANDS AND PDGF In order to determine the sites on the DNA ligands and PDGF that are in close contact, a series of photo-crosslinking experiments were performed with 5'-iodo-2'-deoxyuridine (IdU)-substituted DNA ligands 20t, 36t and 41t (SEQ ID NOS:172-174). Upon monochromatic excitation at 308 nm, 5-iodo- and 5-bromo-substituted pyrimidine nucleotides populate a reactive triplet state following intersystem crossing from the initial n to n* transition. The excited triplet state species then reacts with electron rich amino acid residues (such as Trp, Tyr and His) that are in its close proximity to yield a covalent crosslink. This method has been used extensively in studies of nucleic acid-protein interactions since it allows irradiation with >300 nm light which minimizes photodamage (Willis et al., (1994) Nucleic Acids Res. 22: 4947-4952; Stump, W. and Hall, K. B. (1995) RNA 1: 55-63; Willis et al., (1993) Science 262: 1255-1257; Jensen et al., (1995) Proc. Natl. Acad. Sci.. U. S. A. 22: 12220-12224). Analogs of ligands 20t, 36t and 41t were synthesized in which all thymidine residues were replaced with IdU residues using the solid phase phosphoramidite method. The affinity of these IdU-substituted ligands for PDGF-AB was somewhat enhanced compared to the unsubstituted ligands and based on the appearance of bands with slower electrophoretic mobility on 8% polyacrilamide/7 M urea gels, all three end-labeled IdU-substituted ligands crosslinked to PDGF-AB upon irradiation at 308 nm (data not shown). The highest crosslinking efficiency was observed with IdU-substituted ligand 20t. To identify the specific IdU position(s) responsible for the observed crosslinking, seven singly or multiply IdU-substituted analogs of were tested for their ability to photo-crosslink to PDGF-AB: ligands through 20t-17 (5'-TGGGAGGGCGCGT'T'CT'T'CGT 2

GGT

3

TACT

5 T6T6TAGT 7 CCCG-3' (SEQ ID NOS:178-184) where the numbers indicate IdU substitutions at indicated thymidine nucleotides for the seven ligands). Of these seven ligands, efficient crosslinking to PDGF-AB was observed only with ligand 20t-14. The photo-reactive IdU position corresponds to the 3' proximal thymidine in the loop at the helix junction (Figure 4).

To identify the crosslinked amino acid residue(s) on PDGF-AB, a mixture of 5' end-labeled 20t-14 and PDGF-AB was incubated for 15 min at 37 0 C followed by irradiation at 308 nm. The reaction mixture was then digested with modified trypsin and the crosslinked fragments resolved on an 8% polyacrylamide/7 M urea gel. Edman degradation of the peptide fragment recovered from the band that migrated closest to the free DNA band revealed the amino acid sequence KKPIXKK (SEQ ID NO:185), where X indicates a modified amino acid that could not be identified with the 20 derivatized amino acid standards. This peptide sequence, where X is phenylalanine, corresponds to amino acids 80-86 in the PDGF-B chain (Johnsson et al., (1984) EMBO J. 3: 921-928) which in the crystal structure of PDGF-BB comprises a part of solvent-exposed loop III (Oefner et al., (1992) EMBO J. 11: 3921-3926). In the PDGF A-chain, this peptide sequence does not occur (Betsholtz et al., (1986) Nature 320, 695-699). Together, these data establish a point contact between a specific thymidine residue in ligand and phenylalanine 84 of the PDGF B-chain.

EXAMPLE 13. INHIBITION OF PDGF BY NUCLEIC ACID LIGANDS In order to determine whether the DNA ligands to PDGF were able to inhibit the effects of PDGF isoforms on cultured cells, the effects on binding of 12"I-labeled PDGF isoforms to PDGF a- and B-receptors stably expressed in porcine aortic endothelial (PAE) cells by transfection was determined. Ligands 36t and 41t (SEQ ID NOS:172-174) all efficiently inhibited the binding of 2 I-PDGF-BB to PDGF a-receptors (Figure 7) or PDGF B-receptors (data not shown), with half maximal effects around 1 nM of DNA ligand. DNA ligand T39, directed against thrombin and included as a control, showed no effect. None of the ligands was able to inhibit the binding of 2 I-PDGF-AA to the PDGF a-receptor (Figure consistent with the observed specificity of ligands 20t, 36t and 41t for PDGF-BB and PDGF-AB.

The ability of the DNA ligands to inhibit the mitogenic effects of PDGF-BB on PAE cells expressing PDGF B-receptors was investigated. As shown in Figure 8, the stimulatory effect of PDGF-BB on 3 H]thymidine incorporation was neutralized by ligands 20t, 36t and 41t. Ligand 36t exhibited half maximal inhibition at the concentration of 2.5 nM; ligands 41t was slightly more efficient and 20t slightly less efficient. The control ligand T39 had no effect.

Moreover, none of the ligands inhibited the stimulatory effects of fetal calf serum on ['H]thymidine incorporation in these cells, showing that the inhibitory effects are specific for PDGF.

EXAMPLE 14. POST-SELEX PROCESS NUCLEOTIDE

SUBSTITUTIONS

The stability of nucleic acids to nucleases is an important consideration in efforts to develop nucleic acid-based therapeutics. Experiments have shown that many, and in some cases most of the nucleotides in SELEX-derived ligands can be substituted with modified nucleotides that resist nuclease digestion, without compromising high affinity binding (Green et al., (1995) Chemistry and Biology 2: 683-695; Green et al., (1995) J. Mol. Biol. 242, 60-68). Experiments of this type with the DNA ligands reported here suggest that substitutions with modified nucleotides are tolerated at many positions (Figure 9; SEQ ID NOS:175-176).

Specifically, we have examined the substitution of 2'-O-methyl-2'-deoxy- and 2'-fluoro-2'-deoxyribonucleotides for 2'-deoxyribonucleotides in ligand 36t, by examining the PDGF-AB binding properties of singly or multiply substituted ligand 36t. The substitution pattern indicated in Figure 9 is compatible with high affinity binding to PDGF-AB. Furthermore, this ligand tolerates the substitution of pentaethylene glycol spacers (Glen Research, Sterling, VA) for the trinucleotide loops at the ends of helices II and III (Figure These DNA ligands therefore represent lead compounds for a novel class of high affinity, specific antagonists of PDGF-AB and PDGF-BB.

EXAMPLE 15. EXPERIMENTAL PROCEDURE FOR EVOLVING 2'-FLUORO-2'-DEOXYPYRIMIDINE RNA LIGANDS TO PDGF AND RNA SEOUENCES OBTAINED.

A. 2'-FLUORO-2'-DEOXYPYRIMIDINE RNA SELEX SELEX with 2'-fluoro-2'-deoxypyrimidine RNA targeting PDGF AB was done essentially as described previously (vide supra, and Jellinek et al., (1993, 1994) supra) using the primer template set as shown in Table 12 (SEQ ID NOS:125-127). Briefly, the 2'-fluoro-2'-deoxypyrimidine RNA for affinity selections was prepared by in vitro transcription from synthetic DNA templates using T7 RNA polymerase (Milligan et al., Nucl. Acids Res., 8783 (1987)).

The conditions for in vitro transcription described in detail previously (Jellinek et al., (1994) supra) were used, except that higher concentration (3 mM) of the 2'-fluoro-2'-deoxypyrimidine nucleoside triphosphates (2'-F-UTP and 2'-F-CTP) was used compared to ATP and GTP (1 mM). Affinity selections were done by incubating PDGF AB with 2'-fluoro-2'-deoxypyrimidine RNA for at least 15 min at 37 °C in PBS containing 0.01% human serum albumin. Partitioning of free RNA from protein-bound RNA was done by nitrocellulose filtration as described (Jellinek et al., (1993, 1994) supra). Reverse transcription of the affinity-selected RNA and amplification by PCR were done as described previously (Jellinek et al., (1994) supra). Nineteen rounds of SELEX were performed, typically selecting between 1-12% of the input RNA. For the first eight rounds of selection, suramin (3-15 pM) was included in the selection buffer to increase the selection pressure.

The affinity-enriched pool (round 19) was cloned and sequenced as described (Schneider et al., (1992) supra). Forty-six unique sequences have been identified, and the sequences are shown in Table 13 (SEQ ID NOS:128-170). The unique-sequence ligands were screened for their ability to bind PDGF AB with high affinity. While random 2'-fluoropyrimidine RNA (Table 12) bound to PDGF with a dissociation constant (Kd) of 35 7 nM, many of the affinity-selected ligands bound to PDGF AB with 100-fold higher affinities. Among the unique ligands, clones 9 (Kd 91 16 pM), 11 (Kd 120 21 pM), 16 (Kd= 116 34 pM), 23 (Kd 173 38 pM), 25 (Kd 80 22 pM), 37 (Kd 97 29 pM), 38 (Kd 74 39 pM), and 40 (Kd 91 32 pM) exhibited the highest affinity for PDGF AB (binding of all of these ligands to PDGF AB is biphasic and the Kd for the higher affinity binding component is given).

EXAMPLE 16. EXPERIMENTAL PROCEDURES This Example provides the general procedures followed and incorporated in Examples 17-19 for the evolution of nucleic acid ligands to hKGF.

A. Materials and Methods Recombinant human Keratinocyte Growth Factor (hKGF) and human Epidermal Growth Factor (hEGF) were purchased from Upstate Biotechnology Inc.(Lake Placid, NY). haFGF, hbFGF, PDGF-AB, TGFpl, and anti-KGF neutralizing monoclonal antibody were purchased from R&D Systems (Minneapolis, MN). Recombinant rat KGF was purchased from QED Advanced Research Technologies (San Diego, CA). Human thrombin was purchased from Enzyme Research Laboratories (South Bend, IN). T4 DNA ligase, HpaII methylase, and restriction enzymes were purchased from New England Biolabs (Beverly, MA). pCR-Script Amp cloning kit was purchased from Stratagene (La Jolla, CA). AMV reverse transcriptase was purchased from Life Sciences (St. Petersburg, FL). Taq DNA polymerase was purchased from Perkin Elmer (Foster City, CA). Ultrapure nucleotide triphosphates were purchased from Pharmacia (Piscataway, NJ). a-"-p-ATP, y- 2 P-ATP, and 5' 3 2 P-cytidine (phosphate) (5' 3 2 P-pCp) were from DuPont NEN Research Products (Boston, MA). 'sI-labeled KGF was prepared as described before (Bottaro et al., (1990) J.Biol.Chem. 26:12767-12770). PC-3 prostatic carcinoma cells were obtained from ATCC (catalog number CRL1435). Balb/MK cells and NIH3T3 transfected cells with the human KGF receptor (NIH3T3/KGFR) were a generous gift from S.

Aaronson, Mt. Sinai Medical Center, NY, and have been described elsewhere (Miki et al., (1992) Proc.Natl.Acad.Sci.USA 89:246-250; Miki et al., (1991) Science 251:72-75; Weissman et al., (1983) Cell 2;599-606). T7 RNA polymerase, 2'NH 2 and 2'F-modified CTP and UTP were from NeXstar 64 Pharmaceuticals, Inc. (Boulder, CO). DNA oligonucleotides were obtained from Operon Technologies, Inc. (Alameda, CA). Nitrocellulose/cellulose acetate mixed matrix, 0.45 pm, HA filters were from Millipore (Bedford, MA). Calcium and magnesium containing Dulbeco's Phosphate Buffered Saline (DPBS) was purchased from Life Technologies (Gaithersburg, MD). Chemicals were at least reagent grade and purchased from commercial sources.

B. SELEX The SELEX procedure has been described in detail in US patent 5,270,163 (see also Tuerk and Gold (1990) Science 249:505-510). A single-stranded DNA (ssDNA) pool was used to generate the double-stranded (dsDNA) template for generating the initial random sequence RNA pool by transcription. The DNA template contained 40 random nucleotides, flanked by 5' and 3' constant regions for primer anealing sites for PCR and cDNA synthesis (Table 14; SEQ ID NOS: 186-188). The 5' primer contains the T7 promotor sequence for in vitro transcriptions. The template was PCR amplified following an initial denaturation at 93°C for 3.5 minutes through 15 cycles of 30 second denaturation at 93°C, 1 minute annealing at 60 0 C, and 1 minute elongation at 72 0 C, in 50 mM KCI, Tris-HC1, pH9, 0.1% Triton X-100, 3 mM MgCl 2 0.5 mM of each dATP, dCTP, dGTP, and dTTP, 0.1 units/l Taq DNA polymerase, and 2.5 nM each of 3G7 and 5G7 primers (Table 14; SEQ ID NOS.187-188). SELEX experiments for hKGF were initiated with a random sequence pool of RNA in which all pyrimidines were 2'-NH 2 -modified or 2'-F-modified. Transcription reactions were done with about 5.M DNA template, 5 units/gl T7 RNA polymerase, Tris-HCl (pH8), 12 mM MgCl,, 5mM DTT, ImM spermidine, 0.002% Triton X-100, 4% PEG 8000,2-4 mM each 2'OH ATP, 2'OH GTP, 2'NH2 or 2'F CTP, 2'NH2 or 2'F UTP, and 0.25 fM a 32 P 2'OH ATP (800 Ci/mmole). The full length transcripts were gel-purified prior to use. To prepare binding reactions, the RNA molecules were incubated with recombinant hKGF in Dulbecco's Phosphate-Buffered Saline (DPBS) with calcium and magnesium (Life Technologies, Gaithersburg, MD, Cat. No 21300-025) containing 0.01% human serum albumin. Following incubation at room temperature (ranging from minutes to 10 hours) the protein-RNA complexes were partitioned from unbound RNA by filtering through nitrocellulose. Nitrocellulose filter bound RNA was recovered by phenol/urea extraction. The partitioned RNA was reverse transcribed into cDNA by AMV reverse transcriptase at 48 0 C for 60 min in mM Tris-HCl pH8.3, 60 mM NaCI, 6 mM Mg(OAc),, 10mM DTT, 50 pmol DNA 3' primer (Table 14), 0.4 mM each of dATP, dCTP, dGTP, and dTTP, and 1 unit/Ml AMV RT. The cDNA was PCR amplified and used to initiate the next SELEX cycle.

C. Nitrocellulose Filter Partitioning In order to partition the protein-RNA complexes, the binding reactions were filtered through nitrocellulose/cellulose acetated mixed matrix, 0.45 um pore size (filter disks, Millipore, Co., Bedford, MA). For filtration, the filters were placed onto a vacuum manifold and wetted by aspirating 5 ml of DPBS. The binding reactions were aspirated through the filters, and following a 5 ml wash, the filters were counted in a scintillation counter (Beckmann). Higher wash volumes with DPBS or 0.5 M urea were used as a means to increase selection stringency as shown in Table 15. Gel purified, internally a-"P-ATP labeled transcripts were incubated with various concentrations ofhKGF in DPBS at 37"C for 10 minutes. Oligonucleotide protein mixtures were filtered through prewetted 0.45 pm pore size HA filters, followed by a 5 ml wash with DPBS. The radioactivity retained on the filter was counted and corrected for background binding in the absence of protein. Nonlinear least square method was used to fit the data into monophasic or biphasic binding curves and to obtain the equilibrium dissociation constant Kd (Jellinek et al., (1993) Proc.Natl.Acad.Sci. USA 90:11227-11231) using the software package Kaleidagraph (Synergy Software, Reading, PA). Biphasic binding can be described as the binding of two affinity species that are not in equilibrium.

D. Cloning and Sequencing The RNA recovered from the round 8 filters was reverse transcribed and PCR amplified. Following column purification with QIA-quick spin columns (Qiagen, Inc., Chatsworth, CA) and ethanol precipitation, the amplified DNA was 66 methylated with Hpall methylase (New England Biolabs, Beverly, MA). The methylated DNA was cloned into the Srfl restriction site of pCR-Script Direct plasmid using the pCR-Script Amp cloning kit (Stratagene Cloning Systems, La Jolla, CA). About 80 clones were sequenced with Sequenase sequencing kit (United States Biochemical Corporation, Cleveland, OH).

Sequence analysis and secondary structure prediction was done by using previously described computer software (Feng and Doolittle (1987) J. Mol. Evol 25:351-360; Jaeger et al., (1989) Proc. Natl, Acad. Sci. USA 86:7706-7710; Jaeger et al., (1990) Methods Enzymol. 18:281-306; Zucker (1989) Science 244:48-52).

E. Determination of Minimal Sequences Necessary for Binding Oligonucleotide ligands end labeled at the 5' end with y- 2 P-ATP using T4 polynucleotide kinase, or at the 3' end with 5'- 32 P-pCp and T4 RNA ligase, were used to establish 3' and 5' boundaries respectively (Fitzwater et al., (1996) Methods Enzymol. 262:275-301). After partial alkaline hydrolysis, the radiolabeled oligonucleotide was incubated with 0.1, 0.6, and 3.0 nM hKGF, and the protein bound oligonucleotide was isolated by nitrocellulose filtration. The nitrocellulose retained oligonucleotide truncates were analyzed on a high resolution denaturing polyacrylamide gel. An alkaline hydrolysis ladder and a ladder of radioactively labeled ligands terminated with G-residues, generated by partial RNase Ti digestion, were used as markers to map the 3' and 5' boundaries.

F. Thermal Denaturation Profiles Oligonucleotide melting profiles were obtained with a Cary Model 1E spectrophotometer. Oligonucleotides were heated to 95C in PBS (Sambrook et al., (1989) Molecular Cloning: A Laboratory Manual. 2nd Ed., Cold Spring Harbor, NY) or 10 mM phosphate buffer and cooled to room temperature before recording the melting profile. The melting profiles generated show the change in absorbance at 260 nm as a function of temperature. During recording, the samples were heated at a rate of I C min-' from 20-95*C.

EXAMPLE 17. RNA LIGANDS TO hKGF A. SELEX To generate RNA ligands for hKGF, two parallel SELEX experiments were initiated, one with 2'-NH, and the other 2'-F pyrimidine modified RNA molecules randomized at 40 contiguous positions. The SELEX conditions and results for each round are summarized in Table 15. The starting pool contained 5x10"' (500 pmoles) and 2.5x1014 (250 pmoles) 2'-NH, and 2'-F pyrimidine modified RNA molecules, respectively, and bound to hKGF with an approximate

K

D of 30 nM. After 8 rounds of SELEX, the evolved pools bound with a KD of 0.6 nM. No further improvement in the KD was observed in the subsequent two rounds. The RNA pools from the 8th round were reverse transcribed, PCR amplified and cloned as described.

B. RNA sequences In the 2'-NH 2 SELEX, 29 out of 31 clones were unique. In the 2'-F SELEX all 43 clones sequenced were unique. A unique sequence is defined as one that differs from all others by three or more nucleotides. Table 16 lists the sequences (SEQ ID NOS: 189-262) of all of the clones sequenced in standard single letter code (Comish-Bowden, (1985) Nucleic Acid Res 13:3021-3030).

Computer assisted global and local alignment did not reveal any extensive homologies among the clones, and no obvious families were apparent. The 2'-NH 2 clones are in general purine rich while the 2'-F clones are pyrimidine rich.

When the alignment parameters were relaxed, the Feng/Doolittle algorithm grouped the 2'-NH 2 clones in one family and the 2'-F clones in another. Visual inspection of the sequences suggested two and three possible families for the 2'-NH 2 and the 2'-F ligands, respectively. Using conserved predicted secondary structure, 38 2'F ligands could be assigned into two classes (Figures 12A and 12B). Similarly, 15 2'NH 2 ligands could be assigned into two classes (Figures 12C and 12D). The two proposed classes for the 2'F ligands can be folded into pseudoknot structures (Wyatt et al, (1993) The RNA World 465-496; ten Dam, E.

(1992) Biochemistry 31:1665-1676). These structures are very related and in fact they could be circular permutations of a common structure. Loop 3 (L3) of class 1 pseudoknots presents the conserved sequence 5'RRYuy while loop 1 (LI) of class 2 ligands presents the sequence 5'AaYY. Both of these sequences contain the consensus 5'RRYY. Some of the 2'F ligands contain two to three copies of the RRYY sequence (Figures 12A and 12B). Another feature of these structures is the unequal distribution of purines and pyrimidines in stem I One strand of that stem contains almost exclusively purines while the other strand contains pyrimidines.

Class 1 of the 2'NH, ligands includes 8 members that can be folded into stem-loop structures with internal symmetric or asymmetric loops. The stem contains three consecutive GC base pairs. The terminal loops are long and present the conserved sequence 5'GGAA(N)., 4 YAA(N),.,RCRR (SEQ ID NO:263). Both sides of the internal asymmetric loops of the class 1 ligands contain the sequence Class 2 includes 7 ligands that can be folded into dumbbells with variable sized loops. One loop contains the conserved sequence 5'YGAY while the other loop contains the conserved sequence 5'GGAA(N)oYGA (SEQ ID NO:264).

Clones 2N and 54N are circular permutations of the remaining 5 clones.

C. Affinities The dissociation constants of the hKGF ligands were determined by nitrocellulose filter binding and are listed in Table 17. Eight out of41 2'-F ligands bound biphasically. The remaining of the 2'-F and all the 2'-NH 2 ligands bound monophasically. Under protein excess, biphasic binding suggests that the ligand exists as two affinity species (presumably isoconformers) that are not in equilibrium. The best 2'-F-modified ligand, K14F, binds biphasically with the high and low affinity dissociation constant at about 0.3-3pM and 2-10 nM respectively. There is some observed variability in the KD determinations for the various clones and the random RNA. Despite the experimental variability in the KD determinations, the high affinity species of K14F have a 1,000-5,000 fold better affinity than the random RNA. Among the monophasic 2'-F-modified ligands, K38F had the best KD of about 0.3nM. The best 2'-NH 2 -modified ligands bound with a KD of 0.4nM which represent about 75 fold improvement over the random RNA.

D. Determination of Minimal Sequences Necessary for Binding Two 2'F ligands (6F and 14F) (SEQ ID NOS:223 and 231) were studied further to determine the minimal sequences necessary for binding. Sequence boundaries were determined by allowing an alkaline hydrolysis ladder, labeled at the 3' or 5' end, to bind to hKGF. The partial fragments were affinity purified by nitrocellulose filtration and analyzed on high resolution denaturing gels.

Boundaries were clearly observed only at the 3' ends for both ligands (Figure 13) and are in agreement with the class I proposed folding as shown in Figures 12A and 12B. Truncated templates were then used to confirm the boundaries (Figure 13). Three truncates were tested for 6F because a run of 7 consecutive pyrimidines did not allow the precise mapping of the boundary. From these three truncates, one lost its KGF binding activity as shown in Figure 13. A single 14F truncate, designated 14F3'T, was tested. This truncate was two bases longer than the observed boundary in order to extend stem 2 (S2) of the proposed pseudoknot structure. The 14F3'T truncated ligand retained binding activity with affinity similar to the full length ligand. Like the full length ligand, 14F3'T bound KGF biphasically where the high affinity species represented about 20% of the molecules and showed K values of about 0.3-3 pM. These high affinity species when partially separated from the low affinity species on the basis of differential affinity to KGF, exhibited binding curves with mid points at 0.3-3 pM and maximum plateaus of about 70% (data not shown). Figure 13 shows the predicted folding of the shortest active truncates for 6F and 14F which are 53 and 49 bases long respectively. Both proposed pseudoknot structures contain relatively long stems. The two proposed stems of 6F are separated by a single base forming a non-H-type pseudoknot. The proposed 6F structure resembles the solution structure of a similar pseudoknot motif from a frame-shifting element found in the MMTV RNA (Shen et al., (1995) J.Mol.Biol. 242:963-978). The two stems (SI and S2) of 14F could be drawn as two coaxially stacked helices of 16 base pairs total length (H-type pseudoknot). A Similar pseudoknot structure has been proposed before, based on NMR data (Du et al., (1996) Biochemi=stry 3:4187- 4198). Given the short length of Ll, it is possible that ligand 14F forms a non-H-type pseudoknot where the last GU base pair of S is not formed allowing a more flexible helical region and a longer LI. Temperature melting curves of 14F and 14F3'T suggest a remarkable thermostability for this ligand (data not shown). These melting curves appear to be concentration independent and biphasic in 150 mM salt. Biphasic melting curves have been observed before with tRNA (Hilbers et al., (1976) Biochemistry 15:1874-1882), and have been attributed to the tertiary folding of the RNA molecule. Multiphasic temperature transitions have also been proposed for RNA pseudoknots (Du et al., (1996) Biochemistry 15:4187-4198). The biphasic curves observed include a low Tm at about 55 0 C and a high Tm of greater than 85-90°C. In 10 mM salt the low Tm of 14F is not observed while the high Tm is shifted down to 75-78 0 C. The melting profile for 14F appears to be flatter than 14F3'T even though the Tm values are the same. The data suggest that the observed thermostability is attributable to just the minimal 49-mer.

In an effort to identify shorter KGF ligands that retained binding, the binding activity of various deletions of the shortest truncate of ligand 14F, namely 14F3'T were tested. Deletions were tested in all the structural elements of the proposed pseudoknot structure. The results are summarized in Table 23 (SEQ ID NOS:272-304). RNA transcripts containing 2'F pyrimidines and 2'OH purines were obtained by in vitro transcription using synthetic DNA templates. The activity of each ligand is shown by scoring for both the high and low (L) affinity component of the 14F3'T binding curve with (active) or (not active).

Truncates T35 and T36 represent two complementary halves of 14F3'T molecule and were additionally tested as an equimolar mixture. The structural elements of the proposed pseudoknot structure are separated by and are indicated by symbols S1 (stem S2 (stem LI (loop 1) and L3 (loop The proposed pseudoknot structure for 14F3'T is a non-H-type pseudoknot and lacks L2 (loop The complementary sequences forming Sl (Sl and Sl') and S2 (S2 and S2') are marked by single and double underlines respectfully. In the tabulated sequences, deleted bases were replaced with periods Any deletion attempt in the stems S and S2 of the proposed pseudoknot structure resulted in loss of both the high and low affinity component of the binding curve as observed with the 14F3T ligand. Deletions in loop 3 however, were tolerated as long as one copy of the RRYY box remained intact. The shortest ligand that retained activity is T22 which is a 43-mer. In trying to obtain shorter ligands by truncating L3 further a mutant version of T22 (designated T22mu) was used where the last GC base pair of SI was eliminated by a G to U mutation at position 6. The reasoning for this mutation was to enhance the flexibility of the double stranded region of this ligand by allowing an unpaired base between S1 and S2. Although this mutation did not affect the binding of T22 it did not allow further active truncations in L3.

E. Specificity of RNA Ligands to hKGF The specificity of the K14F ligand was tested by determining its KD against rat hKGF, and the heparin binding human growth factors, aFGF, bFGF, and PDGF (Table 18). The results suggest that the K14F binds all tested targets like random RNA, except hKGF, and it can discriminate between hKGF and other similar proteins by a factor of 400-40,000.

The specificity of ligand 14F3'T was tested by determining its Kd against a variety of heparin binding proteins. The results summarized in Table 22 show that ligand 14F3'T can discriminate KGF from all other heparin binding proteins tested by a factor of 1.2x10'-3x10'. Ligand 14F3T binds only to KGF with high affinity while it binds all other heparin binding proteins tested like random RNA. Binding of 14F3'T to the rat KGF, which is 91% identical to human KGF, is with about a 5-10 fold reduced affinity. Similar specificity was observed during the inhibition of the KGF induced DNA synthesis of Balb/MK cells. Ligand 14F3'T inhibits rat KGF induced DNA synthesis with a K, of 1.8 nM which is 20-50 fold higher than the K observed with the human KGF. Ligand 14F3'T inhibits the DNA synthesis of Balb/MK cells only if it is the result of KGF but not EGF stimulation (data not shown).

EXAMPLE 18. INHIBITION OF hKGF BINDING TO CELL SURFACE

RECEPTORS

A. Receptor Binding Assay To test the ability of the hKGF ligands to competitively inhibit the binding of hKGF to its cell surface receptor, two cell lines were used. The first cell line, PC-3, is an isolate from a grade IV prostatic adenocarcinoma (ATCC CRL 1435).

The second cell line is designated as NIH3T3/FGFR-2 and is a recombinant NIH/3T3 cell line carrying the human hKGF receptor at about,0.5-1X106 high affinity KGF binding sites per cell (Miki et al., (1992) Proc. Natl. Acad. Sci. USA 82:246-250).

PC-3 cells were plated in 24-well plates at about 10' cells per well.

Following growth for 48-36 hours, the cells were serum starved for 24 hours, washed two times with 500 ,l of cold DPBS, and then incubated with 500 ,l binding buffer (BB1; DPBS, 0.5mM MgCI 2 0.2% BSA. 0.02% sodium azide) containing various concentrations of 2 I-labeled KGF ranging from 0 to 0.8 nM.

Following 3-3.5 hour incubation at 4 0 C, the binding mixes were aspirated and the well-adhered cells were washed two times with 1 ml BB1 and once with 1 ml BB supplemented with 0.5M NaCl. The remaining bound labeled hKGF was solubilized in 600 ,l 0.5% SDS/0.1M NaOH and counted in a gamma counter (Beckmann). Nonspecific binding was determined in the presence of 100 fold molar excess of unlabeled hKGF. For competition assays, the labeled hKGF was kept constant at 0.3 nM, and varying concentrations of competitor molecules were included in the binding reactions ranging from 0 -1,000 nM. Binding curves were fitted to the equation: [Bound Tracer]=([Total Tracer]* [Receptor])/(KD+[Total Tracer]) where [Total Tracer] and [Bound Tracer] were fixed and the KD and [Receptor] were determined by regression analysis using the software Kaleidagraph (Synergy Software, Reading, PA).

NIH3T3/KGFR-2 cells were plated in 24-well plates at about 10 s cells per well. Following growth overnight, the cells were serum starved for 1-5 hours, washed two times with 500 ,1 binding buffer (BB2: serum-free MEM growth medium, 0.1% BSA, 25mM HEPES, pH and then incubated with 250 ul BB2 containing 1/g/ml heparin (from bovine lung, SIGMA, St. Louis, MO), "'I-labeled hKGF at 0.03 nM, and varying concentrations of competitor molecules (300nM-OnM). Following 1 hour incubation at room temperature, the binding mixes were aspirated, and the wells were washed two times with 250 ,1 cold DPBS and once with 250 ,l cold DPBS supplemented with 0.5M NaCI. The bound labeled hKGF was solubilized in 500 ,1 0.5% SDS and counted in a scintillation counter (Beckmann).

The inhibition constants (Ki) of the RNA ligands were determined by a nonlinear regression analysis of the data.

In search of KGF receptors on the surface of PC-3 cells, different concentrations of 2 I-hKGF were used, ranging from 0.002 to 0.8 nM, in the presence and absence of 100 fold molar excess of unlabeled hKGF, and saturation binding of the tracer on the surface of PC-3 cells was observed. Figure 10 shows the plot of the concentration of bound tracer as a function of the total concentration of tracer as well as the Scatchard analysis of the same data.

Analysis of the data suggested that there are about 5,000 specific hKGF binding sites per cell with a KD of 100-200 pM. This KD is in good agreement with the reported KD for hKGF of 200 pM (Miki et al., (1992) Proc natl Acad Sci USA 89:246-250).

PC-3 plasma membrane extracts were found to alter the electrophoretic mobility (gel shift) of radiolabeled hKGF upon native gel electrophoresis (Figure 11). For electrophoretic mobility shift gels, about 3x10' PC-3 cells were gently spun and washed with PBS and then lysed by mixing with equal volume of lysis buffer containing 40 mM Hepes, pH 7.4, 150 mM NaCI, 20% glycerol, 2% triton X-100, 0.1% sodium azide, 3 mM MgC1 2 3 mM EGTA, 2 MM aprotinin, 2 ;M leupeptin, 2 mM PMSF, and 400 MM sodium orthovanadate. Following 15 min incubation on ice the extract was spun at 11,000 g at 4 0 C for 30 min to remove debris and nuclei and the supernatant was aliquoted and stored at -70 0 C. For gel analysis, 25ul binding reactions were set in DPBS, 0.01% HSA, 2 mM MgC 2 containing 3 ul of a 10 fold diluted PC-3 membrane extract in 0.01% HSA, and various concentrations of '2I-labeled hKGF. Following a 10 min incubation at room temperature, 6X loading dye was added to achieve 1X concentration, and the samples were loaded onto a 5% or 10% native TBE polyacrylamide gel. The gel was prerun at room temperature at 100 Volts. Following loading, the gel was run at 200 Volts for 5 min and then at 100 Volts for 30-60 min at room temperature.

The radioactive bands were then visualized by autoradiography. The gel shift of radiolabeled hKGF is not observed in the presence of 100 fold molar excess of unlabeled hKGF (Figure 11), demonstrating a specific interaction between a component found in the PC-3 membrane extracts and hKGF. The estimated KD from the gel shift experiment is about 8 nM.

In agreement with the competition experiments reported in the literature (Miki et al., Proc Natl Acad Sci USA ,8:246-250), gel shift competition curves using unlabeled hKGF and bFGF as well as an unrelated small basic protein namely lysozyme were obtained. Table 21 lists the IC50 values obtained in this experiment. In agreement with previous reports, the data presented in Table 21 show that bFGF competes about 20 fold worse than hKGF for binding with the hKGF receptor present in the PC-3 plasma membrane extracts. The interaction observed by the gel shift appears to be a specific interaction for FGF and it is not due to a charge-charge interaction, as lysozyme, another small positively charged molecule, competes for the PC-3 membrane extract:hKGF complex with about 100 fold worse affinity than hKGF alone.

IC50 values for various RNA ligands obtained with the PC-3 assay are shown in Table 19. A subset of these ligands was tested on the NIH3T3/FGFR-2.

Competitive inhibition constants (Ki) were determined from full competition curves and are summarized in Table 20. In determining the Ki values, it was assumed that 3T3 cells have 500,000 binding sites per cell and PC-3 cells have 5,000 binding sites per cell.

The data show that several hKGF ligands can competitively inhibit binding of hKGF to its cell surface receptors. Some of these ligands, such as K14F, have potent competitive activities with Ki's in the low nM range.

This work not only demonstrates that nucleic acid competitors for hKGF were obtained, but also identifies a new assay for screening hKGF competitors including small molecules, antibodies, and peptides. This new assay includes the use of the prostate carcinoma cell line, PC-3.

The two cell lines, PC3 and NIH3T3/FGFR-2, give slightly different results (see Table 20). KGF binding to PC-3 cells is more sensitive to inhibition by several ligands and by heparin. Random RNA, however, does not effectively compete for KGF binding on the PC-3 cells. KGF binding to NIH3T3/FGFR-2 is resistant to inhibition by some RNA ligands and heparin. This is because the NIH3T3/KGFR assay is more stringent since it is done in the presence of 1 ug/ml heparin. The random oligonucleotide competition curve with the NIH3T3/FGFR-2 is completely flat with K, >10" M. Ligands 6F and 14F show the best inhibitory activity with K, values of 100-200 pM and 2-8 nM in the PC-3 and NIH3T3/FGFR-2 assay respectively. Only two 2'NH, ligands, 14N and 29N, show good activity with the PC-3 cells value of 1.4 nM). From these two ligands, only 14N retains its inhibitory activity in the NIH3T3/FGFR-2 assay showing a K, value of 100 nM. The observed inhibition of the KGF mitogenic activity by these ligands is not due to a nonspecific affect in the proliferative ability of the cell lines because these ligands have no antiproliferative activity on cells induced by EGF instead of KGF (data not shown).

This work not only demonstrates that nucleic acid competitors for hKGF were obtained, but also identifies a new assay for screening hKGF competitors including small molecules, antibodies, and peptides. This new assay includes the use of the prostate carcinoma cell line, PC-3.

EXAMPLE 19. INHIBITION OF THE MITOGENIC ACTIVITY OF KGF One of the biological effects of KGF is the stimulation of proliferation of epithelial cells (Rubin et al., (1989) Proc Natl Acad Sci USA M6:802-806). This proliferative effect of KGF can be measured by the stimulation of H-thymidine incorporation in responding cells after exposure to KGF. Three such cell lines have been described before (Rubin et al., (1989) Proc Natl Acad Sci USA 6:802-806). Two cell lines were used to test the anti-mitogenic activity of various ligands. One is 4MBr-5 (ATCC #CCL208), a monkey epithelial, low passage, cell line (Caputo et al., (1979) In Vitro 1:222-223) while the second is Balb/MK, a transformed rat keratinocyte cell line (Weissman and Aaronson (1983) Cll32:599-606). 4-MBr5 cells grown in F12K containing 30 ng/ml, hEGF, and 10% FCS, were trypsinized and resuspended in M199 containing mM HEPES, pH 7.4, and 10 FCS at 1.4x10 5 cells/ml. A 96-well microtiter plate was seeded with 100 p1 of cell suspension and KGF was added at 10 ng/ml (0.5 nM), as well as K14F ligand at various concentrations ranging from 0-1000 nM. Each incubation reaction was set in at least triplicates. Following 24 h incubation at 37 0 C, 'H-thymidine was added at 1 Ci/well along with unlabeled thymidine at 10 nM. The cells were incubated for additional 24 h, the supernatant was aspirated, and the remaining cells were harvested by lysis in 20 ,e of 0.2 N NaOH. The extent of 3 H-thymidine incorporation was determined by TCA precipitation and filtration through GFC filter disks (Whattman, Hillsboro, OR).

Balb/MK cells grown in Low Ca" EMEM with 10% FCS (dialyzed and heat inactivated) and 5 ng/ml rhEGF were trypsinized and resuspended in Low Ca EMEM with 1% FCS (dialyzed and heat inactivated) and 0.5 ng/ml rhEGF and plated on 96 well fibronectin coated culture plates at 4-6x10 4 cells per well in 100 ,l total volume. Following overnight growth, the medium was replaced with Low Ca- EMEM without FCS or rhEGF and serum starved for about 30 hrs.

Human recombinant KGF or EGF was then added at 16 and 49 pM respectively, along with various concentrations of competitors ranging from 0-1000 nM.

Following over-night incubation, 'H-thymidine was added at 0.2 1Ci/well and incubation continued for an additional 7-8 hrs. The extent of 3 H-thymidine incorporation was determined by TCA precipitation and filtration through GFC filter disks.

The inhibition constants of the oligonucleotide ligands were determined by a nonlinear regression analysis of the data as described before (Gill et al., (1991) J.Mol.Biol. 220:307-324).

The two assays give slightly different results. The 4MBr-5 assay was performed in the presence of fetal calf serum, while the Balb/MK was done following serum starvation. The Balb/MK assay is more sensitive and a prototypic assay for the KGF induced mitogenic activity. Similar to the results obtained with the PC-3 cells, the 4MBr-5 assay showed a good activity for ligand 14F value of 9.8 nM but incomplete inhibition). In the same assay, the random oligonucleotides showed K, values of pM while a monoclonal neutralizing antibody showed a K, value of 2.9 nM. It appears that ligand 14F is as good or even better than the monoclonal neutralizing antibody. The competition curves for the neutralizing monoclonal antibody and ligand 14N plateau at about 20-40%, suggesting that these antagonists do not completely abolish the KGF mitogenic activity. In contrast to the monoclonal antibody, ligand 14F completely blocks the KGF mitogenic activity on the 4MBr-5 cells. In the Balb/MK assay, 14N showed K, values of about 10 nM (incomplete inhibition) while the random oligonucleotide showed Ki values of about 300 nM. The Ki values for 6F and 14F are 830 and 92 pM, respectively. Similar to the assay, ligand 14F appears to be as good if not better than the monoclonal neutralizing antibody which shows a K, value of 980 pM. The best inhibitory activity was observed with 14F3T with a K, value of 34 pM.

EXAMPLE Nucleic acid ligands that bind to basic fibroblast growth factor (bFGF) have been derived by the SELEX method as described in U.S. Patent No.

5,459,015 (see also U.S. Patent No. 5,270,163 and Tuerk and Gold (1990) Science 249:505-510). A 2'NH 2 -modified nucleic acid ligand designated 21A having the sequence 78

GGGAGACAAGAAUAACGCUCAAGUAGACUAAUGUGUGGAAGACAGC

GGGUGGUUCGACAGGAGGCUCACAACAGGC (SEQ ID NO:265) was examined by deletion analysis for the minimal sequence information required for high affinity binding to bFGF. This analysis led to truncated ligand 21A-t (GGUGUGUGGAAGACAGCGGGUGGuuc (SEQ ID NO:266) where the underlined G's are guanines added to improve efficiency of transcription and lowercase letters are from the constant region.

In order to increase the stability of ligand 21A-t against degradation by nucleases, short phosphorothioate caps were added to the 5' and the 3' ends. In addition, nine ribopurine positions were identified that can be substituted with 2'deoxy-2'-O-methylpurines without a loss in binding affinity for bFGF, using the method described in Green et al., Chem.Biol. 2:683-695, resulting in the ligand designated as NX-286 (5'-TsTsTsTs mGmGaU rGaUrG aUrGrG mArArG mAaCrA rGaCmG mGmGaU mGmGaU aUaC TsTsTsTsT-3' (SEQ ID NO:267), where s represents phosphorothioate intemucleoside linkage, aU and aC are 2'deoxy-2'-aminouridine and 2'-deoxy-2'-aminocytidine residues, respectively, mA and mG are 2'-deoxy-2'-O-methyladenosine and guanosine residues, respecitvely, rA and rG are adenosine and guanosine residues and T is 2'-deoxythymidine). The modified nucleic acid ligand had a Kd of 0.4 nM as measured by electrophoretic mobility shift assay.

Table I Nucleic Acid Sequences Used in SELEX Experiments described in Examples 1-4 SEQ ID Starting RNAs:NO 64N6 transcript: GGGGGAGAACGCGGAUCC

AAGCUUCGCUCUAGAUCUCCCEJUUAGU

GAGGGUUA 3' 40N6 transcript: GGGGGAGAACGCGGAUCC

AAGCUUCGCUCUJAGAUCUCCCUUUAGU

GAGGG UUA 3' 2 randomized lib2-6-1 transcript*: gaggggggugguucggaga]AAGCUUCGCUCUAGAUCUCCCUUUiAGUGAGGGUJUA 3'3 Starting DNA templates: Z-54 (64N60): S'GGGAGAACGCGGATCC AAGCTTCGCTCTAGA3' 4 (40N6): S'GGGAGAACGCGGATCC AAGCTrCGCTCTAGA3' D- 12 3(randomized lib2-6-1I)*': gggtggttcggaga]AAGCTTCGCTCTAG. 3- 6 PCR and cloning primers: T7SacBam: 5'TAATACGACTCACTATAGGGGTCGGGTC3' 7 Sacl BaznHl T'7SB2N: 5'TAATA'CGACTCACTATAGGGGGAGAACGCGATC3' 8 BamH I 3XH: C GMJ3 9 XbaI Hindll BamHl clonenQ site enineered into UGem9zf to clone SELEX experments 3-9.

SP6 promoter Xbal GGTCGACCCAATATGAT I1 BamH I T7 promoter OGAUC or GATC, these bases only gauc or gact 62.5 specified base, 12.5 the other three bases Table 2 RNA SELEX Experiments described in Examples 1-4: template, pyrimidine nucleotides, and round cloned.

SELEXLn iIN I ib2 lib3 IRA4 IM6 lib7

IMS

lib9 template* 64N6 64N6 40N6+64N6 40N6+64N6 64N6 64N6 64N6 D-123 64N6 2 substitued UTP

OH

OH

F

NH

2 NF4

F

OH

NH

2 2'substituted CTP

OH

OH

F

F

NH,

OH

F

Rond cloned 6 4 13 13 14 6 Sequences of templates are described in Table I.

Table 3 TGFb Binding ligands clone 5'CONSTANT VARIABLE 31CONSTANT SEQ ID Group A gggggagaacgcggaucc [40 or 64N1 aagcuucgcucuagaucucccuuuagugaggguua 1 ib3NO 13 GAGCAAUCCCAGGCGCAUAGCUUCCGAGUAGACAGGAGGGAGGGGUGGAUGUGGCGUCUACUCGGUGUCGUG 3 GACACCGCCUGUCGGAAAGGGGGUGUUGGCAUGUUGG12 4GAGCAACCCCAGGCGCAUAGcCUCAGUAGACAGGGGGAGGGGUGGAUGUGGCGUCUACUCGGAGUCGUG 13 14 1 ib4 32 G GCAACCCCAGGCGCAUAGCUUCCGAGUAGACAGGCGGGAGGGGUGGAUGUGGCGUC

ACOAGO

is 1 ib8 9 GCAAUCCCAGGCGCAUAGCULJCCGAGUAGACAGGAGGGAGGGGUGGAUGUGGCGUCUACUCGGCGUCGUG, 16 0 1 GAGCAAUCCCAGGCGCAUAGCUCCGAGUAGACAGGAGGGAGGGGUGGAUGUGGCGUCU

CGAGG

GAGCAGCCUGGC17 7GGA(CCGCAUAGCUUCCGAGUAGACAGGAGGGAGGGGUGGAUGUGGCGUCUACUCGGUGUCGUG 18 48 G GCAAUCCCAGGCGCAUAGCUUCCGAGUAGACAGGAGGGAGGGGUGGAUGUGGUGU

ACGAGG

19 1 ib2 6- 4 GAGCAAUCCCAGGCGCAUAGCUUCCGAGUAGACAGGAGGGAGGGGUGGAUGUGGUGUCU

CGAG

M i6 23 A AGCUUC GAGUAGACAGGAGGGAGGGGUGGAUGUGGAGUCU

CGAG

4 GAGCAAUCCUAA GCAUAGCUUC GAGUAGACAGGAGGGAGGGGUGGAUGUGGCGUCU CGAG 21 22 1 ib7 21 GAGCAAUCCCGGGCGCAUAGICUCCGAGGAGACAGGCGGGAGGGGUGGAUGUGGCGUCU CGAG 2 43 GAGcAAUCCCAGGCGCAUAGCUUCCGAGUAGACAGGCGGGAGGGGUGGAUGUGGCGUCU CGAG 23 Table 3 (Page 2) clone SICONSTANT VARIABLE 31CONSTANT SQI gggggagaacgcggaucc (40 or 64N] aagcuucgcucuagaucucccuuuagugaggguua SEQOI Group B.

1 ib4 -12 U)GAGAAGGACGUCGGGGUCAACGGGGUGAGGUGCAGCAGAAAGGGCCGGCACCACAUGACGUAA 1 1b3 -44 UGAGAAGGACGUCGGGGU GAGGUGCAGCAGAAAGGGCCGGCACCACAUGACGUAA 28 lib3-42 GGUGGGAAA GUCGGAtnU AUGUGU GUAGAUUU GU GUGCOA 29 Group C.

libi UGCUAGACCGAGGAUGCAAAGGGACAUGCAUUAGGGAAACCUAUGUAUAAGAACGCGGUCGCAG libi -20- 3H .LIGCUAGACCGAGGAUGCAAAOGGACAUGCALJUAGGGAAACCUAUGUAUAAGAACGCGGUCGCAGA 32 lib6 -30 UGCUAGACCGAGGAUGCAAAGCGACAUGCAUUAGGGAAACCUAU UAIIAAGAACGCCCUCGCAG 33 34 Group D. 0 11 b2 UGUCUCCACCGCCGAUACUGGCGUUCCUGCGGGCCCCUCCAUGCAGGAGGGGGUGGUUCGGAGA

N

1 ib2 -6-1-81 UGUCUCCACCGCCGAUACUGGGGUUCCUCGGGCCCCUCCAUGCAGGACGGGGGUGGUUCGGAG 1 ibs -23 UGUCUCCACCGCCGAUACUGGGGUUCCUGGGGCCGCUCCAUGCAGGAGGGGGGUGGUUCGGAGA 36 1 ib9 -10 UGUCUCCACCGCCGAUACUGCGGGUUCCUGGGGCCCCUCCAUGCAGGAGGGGUGGUUCGGAGA 37 38

ORPHANS.

clone# 1 1b3 -45 GGAAGUCUICCUCUUUGGGGAGUCCGCAUGGCCCUGGCGA 1 ibi AAGAAUGUUCGGCCGCACGAGGUGACAGUGGUGCGGAUACGGACCGAUUGGGUUUGCC 39 1 ibi -20-12 GGUCACCCGGGCAUAUAACAAUGCCGACACUGGGGUACCUGGGACGGGUCGGACUGGACCGAAG 1 ib2 AUAACCGGCUGCAUGGGAGGGACAUCCUGGGAAAGGACGGGUCGAGAUGACCUGAGC-AGUUCCGGC 41 Table 3 (Page 3) clone SEQ ID group A Boundary Experiments ]i b3 -13 boundaries 5' GCUUICCGAGUAGACAGGAGGGAGGGGUJGGAUGUGGCGUCUAC 3*2 1 ibS -9 boundaries 51 CUUCCGAGUAGACAGGAGGGAGGGGUGGAUGUGGCGUCUACUC 3, 26 ib4 -32 boundary GGCAACCCCAGGCGCAUAGCUUCCGAGUAGACAGGCGGGAGGGGUGGAUGUGGCGUCACG 3' 26 Group B Boundary Experiments lib4 -12 boundaries 5- UGAGAAGGACGUCGGGGUCAACGGGGUGAGGUGCAGCAGAAAGGGCCGGCACCA 3-'3 Legend: The constant region of the ligand is shown in lower case and variable in tipper. Sequences have been aligned. Deletions with respect to the first sequence in each group are shown by gaps, substitutions are in bold type.

2TNI 12-UT"P, 2'F-CTP: 00 *021 -UTP, 2*F-CTP.

I-UTP, 2'Ol I-CTP Group A and B bind with either 2'NH2- or 2'F- pyrimidines.

Ligands bind with eiher 2'NI12- or 2'F- pyritnidines unless othierwise indicated.

84 Table 4 Dissociation and Inhibition Constants Group Ligand lib3-13 lib3-3 Ib4-32 lib5-7 Iib8-9 lib8-9* libS-48 lib2-6-4 lib6-23 lib721**** lib7430*** lib4-12 lib3-44 lib3-42 libl-20.3** lib 1.20-3-820* lib6.30** lib2-6-1 lib2-6-1-81 libS-23 lib9-10* lib3-45 libl-20-5S* libi.20-12*** liblI-6-8

B,

31 0.60 0.38 0.55 0.44 0.50 0.37 0.33 0.4 0.35 0.32 0.20 0.35 0.18 0.33 0.15 0.08 0.18 0.16 0.67 0.46 0.35 Kd 0.9 nM 0.7 nM 0.9 nM 1.7 nM 0.8n.M 2.4nM 3.4nM 1.7nM 2.8 nM 3.8nM 3.InM 3.4 nLM 2.4 nM 3.3 rLM 0.4 nM 0.2 nM 1.3 nM 0.6 nM 3On.M 6.1 nM 8.8 nM 14.3 nM 10.7 nM 6.6 nM 1.1 nM 1.9 nM 46 nM 3.1 nM 4.7 nM nonbinder nonbinder

'CS

0 9.7 nM 42 nM 18 nM 32 nM

NT

2On.M 157 nM 49 nM 17 nM 124 nM not inhibitory not inhibitory not inhibitory not inhibitory not inhibitory 109 nM 108mnM 69 nM 119 nM 22 nM not inhibitory not inhibitory not inhibitory 112 nM 103 nM 201 nM 298 nM not inhibitory not inhibitory not inhibitory not inhibitory

NT

NT

not inhibitory not inhibitory 0.40 0.39 0.48 0.24 0.08 0.42 0.34 0.12 Orphans Controls lib5-9 random 64N6 ligands are NH2 pyrimidines unless otherwise noted *2'-NH2-UTP, 2'-F-CTP, pyrimidines, pyrimidines, 002'-F-UTrP,2'-NH2-CTP Table DNA oligonucleotides used in Examples 5 and 60SQI SEQ ID

NO,

40N7 Template for RNA SELEX 5N7 5'-primer for PICR 3N7 3'-primer for PCR Seguence

TAATACGACTCACTATAGGGAGGACGATGCGG

TCGGGCGAGTCGTCTG

GGGAGGACCIATGCGG140NICAGACGACTCGCCCGA

GGGAGGACGATGCGG

(biotiin),TCGGGCGAGTCGTCI"G 40137 5D7 3D37 5N8 3N8 40D8 5D8 3D8 Starting material for DNA SEI.EX S-primer for PCR 3'-primer for PCR Template for RNA SEILEX S'-primer for PCR Y-primer for PCR GCCTGI1TGTGAGCCT'CCTGTCGAA(40N IT1GAGCG1TAT-rCTTGTCITCCC

TAATACGACTCACTATAGGGAGACAAGAATAAACGCTCAA

GCCTGTGTGAGCCTCCTGTCGAA

GGGAGACAAGAATAAACGCTCAAI40NJ1TCGACAGGAGGCTCACAACAGGC

GGGAGACAAGAATAAACGCTCAA

(biotini) 1 GCCTG1TGTGAGCCTCCTGTCGAA 49 51 00 Starting material for DNA SELEX 51-primer for PCR 3'-primer for PCR a. DNA ol igonucleol ides. 40N7 and 40N8 were used to generate the double-stranded DNA temnplate for inI Wiow transcription. The Y-primers 3117 and 3N8 were also used to generate cDNA from the RNA repertoire. Synthetically synthesized DNA oligonucleotides 40137 and 40D38 were used directly as the starting repertoire for the two single-stranded DNA SELEX experiments. PCR amplification of the selected repertoires used the appropriate or Y-primer. The symbol O0N indicated a randomized region within the oligonucleotide.

Table 6 TGFBI 40N7 DNA Selex Sequence of fifty randomly chosen clones.

GGGAGGACGATGCGG... 40N... CAGACGACTCGCCCGA 3 I 20(11 clones) 4 8 13 16 48 18 19 32 46 12(3 clones) 1 3.

23 34

CCAGGGGGGGTATGGGGGTGGTGCTACTTACTTGCGTCTT

CCAGGGGGGGTATGGGGGTAGTGCTACTTACTTGCGTCrr

CCAGGGGGGGTATGGGGGTAGTACTACTTACTTACGTCTT

CCAGGGGGGGTATGGGGGTATACTACTTACTTACGTCTT

CCAGGGGGGGTATGGGGGTAATACTACTTACTTACATCTT

CCAGGGGGGGTATGGGGGTAATACTACTTACTTACGTCTT

CCAGGGGGGGTATGGGGGTGGTGTTACTTACTTGCGTCTT

CCAGGGGGGGTATGGGGGTGGTGCTTCTTACTTGCGTCTT

CCAGGGGGGGTATGGGGGTGGTGTACTTTTTCCTGCGTCTTC

CCAGGGGGGGTATGGGGGTGGTTCGTTTTTCTTTGCGGCTT

CCAGGGGGGGTGTGGGGGTGGTGTACTTTTTCTTGTCTrC

CCAGGGGGGGTATGGGGGTGGTTTGGTATGTTGCGTCCGT

CCGGGGTGGGTATGGGGGTAATACTACTTACTTACGTC'I-I

CCGGGGGTGGGTAGGGGGGTAGTGCTACTTACTTACGTCTT

CCAGGGTCGGTGTGGGGGTAGTACTACTTACTTGCGTCTT

CCAGGGTGGGTATGGGGGTAGTGCTACTTACTTGCGTCTT

CCGGGGTGGGTATGGGGGTGGTGCTACTTACTTGCGTCTT

CCTGGGTGGGTATGGGGGTGGTGCTACTTACTTGCGTCTT

CCACGGGTGGGTGTGGGGTAGTGTGTCTCACTTTACATCAC

CCCGGGGTGGGTGTGGGGTAGTGTATTATATTTACAGCCT

CCAGGGTCGGTGTGGGGTGGTGTACTTTTTCCTGTCCTTC

CCAGGGTCGGTATGGGGTAGTGTACTTTTTAATGATCTTC

CCCGGGGGAGAGCGGTGGGTAGTGTTCTATAGTATTCGTGT

CCAGGGGGGGTATGTTTTTAATACTACTTACTTACGTCTT

CCAGGGAGGGTATGGGGGTGGTGTTTCTAGTTTTGCGGCGT

CCAGGGTGGGCATGGGGGTGGTGTGGATTAAT1'CTTCGTCC CCAGGGTCGGTGTGGGGTGGTGTTTrTATTTACTCGTCGC

GGGGCGGCTTGGAAGAGGTTGCCGGTTGGAGTATTCGAGC

CCAGGTGTGGGGTGGTTTGGGTTTTCTTTCGTCGCC

CCAGGGTGGGTATGGGGGTTTAATTAATTCTTCGTCCCA

GGGGCGGCTTGGAAGAGGTTGCCGGTTGGAGTATTCGAGC

CCCGGGGTGGGTGTGGGGTGGTGTGAATTAATTCTTCGTCC

CCCGGGGTGGGTGTGGGGTGGTGTATTATATTTGCGGCCT

CCAGGGTCGGTGTGGGTGGTGTACTTTTTCCTGTCCTTC

GGGGC.GGCTTGGAAGAGGTTGCCGGTTGGAGTATTCGAGC

SEQ ID 56 57 58 59 61 62 63 64 66 67 68 69 71.

72 2 6 &38 7 9 11 17 21 24 28 &30 29 31 36 41 44 &45 Bold typeface indicates a discrepancy with the most common sequence of that group.

Table 7 Starting DNA and PCR primers for the ssDNA SELEX experiment SEQ ID

NO.

Starting ssDNA: 5'-ATCCGCCTGATTAGCGATACT[-40N.IACTTGAGCAAAATCACCTGCAGGGG-3 PCR Primer 3N2*: 5'-BBBCCCCTGCAGGTGATTTGCTCAAGT-3' 91 PCR Primer 5N2**: 5'-CCGAAGCTTAATACGACTCACTATAGGGATCCGCCTGATTAGCGATACT-3' 92 *B=biotin phosphoramidite Glen Research, Sterling, VA) "*For rounds 10, 11, and 12, the truncated PCR primer 5N2 (underlined) was used to amplify the template.

Table 8 Unique Sequences of the ssDNA high affinity ligands to PDGF [40N] ACTTGAGCAAAATCACCTGCAGGGG-3' SEQ ID

NO

14 AGGCTrGACAAAGGGCACCATGGCFTAGTGGTCCTAGT 93.

41 CAGGGCACTGCAAGCAATITGTGGTCCCAATGGGCTGAGT 94 6 CCAGGCAGTCATGGTCATTGTTTACAGTCGTGGAGTAGGT 23 AGGTGATCCCTGCAAAGGCAGGATAACGTCCTGAGCATC 96 2 ATGTGATCCCTGCAGAGGGAGGANACGTCTGAGCATC 97 34 CACGTGATCCCATAAGGGCTGCGCAAAATAGCAGAGCATC 98 8 GGTGGACTAGAGGGCAGCAAACGATCC1TGGTTAGCGTCC 99 1 GGTGCGACGAGGCTTACACAAACGTACACG1TCCCCGC 100 TGTCGGAGCAGGGGCGTACGAAAACTTACAGTTCCCCCG 101 040 AGTGGAACAGGGCACGGAGAGTCAAACTIGG1TrCCCCC 102 47 GTGGGTAGGGATCGGTGGATGCCTCGTCACTTCTAGTCCC 103 18 GGGCGCCCTAAACAAAGGGTGGTCAC1TCTAGTCCCAGGA 104 TCCGGGCTCGGGATTCGTGGTCACMrCAGTCCCGGATATA 105 ATGGGAGGGCGCGTTCTTCGTGGTAC1TrAGTCCCG 106 ACGGGAGGGCACGTTCTITCGTGGUTACTITAGTCCCG 107 13 GCTCGTAGGGGGCGATTC1TCGCCGTTACTTCCAGTCCT 108 16 GAGGCATGrrAACATGAGCATCGTCTCACGATCCTCAGCC 109 *36 CCACAGGCTACGGCACGTAGAGCATCACCATGATCCTGTG 110 GCGGGCATGGCACATGAGCATCTCTGATCCCGCAATCCTC

III

4 ACCGGGCTACTTCGTAGAGCATCTCTGATCCCGGTGCTCG 112 44 AAAGGGCGAACGTAGGTCGAGGCATCCATTGGATCCCTrC 113 24 ACGGGCTCTGTCACTGTGGCACTAGCAATAGTCCCGTCGC 114 7 GGGCAGACCTrCTGGACGAGCATCACCTATGTGATCCCG 115 $26 AGAGGGGAAGTAGGCTGCCTGACTCGAGAGAGTCCTCCCG 116 19 AGGGGTGCGAAACACATAATCCTCGCGGATTCCCATCGCT 117 48 GGGGGGGCAATGGCGGTACCTCTGGTCCCCTAAATAC 118 46 GCGGCTCAAAGTCCTGCTACCCGCAGCACATCTGTGGTC 119 1-rGGGCGTGAATGTCCACGGGTACCTCCGGTCCCAAAGAG 120 31 TCCGCGCAAGTCCCTGGTAAAGGGCAGCCCTAACTGGTC 121 12 CAAGTTCCCCACAAGACTGGGGCTGT-rCAAACCGCTAGTA 122 CAAGTAGGGCGCGACACACGTCCGGGCACCTAAGGTCCCA 123 038 AAAGTCGTGCAGGGTCCCCTGGAAGCATCTCCGATCCCAG 124 0 Indicates a boundary experiment was performed.

Italics indicate the clones that were found to retain high affinity binding as minimal ligands.

Tabl e 9 SEQ ID NO: Group 97 2 112 4 115 7 109 16 110 36 124 38 113 44 ill 50 96 23 9083 =AGGQ rJG-AGC-Nr-Cc 3 5,ATGTcA -CCGCAG= &CCMG~ CAC- TTC---G1~-G-\C C2AX-C

GCTCG

CC-TTICT GG4Q.3- AGC -4lTCAC- -CTAT -=TAJ-C jC3GJ- cQaG- -T'rAA CAT 3-AGC-f -7CTC- -ACG,-CTC

GCC

CCbgC&GP -CJACG-CCA------ CGT0-AGC-4TCA- CCA-

-T-CCG

AAA=TGTGCAGG jC-- -CCT GCj-AGC-4IT2- .TCC GAJ-=CAQacIL Ab&AG C~Gb4C-GPA GGCTCS2.-AGGC4T- ATT--- W A-.CC=

C

GCM.--CAM--GCA CATVG-AGC-4TC TCT QAX-QC(CAATCCTC =zAGP CA GATAAC-----G;TCCT -AGC-1iTCfc3I 5 AGGTAXCC=GCAA=

B

GCTCGTAG3W -GG3=2A-TTCTT -TCGC -GT'r-4CT- G--Ta Lj G Qr -GACA AAi-GGC-4CCA--GG(CTTAGTGy-.QCIgTh CCAG &A-ACAC-----GTCC GGC-14c- TA atg ATCGGAGG ~CGiCi- -TTCTT- G-GTT 4CM -TTT

AJ-CG

CLIrg3= -AATG'rc- -CAC-GGT-4CM TCC QG-C~AA TfCMxr- cQrCM-GAT j=r 3-GrC-4C2 TTC- AJU-CCGQGATATA L 5'TCCD!CAAQ

-CTGGTAA=

C A- CT- GTC L tgc3 5 ACGGGAGG-C.&M- -TTCTT- C-GTT TTT 13(' AGTa3' S 1LaLCAi3-GGC-

-TGCAAGCAATTGTX-CCCAAT=

7 GTGG~PATC3~i -ATG CC.-GTC-4C TCT---AXZ~c SEQ ID 108 93 123 104 106 120 105 121 1 07 941 102 NO: Group 13 14 15 20 3c 34 4 4 Table Frequency of base pairs in the helical regions of the consensus motif shown in Figure 3 Base pair" AT IA GC m other Position' I-i 11-2 11-3 11-4

III-I

111-2 111-3 aHelices are numbered with roman numerals as shown in Figure 3. Individual base pairs are numbered with arabic numerals starting with position 1 at the helix junction and increasing with increased distance from the junction.

bWe have included the TG and GT base pairs to the Watson-Crick base pairs for this analysis. There is a total of 21 sequences in the set.

Table 11 Affinities of the minimal DNA ligands to PDGF AA, PDGF AB and PDGF BB Kd, nM PDGF AWEDGE A PDGF BBb Ligand 47 ±4 72 12 49 8 0.147 0.011 0.094 0.011 0.138 0.009 0.127 0.031 0.093 0.009 0.129 ±0.011 'Data points shown in Figure 5A were fitted to eq 1 (Example 7).

bData points in Figures 5B and 5C were fitted to eq. 2. The dissociation constant (Kd) values shown are for the higher affinity binding component. The mole fraction of DNA that binds to PDGF AB or PDGF BB as the high affinity component ranges between 0.58 to 0.88. The Kd values for the lower affinity interaction range between 13 to 78 nM.

92 Table 12 Starting RNA and PCR primers for the 2'-fluoropyrimidine RNA SELEX experiment SEQ ID

NO

Starting 2'-fluoropyrimidine

RNA:

Starting RNA: N-J UUCGACAGGAGGCUCACAACAGGC-3- 125 PCR Primer 1: 5'TAAGCCCAAGAAAGAACCCA3 126 PCR Primer 2: 5.-GCCTGTTGTGAGCCTCCTGTCGAA-3 127 Table 13 Sequences of the 2'-fluoropyriDmidiDC RNA high affinity ligands to PDGF AB.

SEQ ID

NO.

I CGGUGGCAUUCUCACCCUCUCGCUCUCGCGUUGGGCNCG 128 2 CCAACCUUCUGUCGGCGUUGCUUUGGACGGCACUCAGGCUCC 129 3 UCGAUCGGUUGUGUGCCGGACAGCCUUAACCAGGGCUGGGACCGAGGCC 130 4 CUGAGUAGGGGAGGAAGUGAAUCAGUUGUGGCGCCUCUCAUUCGC 131 CAGCACUUTUCGCUUUTUCAUCAUUUUCUU1JCCACUGLU.GGGCGCGGAA 132 6 UCAGUGCUGGCGUCAUGUCUCGAUGGGGAUUCUUCAGCACUUUGCCA 133 7 UCAUUCUUUUUCUUCCACGUUCGGAC 134 8 CGAUAGUGACUACGAUGACGAAGGCCGCGGGUGGAUGCCCGCATGA 135 GUCGAUACUGGCGACUUGCUCCAUUGGCCGAUUAACGAUUCGGUCAG 136 13 GUGCAAACUUAACCCGGGAACCGCGCGUCGAUCGACCCUCCA 137 ALJUCCGCGUUCCGAUUAAUCCUGUGCUCGGAAAUCGGUAGCCAUAGUGCA 138 16 CGAACGAGGAGGGAGUGGCAAGGGAUGGUUGGAUAGGCUCUACGCUCA 139 17 GCGAAACUGGCGACUGCUCCAUGGCCGAUAUAACGAUCGGUUCAU 14 18 CGAACGAGGAGGGAGUCGCAAGGGAUGGUGGAUAGGCUCUACGCUCAA 141 19 CGAGAAGUGACUACGAUGACGAAGGCCGCGGGUTUGAAUCCCUCAUUGA 14 AGCAACGAGACCUGACGCCUGAUGUGACUGUGCUGCACCCGAUUCUG 143 21 UGAUCUAUUUCAUUUCUCACAACULUUUCCACUIJCAGCGUGA 144 22 AAGCAACGAGACUCGACGCCUGAUGUGACUGUGCUGCACCCGAUUCU 145 23 CGACGGUGUUGCGGCAGCIJUUGACCAUGAGCUGGGACCGAGGCC 146 24 NGACGNGUGGACCUGACUAAUCGACUGAUCAAAGAUCCCGCCCAGAUGGG 147 26 ACGCGCUGCAAACCUUGUGUGGCGGUACCCCCUUUGGCCUCG 148 27 GGUGGCAUUCUUCAUUUCCUUCUCGCUUCUCCGCCGUGGGCGCG 149 29 CCUGAGUAGGGGGGAAAGUGAAUCAGUUGUGGCGCUCUACUCAUUCGCC 150 GUCGAAACUGGCGACUGCUCCAUGGCCGAUAUAACGAUCGGUUCA 152 31 GCGAUACUGGCGACUUGCUCCAUUGGCCGAUAUAACGAUUCGGCUCAG 32 ACGUGGGGCACAGGACCGAGAGUCCCUCCGGCAAUAGCCGCUACCCCACC 153 33 CACAGCCUNANAGGGGGGAAGUUGAAUCAGUUGUGGCGCUCUACUCATCG 154 34 ANGGGNUAUGGUGACUGCUCCAUUGGCCGAUAUAACGAUCGGUCAG 155 CCUGCGUAGGGNGGGAAGUUGAAUCAGUUGUGGCGCUCUACUCAUUCGCC 156 39 CGAACGAGGAGGGAGUGGCAAGGGAUGGUUGGAUAGGCUCUACGCUCA 157 41 GUGCAAACUUAACCCGGGAACCGCGCGCGAUCGCUUCCNUAUUCCA 158 42 CGAACGAGGAGGGAGUGGCAAGGGACGGUNNAUAGGCUCUACGCUCA 159 43 UCGGUGUGGCUCAGAAACUGACACGCGUGAGCUUCGCACACAUCUGC 160 44 UAUCGCUUU1JCAUCAAUUCAUUUAUUUACUGCUC 6 GUGCAAACUUAACCCGGGAACCGCGCGUUUCGAUCCUGCAUCCUCC 162 46 UCGNIJCGGUUGUGUGCCGGCAGCUGUCCAGCGUGGGCCGAGGC 163 47 AGUACCCAUCUCAUCUUCUUCUCUAGAAUAGU 16 49 CCUGAGUAGGGGGGGAAGUUGAACCAGUUGUGGCNGCCUACUCAUUCNCCA 165 51 CCNNCCUNCUGUCGGCGCUUGUCUUGGACGGGCACCCAGGGCUC 166 52 CAACUNCGUCGCGUUGUCUUUUUGGACGAGCAACUCAAGGCUCGU 167 53 CCAGCGCAGAUCCCGGGCUGAAGUGACUGCCGGCAACGGCCGCUCCA 168 54 UUCCCGUAACAACUUU1JCA1)UUCACUUUCAUCCAACCAGUGAGCAGC 169 UAUCGCUUUCAUCAAAUUCCACUCCUUCACUUCUUAACUUGGGCGUC 170 Table 14 Starting RNAs: 40N7: 5'GGGAGGACGAUGCGGII.4ON-]CAGACGACUCGCCCGA 3' (SEQ ID NO: 186) SELEX PCR Primers: G7: 3' T7 Promoter (SEQ ID NO: 187) 3G7: 3' (E DN:18 (SEQ ID NO: 188) Table Conditions and progress of the SELEX against hKGF KDrnM 30.0 Roud RNI tGFI- MA net bound SP-nal/noise 1 1.0013-06 3.0O13-07 4.4 1.

2 4.0013-06 3.0013-07 1.5 4.2 3 1.0011-06 1.0012-07 5.9 20.6 4 1.OO13-06 l.OOE-07 14.3 12.8 3.0013-07 l.OOE-08 2.5 4.5 6 3.7013-08 3.70E-09 0.7 2.6 7 4.I1OE-09 4.10E-1O 1.1 8.2 8 4.6013-10 4.6013-l1 1.5 8.8 9 5.lO1-1l 5.1013-12 0.7 5.9 1.7013- 11 .70E- 12 0.3 2.1 PF, Spin" R-Wash' (ml) U-Wash' (ml) 1 .5812-05 8.52E2-07 3.2 1 E-06 7.64E-08 3.7312-07 2.4613-08 2.04E-09 8.76E- 4.12 2E- 17.0 0.7 0.3 Tr. SELEX 1 L .00-06 2 4.00r.-06 3 3.0013-06 4 2.5013-06 6.7013-07 6 1 .20E-08 7 1 .4013-09 8 1.5012-10 9 1.7013-11 1.70C-I1 3.000-07 3.0013-07 3.0013-07 3.0013-07 3.0011-08 1 .2-3E-09 1 .4013-10 I .5013-11 1.70E- 12 1.7013-12 11.0 9.9 5.7 11.7 5.8 1.8 11.2 4.8 3.1 3.0 9.2813-06 2.1511-06 4.9813-06 3.64E-06 1.59E-07 6.86E-09 5.36E- 5.67E-I10 1.42E-10 30.0 15.0 aprelitered RNA through nitrocellulose to counter select for nitrocellulose binding molecules ImBrief spinning of the binding reactions ,Volume of buffer used to wash the captured complexes 'Volume co .M urea wash following the buffer wash 'Calculated single point K, from the binding data at'each round values obtained from binding curves Table 16 Scquenccs of 2'-N11 1 and 2'-F KGF ligands Clilne 5' conistanit randoin 3'constant SQI NO 2II lg(CA(A(A J3 (;GGACC;AIIAAA( AGGAALJC(,AACAACAAG[IGCCCIJGCC 2N C300AGG;ACGAUGCGG GC(CGGAAGGUCC(GAACiACCCGAGACGALJC 4N GCGAGGACGAUGCGG

GUGGUGAAGAGGUACC(JGAAUUGCLJAAAGAUACCACGGCC

6N GGGACGACGAIJGCGG

GCAGGGACCAAUGAACUCAAGUCAACCCGGUGCACGUGGG

ION GjCGAGGACGAIJOCGG tIAGCtUGCUGLJUCAUGCAAGACACUA(IAA(IAI JUAAGAtIGGGG II N G(;GAGOACGAI JUCCi GGGCCG;GAt)UUGAACCGACGACUUCGGGUt)AUGAGCCCGACGU I4N G(GAGGAC1AIJGCGCI

UCCAGGCAUUGAACEJCUCGGGGUAGGAACAIJAAAGGCGGC

16N GGGAGGACCIAI ICGC AAGI IUCtJAACAACIUUAtIGGAAGGuJtICCACtjtJ(;AAtIGUA 22N C(A6(ACAICGCi At)JUGAGCUGAAAU 24N CGGAGGACCIAIJCGG

GUGC.GAAGAIJGAGCC(;GUCGGCAGUAAUGIJGACACUGCGG

CatJCACuACCAIJGCGG GA(GGAAtJCAGGAAACAACtJA(,CAGAIJAACCGAGCI)G(jC 27N GGGAGUACGAUGCGG

AUCEGAGCUCIAAAU

29N (-GCIAGCIACrsAtI(;CCIO tJLJGCtJCIJACAAtJGACCECGGtJGACLJCCGCAGI

IJCUUGGACA

29N (;(i(oA(;oiAC(iA1 i(i'C; GAG(CmeGA~iAAGAAtI GCAGGAAACAGCOmAAAIJGCGI j(;tx;GCc 34tN G(GAC.GACGAIMICC.i ('Cc(;(;AA(;AGsCtJAAtI(,GAAGtJ(;(iAAI ICAGI I)CACAGI I(C;G C3G;(A(;ACGiA1I(;CCi~i (JCULJAGGGAAAUG(IUUCLI(IAUUUGLtI 36N GGCAGGACCAUGCGO GAAGGGAACAGGAt)AAGACAAGUCOAACAAAGCCGAGGUG 37N GGGCoCACGAIJUC(1G

AUGCIAGCIJGAAAU

42N GGGAGG~ACGAL)GCOjG

GGACGCUAGACGGGAACAUAGAACGAACA(JCAACGCUGC

43N GOACIGACGAUGCOG CAA(ILt)GGAUAGAACAGUCAGAAAUGUAAGCGUGAGGUG 47N GGGAGGACGAUGCGG CAAGGGUAGGAAGGUCAAGAGGAAACAGCGCtiUCGGGGUCG 48N GGGAG(IACGAUCCG

GC

1 CAAAGGAAGUUGGAAUCGGGACUAAGUAGtJGUGUGGC 54N GGGAGGACGAUGCGG AGAACCAACAGaAGCCCCCUGGUGGU(IGG(;GAAGGAUUCU SSN GGGAGGACGALJGCGG ACACACAAGUGAAGGtJCAGACGCGAAUUACGtJGGGUGGG 57N GGGAGGACGAUGCG(3 UCGiUGG(GGUGGGUGGGGGCAGCGUUGGAAUAA(JUAACUGGIJAACG(ICUGG;C 59N GGGAGCACGiAIC~o GGUGGGUGGt)UACCUGUAAUUAUAUUGAUt)CUGGCUUUAG GC3GAGGACOALJGCGG

CCCCUUAGCUCAGUGGULJAGAG

GGGiAC1CACAI ICOCG UAACGUGGAAUACJGGIJUAAACAGCI)JC.CAAAIJAACGIJAGCJUG(GC 69N (GA(KGACCiAU(iC(I (dIJAGiCAGUAGCGACA(;ACAI JAACA(GIJGCAACCALJCCjIJGGC CA(;ACGAC)CGiC(CGCA'

CAGACGACUCGCCC

1

A

CAGACGACUCGCCC3A

CAGACGACUCGCCCGA

CACIACGACtJCGCCCCA

CA(GACGAC(JCGCCCGA

CAGACGACUCGCCCGA

CAGACGACUCGCCCJA

CAGACGACUCGCCCA

CAGACCDACUCGCCCG;A

CAGACGACUCCICCCGA

CAGACGACUCGCCC;A

CAGACGACtJCGCCCGA CA(GACCACtCGCCCGA CAGACGACtJCGCCCG;A

CAG;ACGACUCGCCCU;A

CAGACGACUCGCCCcJA

CAGACGACUCGCCCGA

CAGACGACUCGCCCGA

CAGACGACUCGCCCIA

CAGACGACUCGCCCA

CAGACGACUCGCCCGA

CACIACGACIJCGCCCGA

CAGACGACtJCGCCCGA

CAGACGACUCGCCCGA

CAGACGACUCGCCCGA

CAGACCACUCGCCCGA

CACIACGACO CGCCCCA CA(iACCIACIJCGCCCGA 189 190 191 192 193 19I 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 Table 16 (Page 2) randoin Y constant Clone 5 constant1 SEQ ID)

NO:

71N cGGAGGACGAUGCGG

AAACGGCGUGGCAAAAGUGAGGGGGUAGGAUGUACCAUGGGJ

72N GGGAGGACGAUCJCGC GAGGGGAAAAUGAGACCCACAGAUtJGACGGAAGUACtJGG(i 2F' GGGAGjGACGAIJGCGCI GCAtJUCGUCAAUACCUUGLJUUI JAUUCCUUUI JCUAGCGGCC 31' GC.CAGCACGAIJOCGG

AUCCEUAACCCACUACIJACUUUCCGAACCCGCACG()GGC

51 CGjGAGG(ACGAtJIGCGG CGIJCCCGAcLJCACGCtl(IUCCIJGAI)AACCJJCtJCI

J(UGCC

61: GG( J(C(G G AK JC(CULJI 1;(11icGGCtJC:JGK it IACCC( CUCGUI J( .uCCCCCC 7F GGGA(;6ACcIAIGCG(; CUOOUACtJCLJUCGCCAGCUCC( iCCAAGCGCGACCUGUGCC 81: 1 C.CAC.ACGUIGCG Ut IUCGAAtUAGGGCCAUIIICIJCACtUAGsCtAIICCUACCCIJGCC 917 CICIIAGOACGAtIGCGO AUAALIGGCIJAGAACUAGCI ICGCAUCIJUGGUGI

JCCGCIUGCC

101' C 1 06AGGACGAtIGCGG~ ClACCAGAIJGGCGGAUUULIUICAGCAAIJCCLICCCCCGCtJGCC III, (iGGACGcACCAJCCGG IOGAKIC.CCGACCACUCAAACCGGUGCIJI

JIJUACLIJCCCCCGC

121: GGC'AGGACGAI If CGG GAAIJtAACAC'GGCCA(GAAtJLJCUCAUCI INNCUIJCCCGtJGACC O31 (GGGAGGACGAtJGCGO CACCUUAGACCtJGUCCUCCAGCGIJGA(iJUGCUGLJGGCC 1 4F GGGAG('ACGAIJIGCGG UGGt)CUCCCAAULJCUAAACI t IIJCtJCCAUCGUAUCLJGGGC 131. (1(GAGGACGAIJUCGhO ICAUGGLiIGUCt)UUCCACAGCUCLJUCCCAUGAUCGCCCGGC 1 61' GGGAGGACGAIK'.CGO (IAAU(JCCCAGCGCUUGACt)GAUACAAACNJLJCCCGUGCCC 19F' GGOAGGACCEAI ICOG CAA-NNNNNNNCUCUCUCCIJGGCGLIUCCGCAACCCGCCCC 201: cGOAGdIACGAIIC"CGG AMl JAUUCCAGCCt IGGAI IJCAI IA(oIUCAGUGCUCICCGUGCC 211: GGGAGGACGAIJ(COG LiCCUAGCAGCGAt)UCAtICCCCGULJCt)CUCAGCGtJtJOCCCC 221: GGGAGGACGUVOCO

CCUGAAGUACAGGCUCUAAACUCCAAGCOCGACCGUCCGC

23F OGGAGGACGAUGCGG

CCCUACCACUUUUUCCCUCUACUGUUAUCCUGUCCCC

241' GCICAGGACGAIMICGO UOO~rUCtICCCUAGAtiICIJACAGCACtIJUCCAUICGCAUtJGGGC 261' c,((;A(;(sAc(GAUJ(Ccc I ICAArCIJIJAACAOtiJCtIJGGCAAI JOOBcsCAI IIIAI JGCG(C(C 271: COOAOGiACOAUJOCI CaOItJCtJ(OUAUCUCUAIJIJG( AAUtIJ AGUUCCUCAACUGU(;(Cc 281- GGGAGGACGAUGCGG GAUUCIUUUCGOCAA(OIJGAAAAUAJCCt)UGCULJCCCGAGC 291' GGOAGOACO#AtJ(,CG G OACUUCAACU)AAGtJCCUCAUUUGCCUCGCUCCUCGUGCC 311: GGGAGGACGAIMGO AACOOAGAUOIJlCCCCtJCAAMAULJLIACC(IUCt)CCGIJUIJG(.GCCC 351- GOOAOCJACALJOG CGAAAUUAGCUUCULJAU(sACLJCACGUtJIJCCt)UGCCGCCC 37F' GGGAGOACGIAI JOCM OCCCOiAUCU)ACUGCAUUACCGAAACOeAUUtJCCCCACUtiJG 381: GGCAGGACCAUGCGG NOACt)GAUUUUUCCUUGNCAGUGUAAUUUCCUGGCUGCCC

CAGACGACUCGCCCGA

CAGACGACIJCGCCCGA

CAGACGACUCGCCCGA

CA(;ACGACtJCGCCCGA

CAGACGACUCGCCCGA

CA(;ACGACIJCGCCCGA

CAGACGAC1JCGCCCGA

CAGACGACUCGCCCG;A

CAGACGACUCGCCC(,A

CAGACGACIJCCGCCCGA

CAGACGACUCGCCCGA

CAGACGACtJCGCCCGA

CAGACGACUCGCCCCJA

CAGACO AC! JCGCCCGA

CAGACGACUCCICCCGA

CAGACGACUCGCCCGA

CAGACGACtJCGCCCGA

CAGACGACUCCICCCGA

CAGACGACUCGCCCGA

CAGACGACUCGCCCGA

CAGACGACUCGCCCGA

CAGACGAC1ICGCCCGA CAGACGACI ICGCC('GA

CAGACGACUCGCCC(;A

CAGACGACUCGCCCGA

CAGACGACUCGCCCGA

CAGACGACIJCGCCCGA

CAGACGACUCCJCCCGA

CAGACGACtJCGCCCGA

CAGACGACUCGCCCGA

218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240O 241 242 243 2,14 245 246 247 Table 16 (Page 3) Y constant random Clone S constant SEQ ID

NO:

(;(iGAGuGACGAIJGCGCS( GGACUUUGACAGGCAUUGAUUUCGACCUGUIJCCCCGtJG(iC GGGAGGACGALJGCGG CGACACAAUAGCCtJIJUGAUCCCAUGAUGGCUCGCCGUGCC GGGAGGAC'GAU(GC(iG IJGUAGIJLJUCCCUGUAt)GCCAUtJCUIJUCCCAUGCCGCACGC GGCJACiGACGAtI IGGC' UCGAGUCUUCIJCC(JUCGG()AACUAtiJIINNNNAtlUUCGUGCC (GGGAGGACGALJGCGG

GUCGUAUUCAUCUCCUUGUUCUGUUUCGUUGCACCUGGCC

GGCJAGGACGALJGCGG CIGACUUUGiACAGGCaUU(JAUUUCGACGUGUUCCCCGLJGGC GGCAGzCACGAIJGCGG UGAtJCAAUCGGCGCIJIJUACUCUUGCGCtJCACCGtJGCCC

CC

1 A6CACGALJGC(IG CAGtJCUCCCUACIGUUtJ('AUCtJCU(ICA(CAUtJCCGGGGIJNC 6GC1AG;GACUiAIJ(jCCIG AUICAAAAGCACIJCAIJIJCC('Gt)GCI JCCICIJUCAUUGGI

ICCCC

AAGAI JCI JCCCAACI J(CtIG IJ((;CIJAAUAA 111 JCt IJCCGC(IItCCC (IGUiA(CiGACGC;A(c(; UCCGtiCAt)AACGGCCA(jAAACt)GC(*AALJACtJCCCUGGCC (GGAGGACCiAlI JCOsG CGACAAWYAGCGGtJGIJCIJtIIICAUIJUNKAtJ(CIJCCGACRUCC GGGAG(1AC(;AIJ(ICGG

LCACUAUCUGGC)CGAJCCAAUCACCCGA(;CCCACCGCGC

GG,6AGGiACGAtJGCGGO GAACUAAUGGCCGtJGAUUAACCAAt)GCAGGCUUCCIJGCGC GGCJACGACGAUGCGG tJGACAUGGAAUUUUCUACGGGCCCGAUCCUGCCAGCCGUGUG

CAGACC.ACUCGCCCGA

CAGACGACUCGCCCIA

CAGACGACUCGCCC(;A

CACIACGACUCGCCCCIA

CAGACGACUCGCCCGA

CACIACGACUCGCCCJA

CAGACGACUCGCCCGA

CAGACGACIJCGCCCGA

CAGACGACUCG;CCCGA

CAGACGACt)C(CCC(;A

CAGACGACUCGCCCGA

CAGACGACtJCGCCCGA

CAGACGACUCGCCCGA

CAGACGACUCGCCCGA

CAGACGACUCGCCCGA

248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 Table 17 K, values 1IKGF ligands

K

8 in nM K, in nM

C

2 4 6 2 2 2 2 2 2 3 3 :lone 1 2 Clone I t N 0.51 2F N 0.77 3F N 0.75 5F N 0.71 6F ON 1.10 7F IN 1.28 8F 4N 0.44 9F 6N 1.40 lOP 2N 5.70 Ill 4N 1.16 12F 5N 0.87 13F 7N ND 14F '8N 2.54 1SF 9N 0.43 16F 4N 0.80 19F 5N 2.32 20F 6N 8.27 21F 17N ND 22F 12N 0.78 23F 13N 0.79 24F l7N 1.76 26F 48N 1.34 27F 54N 5.35 28F 1.25 29F 57N 35.8 31F 59N 22.0 3SF 7.38 37F 26.56 38F 69N 15.20 41F 71N 3.52 42F 72N 7.67 43F random 30 44F 4SF 49F

SOF

51F 53F 54F 56F 57F 58F random For biphasic curves, Kd I is for th igh affinity component.

1.77 4.47 2.53 0.05 (37) 3.69 2.63 0.83 0.47 3.74 1.38 0.03 (28) 0.006-0.03 (25-44) 0l.07 (33) 0.83 (49) 1.6 2.05

ND

2.75 2.52 2.02 0.23 (43) 1.52

ND

3.24 1.1 0.46 0.33 1.44 0.9 1.13 1.32 4.7 0.12 (21) 1.27 0.70 1.23 2.52 0.07 (32) 1.20 2.52 2.10 3.25 3.39 0.94-2.57 8.70 44.8 2.55 2.10 3.00 Number in parentheses indicate the per cent of the high affinity component.

Table 18 Binding Specificity of the 2'-F Ligand K14F Ratio: Target KDTargetIKDhKGF human hKGF 1 rat hKGF 1,254 human aFGF .38,65 0 human bFGF 1,071 human PDGF 432 Th e ratios shown are av'erages of at least two determinations Table 19

IC

50 values from the PC-3 assay Competitor IC50, nM hKGF Heparin, 5,000 40N7F >1000 K6F 4 K13F K14F K56FI KIOF K37F K38F 0.6 K43F 40N7N >1000 KIN K2N 200 K4N K6N K14N 6 K29N K42N 800 K43N 800 Table Ki values of hKGF competitors on the PO3 and NIH3T3IFGFR-2 competition assny r"I i i; rnmnPtitrr Ki~ nM 39 U PC-3 hKGF 2'F random 2rNH 2 random Hep5000 K 14F K6F K3 8F K56F K14N 7.700 930.000 673.000 6.500 0.200 0.160 0.220 0.160 1.400 0.95519 0.997 13 0.853 57 0.99984 0.9773 0.95927 0.990 13 0.95927 0.94698 0.9933 0.97856 0.99047 0.96202 0.98659 0.97582 0.97856 NIH3T3/FGFR-2 hKGF 0.034 2'F random 10,000.000 2'NH, random >10,000.000 Hep5000 26.300 K14F 2.700 K6F 6.800 K38F 20.000 K56F 27.400 K14N 10.600 partial comp.

partial comp.

Table 21 values obtained with the gel shift assay Competitor 1C50. nM KGF bFGF 1,500 Lysozyme 10,000 Table 22 Binding Specificity of Ligand K14F3'T random RNA K141737 Protein -Kd, rM b Kd2, nM Kd 1, M Kd2, nM cDF hKGF 20.1 0.0008 10.2 1 rKGF 45.3 0.004 1 70.0 hbFGF 0.0375 10.3 10.0 haFGF 16,000,000 24,000,000 3x10' 0 hPDGF-AB 22.0 50.0 6.2x hTGFp 1 10.4 98.0 1.2xI0 hEGF 2,000 256 IWOx1' Thrombin 7,200,000 22,700,000 2.8x 1010 'High affinity dissociation constant from biphasic binding curves.

'Low affinity dissociation constant from biphasic binding curves or affinity dissociation constant from monophasic binding curves.

'Discrimination factor defined as the ratio of the highest affinity Kd of 141737 for the corresponding protein over the affinity Kd for hKGF.

Table 23 SEQ ID NQ:9 Si LI S2 Si' L3 S2' og, OAC IGAG~QD -I UCXJCI AAUUCUAAACLJUUCU I CAUCGUAI1C 1431T T2 T3 T4 T6 T7

TB

T11 T12 T13 T14 T16 T18 T19 T21 T22 T29 T31 T32 GGGAGG IAC IGAUGCGGUGGIjUCUCCC

IAAUUCUAAACUUUCI

GGGAGG IACI .UGCGGUGG IUCUCCC IAAUUCUAAACUUUCU

I

GGGAGG AC IGAUGCGGUGG IUCUCCCI

CUAAACUUUCUI

GGGAGG AC IGAUGCGGUGGI UCUCCC IAAUUCUA. UUCU I GGGAGG AC IGAUGCGGUGG IJCUCCC IAAUUCUAAACU..

I

GGGAGGIAC (GAUGCGGUGG IUCUCCCIAU AACU

I

GAGG AC IGAUGCGGUGGIUCUCCC IAAUIJCUAAACLJUUCU

I

GGGAGGIAC IGAUGCGGUGG UCUCCC AAUU UUCU I GGGAGG AC IGAUGCGGUGGIjUCUCCC IAAUUCUA

I

GGAGGIACIjGAUGCGGUGG LJCUCCC IAAUUCUA. UUCU I GGAGG ACI GAUGCGGUGGIUCUCC. IAAUUCUA"

UUCUI

*..GAGG IACj GAUGCGGUGG UCUCCC IAAUUCJA.. .UUCU I *..GAGGIACI GAUGCGGUGG UCUC. AAUUCUA...

.UUCUI

GGGAGG CIGAUGCGGUGGIjUCUCCC IAAUUCUA.... UUICU I GGGAGG AC IGAUGCGGUGGIUCUCCC IAAUUCUA UCU I GGGAGG IAC IGAUGCGGUGG UCUCCC IAAUUCUA

CU

GGGAGGIAC IGAUGCGGUGG UCLJCCC AAUUCUA

U

GGGAGGIAC IGAUGCGGUGGIUCUCCC IAAUUCU

UUCU

GGGAGGIAC IGAUGCGGUGGIUCIJCCC AAUUC UUqu CCC IAC IGAUGCGGUGGI. C. CCC IAAUUCUA

UUCUJ

GGG... IACIGAUGCGGUGGI CCCjAAUUCUA.

.UUCU

GACIGAUGCGGUGGI... .CCCIAAUUCIJA....

.UUCU

GGIACIGAUGCGGUGGI CCCIAAUUCUA.

.UUCU

CCAUCGUA..

CCAUCGUA..

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

CCAUCGUAUC

I CCAUCGUAUC I CCAUCGUAUC I CCAUCGIJAUC I CCAUCGUAUC I CCAUCGUAUC I CCAUCGUAUC I CCAUCGUAUC I CCAUCGUJAUC H L 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 Table 23 (Page 2) SEQ ID SI 1.1 I I L T3 3 T3 4 T3 5 T3 6 T3 7 T39 T4 0 T4 1 T22mu T35/36 IGAUGCGGUGGI CCC AUCU.

.jGAUGCGGUGG IUCUCCCIAAUUiCUA..

.j....GGUGGIUCUCCCIAAUUCUA..

GGGAGGIACIGAUG.............. GGGAGGIACIGAUGCGGUGGIUCJCCCIAAUUC GGGAGGIACIGAUGCGGUG. IUCUCCCIAAUUC GGGAGG IAC IGAUGCGG .GG IUCUCCC jAAUUC GGGAGGIACIGAUGCGG.G. IUCUCCCIAAUUC GGGAGUIAC IGAUGCGGUGG jUCUCCC IAAUUC equimolar amounts of T35 and T36

.IJTUCUICCAUCGUAUC

.UUCUICCAUCGUAUC

.UUCUICCAUCGUAUC

UCUICCAUCGUAUC

.UUCulJ.CAUCGUAUC

.UUCUICC.UCGUAUC

.UtJCUI .C.UCGUAUC

.UUCUICCAUCGUAUC

296 297 298 299 300 301 302 303 304

Claims (12)

1. A method of identifying nucleic acid ligands of PDGF, comprising: a) contacting a candidate mixture of nucleic acids with PDGF, wherein nucleic acids having an increased affinity to PDGF relative to the candidate mixture may be partitioned from the remainder of the candidate mixture; and b) partitioning the increased affinity nucleic acids from the remainder of the candidate mixture; and c) amplifying the increased affinity nucleic acids to yield a mixture of nucleic acids enriched for nucleic acid sequences with relatively higher affinity and specificity for binding to PDGF, whereby nucleic acid ligands of PDGF may be identified.

2. The method of claim 1, further comprising: d) repeating steps and c).

3. The method of claim 1, wherein said candidate mixture of nucleic acids is comprised of single stranded nucleic acids.

4. The method of claim 3, wherein said single stranded nucleic acids are ribonucleic acids. The method of claim 3, wherein said single stranded nucleic acids are deoxyribonucleic acids.

6. The method of claim 4, wherein said nucleic acids comprise modified nucleic acids.

7. The method of claim 6, wherein said nucleic acids are 2'-amino (2'-NH 2 modified ribonucleic acids. P QOPERJEiH\Rcs C:MS\-1(9J 9.70fl cm, dccs 180

8. The method of claim 6, wherein said nucleic acids are 2'-fluoro modified ribonucleic acids.

9. A purified and isolated non-naturally occurring nucleic acid ligand to PDGF. The purified and isolated non-naturally occurring nucleic acid ligand of claim 9, wherein said nucleic acid ligand is single-stranded.

11. The purified and isolated non-naturally occurring nucleic acid ligand of claim 10, wherein said nucleic acid ligand is ribonucleic acid.

12. The purified and isolated non-naturally occuring nucleic acid ligand of claim 10, wherein said nucleic acid ligand is deoxyribonucleic acid.

13. A nucleic acid ligand to PDGF identified according to the method comprising: a) contacting a candidate mixture of nucleic acids with PDGF, wherein nucleic acids having an increased affinity to PDGF relative to the candidate mixture may be partitioned from the remainder of the candidate mixture; and b) partitioning the increased affinity nucleic acids from the remainder of the candidate mixture; and c) amplifying the increased affinity nucleic acids to yield a mixture of nucleic acids enriched for nucleic acid sequences with relatively higher affinity and specificity for binding to PDGF, whereby nucleic acid ligands of PDGF may be identified.

14. A method according to any one of claims 1 to 8 or a nucleic acid ligand according to any one of claims 9 to 13 substantially as hereinbefore described with reference to the Figures and/or Examples.