AU6450796A

AU6450796A - High affinity nucleic acid ligands to lectins

Info

Publication number: AU6450796A
Application number: AU64507/96A
Authority: AU
Inventors: Philippe Bridonneau; Larry Gold; Brian Hicke; David H. Parma
Original assignee: Nexstar Pharmaceuticals Inc
Current assignee: Nexstar Pharmaceuticals Inc
Priority date: 1995-06-07
Filing date: 1996-06-05
Publication date: 1996-12-30
Anticipated expiration: 2016-06-05
Also published as: JPH11507526A; EP0840739A4; EP0840739A1; CA2223275A1; JP2009039126A; WO1996040703A1; AU725590B2

Description

High Affinity Nucleic Acid Ligands to Lectins

FIELD OF THE INVENTION

Described herein are methods for identifying and preparing high-affinity nucleic acid ligands to lectins. Lectins are carbohydrate binding proteins. The method utilized herein for identifying such nucleic acidligands is called SELEX, an acronym for Systematic Evolution ofLigands by Exponential enrichment.

Specifically disclosed herein are high-affinity nucleic acid ligands to wheatgerm agglutinin (WGA), L-selectin, E-selectin, and P-selectin.

BACKGROUND OF THE INVENTION

The biological role oflectins (non-enzymatic carbohydrate-bindingproteins ofnon-immune origin; I. J. Goldstein et al., 1980, Nature 285:66) is inextricably linked to that ofcarbohydrates. One function ofcarbohydrates is the modification of physical characteristics ofglyco-conjugates (i.e., solubility, stability, activity, susceptibility to enzyme or antibody recognition), however, amore interesting and relevant aspect ofcarbohydrate biology has emerged in recentyears; the

carbohydrate portions ofglyco-conjugates are information rich molecules (N.

Sharon and H. Lis, 1989, Science 246:227-234; K. Drickamer and M. Taylor, 1993, Annu. Rev. Cell Biol.9:237-264; A. Varki, 1993, Glycobiol.3:97-130). Within limits, the binding ofcarbohydrates by lectins is specific (i.e., there are lectins thatbind only galactose or N-acetylgalactose; other lectins bind mannose; still others bind sialic acid and so on; K. Drickamer and M. Taylor, supra). Specificity ofbinding enables lectins to decode information contained in the carbohydrate portion ofglyco-conjugates and thereby mediate many importantbiological functions.

Numerous mammalian, plant, microbial and viral lectins have been described (I. Ofek and N. Sharon, 1990, Current Topics in Microbiol.and Immunol.151:91- 113; K. Drickamer and M. Taylor, supra; I. J. Goldstein and R. D. Poretz, 1986, in The Lectins, p.p.33-247; A. Varki, supra). These proteins mediate a diverse array ofbiological processes which include: trafficking oflysosomal enzymes, clearance ofserum proteins, endocytosis, phagocytosis, opsonization, microbial and viral infections, toxinbinding, fertilization, immune and inflammatory responses, cell adhesion and migration in development and in pathological conditions such as metastasis. Roles in symbiosis and host defense have been proposed for plant lectins but remain controversial. While the functional role ofsome lectins is well understood, that ofmany others is understood poorly or not at all.

The diversity and importance ofprocesses mediated by lectins is illustrated by two well documentedmammalian lectins, the asialoglycoprotein receptor andthe serummannose bindingprotein, andby the viral lectin, influenzavirus

hemagglutinin. The hepatic asialoglycoproteinreceptor specifically binds galactose andN-acetylgalactose and thereby mediates the clearance ofserumglycoproteins that present terminal N-acetylgalactose or galactose residues, exposedby the prior removal ofa terminal sialic acid. The human mannose-binding protein (MBP) is a serum protein thatbinds terminal mannose, fucose and N-acetylglucosamine residues. These terminal residues are common on microbes butnot mammalian glyco-conjugates. The binding specificity ofMBP constitutes anon-immune mechanismfordistinguishing selffrom non-selfand mediates host defense through opsonization andcomplement fixation. Influenzavirus hemagglutinin mediates the initial step ofinfection, attachment to nasal epithelial cells, by binding sialic acid residues ofcell-surface receptors.

The diversity oflectin mediated functions provides avast array ofpotential therapeutic targets for lectin antagonists. Both lectins thatbindendogenous carbohydrates andthose thatbind exogenous carbohydrates are target candidates. Forexample, antagonists to the mammalian selectins, afamily ofendogenous carbohydratebinding lectins, may have therapeutic applications in avariety of leukocyte-mediated disease states. Inhibition ofselectin binding to its receptor blocks cellular adhesion and consequently may be useful in treating inflammation, coagulation, transplant rejection, tumor metastasis, rheumatoid arthritis, reperfusion injury, stroke, myocardial infarction, burns, psoriasis, multiple sclerosis, bacterial sepsis, hypovolaemic andtraumatic shock, acute lung injury, andARDS.

The selectins, E-, P- and L-, are three homologous C-type lectins that recognize the tetrasaccharide, sialyl-Lewis^x (C. Foxall et al, 1992, J. Cell Biol. 117,895-902). Selectins mediate the initial adhesion ofneutrophils andmonocytes to activated vascular endothelium at sites ofinflammation (R. S. Cotran et al., 1986, J. Exp. Med.164, 661-; M. A. Jutila et al., 1989, J. Immunol.143,3318-; J. G. Geng et al., 1990, Nature, 757; U. H. Von Adrian et al., 1994, Am. J. Physiol. Heart Circ. Physiol.263, H1034-H1044). In addition, L-selectin is responsible for the homing oflymphocytes to peripheral andmesenteric lymphnodes (W. M.

Gallatin et al., 1983, Nature 304,30; T. K. Kishimoto et al., 1990, Proc. Natl. Acad. Sci.87,2244-) and P-selectin mediates the adherence ofplatelets to

neutrophils and monocytes (S-C. Hsu-Lin et al., 1984, J. BioL Chem.259,9121). Selectin antagonists (antibodies and carbohydrates) have been shown to blockthe extravasation ofneutrophils at sites ofinflammation (P. Piscueta andF. W. Luscinskas, 1994, Am. J. Pathol.145, 461-469), to be efficacious in animal models ofischemia/reperfusion (A.S. Weyrich et al., 1993, J. Clin. Invest.

91,2620-2629; R.K. Winn et al., 1993, J. Clin. Invest.92, 2042-2047), acute lung injury (M.S. Mulligan et al., 1993, J. Immunol.151, 6410-6417; A. Seekamp et al., 1994, Am. J. Pathol.144, 592-598), insulitis/diabetes (X.D. Yang et al., 1993, Proc. Natl. Acad. Sci.90,10494-10498), meningitis (C. Granet et al., 1994, J. Clin. Invest.93, 929-936), hemorrhagic shock (R.K. Winn et al., 1994, Am J. Physiol. Heart Circ. Physiol.267, H2391-H2397) and transplantation. In addition, selectin expression has been documented in models ofarthritis (F. Jamar et al., 1995, Radiology 194, 843-850), experimental allergic encephalomyelitis (J.M. Dopp et al., 1994, J. Neuroimmunol.54, 129-144), cutaneous inflammation (A. Siber et al., 1994, Lab. Invest.70, 163-170) glomerulonephritis (P.G. Tipping et al., 1994, Kidney Int.46, 79-88), on leukaemic cells and colon carcinomas (R.M. Lafrenie et al., 1994, Eur. J. Cancer [A] 30A, 2151-2158) andL-selectin receptors have been observed in myelinated regions ofthe central nervous system (K. Huang et al., 1991, J. Clin. Invest.88, 1778-1783). These animal model data strongly support the expectation ofatherapeutic role for selectin antagonists in a wide variety ofdisease states in whichhost tissue damage is neutrophil-mediated.

Otherexamples oflectins thatrecognize endogenous carbohydrates are

CD22β, CD23, CD44 and sperm lectins (A. Varki, 1993, Glycobiol.3, 97-130; P.M. Wassarman, 1988, Ann. Rev. Biochem.57, 415-442). CD22β is involved in early stages ofB lymphocyte activation; antagonists may modulate the immune response. CD23 is the low affinity IgE receptor; antagonists may modulate the IgE response in allergies and asthma. CD44 binds hyaluronic acid and thereby mediates cell/cell and cell/matrix adhesion; antagonists may modulate the inflammatory response. Sperm lectins are thought to be involved in sperm/egg adhesion and in the acrosomal response; antagonists may be effective contraceptives, eitherby blocking adhesion orby inducing apremature, spermicidal acrosomal response.

Antagonists to lectins thatrecognize exogenous carbohydrates may have wide application forthe prevention ofinfectious diseases. Many viruses (influenza A, B and C; Sendhi, Newcastle disease, coronavirus, rotavirus, encephalomyelitis virus, enchephalomyocarditis virus, reovirus, paramyxovirus) use lectins on the surface ofthe viral particle for attachment to cells, aprerequisite for infection;

antagonists to these lectins are expected to prevent infection (A.Varki, 1993,

Glycobiol.3, 97-130). Similarly colonization/infection strategies ofmany bacteria utilize cell surface lectins to adhere to mammalian cell surface glyco-conjugates. Antagonists to bacterial cell surface lectins are expected to have therapeutic potential for a wide spectrum ofbacterial infections, including: gastric (Helicobacterpylori), urinary tract (E. coli), pulmonary (Klebsiellapneumoniae, Stretococcus

pneumoniae, Mycoplasmapneumoniae) and oral (Actinomyces naeslundi and Actinomyces viscosus) colonization/infection (S.N. Abraham, 1994, Bacterial Adhesins, in The Handbook ofImmunopharmacology: Adhesion Molecules, CD. Wegner, ed; B.J. Mann et al., 1991, Proc. Natl. Acad. Sci.88, 3248-3252). A specific bacterial mediated disease state is Pseudomonas aeruginosa infection, the leading cause ofmorbidity and mortality in cystic fibrosis patients. The expectation that high affinity antagonists will have efficacy in treating P. aeruginosa infection is based on three observations. First, abacterial cell surface, GalNAcβ1-4Gal binding lectin mediates infection by adherence to asialogangliosides (αGM1 and αGM2) ofpulmonary epithelium (L. Imundo et al., 1995, Proc. Natl. Acad. Sci 92, 3019-3023). Second, in vitro, the binding ofP. aeruginosa is competed by the gangliosides' tetrasaccharide moiety, Galβ1-3GalNAcβ1-4Galβ1-4Glc. Third, in vivo, instillation ofantibodies to Pseudomonas surface antigens can prevent lung and pleural damage (J.F. Pittet et al., 1993, J. Clin. Invest.92, 1221-1228).

Non-bacterial microbes thatutilize lectins to initiate infection include

Entamoebahistalytica (aGal specific lectin thatmediates adhesion to intestinal mucosa; W.A. Petri, Jr., 1991, AMS News 57:299-306) and Plasmodium faciparum (a lectin specific for the terminal Neu5Ac(a2-3)Gal ofglycophorin A of erthrocytes; PA. Orlandi et al., 1992, J. Cell Biol.116:901-909). Antagonists to these lectins are potential therapeutics for dysentery and malaria.

Toxins are another class ofproteins that recognize exogenous carbohydrates (K-A Karlsson, 1989, Ann. Rev. Biochem.58:309-350). Toxins are complex, two domain molecules, composed ofa functional and a cell recognition/adhesion domain. The adhesion domain is often a lectin (i.e., bacterial toxins: pertussis toxin, cholera toxin, heat labile toxin, verotoxin and tetanus toxin; plant toxins: ricin and abrin). Lectin antagonists are expected to prevent these toxins frombinding their target cells and consequently to be useful as antitoxins.

There are still otherconditions for which the role oflectins is currently speculative. For example, genetic mutations result in reduced levels ofthe serum mannose-binding protein (MBP). Infants who have insufficient levels ofthis lectin suffer from severe infections, but adults do not. The high frequency ofmutations in both oriental and Caucasian populations suggests a condition may exist in which low levels ofserum mannose-binding protein are advantageous. Rheumatoid arthritis (RA) may be such a condition. The severity ofRA is correlated with an increase in IgG antibodies lacking terminal galactose residues on Fc region carbohydrates (A. Young et al., 1991, Arth. Rheum.34, 1425-1429; I.M. Roitt et al., 1988, J.

Autoimm.1, 499-506). Unlike theirnormal counterpart, these gal-deficient carbohydrates are substrates forMBP. MBP/IgG immunocomplexes may contribute to host tissue damage through complement activation. Similarly, the eosinophil basic protein is cytotoxic. Ifthe cytotoxicity is mediatedby the lectin activity ofthis protein, then a lectin antagonistmay have therapeutic applications in treating eosinophilmediated lung damage.

Lectin antagonists may also be useful as imaging agents or diagnostics. For example, E-selectin antagonists may be used to image inflamed endothelium.

Similarly antagonists to specific serum lectins, i.e. mannose-binding protein, may also beuseful in quantitatingproteinlevels.

Lectins are often complex, multi-domain, multimeric proteins. However, the carbohydrate-binding activity ofmammalian lectins is normally the property ofa carbohydrate recognition domain or CRD. The CRDs ofmammalian lectins fall into three phylogenetically conservedclasses: C-type, S-type and P-type (K. Drickamer and M.E. Taylor, 1993, Annu. Rev. Cell Biol.9, 237-264). C-type lectins require Ca⁺⁺ forligandbinding, are extracellular membrane and soluble proteins and, as a class, bind a variety ofcarbohydrates. S-type lectins are most active underreducing conditions, occurboth intra- and extracellularly, bind β-galactosides and do not require Ca⁺⁺. P-type lectins bind mannose 6-phosphate as their primary ligand.

Although lectin specificity is usually expressed in terms ofmonosaccharides and/or oligosacchrides (i.e., MBP binds mannose, fucose andN- acetylglucosamine), the affinity formonosaccharides is weak. The dissociation constants formonomeric saccharides are typically in the millimolarrange (Y.C. Lee, 1992, FASEB J.6:3193-3200; G.D. Glick et al., 1991, J Biol.Chem.266:23660- 23669; Y. Nagata and M.M. Burger, 1974, J. Biol. Chem.249:116-3122).

Co-crystals ofMBP complexed with mannose oligomers offer insight into the molecular limitations on affinity and specificity ofC-type lectins (W.I. Weis et al., 1992, Nature 360:127-134; K. Drickamer, 1993, Biochem. Soc. Trans.21:456- 459). The 3- and 4-hydroxyl groups ofmannose form coordination bonds with bound Ca⁺⁺ ion #2 and hydrogen bonds with glutamic acid (185 and 193) and asparagine (187 and 206). The limited contacts between the CRD and bound sugar are consistent with its spectrum ofmonosaccharide binding; N-acetylglucosamine has equatorial 3- and 4-hydroxyls while fucose has similarly configured hydroxyls at the 2 and 3 positions. The affinity ofthe mannose-binding protein and other lectins fortheirnatural ligands is greater than that for monosaccharides. Increased specificity and affinity can be accomplished by establishing additional contacts between aprotein and its ligand (K. Drickamer, 1993, supra) eitherby 1) additional contacts withthe terminal sugar (i.e., chickenhepatic lectin binds N-acetylglucose amine with greater affinity than mannose or fucose suggesting interaction with the 2-substituent); 2) clustering ofCRDs forbinding complex oligosaccharides (i.e., the mammalian asialylglycoprotein receptor); 3) interactions with additional saccharide residues (i.e., the lectin domain ofselectins appears to interactwithtwo residues ofthe tetrasaccharide sialyl-Lewis^x : with the chargedterminal residue, sialic acid, and withthe fucose residue; wheat germ agglutinin appears to interact with all three residues oftrimers ofN-acetylglucosamine); orby 4) contacts with anon- carbohydrate portion ofa glyco-protein.

The low affinity oflectins for mono- and oligo-saccharides presents major difficulties in developing high affinity antagonists that may be useful therapeutics. Approaches that have been used to develop antagonists are similarto those that occur in nature: 1) addition ormodification ofsubstituents to increase the number of interactions; and 2) multimerization ofsimple ligands.

The first approach has had limited success. For example, homologues of sialic acid have been analyzed for affinity to influenzavirus hemagglutinin (SJ. Watowich et al.1994, Structure 2:719-731). The dissociation constants ofthe best analogues are 30 to 300 μM which is only 10 to 100-foldbetter than the standard monosaccharide.

Modifications ofcarbohydrate ligands to the selectins have also hadlimited success. In static ELISA competition assays, sialyl-Lewis^a and sialyl-Lewis^x have IC50S of220 μM and750 μM, respectively, for the inhibition ofthe binding ofan

E-selectin/IgG chimerato immobilized sialyl-Lewis^x (R.M. Nelson et al., 1993, J.

Clin. Invest.91,1157-1166). The IC50 ofa sialyl-Lewis^a derivative (addition of an aliphatic aglycone to the GlcNAc and replacementofthe N-acetyl with anNH2 group) improved 10-fold to 21 μM. Similarly, removal ofthe N-acetyl from sialyl- lewis^x improves inhibition in an assay dependent manner (C. Foxall et al., 1992, J.

Cell Biol.117, 895-902; S.A. DeFrees et al., 1993, J. Am. Chem. Soc.115, 7549-

7550).

The second approach, multimerization ofsimple ligands, can lead to dramatic improvements in affinity for lectins that bind complex carbohydrates (Y.C. Lee, supra). On the other hand, the approach does not show great enhancement for lectins that bind simple oligosaccharides; dimerizing sialyl-Lewis^x, a minimal carbohydrate ligand for E-selectin, improves inhibition approximately 5-fold (S.A. DeFrees et al., supra).

An alternative approach is to design compounds that are chemically unrelated to the natural ligand. In the static ELISA competition assays inositol polyanions inhibit L- andP-selectin binding with IC50S that range from 1.4 μM to 2.8 mM (O. Cecconi et al., 1994, J. Biol. Chem.269, 15060-15066). Synthetic ohgopeptides, based on selectin amino acid sequences, inhibit neutrophil binding to immobilized P- selectin with IC50S ranging from 50 μM to 1 mM (J-G Geng et al., 1992, J ofBiol.

Chem.267, 19846-19853).

Lectins are nearly ideal targets for isolation ofantagonists by SELEX technology described below. The reason is that oligonucleotide ligands that are boundto thecarbohydratebinding sitecanbe specifically elutedwiththerelevant sugar(s). Oligonucleotide ligands with affinities that are several orders ofmagnitude greaterthan that ofthe competing sugarcan be obtainedby the appropriate manipulation ofthenucleic acidligand to competitorratio. Since the carbohydrate binding site is the active site ofa lectin, essentially all ligands isolatedby this procedure will be antagonists. In addition, these SELEX ligands will exhibit much greater specificity than monomeric and oligomeric saccharides.

A method for the in vitro evolution ofnucleic acid molecules with highly specific binding to target molecules has been developed. This method, Systematic Evolution ofLigands by Exponential enrichment, termed SELEX, is describedin United States Patent Application Serial No.07/536,428, entitled "Systematic Evolution ofLigands by Exponential Enrichment," now abandoned, United States Patent Application Serial No.07/714,131, filed June 10, 1991, entitled "Nucleic Acid Ligands," now United States Patent Number 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 ofwhich is herein specifically incorporated by reference. Each ofthese applications, collectively referred to herein as the SELEX Patent Applications, describes afundamentally novel methodformaking anucleic acidligandto any desiredtargetmolecule.

The SELEX method involves selection from a mixture ofcandidate oligonucleotides and step-wise iterations ofbinding, partitioning and amplification, using the same general selection scheme, to achieve virtually any desiredcriterion of binding affinity and selectivity. Starting from a mixture ofnucleic acids, preferably comprising a segment ofrandomized sequence, the SELEX method includes steps ofcontacting the mixture with the targetunder conditions favorable forbinding, partitioning unboundnucleic acids from those nucleic acids which have bound specifically to targetmolecules, dissociating the nucleic acid-targetcomplexes, amplifying the nucleic acids dissociatedfrom the nucleic acid-target complexes to yield aligand-enrichedmixture ofnucleic acids, then reiterating the steps ofbinding, partitioning, dissociating and amplifying through as many cycles as desired to yield highly specific, high affinity nucleic acid ligands to the targetmolecule.

The basic SELEX method has been modified to achieve anumber ofspecific objectives. Forexample, 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 ofSELEX in conjunction with gel electrophoresis to select nucleic acid molecules with specific structural characteristics, such as bent DNA. United States Patent Application SerialNo.08/123,935, filed September 17, 1993, entitled "Photoselection ofNucleic AcidLigands" describes a SELEXbased method for selecting nucleic acidligands containingphotoreactive groups capable of binding and/or photocrosslinking to and/orphotoinactivating atargetmolecule. United States Patent Application Serial No.08/134,028, filed October 7, 1993, entitled "High-Affinity Nucleic AcidLigandsThatDiscrirninateBetween

Theophylline and Caffeine," describes amethod foridentifying highly specific nucleic acid ligands able to discriminate between closely relatedmolecules, termed Counter-SELEX. United States Patent Application SerialNo.08/143,564, filed October 25, 1993, entitled "Systematic Evolution ofLigands by Exponential

Enrichment: Solution SELEX," describes a SELEX-based method which achieves highly efficientpartitioning between oligonucleotides having high and low affinity for atargetmolecule. United States Patent Application SerialNo.07/964,624, filed October 21, 1992, entitled "Methods ofProducing Nucleic AcidLigands" describes methods for obtaining improvednucleic acidligands afterSELEX has been performed. United States Patent Application Serial No.08/400,440, filed March 8, 1995, entitled "Systematic Evolution ofLigands by Exponential Enrichment:

Chemi-SELEX," describes methods for covalently linking aligand to its target.

The SELEX methodencompasses the identification ofhigh-affinity nucleic acid ligands containing modified nucleotides conferring improved characteristics on the ligand, such as improved in vivo stability orimproved delivery characteristics. Examples ofsuch modifications include chemical substitutions at the ribose and/or phosphate and/orbase positions. SELEX-identifiednucleic acid ligands containing modified nucleotides are described in United States Patent Application Serial No. 08/117,991, filed 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 ormore nucleotides modified with 2'- amino (2'-NH2), 2'-fluoro (2'-F), and/or 2'-O-methyl (2'-OMe). United States

Patent Application Serial No.08/264,029, filed June 22, 1994, entitled "Novel Method ofPreparation of2' ModifiedPyrimidine IntramolecularNucleophilic Displacement," describes novel methods formaking 2'-modified nucleosides.

The SELEXmethodencompasses combining selected oligonucleotides with otherselected oligonucleotides as describedinUnited States Patent Application Serial No.08/284,063, filed August 2, 1994, entitled "Systematic Evolution of Ligands by ExponentialEnrichment: Chimeric SELEX". The SELEX method also includes combining the selectednucleic acidligands withnon-oligonucleotide functional units and United States Patent Application Serial No.08/234,997, filed April 28, 1994, entitled "Systematic Evolution ofLigands by Exponential

Enrichment: Blended SELEX" andUnited States Patent Application Serial No. 08/434,465, filed May 4, 1995, entitled "Nucleic Acid Ligand Complexes". These applications allow the combination ofthe broad array ofshapes and otherproperties, and the efficient amplification and replication properties, ofoligonucleotides with the desirable properties ofothermolecules. Each ofthe above described patent applications which describe modifications ofthe basic SELEX procedure are specifically incorporated by reference herein in their entirety.

The presentinvention applies the SELEX methodology to obtain nucleic acid ligands to lectin targets. Lectin targets, or lectins, include all the non-enzymatic carbohydrate-binding proteins ofnon-immune origin, which include, but are not limited to, those described above.

Specifically, high affinity nucleic acid ligands towheat germ agglutinin, and various selectin proteins have been isolated. For the purposes ofthe invention the terms wheat germ agglutinin, wheat germ lectin andWGA are usedinterchangeably. Wheat germ agglutinin (WGA) is widely used for isolation, purification and structural studies ofglyco-conjugates. As outlined above, the selectins are important anti-inflammatory targets. Antagonists to the selectins modulate extravasion of leukocytes at sites ofinflammation andthereby reduce neutrophil causedhost tissue damage. Using the SELEX technology, high affinity antagonists ofL-selectin, E- selectin and P-selectin mediated adhesion are isolated. BRIEF SUMMARY OFTHE INVENTION

The present invention includes methods ofidentifying and producing nucleic acidligands to lectins andthe nucleic acid ligands so identified and produced. More particularly, nucleic acid ligands are provided that are capable ofbinding specifically toWheatGerm Agglutinin (WGA), L-Selectin, E-selectin andP-selectin.

Further included in this invention is amethod ofidentifying nucleic acid ligands and nucleic acid ligand sequences to lectins comprising the steps of(a) preparing acandidate mixture ofnucleic acids, (b) partitioningbetween members of said candidate mixture on thebasis ofaffinity to said lectin, and (c) amplifying the selectedmoleculesto yield amixture ofnucleic acids enrichedfornucleic acid sequences with arelatively higher affinity forbinding to saidlectin.

More specifically, thepresentinvention includes thenucleic acidligandsto lectins identified according to the above-described method, including those ligands to Wheat Germ Agglutinin listed in Table 2, those ligands to L-selectin listed in Tables 8,12 and 16, and those ligands to P-selectin listed inTables 19 and 25. Additionally, nucleic acid ligands to E-selectin and serummannose binding protein areprovided. Also included arenucleic acidligands to lectins that are substantially homologous to any ofthe given ligands and that have substantially the same ability to bind lectins and antagonize the ability ofthe lectin tobind carbohydrates. Further includedin this invention are nucleic acid ligands to lectins thathave substantially the same structural form as the ligands presentedherein and thathave substantially the same ability tobind lectins and antagonize the ability ofthe lectin tobind

carbohydrates.

Thepresentinvention also includes modifiednucleotide sequences based on the nucleic acid ligands identified herein and mixtures ofthe same.

The present invention also includes the use ofthe nucleic acidligandsin therapeutic, prophylactic and diagnostic applications.

BRIEF DESCRIPTION OF THE FIGURES

Figure 1 shows consensus hairpin secondary structures for WGA 2'-NH₂ RNA ligands: (a) family 1, (b) family 2 and (c) family 3. Nucleotide sequence is in standard one letter code. Invariantnucleotides are in bold type. Nucleotides derived from fixed sequence are in lowercase.

Figure 2 shows binding curves for the L-selectin SELEX second and ninth round 2'-NH₂ RNA pools to peripheral blood lymphocytes (PBMCs). Figure 3 shows binding curves for random 40N72-NH₂ RNA (SEQ ED

NO: 64) and the cloned L-selectin ligand, F14.12 (SEQ ID NO: 78), to peripheral blood lymphocytes (PBMC).

Figure 4 shows the results ofa competition experiment in which the binding of5 nM ³²P-labeled F14.12 (SEQ ID NO: 78) to PBMCs (10⁷/ml) is competed with increasing concentrations ofunlabeled F14.12 (SEQ ID NO: 78). RNA Bound equals 100 x (netcounts bound in the presence ofcompetitor/net counts boundin the absence ofcompetitor).

Figure 5 shows the results ofa competition experimentin which the binding of5 nM ³²P-labeled F14.12 (SEQ ID NO: 78) to PBMCs (10⁷/ml) is competed withincreasing concentrations oftheblockingmonoclonal anti-L-selectin antibody, DREG-56, or an isotype matched, negative control antibody. RNABound equals 100 x (net counts bound in the presence ofcompetitor/net counts bound in the absence ofcompetitor).

Figure 6 shows the results ofacompetitive ELISA assay in which the binding ofsolubleLS-Rg to immobilized sialyl-Lewis^x/BSAconjugates is competed with increasing concentrations ofunlabeledF14.12 (SEQID NO: 78). Binding of LS-Rg was monitored with an HRP conjugated anti-human IgG antibody. LS-Rg Bound equals 100x (OD₄₅₀ in the presence ofcompetitor)/(OD₄₅₀ in the absence ofcompetitor). The observed OD450 was corrected fornonspecific binding by subtracting the OD₄₅₀ in the absence ofLS-Rg from the experimental values. In the absence ofcompetitorthe OD₄₅₀ was 0.324 and in the absence ofLS-Rg 0.052.

Binding ofLS-Rg requires divalent cations; in the absence ofcompetitor, replacement of Ca++/Mg⁺⁺ with4 mM EDTA reduced the OD₄₅₀ to 0.045.

Figure 7 shows hairpin secondary structures for representative L-selectin 2'NH₂ RNA ligands: (a) F13.32 (SEQ. ID NO: 67), family I; (b) 6.16 (SEQ. ID

NO: 84), family IIl; and (c) F14.12 (SEQ. ID NO: 78), family H Nucleotide sequence is in standard one letter code. Invariant nucleotides are in boldtype.

Nucleotides derived from fixed sequence are in lower case.

Figure 8 shows a schematic representation ofeach dimeric and mutimeric oligonucleotide complex: (a) dimeric branched oligonucleotide; (b) multivalent streptavidin/bio-oUgonucleotide complex (A: streptavidin; B: biotin); (c) dimeric dumbell oligonucleotide; (d) dimeric fork oligonucleotide.

Figure 9 shows binding curves for the L-selectin SELEX fifteenth round ssDNA pool to PBMCs (10⁷/ml).

Figure 10 shows the results ofacompetition experimentin whichthe binding of2 nM ³²P-labeled round 15 ssDNA to PBMCs (10⁷/ml) is competed with increasing concentrations ofthe blocking monoclonal anti-L-selectin antibody, DREG-56, or an isotype matched, negative control antibody. RNA Bound equals 100 x (net counts bound in the presence ofcompetitor/net counts bound in the absence ofcompetitor).

Figure 11 shows L-selectin specific binding ofLD201T1 (SEQ ID NO: 185) to human lymphocytes and granulocytes in whole blood, a, FITC-LD201T1 binding to lymphocytes is competedby DREG-56, unlabeled LD201T1, and inhibitedby EDTA. b, F1TC-LD201T1 binding to granulocytes is competed by DREG-56, unlabeled LD201T1, and inhibited by EDTA. All samples were stained with 0.15 mMFITC-LD201T1; thickline: FTTC-LD201T1 only; thickdashedline: FTTC- LD201T1 with 0.3 mM DREG-56; medium thick line: FLTC-LD201T1 with 7 mM unlabeled NX280; thin line: FTTC-LD201T1 stained cells, reassayed after addition of4 mMEDTA; thin dashed line: autofluorescence.

Figure 12 shows the consensus hairpin secondary structures for family 1 ssDNA ligands to L-selectin. Nucleotide sequence is in standard one letter code. Invariant nucleotides are in bold type. The base pairs at highly variable positions are designatedN-N'. To the right ofthe stem is a matrix showing the number of occurances ofparticularbase pairs for the position in the stem that is on the same line.

Figure 13 shows that in vitro pre-treatment ofhuman PBMC withNX288 (SEQ ID NO: 193) inhibits lymphocyte trafficking to SOD mouse PLN. Human PBMC were purified from heparinised blood by a Ficoll-Hypaque gradient, washed twice with HBSS (calcium/magnesium free) and labeled with ⁵¹Cr (Amersham). After labeling, the cells were washed twice with HBSS (containing calcium and magnesium) and 1% bovine serum albumin (Sigma). Female SCID mice (6-12 weeks ofage) were injected intravenously with 2x10'-' cells. The cells were either untreated ormixed with either 13 pmol ofantibody (DREG-56 orMEL-14), or4, 1, or 0.4 nmol ofmodified oligonucleotide. One hour later the animals were anaesthetised, a blood sample taken and the mice were euthanised. PLN, MLN, Peyer's patches, spleen, liver, lungs, thymus, kidneys and bone marrow were removed and the counts incorporated into the organs determined by a Packard gamma counter. Values shown represent the mean ± s.e. oftriplicate samples, and are representative of3 experiments.

Figure 14 shows that pre-injection ofNX288 (SEQ ID NO: 193) inhibits human lymphocyte trafficking to SCID mouse PLN and MLN. Human PBMC were purified, labeled, and washed as described above. Cells were prepared as described in Figure 13. Female SCID mice (6-12 weeks of age) were injected intravenously with 2x10⁶ cells. One to 5 min prior to injecting the cells, the animals were injected with either 15 pmol DREG-56 or4 nmol modified oligonucleotide. Animals were scarificed 1 hour after injection ofcells. Counts incorporated into organs were quantified as described in Figure 13. Values shown represent the mean ± s.e. of triplicate samples, and are representative of2 experiments.

Figure 15 shows the consensus hairpin secondary structures for 2'-F RNA ligands to L-selectin. Nucleotide sequence is in standard one letter code. Invariant nucleotides are in bold type. The base pairs at highly variablepositions are designated N-N'. To the right ofthe stem is a matrix showing the number of occurances ofparticularbasepairs forthe position in the stemthat is on the same line.

Figure 16 shows the consensus hairpin secondary structures for 2'-F RNA ligands toP-selectin. Nucleotide sequence is in standard one lettercode. Invariant nucleotides are in bold type. Thebasepairs athighly variablepositions are designatedN-N'. To the right ofthe stem is amatrix showing the number of occurances ofparticularbase pairs forthe position in the stemthatis on the same line.

DETAILEDDESCRIPTION OFTHEINVENTION

This.application describes high-affinity nucleic acid ligands tolectins identified through the method known as SELEX. SELEX is described in U.S.

PatentApplication Serial No.07/536,428, entitled "Systematic Evolution ofLigands by ExponentialEnrichment", now abandoned; U.S. Patent Application Serial No. 07/714,131, filedJune 10, 1991, entitled "Nucleic Acid Ligands", now United States PatentNo.5,475,096; United States Patent Application Serial No.

07/931,473, filed August 17, 1992, entitled "Nucleic AcidLigands", now United States Patent No.5,270,163, (see also PCT/US91/04078). These applications, each specifically incorporatedherein byreference, are collectively calledthe SELEX Patent Applications.

In its mostbasic form, the SELEX process may be definedby the following series ofsteps:

1) A candidate mixture ofnucleic acids ofdiffering sequence is prepared. The candidate mixture generally includes regions offixed sequences (i.e., each of the members ofthe candidate mixture contains the same sequences in the same location) and regions ofrandomized sequences. The fixed sequence regions are selected either: (a) to assist in the amplification steps describedbelow, (b) to mimic a sequence known to bind to the target, or (c) to enhance the concentration ofa given structural arrangementofthe nucleic acids in the candidate mixture. The randomized sequences can be totally randomized (i.e., the probability offinding abase at any position being one in four) or only partially randomized (e.g., the probability of finding abase at any location can be selected at any level between 0 and 100 percent).

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

3) The nucleic acids with the highest affinity forthe target arepartitioned fromthose nucleic acids with lesser affinity to the target. Because only an extremely small number ofsequences (andpossibly only one molecule ofnucleic acid) corresponding to the highest affinity nucleic acids existin the candidate mixture, it is generally desirable to setthe partitioning criteria so that asignificantamountofthe nucleic acids in the candidate mixture (approximately .05-50%) are retained during partitioning.

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

5) By repeating the partitioning and amplifying steps above, the newly formed candidate mixture contains fewer and fewer unique sequences, and the average degree ofaffinity ofthe 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 ofunique nucleic acids representing those nucleic acids from the original candidate mixture having the highest affinity to the targetmolecule.

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 partitionednucleic acids to generate enriched candidate mixture. The SELEX Patent Applications also describe ligands obtained to anumber oftarget species, including bothprotein targets where the protein is and is not a nucleic acid binding protein.

This invention also includes the ligands as described above, wherein certain chemical modifications are made in order to increase the in vivo stability ofthe ligand or to enhance or mediate the delivery ofthe ligand. Examples ofsuch modifications include chemical substitutions at the sugar and/ orphosphate and/or base positions ofa given nucleic acid sequence. See, e.g., U.S. Patent Application Serial No.08/117,991, filed September 9, 1993, entitled "High Affinity Nucleic Acid Ligands Containing ModifiedNucleotides" which is specifically incorporated herein by reference. Additionally, in co-pending and commonly assignedU.S. Patent Application Serial No.07/964,624, filed October 21, 1992 ('624), 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 ofProducing Nucleic AcidLigands," is specifically incorporated herein by reference. Further included in the '624 patent are methods for determining the three- dimensional structures ofnucleic acid ligands. Such methods include mathematical modeling and structure modifications ofthe SELEX-derived ligands, such as chemical modification and nucleotide substitution. Othermodifications are known to one ofordinary skill in the art. Suchmodifications may be made post-SELEX (modification ofpreviously identifiedunmodified ligands) orby incorporation into the SELEX process. Additionally, the nucleic acid ligands ofthe invention can be complexed with various other compounds, including butnot limited to, lipophilic compounds ornon-immunogenic, highmolecularweightcompounds. Lipophilic compounds include, but are not limited to, cholesterol, dialkyl glycerol, and diacyl glycerol. Non-immunogenic, high molecularweight compounds include, but are notUmitedto, polyethylene glycol, dextran, albumin andmagnetite. The nucleic acid ligands described herein can be complexed with a lipophilic compound (e.g., cholesterol) or attached to orencapsulated in a complex comprised oflipophilic components (e.g., aliposome). The complexed nucleic acid ligands can enhance the cellularuptake ofthe nucleic acid ligands by a cell for delivery ofthe nucleic acid ligands to an intracellulartarget. The complexed nucleic acid ligands can also have enhancedpharmacokinetics and stability. United States PatentApplication Serial Number 08/434,465, filed May 4, 1995, entitled "Nucleic Acid Ligand Complexes," which is herein incorporated by reference describes a method for preparing a therapeutic or diagnostic complex comprised ofa nucleic acid ligand and alipophilic compound or a non-immunogenic, high molecular weight compound.

The methods described herein and the nucleic acid ligands identifiedby such methods are useful forboth therapeutic and diagnostic purposes. Therapeutic uses include the treatment orprevention ofdiseases or medical conditions in human patients. Many ofthe therapeutic uses are described in the background ofthe invention, particularly, nucleic acid ligands to selectins are useful as anti- inflammatory agents. Antagonists to the selectins modulate extravasion of leukocytes at sites ofinflammation and thereby reduce neutrophil caused host tissue damage. Diagnostic utilization may include both in vivo or in vitro diagnostic applications. The SELEX method generally, and the specific adaptations ofthe 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, ofcourse, the desired properties one skilled in the art would seek in a diagnostic ligand.

The nucleic acid ligands ofthe presentinventionmayberoutinely adapted for diagnostic purposes according to any number oftechniques employed by those skilled in the art. Diagnostic agents need only be able to allow the user to identify the presence ofagiven targetat aparticularlocale orconcentration. Simply the ability to formbinding pairs with the target may be sufficientto trigger apositive signal for diagnostic purposes. Those skilled in the art would also be able to adapt any nucleic acid ligandby procedures knowninthe art to incorporate alabeling tag in order to track the presence ofsuch ligand. Such atag couldbe used in anumber ofdiagnostic procedures. The nucleic acid ligands to lectin, particularly selectins, describedherein may specifically be used foridentification ofthe lectinproteins.

SELEX provides high affinity ligands ofa target molecule. This represents a singular achievement that is unprecedented in the field ofnucleic acids research. The presentinvention applies the SELEXprocedure to lectin targets. Specifically, the presentinvention describes the identification ofnucleic acid ligands to Wheat Germ Agglutinin, and the selectins, specifically, L-selectin, P-selectin and E-selectin. In the Example section below, the experimental parameters used to isolate andidentify the nucleic acid ligands to lectins are described.

In orderto produce nucleic acids desirable for use as apharmaceutical, it is preferredthatthe nucleic acidligand (1) binds to the targetin amanner capable of achieving the desired effect on the target; (2) be as small as possible to obtain the desired effect; (3) be as stable as possible; and (4) be a specific ligand to the chosen target. In most situations, it is preferred that the nucleic acid ligandhave the highest possible affinity to the target.

In the present invention, a SELEX experiment was performed in search of nucleic acid ligands with specific high affinity forWheat Germ Agglutinin from a degenerate library containing 50 random positions (50N). This invention includes the specific nucleic acid ligands to Wheat Germ Agglutinin shown in Table 2 (SEQ ID NOS: 4-55), identified by the methods described in Examples 1 and 2.

Specifically, RNA ligands containing 2'-NH₂ modified pyrimidines are provided. The scope ofthe ligands covered by this invention extends to all nucleic acid ligands ofWheat Germ Agglutinin, modified and unmodified, identified according to the SELEX procedure. More specifically, this invention includes nucleic acid sequences that are substantially homologous to the ligands shown in Table 2. By substantially homologous it is meant a degree ofprimary sequence homology in excess of70%, mostpreferably in excess of80%. A review ofthe sequence homologies ofthe ligands ofWheat Germ Agglutinin shown in Table 2 shows that sequences with little orno primary homology may have substantially the same ability to bind Wheat Germ Agglutinin. Forthese reasons, this invention also includes nucleic acid ligands that have substantially the same ability to bind Wheat Germ Agglutinin as the nucleic acid ligands shown in Table 2. Substantially the same ability to bindWheat GermAgglutinin means that the affinity is within a few orders ofmagnitude ofthe affinity ofthe ligands described herein. It is well within the skill ofthose of ordinary skill in the art to determine whether agiven sequence --substantially homologous to those specifically described herein --has substantially the same ability to bindWheat Germ Agglutinin.

In the present invention, SELEX experiments were performed in search of nucleic acid ligands with specific high affinity for L-selectin from degenerate libraries containing 30 or40 random positions (30N or 40N). This invention includes the.specific nucleic acid ligands to L-selectin shown in Tables 8, 12 and 16 (SEQ ID NOS: 67-117, 129-180, 185-196 and 293-388), identified by the methods described in Examples 7, 8, 13, 14, 22 and 23. Specifically, RNA ligands containing 2'-NH2 or 2'-F pyrimidines and ssDNA ligands are provided. The scope ofthe ligands covered by this invention extends to all nucleic acid ligands of L-selectin, modified and unmodified, identified according to the SELEX procedure. More specifically, this invention includes nucleic acid sequences that are

substantially homologous to the ligands shown in Tables 8, 12 and 16. By substantially homologous it is meant a degree ofprimary sequence homology in excess of70%, most preferably in excess of80%. A review ofthe sequence homologies ofthe ligands ofL-selectin shown in Tables 8, 12 and 16 shows that sequences with little or no primary homology may have substantially the same ability to bind L-selectin. For these reasons, this invention also includes nucleic acid ligands that have substantially the same ability to bind L-selectin as the nucleic acid ligands shown in Tables 8, 12 and 16. Substantially the same ability to bind L- selectin means thatthe affinity is within a few orders ofmagnitude ofthe affinity of the ligands described herein. It is well within the skill ofthose ofordinary skill in the art to determine whether a given sequence -- substantially homologous to those specifically described herein -- has substantially the same ability to bind L-selectin.

In the present invention, SELEX experiments were performed in search of nucleic acid ligands with specific high affinity for P-selectin from degenerate libraries containing 50 random positions (50N). This invention includes the specific nucleic acid ligands to P-selectin shown in Tables 19 and 25 (SEQ ID NOS: 199- 247 and 251-290), identifiedby the methods described in Examples 27, 28, 35 and 36. Specifically, RNA ligands containing 2'-NH₂ and 2'-F pyrimidines are provided. The scope ofthe ligands covered by this invention extends to all nucleic acidligands ofP-selectin, modified andunmodified, identified according to the SELEX procedure. More specifically, this invention includes nucleic acid sequences that are substantially homologous to the ligands shown in Tables 19 and 25. By substantially homologous it is meant a degree ofprimary sequence homology in excess of70%, most preferably in excess of80%. A review ofthe sequence homologies ofthe ligands ofP-selectin shown in Tables 19 and 25 shows that sequences with little orno primary homology may have substantially the same ability to bindP-selectin. Forthese reasons, this invention also includes nucleic acid ligands that have substantially the same ability to bind P-selectin as the nucleic acid ligands shown in Tables 19 and 25. Substantially the same ability to bind P-selectin means that the affinity is within a few orders ofmagnitude ofthe affinity ofthe ligands described herein. It is well within the skill ofthose ofordinary skill in the art to determine whether a given sequence--substantially homologous to those specifically described herein-- has substantially the same ability to bind P-selectin.

In the present invention, a SELEX experimentwas performed in search of nucleic acid ligands with specific high affinity for E-selectin from a degenerate library containing 40 random positions (40N). This invention includes specific nucleic acid ligands to E-selectin identified by the methods described in Example 40. The scope ofthe ligands coveredby this invention extends to all nucleic acidligands ofE-selectin, modified and unmodified, identified according to the SELEX procedure.

Additionally, the presentinvention includes multivalent Complexes comprising the nucleic acid ligands ofthe invention. The mulivalent Complexes increase the binding energy to facilitate betterbinding affinities through slower off- rates ofthe nucleic acid ligands. The multivalent Complexes may be useful at lower doses than their monomeric counterparts. In addition, high molecularweight polyethylene glycol was included in some ofthe Complexes to decrease the in vivo clearance rate ofthe Complexes. Specifically, nucleic acid ligands to L-selectin were placed in multivalent Complexes.

As described above, because oftheirability to selectively bind lectins, the nucleic acid ligands to lectins described herein are useful as pharmaceuticals. This invention, therefore, also includes amethod for treating lectin-mediated diseases by administration ofa nucleic acid ligand capable ofbinding to alectin.

Therapeutic compositions ofthe nucleic acid ligands may be administered parenterally by injection, although othereffective administration forms, such as intraarticularinjection, inhalant mists, orally active formulations, transdermal iontophoresis or suppositories, are also envisioned. One preferred carrieris physiological saline solution, but it is contemplated that otherpharmaceutically acceptable carriers may alsobe used. In onepreferred embodiment, it is envisioned that the carrier and the ligand constitute aphysiologically-compatible, slow release formulation. The primary solvent in such acarrier may be either aqueous ornon- aqueous in nature. In addition, the carriermay contain otherpharmacologically- acceptable excipients for modifying ormaintaining the pH, osmolarity, viscosity, clarity, color, sterility, stability, rate ofdissolution, or odor ofthe formulation. Similarly, the carriermay contain still otherpharmacologically-acceptable excipients formodifying ormaintaining the stability, rate ofdissolution, release, or absorption ofthe ligand. Such excipients are those substances usually and customarily employedto formulate dosages for parental administration in eitherunitdose 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 aready to use form orrequiring reconstitution immediatelypriorto administration. The mannerof administering formulations containing nucleic acid ligands for systemic delivery may be via subcutaneous, intramuscular, intravenous, intranasal orvaginal orrectal suppository.

Well established animal models exist for many ofthe disease states which are candidates for selectin antagonist therapy. Models available for testing the efficacy ofoligonucleotide selectin antagonists include:

1) mouse models forperitoneal inflammation (P. Pizcueta andF.W.

Luscinskas, 1994, Am. J. Pathol.145, 461-469), diabetes (A.C. Hanninen et al., 1992, J. Clin. Invest.92, 2509-2515), lymphocyte trafficking (L.M. Bradley et al., 1994, J. Exp. Med., 2401-2406), glomerulonephritis (P.G. Tipping et al., 1994, Kidney Int.46, 79-88), experimental allergic encephalomyelitis ( J.M. Dopp et al., 1994, J. Neuroimmunol.54: 129-144), acute inflammation in human/SCID mouse chimera (H.-C. Yan et al., 1994, J. Immunol.152, 3053-3063), endotoxin- mediated inflammation (W.E. Sanders et al., 1992, Blood 80, 795-800);

2) rat models for acute lung injury (M.S. Milligan et al., 1994, J. Immunol. 152, 832-840), hind limb ischemia/reperfusion injury (A. Seekamp et al., 1994, Am. J. Pathol 144, 592-598), remote lung injury (A. Seekamp et al., 1994, supra; D.L. Carden et al., 1993, J. Appl. Physiol 75, 2529-2543), neutrophil rolling on mesenteric venules (K. Ley et al., 1993, Blood 82, 1632-1638), myocardial infarction ischemiareperfusion injury (D. Altavilla et al., 1994, Eur. J. Pharmacol. Environ. Toxicol. Pharmacol.270, 45-51);

3) rabbit models for hemorrhagic shock (R.K. Winn et al., 1994, Am. J.

Physiol. Heart Circ. Physiol.267, H2391-H2397), ear ischemia reperfusion injury (D. Mihelcic et al., 1994, Bollod 84, 2333-2328) neutrophil rolling on mesenteric venules (A.M. Olofsson et al., Blood 84, 2749-2758), experimental meningitis (C. Granert et al., 1994, J. Clin. Invest.93, 929-936); lung, peritoneal and

subcutaneous bacterial infection (S.R. Sharer et al., 1993, J. Immunol.151, 4982- 4988), myocardial ischemia/repefusion (G. Montrucchio et al., 1989, Am. J.

Physiol.256, H1236-H1246), central nervous system ischemic injury (W.M. Clark et al., 1991, Stroke 22, 877-883);

4) cat models for myocardial infraction ischemia reperfusion injury

(M.Buerke et al., 1994, J. Pharmacol. Exp. Ther.271, 134-142);

5) dog models for myocardial infarction ischemia reperfusion injury(D.J. Lefer et al., 1994, Circulation 90, 2390-2401);

6) pig models for arthritis (F. Jamar et al., 1995, Radiology 194, 843-850);

7) rhesus monkey models for cutaneous inflammation (A. Silber et al., Lab. Invest.70, 163-175);

8) cynomolgus monkey models for renal transplants (S.-L. Wee, 1991, Transplant. Prod.23, 279-280); and

9) baboon models for dacron grafts (T. Palabrica et al, 1992, Nature 359, 848-851), septic, traumatic and hypovolemic shock (H. Redl et al., 1991, Am. J. Pathol.139, 461-466).

The nucleic acid ligands to lectins described herein are useful as

pharmaceuticals and as diagnostic reagents.

Examples

The following examples are illustrative ofcertain embodiments ofthe invention and are not to be construed as limiting the present invention in any way. Examples 1-6 describe identification and characterization of2'-NH₂ RNA ligands to Wheat Germ Agglutinin. Examples 7-12 described identification and

characterization of2-NH₂ RNA ligands to L-selectin. Examples 13-21 describe identification and characterization ofssDNA ligands to L-selectin. Examples 22-25 describe identification and characterization of2'-F RNA ligands to L-selectin.

Example 26 describes identification ofssDNA ligands to P-selectin. Examples 27- 39 describes identification and characterization of2-NH₂ and 2'-F RNA ligands to

P-selectin. Example 40 describes identification ofnucleic acid ligands to E-selectin.

Example 1

Nucleic Acid Ligands to Wheat Germ Agglutinin The experimental procedures outlined in this Example were used to identify and characterize nucleic acid ligands to wheatgerm agglutinin (WGA) as described in Examples 2-6.

Experimental Procedures

A) Materials

WheatGermLectin (Triticum vulgare) Sepharose 6MB beads were purchased from PharmaciaBiotech. Wheat GermLectin, Wheat Germ Agglutinin, andWGA are usedinterchangeably herein. FreeWheat GermLectin (Triticum vulgare) and all otherlectins were obtained from E YLaboratories; methyl-oc-D- mannopyranoside was from Calbiochem andN-acetyl-D-glucosamine, GlcNAc, and the trisaccharide N N N'-triacetylchitotriose, (GlcNAc)3, were purchased from

Sigma Chemical Co. The 2'-NH₂ modified CTP and UTP were prepared according to Pieken et. al. (1991, Science 253:314-317). DNA oligonucleotides were synthesizedby Operon. All otherreagents and chemicals were purchased from commercial sources. Unless otherwise indicated, experiments utilized Hanks' Balanced Salt Solutions (HBSS; 1.3 mM CaCl₂, 5.0 mM KCl, 0.3 mM KH₂PO₄, 0.5 mM MgCl2-6H₂O, 0.4 mM MgSO₄.7H₂O, 138 mM NaCl, 4.0 mM NaHCO₃, 0.3 mM Na₂HPO₄, 5.6 mM D-Glucose; GibcoBRL).

B) SELEX

The SELEX procedure is described in detail in United States Patent

5,270,163 and elsewhere. In the wheat germ agglutinin SELEX experiment, the DNA template forthe initial RNA pool contained 50 random nucleotides, flanked by N95' and 3' fixed regions (50N9) 5' gggaaaagcgaaucauacacaaga-50N- gcuccgccagagaccaaccgagaa 3' (SEQ ID NO: 1). All C and U have 2-NH2 substituted for 2'-OH for ribose. The primers for the PCR were the following: 5' Primer 5' taatacgactcactatagggaaaagcgaatcatacacaaga 3' (SEQ ID NO: 2) and 3' Primer 5' ttctcggttggtctctggcggagc 3' (SEQ ED NO: 3). The fixed regions ofthe starting randompool include DNA primer annealing sites for PCR and cDNA synthesis as well as the consensus T7 promoter region to allow in vitro

transcription. These single-stranded DNA molecules were converted into double- stranded transcribable templates by PCR amplification. PCR conditions were 50 mM KCl, 10 mM Tris-Cl, pH 8.3, 0.1% Triton X-100, 7.5 mM MgCl₂, 1 mM ofeach dATP, dCTP, dGTP, and dTTP, and 25 U/ml ofTaq DNA polymerase.

Transcription reactions contained 5 mM DNA template, 5 units/μl T7 RNA polymerase, 40 mM Tris-Cl (pH 8.0), 12 mM MgCl₂, 5 mM DTT, 1 mM spermidine, 0.002% Triton X-100, 4 % PEG 8000, 2 mM each of 2'-OH ATP, 2'- OH GTP, 2'-NH2 CTP, 2'-NH₂ UTP, and 0.31 mM α-³²P 2'-OH ATP.

The strategy forpartitioningWGA/RNA complexes fromunbound RNA was 1) to incubate the RNApool with WGA immobilized on sepharose beads; 2) to remove unboundRNA by extensive washing; and 3) to specifically elute RNA molecules bound at the carbohydrate binding site by incubating the washedbeads in buffer containing high concentrations of(GlcNAc)3. The SELEX protocol is outlined inTable 1.

TheWGA density on Wheat GermLectin Sepharose 6MB beads is approximately 5 mg/ml ofgel or 116 μM (manufacturer's specifications). Mter extensive washing in HBSS, the immobilized WGA was incubated with RNA at roomtemperature for 1 to 2 hours in a2 ml siliconized column with constantrolling (Table 1). Unbound RNA was removed by extensive washing with HBSS. Bound RNA was eluted as two fractions; first, nonspecifically eluted RNAwas removed by incubating and washing with 10 mMmethyl-α-D-mannopyranoside in HBSS (Table

1; rounds 1-4) or with HBSS (Table 1; rounds 5-11); second, specifically eluted RNA was removed by incubating and washing with 0.5 to 10 mM (GlcNAc)3 in HBSS (Table 1). The percentage ofinputRNA thatwas specifically eluted is recorded in Table 1.

The specifically eluted fraction was processed foruse in the following round. Fractions eluted from immobilizedWGA were heated at 90°C for 5 minutes in 1% SDS, 2% β-mercaptoethanol and extracted withphenol/chloroform. RNA was reverse transcribed into cDNAby AMV reversetranscriptase at48°C for 60 min in 50 mM Tris-Cl pH (8.3), 60 mM NaCl, 6 mM Mg(OAc)₂, 10 mM DTT, 100 pmol DNAprimer, 0.4 mM each ofdNTPs, and 0.4 unit/μl AMV RT. PCR amplification ofthis cDNAresulted in approximately 500pmol double-stranded

DNA, transcripts ofwhich were used to initiate the next round ofSELEX. D) Nitrocellulose FilterBinding Assay

As described in SELEX PatentApplications, anitrocellulose filter partitioning method was used to determine the affinity ofRNA ligands forWGA and for other proteins. Filter discs (nitrocellulose/cellulose acetate mixed matrix, 0.45 μm pore size, Millipore; or pure nitrocellulose, 0.45 μm pore size, Bio-Rad) were placed on a vacuum manifold and washed with 4 ml ofHBSS bufferunder vacuum. Reaction mixtures, containing ³²P labeled RNA pools and unlabeledWGA. were incubated in HBSS for 10 min at room temperature, filtered, and then immediately washed with4 ml HBSS. The filters were air-dried and counted in aBeckman LS6500 liquid scintillation counter without fluor.

WGA is a homodimer, molecular weight 43.2 kD, with 4 GlcNAc binding sites per dimer. For affinity calculations, we assume one RNA ligandbinding site permonomer (two per dimer). The monomerconcentration is defined as 2 times the dimer concentration. The equilibrium dissociation constant, Kd, for an RNA pool or specific ligand thatbinds monophasically is given by the equation

K_d = [P_f][R_f]/[RP]

where, [Rf] = free RNA concentration

[Pf] = free WGAmonomerconcentration

[RP]= concentration ofRNA/WGA monomer complexes

K_d = dissociation constant

A rearrangement ofthis equation, in which the fraction ofRNA bound at equilibrium is expressed as a function ofthe total concentration ofthe reactants, was used to calculate Kds ofmonophasic binding curves:

q = (P_τ + R_τ + K_d - ((P_τ + R_τ + K_d)² - 4 P_τ R_τ)^1/2)

q = fraction ofRNA bound

[P_T] = total WGA monomerconcentration

[R_T] = total RNA concentration

Kds were determined by least square fitting ofthe data points using the graphics program Kaleidagraph (Synergy Software, Reading , PA).

E) Cloning and Sequencing

The sixth and eleventh round PCR products were re-amplified with primers which contain a BamH1 or a EcoR1 restriction endonuclease recognition site. Using these restriction sites the DNA sequences were inserted directionally into the pUC18 vector. These recombinant plasmids were transformedintoE. coli strain JM109

(Stratagene, LaJolla, CA). Plasmid DNA was prepared according to the alkaline hydrolysis method (Zhou et al., 1990 Biotechniques 8:172-173) and about 72 clones were sequenced using the Sequenase protocol (United States Biochemical

Corporation, Cleveland, OH). The sequences are provided in Table 2.

F) Competitive Binding Studies

Competitive binding experiments were performed to determine ifRNA ligands and (GlcNAc)3 bind the same site on WGA. A set ofreaction mixtures containing α ³²P labeled RNA ligand and unlabeledWGA, each at a fixed concentration (Table 5), was incubated in HBSS for 15 min at room temperature with (GlcNAc)₃. Individual reaction mixtures were then incubated with a

(GlcNAc)₃ dilution from a 2-fold dilution series for 15 minutes. The final

(GlcNAc)₃ concentrations ranged from 7.8 μM to 8.0 mM (Table 5). The reaction mixtures were filtered, processed and counted as described in "Nitrocellulose Filter Binding Assay," paragraph D above.

Competition titration experiments were analyzed by the following equation to determine the concentration offree protein [P] as afunction ofthe total concentration ofcompetitor added, [C_τ]:

0 = [P](l+K_L[L_τ]/(l+K_L[P])+Kc[C_τ]/(l+Kc[P]))-Pτ where L_Tis the concentration ofinitial ligand, KL is the binding constant ofspecies L to the protein (assuming 1:1 stiochiometry) and Kc is the binding constant of species C to the protein (assuming 1:1 stiochiometry). Since it is difficult to obtain a direct solution for equation 1, iteration to determine values of [P] to aprecision of 1x10¹⁵ was used. Using these values of [P], the concentration ofprotein-ligand complex [PL] as a function of[C_τ] was determinedby the following equation:

[PL] = K_L[L_T][P](1+K_L[P])

Since the experimental data is expressed in terms of%[PL], the calculated concentration of [PL] was normalized by the initial concentration of [PLo] before addition ofthe competitor. ([PLo] was calculated using the quadratic solution for the standardbinding equation for the conditions used. The maximum (M) and minimum (B) %[PL] was allowed to float during the analysis as shown in the following equation.

%[PL] = [PL]/[PLo]*(M-B)+B A non-linear least-squares fitting procedure was used as described by P.R.

Bevington (1969) Data Reduction andError Analysis for the Physical Sciences, McGraw-Hill publishers. The program used was originally written by Stanley J. Gill in MatLab and modified forcompetition analysis by Stanley C. Gill. The data were fit to equations 1-3 to obtain best fit parameters for K_C, M and B as a function of [CT] while leaving K_L and P_Tfixed.

G) Inhibition ofWGA Agglutinating Activity

Agglutination is areadily observed consequence ofthe interaction ofalectin with cells andrequires that individual lectin molecules crosslinktwo ormore cells. Lectin mediated agglutination canbe inhibitedby sugars with appropriate specificity. Visual assay ofthe hemagglutinating activity ofWGA and the inhibitory activity of RNA ligands, GlcNAc and (GlcNAc)3 was made in Falcon roundbottom 96 well microtiterplates,-using sheep erythrocytes. Each well contained 54 μl of erythrocytes (2.5 x 10⁸ cells/ml) and 54 μl oftest solution.

To titrateWGA agglutinating activity, each test solution contained aWGA dilution from a4-fold dilution series. The final WGA concentrations ranged from 0.1 pM to 0.5 μM. Forinhibition assays, the test solutions contained 80 nMWGA (monomer) and a dilution from a4-fold dilution series ofthe designated inhibitor. Reaction mixtures were incubated at room temperature for 2 hours, afterwhich time no changes were observed in the precipitation patterns oferythrocytes. These experiments were carried out in Gelatin Veronal Buffer (0.15 mM CaCl₂, 141 mM NaCl, 0.5 mM MgCl₂, 0.1% gelatin, 1.8 mM sodium barbital, and 3.1 mM barbituric acid, pH 7.3 -7.4; Sigma #G-6514).

In the absence ofagglutination, erythrocytes settle as a compactpellet.

Agglutinated cells form amore diffuse pellet. Consequently, in visual tests, the diameter ofthe pellet is diagnostic for agglutination. The inhibition experiments includedpositive and negative controls for agglutination and appropriate controls to show that the inhibitors alone did not alter the normal precipitation pattern. Example 2

RNA Ligands toWGA

A. SELEX

The starting RNA library for SELEX, randomized 50N9 (SEQ ID NO: 1), contained approximately 2 x 10¹⁵ molecules (2 nmol RNA). The SELEX protocol is outlined in Table 1. Binding ofrandomized RNA to WGA is undetectable at 36 μM WGA monomer. The dissociation constant ofthis interaction is estimated to be >4mM. The percentage ofinput RNA eluted by (GlcNAc)3 increased from 0.05 % in the first round, to 28.5 % in round 5 (Table 1). The bulk Kd ofround 5 RNA was 600 nM (Table 1). Since an additional increase in specifically eluted RNA was not observed in round 6a (Table 1), round 6 was repeated (Table 1, round 6b) with two modifications to increase the stringency ofselection: the volume ofgel, and hence the mass ofWGA, was reduced ten fold; and RNA was specifically eluted with increasing concentrations of (GlcNAc)3, in stepwise fashion, with only the last eluted RNA processed for the following round. The percentage ofspecifically eluted RNA increased from 5.7 % in round 6b to 21.4 % in round 8, with continued improvement in the bulk K_d (260 nM, round 8 RNA, Table 1).

Forrounds 9 through 11, the WGA mass was again reduced ten fold to further increase stringency. The K_d ofround 11 RNA was 68 nM. Sequencing of the bulk starting RNA pool and sixth and eleventh round RNA revealed some nonrandomness in the variable region at the sixthround andincreased

nonrandomess at round eleven.

To monitor the progess ofSELEX, ligands were cloned and sequenced from round 6b and round 11. From each ofthe two rounds, 36 randomly picked clones were sequenced. Sequences were aligned manually and are shown in Table 2.

B. RNA Sequences

From the sixth and eleventh rounds, respectively, 27 of29 and 21 of35 sequenced ligands were unique. The numberbefore the "." in the ligand name indicates whether it was cloned from the round 6 or round 11 pool. Only aportion ofthe entire clone is shown in Table 2 (SEQ ID NOS: 4-55). The entire evolved randomregion is shown in upper case letters. Any portion ofthe fixed region is shown in lower case letters. By definition, each clone includes both the evolved sequence and the associated fixed region, unless specifically stated otherwise. A unique sequence is operationally defined as one that differs from all others by three or more nucleotides. In Table 2, ligands sequences are shown in standard single letter code (Comish-Bowden, 1985 NAR 13: 3021-3030). Sequences that were isolated more than once are indicated by the parenthetical number, (n), following the ligand isolate number. These clones fall into nine sequence families (1-9) and a group ofunrelated sequences (Orphans).

The distribution offamilies from round six to eleven provides a clear illustration ofthe appearance and disappearance ofligand families in response to increased selective pressure (Table 2). Family 3, predominant (11/29 ligands) in round 6, has nearly disappeared (2/35) by round 11. Similarly, minor families 6 through 9 virtually disappear. In contrast, only one (family 1) ofround eleven's predominant families (1, 2, 4 and 5) was detected in round six. The appearance and disappearance offamilies roughly correlates with theirbinding affinities.

Alignment (Table 2) defines consensus sequences for families 1-4 and 6-9 (SEQ ID NOS: 56-63). The consensus sequences offamilies 1-3 are long (20, 16 and 16, respectively) and very highly conserved. The consensus sequences of families 1 and 2 contain two sequences in common: the trinucleotide TCG andthe pentanucleotide ACGAA. A related tetranucleotide, AACG, occurs in family 3. The variation in position ofthe consensus sequences within the variable regions indicates that the ligands do notrequire a specific sequence from either the 5' or 3' fixed region.

The consensus sequences offamily 1 and 2 are flanked by complementary sequences 5 ormore nucleotides in length. These complementary sequences are not conserved and the majority include minor discontinuities. Family 3 also exhibits flanking complementary sequences, but these are more variable in length and structure and utilize two nucleotide pairs ofconserved sequence.

Confidence in the family 4 consensus sequence (Table 2) is limited by the small number ofligands, the variability ofspacing and the high G content. The pentanucleotide, RCTGG, also occurs in families 5 and 8. Ligands offamily 5 show other sequence similarities to those offamily4, especially to ligand 11.28.

C. Affinities

The dissociation constants forrepresentative members offamilies 1-9 and orphan ligands were determined by nitrocellulose filter binding experiments and are listed in Table 3. These calculations assume one RNA ligandbinding site perWGA monomer. Atthe highestWGA concentration tested (36 μM WGA monomer), binding ofrandom RNA is not observed, indicating aK_d atleast 100-fold higher than the protein concentration or > 4 mM.

The data in Table 3 define several characteristics of hgand binding. First, RNA ligands to WGAbind monophasically. Second, the range ofmeasured dissociation constants is 1.4 nM to 840 nM. Third, the binding for a number of ligands, most ofwhich were sixth round isolates, was less than 5% at the highest WGA concentration tested. The dissociation constants ofthese ligands are estimated to be greater than 20 μM. Fourth, on average eleventh round isolates have higher affinity than those from the sixth round. Fifth, the SELEX probably was not taken to completion; the best ligand (11.20)(SEQ ID NO: 40) is not the dominant species. Since the SELEX was arbitrarily stopped at the 11th round, it is not clearthat 11.20 would be the ultimate winner. Sixth, even though the SELEX was not taken to completion, as expected, RNA ligands were isolated that bind WGA with much greater affinity than do mono- or oligosaccharides (ie., the affinity of 11.20 is 5x10⁵ greater than that ofGlcNAc, Kd = 760 μM, and 850 betterthan that of(GlcNAc)3,

Kd = 12 μM; Y.Nagata and M.Burger, 1974, supra). This observation validates the proposition that competitive elution allows the isolation ofoHgonucleotide ligands with affinities that are several orders ofmagnitude greaterthan that ofthe competing sugar.

In addition these data show that even under conditions ofhigh target density,

116 pmol WGA dimer/μl ofbeads, it is possible to overcome avidity problems and recover ligands with nanomolar affinities. From the sixth to the eleventh round

(Table 2), in response to increased selective pressure as indicated by the

improvement in bulk Kd (Table 1), sequence families with lower than average affinity (Table 3) are eUminated from the pool.

Example 3

Specificity ofRNA Ligands to WGA

The affinity ofWGA ligands 6.8, 11.20 and 11.24 (SEQ ID NOS: 13, 40, and 19) for GlcNAc binding lectins from Ulexeuropaeus, Datura stramonium and Canavalia ensiformis were determined by nitrocellulose partitioning. The results of this determination are shown in Table 4. The ligands are highly specific forWGA. For example, the affinity ofligand 11.20 for WGA is 1,500, 8,000 and >15,000 fold greater than it is for the U. europaeus, D. stramonium and C. ensiformis lectins, respectively. The 8,000 fold difference in affinity for ligand 11.20 exhibited by T. vulgare and D. stramonium compares to a 3 to 10 fold difference in their affinity for oligomers ofGlcNAc and validates the proposition that competitive elution allows selection ofoligonucleotide ligands with much greater specificity than monomeric and oligomeric saccharides (J.F.Crowley et al., 1984, Arch. Biochem. and Biophys.231:524-533; Y.Nagata and M.Burger, 1974, supra; J-P.Privat et al., FEBS Letters 46:229-232).

Example 4

Competitive Binding Studies

Ifan RNA ligand and a carbohydrate bind a common site, then binding of the RNA ligand is expected to be competitively inhibited by the carbohydrate.

Furthermore, ifthe oligonucleotide ligands bind exclusively to carbohydrate binding sites, inhibition is expected to be complete at high carbohydrate concentrations. In the experiments reported in Table 5, dilutions ofunlabeled (GlcNAc)₃, from a 2- fold dilution series, were added to three sets ofbinding reactions that contained WGA and an α-³²P labeled RNA ligand (6.8, 11.20 or 11.24 (SEQ ID NOS: 13, 40 andl9); [RNA] final = [WGA]final = 15 nM). After a 15 minute incubation at room temperature, the reactions were filtered and processed as in standard binding experiments.

Qualitatively, it is clear that RNA ligands bind only to sites at which

(GlcNAc)3 binds, since inhibition is complete at high (GlcNAc)₃ concentrations

(Table 5). These data do not rule out the possibility that (GlcNAc)₃ binds one or more sites that are not bound by these RNA ligands.

Quantitatively, these datafit a simple model ofcompetitive inhibition (Table 5) and give estimates of 8.4, 10.9 and 19.4 μM for the Kd of(GlcNAc)₃. These estimates are in good agreement with Uterature values (12 μM @ 4 C, Nagata and Burger, 1974, supra; 11 μM @ 10.8 C, Van Landschoot et al., 1977, Eur. J. Biochem.79:275-283; 50 μM, M.Monsigny et al., 1979, Eur J. Biochem.98:39- 45). These dataconfirm the proposition that competitive elution with a specific carbohydrate targets the lectin's carbohydrate binding site.

Example 5

Inhibition ofWGA Agglutinating Activity

At 0.5 μM, RNA ligands 6.8 and 11.20 (SEQ ID NO: 13 and 40) completely inhibit WGA mediated agglutination ofsheep erythrocytes (Table 6). Ligand 11.24 (SEQ ID NO: 19) is not as effective, showing only partial inhibition at 2 μM, the highest concentration tested (Table 6). (GlcNAc)3 and GlcNAc completely inhibit agglutination at higher concentrations, 8 μM and 800 μM, respectively, (Table 6; Monsigny et al., supra). The inhibition ofagglutination varifies the proposition that ligands isolated by this procedure will be antagonists oflectin function. Inhibition also suggests that more than one RNA ligand is bound perWGA dimer, since agglutination is a function ofmultiple carbohydrate binding sites.

An alternative interpretation for the inhibition ofagglutination is that charge repulsion prevents negatively charged WGA/RNA complexes frombinding carbohydrates (a necessary condition for agglutination) on negatively charged cell surfaces. This explanation seems unlikely for two reasons. First, negatively charged oHgonucleotide ligands selected against an immobilized purified protein are known to bind to the protein when it is presented in the context ofa cell surface (see Example 10, L-selectin cell binding). Second, negatively charged (pi = 4) succinylatedWGA is as effective as native WGA (pI = 8.5) in agglutinating erythrocytes (M.Monsigny et al., supra). Example 6

Secondary Structure ofHigh Affinity WGA Ligands

In favorable instances, comparative analysis ofaligned sequences allows deduction ofsecondary structure and structure-function relationships. Ifthe nucleotides at two positions in a sequence covary according to Watson-Crick base pairing rules, then the nucleotides at these positions are apt to be paired.

Nonconserved sequences, especially those that vary in length are not apt to be directly involved in function, while highly conserved sequence are likely to be directly involved.

Comparative analyses ofboth family 1 and 2 sequences each yield a hairpin structure with a large, highly conservedloop (Figures la and lb). Interactions between loop nucleotides are likely but they are not definedby these data. The stems ofindividual ligands vary in sequence, length and structure (i.e., a variety of bulges and internal loops are allowed; Table 2). Qualitatively itis clearthat the stems are vaUdated by Watson/Crick covariation and thatby the rules ofcomparative analysis the stems are not directly involved in binding WGA. Family 3 can form a similarhairpin in which 2 pairs ofconserved nucleotides are utilized in the stem (Figure 1c).

Ifit is not possible to fold the ligands ofa sequence family into homologous structures, their assignment to a single family is questionable. Both ligand 11.7, the dominant member offamily 4, and ligand 11.28 can be folded into two plane G-quartets. However, this assignment is speculative: 1) 11.28 contains five GG dinucleotides and one GGGG tetranucleotide allowing other G-quartets; and 2) ligands 11.2 and 11.33 cannot form G-quartets. On the other hand, all ligands can form a hairpin with the conserved sequence GAGRFTNCRT in the loop. However, the conserved sequence RCTGGC (Table 2) does not have a consistent role in these hairpins.

Multiple G-quartet structures are possible for Family 5. One ofthese resembles the ligand 11.7 G-quartet. No convincing hairpin structures are possible for ligand 11.20.

Example 7

2'-NH₂ RNA Ligands to Human L-Selectin

The experimental procedures outlined in this Example were used to identify and characterize the 2'-NH₂ RNA ligands to human L-selectin in Examples 8-12. Experimental Procedures A) Materials

LS-Rg is a chimeric protein in which the extracellular domain ofhuman L- selectin isjoined to the Fc domain ofa human G2 immunoglobulin (Norgard et al.,

1993, PNAS 90:1068-1072). ES-R_g, PS-R_g and CD22β-R_g are analogous constructs ofE-selectin, P-selectin and CD22βjoined to a human G1

immunoglobulin Fc domain (R.M. Nelson et al., 1993, supra; I. Stamenkovic et al., 1991, Cell 66, 1133-1144). Purified chimera were provided by A.Varki. Soluble P-selectin was purchased from R&D Systems. Protein A Sepharose 4 Fast Flow beads were purchased from PharmaciaBiotech. Anti-L-selectin monoclonal antibodies: SKI1 was obtained from Becton-Dickinson, San Jose, CA; DREG-56, an L-selectin specific monoclonal antibody, was purchased from Endogen,

Cambridge, MA. The 2'-NH₂ modified CTP and UTP were prepared according to Pieken et. al. (1991, Science 253:314-317). DNA oligonucleotides were synthesizedby Operon. All otherreagents and chemicals were purchased from commercial sources. Unless otherwise indicated, experiments utilized HSMC buffer (1 mM CaCl₂, 1 mM MgCl₂, 150 mM NaCl, 20.0 mM HEPES, pH 7.4).

B) SELEX

The SELEX procedure is described in detail in United States Patent

5,270,163 and elsewhere. The nucleotide sequence ofthe synthetic DNA template for the LS-Rg SELEX was randomized at 40 positions. This variable region was flanked by N75' and 3' fixed regions (40N7).40N7 transcript has the sequence 5' gggaggacgaugcgg-40N-cagacgacucgcccga 3' (SEQ ID NO: 64). All C and U have 2-NH₂ substituted for 2'-OH on the ribose. The primers for the PCR were the following:

N75' Primer 5' taatacgactcactatagggaggacgatgcgg 3' (SEQ ID NO: 65)

N73' Primer 5' tcgggcgagtcgtcctg 3' (SEQ ID NO: 66)

The fixed regions include primer annealing sites for PCR and cDNA synthesis as well as a consensus T7 promoter to allow in vitro transcription. The initial RNA pool was made by first Klenow extending 1 nmol ofsynthetic single stranded DNA and then transcribing the resulting double stranded molecules with T7 RNA polymerase. Klenow extension conditions: 3.5 nmols primer 5N7, 1.4 nmols 40N7, IX Klenow Buffer, 0.4 mM each ofdATP, dCTP, dGTP and dTTP in a reaction volume of 1 ml. For subsequent rounds, eluted RNA was the template for AMV reverse transcriptase mediated synthesis ofsingle-stranded cDNA. These single-stranded DNA molecules were converted into double-stranded transcription templates by PCR amplification. PCR conditions were 50 mM KCl, 10 mM Tris-Cl, pH 8.3, 7.5 mM MgCl2, 1 mM ofeach dATP, dCTP, dGTP, and dTTP, and25 U/ml ofTaq DNA polymerase. Transcription reactions contained 0.5 mM DNA template, 200 nM T7 RNA polymerase, 80 mM HEPES (pH 8.0), 12 mMMgCl₂, 5 mM DTT, 2 mM spermidine, 2 mM each of 2-OH ATP, 2'-OH GTP, 2'-NH₂ CTP, 2'-NH₂ UTP, and 250 nM α-³²P 2'-OH ATP.

The strategy forpartitioning LS-Rg/RNA complexes from unbound RNA is outUned inTables 7a and 7b. First, the RNA pool was incubated withLS-Rg immobiUzed on protein A sepharose beads in HSMC buffer. Second, the unbound RNA was removed by extensive washing. Third, the RNA molecules bound atthe carbohydrate binding site were specifically elutedby incubating the washedbeads in HMSC buffer containing 5 mMEDTA in place ofdivalent cations. The 5 mM elution was followed by anon-specific 50 mM EDTA elution. LS-Rg was coupled to protein A sepharose beads according to the manufacturer's instructions

(PharmaciaBiotech).

The 5 mM EDTA elution is avariation ofa specific site elution strategy. Although it is not apriori as specific as elution by carbohydrate competition, itis a general strategy for C-type (calcium dependent binding) lectins and is apractical alternative when the cost and/or concentration ofthe requiredcarbohydrate competitor is unreasonable (as is the case with sialyl-Lewis^x). This scheme is expected to be fairly specific forligands that formbonds withthe lectin's bound

Ca⁺⁺because the low EDTA concentration does not appreciably increase the buffer's ionic strength and the conformation ofC-type lectins is only subtly altered in the absence ofbound calcium (unpublished observations cited by K. Drickamer, 1993, Biochem. Soc. Trans.21:456-459).

In the initial SELEX rounds, which were performed at 4 °C, the density of immobilized LS-Rg was 16.7 pmols/μl ofProtein A Sepharose 4 Fast Flow beads. In later rounds, the density ofLS-Rg was reduced (Tables 7a and 7b), as needed, to increase the stringency ofselection. At the seventhround, the SELEX was branched and continued in parallel at 4°C (Table 7a) and at roomtemperature (Table 7b). Wash and elution buffers were equilibrated to the relevant incubation temperature. Beginning with the fifth round, SELEX was often done at more than one LS-Rg density. In each branch, the eluted material from only one LS-Rg density was carried forward.

Before eachround, RNA was batch adsorbed to 100 μl ofprotein A sepharose beads for 1 hour in a 2 ml siliconized column. Unbound RNA and RNA eluted with minimal washing (two volumes) were combined andused for SELEX input material. For SELEX, extensively washed, immobilized LS-Rg was batch incubated with pre-adsorbed RNA for 1 to 2 hours in a 2 ml siliconized column with constant rocking. Unbound RNA was removed by extensive batch washing (200 to 500 μl HSMC/wash). Bound RNA was eluted as two fractions; first, bound RNA was eluted by incubating and washing columns with 5 mM EDTA in HSMC without divalent cations; second, the remaining elutable RNA was removed by incubating and/orwashing with 50 mM EDTA in HSMC without divalents. The percentage of input RNA that was eluted is recorded in Tables 7a and 7b. In every round, an equal volume of protein A sepharose beads without LS-Rg was treated identically to the SELEX beads to determine background binding. All unadsorbed, wash and eluted fractions were countedin aBeckman LS6500 scintillation counter in orderto monitor each round ofSELEX.

The eluted fractions were processed for use in the foUowing round (Tables 7aand7b). Afterextractingwithphenol/chloroform andprecipitating with isopropanol/ethanol (1:1, v/v), the RNA was reverse transcribed into cDNAby AMV reverse transcriptase either 1) at48°C for 15 minutes andthen 65°C for 15 minutes or 2) at 37°C and 48°C for 15 minutes each, in 50 mM Tris-Cl pH (8.3), 60 mM NaCl, 6 mM Mg(OAc)2, 10 mM DTT, 100 pmol DNA primer, 0.4 mM each ofdNTPs, and 0.4 unit/μl AMV RT. Transcripts ofthe PCRproduct were used to initiate the next round ofSELEX.

C) Nitrocellulose FilterBinding Assay

As described in SELEX Patent AppHcations, anitrocellulose filter partitioning method was used to determine the affinity ofRNA ligands forLS-Rg and for otherproteins. Filter discs (nitrocellulose/cellulose acetate mixed matrix, 0.45 μm pore size, MilHpore) were placed on a vacuum manifold and washed with 2 ml ofHSMC buffer under vacuum. Reaction mixtures, containing ³²P labeled RNA pools and unlabeled LS-Rg, were incubated in HSMC for 10 - 20 min at 4°C, room temperature or 37°C, filtered, and then immediately washed with 4 ml HSMC at the same temperature. The filters were air-dried and counted in aBeckman LS6500 liquid scintillation counter without fluor. LS-Rg is a dimeric protein that is the expression product ofa recombinant gene constructedby fusing the DNA sequence that encodes the extracellulardomains ofhuman L-selectin to the DNA that encodes a human IgG₂ Fc region. For affinity calculations, we assume one RNA ligandbinding site perLS-Rg monomer (two per dimer). The monomer concentration is defined as 2 times the LS-Rg dimer concentration. The equilibrium dissociation constant, Kd, for an RNA pool or specific ligandthatbinds monophasically is given by the equation

Kd = [Pf][Rf]/[RP]

where, [Rf] = free RNA concentration

[Pf] = free LS-Rg monomer concentration

[RP]= concentration ofRNA/LS-Rg complexes

K_d = dissociation constant

q = (P_T + R_T + K_d - ((P_T + R_T + K_d)² - 4 P_T R_T)^1/2 )

q = fraction ofRNA bound

[P_T] = 2 x (total LS-Rg concentration)

[R_T] = total RNA concentration

Many ligands and evolved RNA pools yield biphasic binding curves. Biphasic binding can be described as the binding oftwo affinity species that are not in equilibrium. Biphasic binding data were evaluated with the equation

q = 2P_t+R_t+Kd₁+Kd₂-[(P_t+X₁R₁+Kd₁)²-4P_tX₁Rt]^1/2

-[(P_t+X₂R_t+Kd₂)²-4P_tX₂R_t]^1/2,

where X₁ and X₂ are the mole fractions ofaffinity species R₁ and R₂ and Kd₁ and Kd₂ are the corresponding dissociation constants. Kds were determined by least square fitting Kds were determined by least square fitting ofthe data points using the graphics program Kaleidagraph (Synergy Software, Reading , PA).

D) Cloning and Sequencing

Sixth, thirteenth (RT) and fourteenth (4°C) round PCR products were re- amplified with primers which contain either aBamHl or aHinDIII restriction endonuclease recognition site. Using these restriction sites, the DNA sequences were inserted directionally into the pUC9 vector. These recombinant plasmids were transformed into E. coli strain DH5a (Life Technologies, Gaithersburg, MD). Plasmid DNA was prepared according to the alkahne hydrolysis method

(PERFECTprep, 5'-3', Boulder, CO). Approximately 150 clones were sequenced using the Sequenase protocol (Amersham, ArHngton Heights, IL). The resulting ligand sequences are shown in Table 8. E) Cell Binding Studies

The ability ofevolved ligandpools andcloned ligands to bindtoL-selectin presented in the context ofa cell surface was tested in experiments with isolated humanperipheral blood mononuclear cells (PBMCs). Whole blood, collected from normal volunteers, was anticoagulated with 5 mM EDTA. Six milliliters ofblood were layered on a 6 ml Histopaque gradientin 15 ml polypropylene tube and centrifuged (700 g) at room temperature for 30 minutes. The mononuclear cell layer was coUected, diluted in 10 ml ofCa⁺⁺/Mg⁺⁺-free DPBS (DPBS(-); Gibco 14190- 029) and centrifuged (225 g) for 10 minutes atroom temperature. Cell pellets from two gradients were combined, resuspended in 10 ml ofDPBS(-) and recentrifuged as described above. These pellets were resuspendedin 100 μl of SMHCK buffer supplemented with 1% BSA. CeUs were counted in a hemocytometer, diluted to

2x10⁷ cells/ml in SMHCK/1% BSA and immediately addedto binding assays. Cell viability was monitored by trypan blue exclusion.

Forcell binding assays, aconstantnumber ofcells were titrated with increasing concentrations ofradiolabeled ligand. The test ligands were serially diluted in DPBS(-)/1%BSA to 2-times the desired final concentration approximately 10 minutes before use. Equal volumes (25 μl) ofeach ligand dilution and the cell suspension (2x10⁷ cells/ml) were added to 0.65 ml eppendorftubes, gently vortexed and incubated on ice for 30 minutes. At 15 minutes the tubes were revortexed. The ligand/PBMC suspension was layered over 50 μl ofice cold phthalate oil (1:1 = dinonyhdibutyl phthalate) and microfuged (14,000 g) for 5 minutes at 4 °C. Tubes were frozen in dry ice/ethanol, visible pellets amputated into scintillation vials and counted in Beckman LS6500 scintilation counter as described in Example 7, paragraph C.

The specificity ofbinding to PBMCs was testedby competitionwiththe L- selectin specific blocking monoclonal antibody, DREG-56, while saturability of binding was testedby competition with unlabeled RNA. Experimental procedure and conditions were like those for PBMC binding experiments, except that the radiolabeled RNA ligand (final concentration 5 nM) was added to serial dilutions of the competitor before mixing with PBMCs. F) Inhibition ofSelectin Binding to sialyl-Lewis^x

The ability ofevolved RNA pools or cloned ligands to inhibit the binding of

LS-Rg to sialyl-Lewis^x was tested in competive ELISA assays (C. Foxall et al., 1992, supra). For these assays, the wells ofCorning (25801) 96 well microtiter plates were coatedwith 100 ng ofa sialyl-Lewis^x/BSA conjugate, air dried overnight, washed with 300 μl ofPBS(-) and then blocked with 1% BSA in

SHMCK for 60 min at room temperature. RNA ligands were incubated with LS-Rg in SHMCK/1% BSA at room temperature for 15 min. Afterremoval ofthe blocking solution, 50 μl ofLS-Rg (10nM) or a LS-Rg (10nMVRNA ligand mix was added to the coated, blocked wells and incubated at roomtemperature for 60 minutes. The binding solution was removed, wells were washed with 300 μl ofPBS(-) and then probed with HRP conjugated anti-human IgG, at room temperature to quantitate LS- Rg binding. After a 30 minute incubation at roomtemperature in the darkwith OPD peroxidase substrate (Sigma P9187), the extent ofLS-Rg binding and percent inhibition was determined from the OD450.

Example 8

2'-NH₂ RNA Ligands to Human L-selectin

A. SELEX

The starting RNA pool for SELEX, randomized 40N7 (SEQ ID NO: 63), contained approximately 10¹⁵ molecules (1 nmol RNA). The SELEX protocol is outlined in Tables 7a and 7b and Example 7. The dissociation constant of randomized RNA to LS-Rg is estimated to be approximately 10 μM. No difference was observed in the RNA elution profiles with 5 mM EDTA from SELEX and background beads for rounds 1 and 2, while the 50 mM elution produced a 2-3 fold excess overbackground (Table 7a). The 50 mM eluted RNA from rounds 1 and 2 were amplified for the input material for rounds 2 and 3, respectively. Beginning in round 3, the 5 mM elution from SELEX beads was significantly higher than background and was processed for the next round's input RNA. The percentage of input RNA eluted by 5 mM EDTA increased from 0.5 % in the first round to 8.4 % in round 5 (Table 7a). An additional increase in specifically eluted RNA from the 10 μM LS-Rg beads was not observed in round 6 (Table 7a). To increase the stringency ofselection, the density ofimmobilized LS-Rg was reduced ten fold in round 5 with furtherreductions in protein density at laterrounds. The affinity ofthe selected pools rapidly increased and the pools gradually evolvedbiphasic binding characteristics. Binding experiments with 6th round RNA revealed that the affinity ofthe evolving pool for L-selectin was temperature sensitive. Beginning with round 7, the SELEX was branched; one branch was continued at 4 °C (Table 7a) while the other was conducted at room temperature (Table 7b). Bulk sequencing of6th, 13th (rm temp) and 14th (4 °C) RNA pools revealed noticeable non-randomness atround six and dramatic non-randomess at the later rounds. The 6th round RNA bound monophasically at4°C with a dissociation constant ofapproximately 40 nM, while the 13th and 14th round RNAs bound biphasically with high affinity Kds of approximately 700 pM. The molar fraction ofthe two pools thatbound with high affinity were 24 % and 65 %, respectively. The binding ofall tested pools required divalent cations. In the absence ofdivalent cations, the Kds ofthe 13th and 14th round pools increased to 45 nM and 480 nM, respectively (HSMC, minus Ca^{+ +}

/Mg⁺⁺, plus 2 mM EDTA).

To monitor the progress of SELEX, ligands were cloned and sequenced from rounds 6, 13 (rm temp) and 14 (4 °C). Sequences were aligned manually and with the aid of a computer program that determines consensus sequences from frequently occurring local alignments.

B. Sequences

In Table 8, ligand sequences are shown in standard single lettercode (Cornish-Bowden, 1985 NAR 13: 3021-3030). The letter/number combination before the "." in the ligandname indicates whether it was cloned from the round 6, 13 or 14 pools. Only the evolved random region is shown in Table 8. Any portion ofthe fixed region is shown in lower case letters. By definition, each clone includes boththe evolved sequence and the associated fixedregion, unless specifically stated otherwise. From the sixth, thirteenth and fourteenth rounds, respectively, 26 of48, 8 of24 and 9 of70 sequenced ligands were unique. A unique sequence is operationally defined as one that differs from all others by three ormore nucleotides. Sequences that were isolated more than once, are indicated by the parenthetical number, (n), following the ligand isolate number. These clones fall into thirteen sequence families (I - XHI) and a group ofunrelated sequences (Orphans)(SEQ ID NOs: 67-117).

Two families, I and DI, are defined by ligands from multiple lineages. Both families occur frequently in round 6, but only one family UL ligandwas identified in the final rounds. Six families (IV, V, VI, VII, VHI, and possibly II) are each defined byjust two lineages which limits confidence in their consensus sequences. Five families (DC through XIII) are defined by a single lineage which precludes determination ofconsensus sequences.

Ligands from family II dominate the final rounds: 60/70 ligands in round 14 and 9/24 in round 13. Family II is representedby three mutational variations ofa single sequence. One explanation for the recovery ofa single lineage is that the ligand's information content is extremely high and was therefore represented by a unique species in the starting pool. Family II ligands were not detected in the sixth round which is consistent with a low frequency in the initialpopulation. An alternative explanation is sampling error. Note that a sequence ofquestionable relationship was detected in the sixth round.

The best defined consensus sequences are those offamily I, AUGUGUA

(SEQ ID NO: 118), and offamily m, AACAUGAAGUA (SEQ ID NO: 120), as shown in Table 8. Family HI has two additional, variably spaced sequences, AGUC and ARUUAG, that may be conserved. The tetranucleotide AUGW is found in the consensus sequence offamiHes I,III, and VII and in families II, VIII and IX. Ifthis sequence is significant, it suggests that the conserved sequences of ligands offamily VIII are circularly permuted. The sequence AGAA is found in the consensus sequence offamiHes IV and VI and in famiHes X and XIII.

C. Affinities

The dissociation constants for representative ligands from rounds 13 and 14, including all orphans, were determined by nitrocellulose filterbinding experiments are described in Example 7 and the results are listed in Table 9. These calculations assume two RNA ligand binding sites per chimera. The affinity ofrandom RNA cannot be reHably determined but is estimated to be approximately 10 μM.

In general, ligands bind monophasically with dissociation constants ranging from 50 pM to 15 nM at 4 °C. Some ofthe highest affinity ligands bind

biphasically. Although ligands offamilies I, VII, X and orphan F14.70 bind about equaHy well at 4 °C and room temperature, in general the affinities decrease with increasing temperature. The observed affinities substantiate the proposition that it is possible to isolate oligonucleotide ligands with affinities that are several orders of magnitude greaterthan that ofcarbohydrate ligands. Example 9

Specificity of 2'-NH₂ RNA Ligands to L-Selectin

The affinity ofL-selectin ligands to ES-Rg, PS-Rg and CD22β-Rg were determined by nitrocellulose partitioning as described in Example 7. As indicated in Table 10, the ligands are highly specific for L-selectin. In general, a ligand's affinity forES-Rg is 10³-fold lower and that forPS-Rg is about 10⁴-fold less than for LS-Rg. Binding above background is not observed for CD22β-Rg at the highest protein concentration tested (660 nM), indicating that ligands do notbind the Fc domain ofthe chimeric constructs nor do they have affinity for the sialic acidbinding site ofan unrelated lectin. The specificity ofoHgonucleotide ligandsbinding contrasts sharply withthe binding ofcognate carbohydrates by the selectins andconfirms the proposition that SELEX ligands will have greater specificity than carbohydrate ligands.

Example 10

Binding ofL-Selectin 2'-NH₂ RNA Ligands to Human PBMCs

Since the L-selectin ligands were isolated againstpurified, immobilized protein, it is essential to demonstrate that they bindL-selectin presented in the context ofa cell surface. Comparison of2nd and 9th round RNAs (Figure 2) shows that the evolved (9th round) ligand pool binds isolated PBMCs with high affinity and, as expected for specific binding, in a saturable fashion. The binding ofround 2 RNA appears to be non-saturable as is characteristic ofnon-specific binding. The cloned ligand, F14.12 (SEQ ID NO: 78), also binds in a saturable fashion with a dissociation constant of 1.3 nM, while random 40N7 (SEQ ID NO: 64) resembles round 2 RNA (Figure 3). The saturability ofbinding is confirmedby the data in Figure 4; > 90% of5 nM ³²P-labeled F14.12 RNA binding is competed by excess cold RNA. Specificity is demonstrated by the results in Figure 5; binding of5 nM

³²P-labeled F14.12 RNA is completely competed by the anti-L-selectin blocking monoclonal antibody, DREG-56, but is unaffected by an isotype-matched irrelevant antibody. These datavaHdate the feasibility ofusing immobilized, purifiedprotein to isolate ligands against a cell surface protein and the binding specificity ofF14.12 to L-selectin in the context ofa cell surface. Example 11

Inhibition ofBinding to Sialyl-Lewis^x

OHgonucleotide ligands, eluted by 2-5 mM EDTA, are expected to derive part oftheirbinding energy from contacts with the lectin domain's bound Ca⁺⁺ and consequently, are expected to compete with sialyl-Lewis^x forbinding. The ability ofligandF14.12 (SEQ ID NO: 78) to inhibit LS-Rg binding to immobilized sialyl-

Lewis^x was determined by competition ELISA assays. As expected, 4 mM EDTA reduced LS-Rg binding 7.4-fold, while 20 mM round 2 RNA did not inhibit LS-Rg binding. Carbohydrate binding is known to be Ca⁺⁺ dependent; the affinity of round 2 RNA is too low to bind 10 nM LS-Rg (Table 7).

In this assay F14.12 RNA inhibits LS-Rg binding in a concentration dependentmannerwith an IC₅₀ ofabout 10 nM (Figure 6). Complete inhibition is observed at 50 nM F14.12. The observed inhibition is reasonable under the experimental conditions; the Kd ofF14.12 atroom temperature is about 1 nM (Table 9) and 10 nM LS-Rg is 20 nM binding sites. These data verify that RNA ligands compete with sialyl-Lewis^x for LS-Rg binding and support the contention that low concentrations ofEDTA specificaUy elute ligands thatbindthe lectin domain's carbohydrate binding site.

Example 12

Secondary Structure of High Affinity 2'- NH₂ Ligands to L-Selectin

In favorable instances, comparative analysis ofaligned sequences allows deduction ofsecondary structure and structure-function relationships. Ifthe nucleotides at two positions in a sequence covary according to Watson-Crickbase pairing rules, then the nucleotides at these positions are apt to be paired.

Nonconserved sequences, especially those that vary in length are not apt tobe directly involved in function, while highly conserved sequence are likely to be directly involved.

Comparative analysis ofthe family I alignment suggests a hairpin structure in which the consensus sequence, AUGUGUGA, is contained within a variable size loop (Figure 7a). The stem sequences are not conserved and may be either 5' or 3'- fixed or variable sequence. The one ligand that does not form a stem, F14.25 (SEQ ID NO: 73), has a significantly lower affinity than the other characterized ligands (Table 9).

The proposed structure for family El is also a hairpin with the conserved sequence, AACAUGAAGUA, contained within a variable length loop (Figure 7b). The 5'-halfofthe stem is 5'-fixed sequence which may account in part for the less highly conserved sequence, AGUC.

Although there is no alignment data for family II, the sequence folds into a pseudoknot (Figure 7c). Three attractive features ofthis model are 1) the helices stack on one another, 2) the structure utilizes only variable sequence and 3) the structure is compatible with the majorvariant sequences.

Example 13

ssDNA Ligands to Human L-Selectin

The experimental procedures outlined in this Example were used to identify and characterize ssDNA ligands to human L-selectin as described in Examples 14-21.

Experimental Procedures

A) Materials

Unless otherwise indicated, all materials used in the ssDNA SELEX against the L-selectin/IgG2 chimera, LS-Rg, were identical to those ofExample 7, paragraph A. The buffer for SELEX experiments was 1 mM CaCl₂, 1 mMMgCl₂, 100 mM NaCl, 10.0 mM HEPES, pH 7.4. The buffer for all binding affinity experiments differed from the above in containing 125 mM NaCl, 5 mM KCl, and 20 mM HEPES, pH 7.4.

B) SELEX

The SELEX procedure is described in detail in United States Patent

5,270,163 and elsewhere. The strategy used for this ssDNA SELEX is essentially identical to that described in Example 7, paragraph B except as noted below. The nucleotide sequence ofthe synthetic DNA template forthe LS-Rg SELEX was randomized at 40 positions. This variable region was flanked by BH 5' and 3' fixed regions. The random DNA template was termed 40BH (SEQ ID NO: 126) and had the following sequence: 5'-ctacctacgatctgactagc<40N>gcttactctcatgtagttcc-3'. The primers for the PCR were the following: 5' Primer: 5'-ctacctacgatctgactagc-3' (SEQ ED NO: 127) and 3' Primer: 5'-ajajaggaactacatgagagtaagc-3';j=biotin (SEQ ED NO: 128). The fixed regions include primer annealing sites for PCR

amplification. The initial DNA pool contained 500 pmols ofeach oftwo types of single-stranded DNA: 1) synthetic ssDNA and 2) PCR amplified, ssDNA from 1 nmol ofsynthetic ssDNA template.

For subsequent rounds, eluted DNA was the template for PCR

amplification. PCR conditions were 50 mM KCl, 10 mM Tris-Cl, pH 8.3, 7.5 mM MgCl2, 1 mM ofeach dATP, dCTP, dGTP, and dTTP and 25 U/ml ofthe Stoffei fragment ofTaq DNA polymerase. After PCR amplification, double stranded DNAs were end-labeled using γ³²P-ATP. Complementary strands were separated by electrophoresis through an 8% polyacrylamide/7M urea gel. Strand separation results from the molecular weight difference ofthe strands due to biotintylation of the 3' PCR primer. In the final rounds, DNA strands were separated prior to end labelling in order to achieve high specific activity. Eluted fractions were processed by ethanol precipitation.

The strategy forpartitioning LS-Rg/ssDNA complexes from unbound ssDNA was as describedin Example 7, paragraphB, exceptthat2 mM EDTA was utilized for specific elution. The SELEX strategy is outlined in Table 11.

C) Nitrocellulose Filter Binding Assay

As described in SELEX Patent Applications and in Example 7, paragraph C, anitrocellulose filterpartitioning method was used to determine the affinity of ssDNA ligands forLS-Rg and for otherproteins. For these experiments a Gibco BRL 96 well manifold was substituted for the 12 well MilHpore manifold used in Example 7 and radioactivity was determined with a Fujix BAS100 phosphorimager. Binding data were analyzed as described in Example 7, paragraph C. D) Cloning and Sequencing

Thirteenth, fifteenth and seventeenth round PCR products were re-amplified withprimers which contain eitheraBamΑl or aHinDEIII restriction endonuclease recognition site. Approximately 140 ligands were cloned and sequenced using the procedures described in Example 7, paragraph D. The resulting sequences are shown in Table 12.

E) Cell Binding Studies

The ability ofevolved ligandpools to bind to L-selectin presented in the context ofa cell surface was tested in experiments with isolated human peripheral blood mononuclear cells (PBMCs) as described in Example 7, paragraphE

Flow Cytometry

Binding ofoligonucleotides to leukocytes was tested in flow cytometry applications. Briefly, peripheral blood mononuclearcells (PBMC) were purified on histoplaque by standard techniques. Cells (500 cells/mL) were incubated with fluorescein labeled oligonucleotide in 0.25 mL SMHCK buffer (140 mM NaCl, 1 mM MgCl₂, 1 mM CaCl₂, 5 mM, KCl, 20 mM HEPES pH 7.4, 8.9 mM NaOH, 0.1% (w/v) BSA, 0.1% (w/v) sodium azide) at room temperature for 15 minutes. Fluorescent staining ofcells was quantified on a FACSCaliber fluorescent activated cell sorter (Becton Dickinson, San Jose, CA).

To examine the ability ofoligonucleotides to bind leukocytes in whole blood,

25 μl aliquots ofheparinised whole blood were stained for 30 min at 22°C with 2 μg Cy5PE labeled anti-CD45 (generous gift ofKen Davis, Becton-Dickinson) and 0.15 μM FITC-LD201T1 (synthesized with a5'-Fluorescein phosphoramidite by Operon Technologies, Alameda, CA; SEQ ED NO: 185). To determine specificity, other samples were stained with FTTC-LD201T1 in the presence of0.3 μMDREG- 56 or 7 μM unlabeled LD201T1; or cells were reassayed after addition of4 mM EDTA. The final concentration ofwhole blood was at least70% (v/v). Stained, concentrated whole blood was diluted 1/15 in 140 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 1 mM CaCl₂, 20 mM HEPES pH 7.4, 0.1% bovine serum albumin and 0.1% NaN₃ immediately prior to flow cytometry on aBecton-Dickinson FACS CaHbur. Lymphocytes and granulocytes were gated using side scatter and

CD45CyPE staining.

F) Synthesis and Characterization ofMultimeric Oligonucleotide Ligands

Synthesis ofBranched Dimeric Oligonucleotide Complexes

Dimeric oligonucleotides were synthesized by standard solid state processes, with initiation from a 3'-3' Symmetric Linking CPG (Operon, Alameda, CA).

Branched complexes contain two copies ofa truncated L-selectin DNA ligand, each ofwhich is linked by the 3' end to the above CPG via a five unit ethylene glycol spacer (Figure 8A). Each ligandis labeled with a fluorescein phosphoramidite at the 5' end (Glen Research, Sterling, VA). Branched dimers were made for 3 truncates ofLD201T1 (SEQ ID NO: 142). The truncated ligands used were LD201T4 (SEQ ID NO: 187), LD201T10 (SEQ ED NO: 187) and LD201T1 (SEQ ED NO: 185). Branched dimers were purified by gel electrophoresis. Synthesis ofMultivalentBiotintylated-DNA Ligand/Streptavidin Complexes

Multivalent oligonucleotide complexes were produced by reacting

biotintylated DNAligands with eitherfluorescein orphycoerythrin labeled streptavidin (SA-FITC, SA-PE, respectively) (Figure 8B). Streptavidin (SA) is a tetrameric protein, each subunit ofwhich has a biotin binding site.5' and 3' biotintylated DNAs were synthesized by Operon Technologies, Inc (Alameda. CA) using BioTEG and BioTEG CPG (Glen Research, Sterling, VA), respectively. The expected stoichiometry is 2 to 4 DNA molecules per complex. SA/bio-DNA complexes were made for 3 truncates ofLD201(SEQ ID NO: 142). The truncated ligands were LD201T4 (SEQ ID NO: 187), LD201T10 (SEQ ID NO: 188) and LD201T1 (SEQ ID NO: 185). The bio-DNASA multivalent complexes were generated by incubating biotin modified oligonucleotide (1 mM) and fluoroscein labeled streptavidin (0.17 mM) in 150 mM NaCl, 20 mM HEPES pH 7.4 at room temperature for at least 2 hours. Oligonucleotide-streptavidin complexes were used directly fromthe reaction mixture without additional purification ofthe Complex from free streptavidin or oligonucleotide.

Synthesis of a Dumbell DimerMultivalent Complex

A "dumbell" DNA dimercomplex was formulated from ahomobifunctional N-hydroxysuccinimidyl (orNHS) active esterofpolyethelene glycol, PEG 3400 MW, and a 29mer DNA oligonucleotide, NX303 (SEQ ID NO: 196), having a 5' terminal Amino Modifier C6 dT (Glen Research) and a 3'-3' terminal

phosphodiesterlinkage (Figure 8C). NX303 is a truncate ofLD201 (SEQ ID NO: 142). The conjugation reaction was in DMSO with 1% TEA with excess equivalents ofthe DNA ligand to PEG. The PEG conjugates were purified from the free oligonucleotideby reversephase chromatography. The dimerwas then purified from the monomerby anion exchange HPLC. The oligonucleotide was labeled at the 5' terminus with fluorescein as previously described.

Synthesis ofa Fork DimerMultivalent Complex

To synthesize the fork dimer multivalent complex (Figure 8D), a glycerol was attachedby its 2-position to one terminus ofa linearPEGmolecule (MW20 kD) to give the bis alcohol. This was further modified to the bis succinate ester, which was activated to the bis N-hydroxysuccinimidyl active ester. The active ester was conjugated to the primary amine at the 5' terminus ofthe truncatedDNA nucleic acid ligand NX303 (SEQ ID NO: 196). The conjugation reaction was in DMSO with 1% TEA with excess equivalents ofthe DNA ligand to PEG. The PEG conjugates were purified away from the free oligonucleotide by reverse phase chromatography. The dimer was then purified away from the monomer by anion exchange HPLC. The oligonucleotide was labeled at the 5' terminus with fluorescein as previously described. Characterization ofMultimeric Oligonucleotide Ligands

The binding ofdimeric and multimeric oligonucleotide complexes to human peripheral blood mononuclearcells was analyzed by flow cytometry as described in Example 13, paragraph D. G) Photo-Crosslinking

A photo-crossHnking version ofDNA ligandLD201T4 (SEQ ID NO: 187) was synthesized by replacing nucleotide T15 (Figure 12) with 5-bromo-deoxyuracil.

4 nmol of³²P-labeled DNA was incubated with 4 nmol L-selectin-Rg in 4 ml IX SHMCK + 0.01 % human serum albumin (w/v), then irradiated at ambient temperature with 12,500 pulses from an excimer laser at a distance of50 cm and at 175 mJ/pulse. Protein andDNA were precipitated with400 μl 3 M sodium acetate and 8.4 ml ethanol followed by incubation at -70 degrees C. Precipitated material was centrifuged, vacuum dried and resuspended in 100 μl 0.1 M Tris pH 8.0, 10 mM CaCl2- Fourty-five μg chymotrypsin were added and after 20 min at 37 degrees C, the material was loaded onto an 8% polyacrylamide/7 M urea/ IXTBE gel and electrophoresed until the xylene cyanole had migrated 15 cm. The gel was soaked for 5 min in IX TBE and then blotted for 30 min at 200 mAmp in IXTBE onto Immobilon-P (MiUipore). The membrane was washed for 2 min in water, air dried, and an autoradiograph taken. A labeled band running slower than the free DNA band, representing a chymotryptic peptide crosslinked to LD201T4, was observed and the autoradiograph was used as a template to excise this band from the membrane. The peptide was sequenced by Edman degradation, and the resulting sequence was LEKTLP_SRSYY. The blank residue corresponds to the crossHnked amino acid, F82 ofthe lectin domain.

H) Lymphocyte Trafficking Experiments

Human PBMC were purified from heparinised blood by a Ficoll-Hypaque gradient, washed twice with HBSS (calcium/magnesium free) and labeled with 51Cr (Amersham). After labeling, the cells were washed twice with HBSS (containing calcium and magnesium) and 1% bovine serum albumin (Sigma). Female SCED mice (6-12 weeks ofage) were injected intravenously with 2x10⁶ cells. The cells were eitheruntreated or mixed with either 13 pmol ofantibody (DREG-56 or MEL- 14), or4, 1, or 0.4 nmol ofmodified oligonucleotide (synthesis described below). One hour laterthe animals were anesthetized, a blood sample taken and the mice were euthanised. PLN, MLN, Peyer's patches, spleen, liver, lungs, thymus, kidneys and bone marrow were removed and the counts incorporated into the organs determined by aPackard gamma counter. In a second protocol, 2x10*5 human

PBMC, purified, labeled, and washed as described above, were injected

intravenously into female SCED mice without antibody or oligonucleotide

pretreatment. One to 5 min prior to injecting the cells, the animals were injected with either 15 pmol DREG-56 or4 nmol modified oHgonucleotide. Counts incorporated into organs were quantified as described above.

Synthesis ofmodified nucleotides NX288 (SEQ ID NO: 193) and NX303 (SEQ ED NO: 196) was initiated by coupling to a dT-5'-CE polystyrene support (Glen Research), resulting in a 3'-3' terminal phosphodiester linkage, and having a 5' terminal an Amino Modifier C6 dT (Glen Research). Once NX288 andNX303 were synthesized, a 20,000 MW PEG2-NHS ester (ShearwaterPolymers,

Huntsville, AL) was then coupled to the oligonucleotide through the 5' amine moiety. The molarratio, PEG:oligo, in the reactions was from 3:1 to 10:1. The reactions were performed in 80:20 (v:v) 100 mM borate bufferpH 8: DMFat 37° C for one hour.

I) Inhibition of L-selectin Binding to Sialyl Lewis^x

SLe^x-BSA (Oxford GlycoSystems, Oxford, UK) in IX PBS, without CaCl₂ and MgCl₂, (GEBCO/BRL) was immobilized at 100 ng/well onto a microtiterplate by overnight incubation at 22°C. The wells were blocked for 1 h with the assay buffer consisting of20mM HEPES, 111 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 5 mM KCl, 8.9 mM NaOH, final pH 8, and 1% globulin-free BSA (Sigma). The reaction mixtures, incubated for90 min with orbital shaking, contained 5 nML-Selectin-Rg, a 1:100 dilution ofanti-human IgG-peroxidase conjugate (Sigma), and 0-50 nM ofcompetitor in assay buffer. Afterincubation, the plate was washed with BSA-free assay buffer to remove unbound chimera-antibody complex and incubated for 25 min with O- phenylenediarnine dihydrochloride peroxidase substrate (Sigma) by shaking in the dark at 22°C. Absorbance was read at 450 nm on a Bio-Kinetics Reader, Model EL312e (Bio-Tek Instruments, Laguna Hills, CA). Values shown represent the mean ± s.e from duplicate, ortriplicate, samples from one representative experiment. Example 14

ssDNA Ligands to L-Selectin

A. SELEX

The starting ssDNA pool for SELEX, randomized 40BH (SEQ ID NO:

126), contained approximately 10¹⁵ molecules (1 nmol ssDNA). The dissociation constant ofrandomized ssDNA to LS-Rg is estimated to be approximately 10 μM. The SELEX protocol is outlined in Table 11.

The initial round ofSELEX was performed at 4 °C with an LS-Rg density of 16.7 pmol/μl ofprotein A sepharose beads. Subsequent rounds were at room temperature except as noted in Table 11. The 2 mM EDTA elution was omitted from rounds 1-3. The signal to noise ratio ofthe 50 mM EDTA elution in these three rounds was 50, 12 and 25, respectively (Table 11). These DNAs were amplified for the inputmaterials ofrounds 2-4. Beginning with round 4, a 2 mM EDTA elution was added to the protocol. In this and all subsequent rounds, the 2 mM EDTA eluted DNA was ampHfied forthe nextround's inputmaterial.

To increase the stringency ofselection, the density ofimmobilized LS-Rg was reduced ten fold in round 4 with furtherreductions in protein as needed to increase the stringency ofselectin (Table 11). Under these conditions arapid increase in the affinity ofthe selectedpools was observed (Tables 11); at4 °C, the dissociation constant ofround 7 ssDNA was 60 nM.

Binding experiments with 7th round DNA revealed that the affinity ofthe evolving pool forL-selectin was weakly temperature sensitive (Kds: 60 nM, 94 nM and 230 nM at 4 °C, room temperature and 37 °C, respectively). To enhance the selection ofligands that bind at physiological temperature, rounds 8, 13, 16 and 17 were performed at 37 °C. Although temperature sensitive, the affinity ofround 15 ssDNA was optimal at room temperature (160 pM), with 3-fold higher Kds at4 °C and 37 °C.

Bulk sequencing ofDNA pools indicates some non-randomness at round 5 and dramatic non-randomness at round 13. Ligands were cloned and sequenced from rounds 13, 15, and 17. Sequences were aligned manually and with the aid of a NeXstar computer program that determines consensus sequences from frequently occurring local alignments.

B. Sequences

In Table 12, ligand sequences are shown in standard single letter code (Comish-Bowden, 1985 NAR 13: 3021-3030). Only the evolved random region is shown in Table 12. Any portion ofthe fixed region is shown in lower case letters. By definition, each clone includes both the evolved sequence and the associated fixed region, unless specifically stated otherwise A unique sequence is

operationally defined as one that differs from all others by three or more nucleotides. Sequences that were isolated more than once are indicated by the parenthetical number, (n), following the ligand isolate number. These clones fall into six families and a group ofunrelated sequences or orphans (Table 12)(SEQ ID NOs: 129-180).

Family 1 is defined by ligands from 33 lineages and has a well defined consensus sequence, TACAAGGYGYTAVACGTA (SEQ ID NO: 181). The conservation ofthe CAAGG and ACG and their 6 nucleotide spacing is nearly absolute (Table 12). The consensus sequence is flanked by variable but

complementary sequences that are 3 to 5 nucleotides in length. The statistical dominance offamily 1 suggests that the properties ofthe bulk population are a reflection ofthose offamily 1 ligands. Note that ssDNA family I and 2'-NH2 family I share a common sequence, CAAGGCG and CAAGGYG, respectively.

Family 2 is represented by a single sequence and is related to family 1. The ligandcontains the absolutely conserved CAAGG and highly conserved ACG of family 1 although the spacing between the two elements is strikingly different (23 compared to 6 nucleotides).

Families 4-6 are each defined by a small number ofligands which limits confidence in their consensus sequence, while family 7 is defined by a single sequence which precludes determination ofa consensus. Family 5 appears to contain two conserved sequences, AGGGT and RCACGAYACA, the positions of which are circularly permuted.

C. Affinities

The dissociation constants ofrepresentative ligands from Table 12 are shown in Table 13. These calculations assume two ssDNA ligand binding sites per chimera. The affinity ofrandom ssDNA cannot be reliably determined but is estimated to be approximately 10 μM.

At room temperature, the dissociation constants range from43 pM to 1.8 nM which is at least a 5x10³ to 2x10⁵ fold improvement overrandomized ssDNA (Table 13). At 37 °C, the Kds range from 130 pM to 23 nM. The extent of temperature sensitivity varies from insensitive (ligands LD122 and LD127 (SEQ ID NO: 159 and 162)) to 80-fold (ligand LD112 (SEQ ID NO: 135)). In general, among family 1 ligands the affinity ofthose from round 15 is greater than that of those from round 13. For the best ligands (LD208, LD227, LD230 and LD233 (SEQ ID NOS: 133, 134, 132, and 146)), the difference in affinity at room temperature and 37°C is about 4-fold.

The observed affinities ofthe evolved ssDNA ligand pools reaffirm our proposition that it is possible to isolate oligonucleotide ligands with affinities that are several orders ofmagnitude greaterthan that ofcarbohydrate ligands.

Example 15

Specificity ofssDNA Ligands to L-Selectin

The affinity ofrepresentative cloned ligands forLS-Rg, ES-Rg, PS-Rg, CD22β-Rg and WGA was determined by nitrocellulose partitioning and the results shown in Table 14. The ligands are highly specific forL-selectin. The affinity for ES-Rg is about 10³-fold lower and that for PS-Rg is about 5x10³-fold less than for LS-Rg. Binding above background is not observed for CD22β-Rg or forWGA at

0.7 and 1.4 μM protein, respectively, indicating that ligands neitherbind the Fc domain ofthe chimeric constructs norhave affinity forunrelated sialic acid binding sites.

The specificity ofoHgonucleotide ligandbinding contrasts sharply with the binding ofcognate carbohydrates by the selectins and reconfirms the proposition that SELEX ligands will have greater specificity than carbohydrate ligands. Example 16

Cell Binding Studies

Round 15 ssDNA pool was tested forits ability to bind to L-selectin presented in the context ofaperipheral bloodmononuclearcell surface as described in Example 13, paragraph E. The evolved pool was tested both for affinity and for specificity by competition with an anti-L-selectin monoclonal antibody. Figure 9 shows that the round 15 ssDNA pool binds isolated PBMCs with a dissociation constant ofapproximately 1.6 nM and, as is expected for specific binding, in a saturable fashion. Figure 10 directly demonstrates specificity ofbinding; in this experiment, binding of2 nM ³²P-labeled round 15 ssDNA is completely competed by the anti-L-selectin blocking monoclonal antibody, DREG-56, but is unaffected by an isotype-matched irrelevant antibody. In analogous experiments, LD201T1 (SEQ ID NO: 185) was shown to bind human PBMC with high affinity. Binding was saturable, divalent cation dependent, and blocked by DREG-56.

These data validate the feasibility ofusing immobilized, purified protein to isolate ligands against a cell surface protein and demonstrate the specific binding of the round 15 ssDNA pool and ofligand LD201T1 to L-selectin in the context ofa cell surface.

The binding ofLD201T1 to leukocytes in whole blood was examined by flow cytometry. Fluorescein isothiocyanate (FTFQ-conjugated LD201T1 specifically bind human lymphocytes and neutrophils (Figure 11A/B); binding is inhibited by competition with DREG-56, unlabeledLD201, and by the addition of4 mM EDTA (Figure 11A/B). These cell binding studies demonstrate that LD201T1 bind saturably and specifically to human L-selectin on lymphocytes and neutrophils.

Example 17

Secondary Structure ofHigh Affinity ssDNA Ligands to L-Selectin

Comparative analysis of24 sequences from family 1 strongly supports a hairpin secondary structure for these ligands (Figure 12). In the figure, consensus nucleotides are specified, with invariant nucleotides in bold type. To the right ofthe stem is a matrix showing the number ofoccurrences ofparticularbase pairs for the positions in the stem that are on the same line. The deduced structure consists of a GYTA tetraloop, a 3 nucleotide-pair upper stem and a 6 to 7 nucleotide-pair lower stem. The upper and lower stems are separated by an asymmetrical, AA internal loop or "bulge." Two ofthe three base pairs in the upper stem and 6 of7 in the lower stem are validated by covariation. The two invariant pairs, positions 7/20 and 10/19 are both standard Watson/Crick basepairs. This structure provides a plausible basis for the direct involvement ofinvariant nucleotides (especially, A8, A9 and T15) in binding the target protein.

The site ofoligonucleotide binding on L-selectin can be deduced from a set ofcompetition experiments. DREG56 is an anti-L-selectin, adhesion blocking monoclonal antibody that is known to bind to the lectin domain. Binding ofthree unrelated ligands, LD201T1 (SEQ ED NO: 185), LD174T1 (SEQ ID NO: 194) and LD196T1 (SEQ ID NO: 195), to LS-Rg was blocked by DREG-56, but not by an isotype-matched control. In cross-competition experiments, LD201T1, LD174T1, orLD196T1 prevented radio-labeled LD201T1 from binding to LS-Rg, consistent with the premise that the ligands bind the same or overlapping sites. The blocking and competition experiments, taken together with divalent cation-dependence of binding, suggest that all three ligands bind to the lectin domain. This conclusion has been verified forLD201 by photo-crosslinking experiments.

IfT15 ofLD201T4 (SEQ ID NO: 187; Figure 12) is replaced with 5-bromo- uracil, the resulting DNA photo-crosslinks at high yield (17%) to LS-Rg following irradiation with an excimer laser as described in Example 13, paragraph G. The high yield ofcrosslinking indicates apoint contact between the protein and T15. Sequencing ofthe chymotryptic peptide corresponding to this point contact revealed apeptide deriving from the lectin domain; F82 is the crossHnking amino acid. Thus, F82 contacts T15 in a stacking arrangement that permits high yieldphoto- crossHnking. By analogy to the structure ofthe highly related E-selectin (Graves et al, Nature 367, 532-538, 1994), F82 is adjacent to the proposed carbohydrate binding site. Thus, this photo-crosslink provides direct evidence that ligandLD201 makes contact with the lectin domain ofLS-Rg and provides an explanation for the function ofthe oligonucleotides in either stericaUy hindering access to the carbohydrate binding site or in altering the conformation ofthe lectin domain upon DNA binding.

Example 18

L-Selectin ssDNALigandTruncateData

Initial experiments to define the minimal high affinity sequence offamily 1 ligands show thatmore than the 26 nucleotide hairpin (Figure 12; Table 13) is required. Ligands corresponding to the hairpin, LD201T4 (SEQ ID NO: 187) and

LD227T1 (SEQ ID NO: 192) derived from LD201 (SEQ ID NO: 173) and LD227 (SEQ ID NO: 134), respectively, bind with 20-fold and 100-fold lower affinity than their full length progenitors. The affinity ofLD201T3 (SEQ ID NO: 186), a41 nucleotide truncate ofligandLD201, is reduced about 15-fold compared to the full length ligand, while the affinity ofthe 49-mer LD201T1 (SEQ ID NO: 185) is not significantly altered (Tables 12 and 13).

Additional experiments show that truncates LD201T10 (SEQ ID NO: 188) and LD227X1 (SEQ ID NO: 191) bind with affinities similar to their full length counterparts. Both ofthese ligands have stems that are extended at the base ofthe consensus stem. Alterations in the sequence ofthe added stem have little, ifany, effect on binding, suggesting that it is not directly involved in binding

The added stem is separated from the consensus stem by a single stranded bulge. The two ligands single stranded bulges differ in length and have unrelated sequences. Furthermore, LD201's bulge is at the 5'-end ofthe original stem base while that ofLD227 is at the 3'-end. Thus, the two ligands do notpresent an obvious consensus structure. Removal ofthe loop (LD201) or scrambling or truncating the sequence (LD227) diminishes affinity, suggesting that the bulged sequences may be directly involved in binding. Note that althoughLD201T3 is longer than LD201T10, it is unable to form the single stranded loop and extended stembecause ofthe position ofthe truncated ends.

Example 19

Inhibition ofBinding to Sialyl Lewis^x

Sialyl Lewis^x is the minimal carbohydrate ligand boundby selectins. The ability ofssDNA ligands to inhibit the binding ofL-selectin to Sialyl Lewis^x was determinedin competition ELISA assays as described in Example 13, paragraph I. LD201T1 (SEQ ID NO: 185 ), LD174T1 (SEQ ED NO: 194) andLD196T1 (SEQ ID

NO: 195) inhibited LS-Rg binding to immobilized SLe^x in a dose dependentmanner with IC50S ofapproximately 3 nM. This is a 10⁵-10⁶-fold improvement over the published IC50 values for SLe^x in similarplate-binding assays. A scrambled sequence based on LD201T1 showed no activity in this assay. These data verify that DNA ligands compete with sialyl-Lewis^x forLS-Rg binding and support the contention thatlow concentrations ofEDTA specifically eluteligands thatbindthe lectin domain's carbohydrate binding site.

Example 20

Inhibition OfLymphocyte Trafficking by L-Selectin ssDNA Ligands

Lymphocyte trafficking to peripheral lymph nodes is exquisitely dependent on L-selectin. Since the ssDNA ligands binds to human but not rodentL-selectin, a xenogeneic lymphocyte trafficking system was established to evaluate in vivo efficacy. Human PBMC, labeled with ⁵¹ Cr, were injected intravenously into SCED mice. Cell trafficking was determined 1 hour later. In this system, human cells traffic to peripheral and mesenteric lymph nodes (PLN and MLN). This

accumulation is inhibited by DREG-56 (Figure 13) but notMEL-14, amonoclonal antibody that blocks murine L-selectin-dependent trafficking. In initial experiments cells were incubated with either DREG-56 or 3 capped and PEG-modified oligonucleotide before injection. NX288 (SEQ ID NO: 193) inhibited trafficking of cells to PLN (Figure 13) and MLN in a dose-dependent fashion but had no effect on the accumulation ofcells in other organs. At the highest dose tested (4 nmol), inhibition by the DNA ligand was comparable to that ofDREG-56 (13 pmol), while ascrambled sequence had no significanteffect (Figure 13). The activity ofLD174T1 (SEQ ID NO: 194) was similar to that ofNX288.

To determine ifthe modifiedoHgonucleotide was effective when itwas not pre-incubated with cells, DREG-56 (13 pmol/mouse) orthe modified

oligonucleotide (4 nmol/mouse) was injectedintravenously into animals and 1-5 min laterthe radio-labeled human cells were given intravenously. Again, both NX288 (SEQ ID NO: 193) and DREG-56 inhibited trafficking to PLN and MLN while the scrambled sequence had no effect (Figure 14). Therefore, the modified

oligonucleotide did not require pre-incubation with the cells to effectively block trafficking. These experiments demonstrate, in vivo, the efficacy ofoligonucleotide ligands in inhibiting aL-selectin dependent process.

Example 21

L-Selectin Nucleic Acid Ligand Multimers

Multivalent Complexes were made in which two nucleic acid ligands to L- selectin were conjugated together. Multivalent Complexes ofnucleic acid ligands are described in copending United States Patent Application Serial Number08/434,465, filedMay 4, 1995, entitled "Nucleic AcidLigand Complexes" which is herein incorporatedby reference in its entirety. These multivalent Complexes were intended to increase the binding energy to faciHtate betterbinding affinities through slower off-rates ofthe nucleic acid ligands. These multivalent Complexes may be useful atlowerdoses than theirmonomeric counterparts. In addition, high molecular weight (20kD) polyethylene gylcol (PEG) was included in some ofthe Complexes to decrease the in vivo clearance rate ofthe complexes. Specifically, the nucleic acid ligands incorporated into the Complexes were LD201T1 (SEQ ID NO: 185), LD201T4 (SEQ ID NO: 187), LD201T10 (SEQ ID NO: 188) andNX303 (SEQ ED NO: 196). Multivalent selectin nucleic acid ligand Complexes were produced as described in Example 13, paragraphF.

A variety ofmonomeric nucleic acid ligands and multivalent Complexes have been examined in flow cytometry. The multivalent Complexes exhibited siirύlar specificity to the monomeric forms, but enhanced affinity as well as improved (i.e., slower) off-rate for human lymphocytes. Titration curves, obtained from incubating fluorescently labeled monomeric FTTC-LD201T1 with peripheral blood mononuclear cells (PBMC) purified human lymphocytes, indicated that binding to cells is saturable. Half-saturation fluorescence occurred at 3 nM oligonucleotide. In contrast, the branched dimeric FEEC-LD201T1 andbio-LD201Tl/SA multivalent Complexes exhibited half-saturation at approximately 0.15 nM, corresponding to an apparent 20-fold increase in affinity. In similar experiments, half saturation ofthe dumbell and fork dimers ofLD201T4 was observed at0.1 and 0.6 nM,

respectively, compared to 20 nM formonomeric LD201T4.

Kinetic competition experiments were performed on monomeric nucleic acid ligands and multivalent Complexes. Kinetic competition experiments were performed with PBMC purified lymphocytes. Cells were stained as described above butused 10 nM oligonucleotide. The off-rate formonomeric, dimeric and multivalent Complexes was determinedby addition of500 nM unlabeled

oHgonucleotide to cells stained with fluorescently labeled ligandsand measurement of the change in the mean fluorescence intensity as afunction oftime. The dissociation rate ofa monomeric LD201T1 fromL-selectin expressing human lymphocytes was approximately 0.005 sec-1, corresponding to ahalf-life ofroughly 2.4 minutes. TheLD201T1 branched dimer andbiotin conjugate multivalentComplexesexhibited apparent off-rates several times slower than that observedfor the monomeric ligand and as slow or slower than that observed for the anti-L-selectin blocking antibody DREG56, determined under the same conditions. A multivalent Complex containing anon-bindingnucleic acid sequence did not stain cells underidentical conditions and didnotcompete in the off-rate experiments. The off-rate ofthe LD201T4 dumbell and forkdimers is faster than the LD201T1 branched dimer and is betterthan all monomers tested. These results confirm the proposition that dimeric and multimeric ligands bind with higher affinities than do monomeric ligands and that the increased affinity results from slower off-rates.

Example 22

2'-F RNA Ligands to Human L-Selectin

The experimental procedures outlined in this Example were used to identify and characterize 2'-F RNAligands to human L-selectin as described in Examples

23-25.

Experimental Procedures

A)Materials

Unless otherwise indicated, all materials used in the 2'-F RNA SELEX againstthe L-selectin/IgG2 chimera, LS-Rg, were identical to those ofExamples 7, paragraph A and 13, paragraph A. SHMCK-140 buffer, used for all SELEX and binding experiments, was 1 mM CaCl₂, 1 mM MgCl₂, 140 mM NaCl, 5 mM KCl, and 20 mM HEPES, pH 7.4. A soluble form ofL-selectin, corresponding to the extracellulardomains, waspurchased from R&D Systems and used for some nitrocellulose filterbinding experiments.

B) SELEX

The SELEX procedure is described in detail in United States Patent

5,270,163 and elsewhere. Procedures are essentially identical to those in Examples 7 and 13 except as noted. The variable regions ofsynthetic DNA templates were randomized at either 30 or40 positions and were flanked by N75' and 3' fixed regions producing transcripts 30N7 (SEQ ID NO: 292) and40N7 (SEQ ID NO: 389). The primers for the PCR were the following:

N75" Primer 5' taatacgactcactatagggaggacgatgcgg 3' (SEQ ID NO: 65)

N73' Primer 5' tcgggcgagtcgtcctg 3' (SEQ ID NO: 66)

The initial RNA pool was made by first Klenow extending 3 nmol of synthetic single stranded DNA and then transcribing the resulting double stranded molecules withT7 RNApolymerase. Klenow extension conditions: 6 nmols primer 5N7, 3 nmols 30N7 or 40n7, IX Klenow Buffer, 1.8 mM each ofdATP, dCTP, dGTP and dTTP in a reaction volume of0.5 ml.

For subsequent rounds, eluted RNA was the template for AMV reverse transcriptase mediated synthesis ofsingle-stranded cDNA. These single-stranded DNA molecules were converted into double-stranded transcription templatesby PCR amplification. PCR conditions were 50 mM KCl, 10 mM Tris-Cl, pH 8.3, 7.5 mM MgCl2, 0.2 mM ofeach dATP, dCTP, dGTP, and dTTP, and 100 U/ml ofTaq DNA polymerase. Transcription reactions contained one third ofthe purified PCR reaction, 200 nM T7 RNA polymerase, 80 mM HEPES (pH 8.0), 12 mM MgCl2, 5 mM DTT, 2 mM spermidine, 1 mM each of2'-OH ATP, 2'-OH GTP, 3 mM each of2'-F CTP, 2'-F UTP, and 250 nM α-³²P 2'-OH ATP. Note that in all transcription reactions 2'-F CTP and 2'-F UTP replaced CTP and UTP.

The strategy forpartitioning LS-Rg/RNA complexes from unbound RNA is outlined in Table 15 and is essentially identical to that ofExample 7, paragraph B. In the initial SELEX rounds, which were performed at 37 °C, the density of immobilized LS-Rg was 10 pmols/μl ofProtein A Sepharose 4 Fast Flow beads. LS-Rg was coupled to protein A sepharose beads according to the manufacturer's instructions (Pharmacia Biotech). In laterrounds, the density ofLS-Rg was reduced (Table 15), as needed, to increase the stringency ofselection. At the seventh round, both SELEXes were branched. One branch was continued as previously described (Example 7, paragraph B). In the second branch ofboth SELEXes, the RNA pool was pre-annealed to oHgonucleotides that are

complementary to the 5' and 3' fixed sequences. These rounds are termed "counter- selected" rounds. Before each round, RNA was batch adsorbed to 100 μl ofprotein A sepharose beads for 15 minutes in a2 ml siliconizedcolumn. Unbound RNA and RNA eluted with minimal washing (two volumes) were combined and used for SELEX input material. For SELEX, extensively washed, immobilized LS-Rg was batch incubated withpre-adsorbed RNA for 1 to 2 hours in a 2 ml column with constant rocking. UnboundRNA was removedby extensive batch washing (500 μl SHMCK 140/wash). In addition, the counter selected rounds were extensively washed with buffercontaining 200 nM ofboth complementary oligos. BoundRNA was eluted as two fractions; first, bound RNA was eluted by incubating and washing columns with 100 μL 5 mM EDTA in SHMCK 140 withoutdivalent cations; second, the remaining elutable RNAwas removedbyincubating and/or washingwith 500 μL 50 mMEDTAin SHMCK 140 withoutdivalents. The percentage ofinput RNA thatwas elutedis recorded in Table 22. In every round, an equal volume of protein A sepharose beads withoutLS-Rg was treated identicaUy to the SELEX beads to determine backgroundbinding. All unadsorbed, wash andeluted fractions were counted in aBeckman LS6500 scintillation counter in orderto monitoreach round ofSELEX.

The 5 mMEDTA eluates were processedforuse in the following round (Table 15). After precipitating with isopropanol/ethanol (1:1, v/v), the RNA was reverse transcribed into cDNA by AMV reverse transcriptase either at48°C for 15 minutes andthen 65°C for 15 minutes in 50 mM Tris-Cl pH (8.3), 60 mMNaCl, 6 mMMg(OAc)₂, 10 mM DTT, 200 pmol DNAprimer, 0.5 mM each ofdNTPs, and 0.4 unit/μL AMV RT. Transcripts ofthe PCR product were used to initiate the next round of SELEX.

C) Nitrocellulose FilterBinding Assay

As described in SELEX Patent Applications, anitro cellulose filter partitioning method was used to determine the affinity ofRNA ligands forLS-Rg and for otherproteins. Filter discs (nitrocellulose/cellulose acetate mixed matrix, 0.45 μmpore size, Millipore) were placed on avacuummanifold and washed with 3 ml ofSHMCK 140 buffer under vacuum. Reaction mixtures, containing ³²P labeled RNA pools and unlabeled LS-Rg. were incubated in SHMCK 140 for 10 - 20 min at 37 °C, and then immediately washedwith 3 ml SHMCK 140. The filters were air-dried and counted in aBeckmanLS6500 liquid scintillation counterwithout fluor. Alternatively, binding studies employed 96 well micro-titer manifolds essentially as described in Example 13, paragraph E.

D) Cloning and Sequencing

12throundPCRproducts were re-amplified with primers which contain either aBamHl or aHinDIII restriction endonuclease recognition site. Using these restriction sites, the DNA sequences were inserted directionally into the pUC9 vector. These recombinantplasmids were transformed into E. coli strain DΗ5a (Life Technologies, Gaithersburg, MD). Plasmid DNA was prepared according to the alkaline lysis method (Quiagen, QIAwell, Chattsworth CA). Approximately 300 clones were sequenced using the ABI Prism protocol (Perkin Elmer, Foster City, CA). Sequences are shown in Table 16.

E) Cell Binding Studies

Binding ofevolved ligands to L-selectin presented in the context ofa cell surface was testedby flow cytometry experiments with human lymphocytes.

Briefly, peripheral blood mononuclear cells (PBMC) were purified on histoplaque by standard techniques. To evaluate leukocyte binding by unlabeled 2'-F ligands, cells (500 cells/mL) were incubated with fluorescein labeledFTTC-LD201T1 (SEQ ED NO: 185) in the presence ofincreasing concentrations ofindividual, unlabeled 2'- F ligands in 0.25 mL SMHCKbuffer (140 mM NaCl, 1 mM MgCl₂, 1 mM CaCl₂, 5 mM, KCl, 20 mM HEPES pH 7.4, 8.9 mM NaOH, 0.1% (w/v) BSA, 0.1% (w/v) sodium azide) at roomtemperature for 15 minutes. Fluorescent staining of cells was quantified on aFACSCaliberfluorescent activated cell sorter (Becton Dickinson, San Jose, CA). The affinity ofthe 2'-F competitorwas calculated from the flurorescence inhibition curves.

Example 23

2'-F RNA Ligands to L-Selectin

A. SELEX

The starting RNA pools for SELEX, randomized 30N7 (SEQ ID NO: 292) or40N7 (SEQ ED NO: 389) contained approximately 10¹⁴ molecules (0.7 nmol RNA). The SELEX protocol is outlined in Table 15 and Example 22. All rounds were selected at 37°C. The dissociation constant ofrandomized RNA to LS-Rg is estimated tobe approximately 10 μM. After six rounds the pool affinities had improved to approximately 300 nM. An aliquot ofthe RNA recovered fromthe seventh round was used as the starting material forthe first counter-selected rounds. Five rounds ofcounter-selection and five additional standard rounds were performed in parallel. Thus, a total oftwelve rounds were performed in both branches ofboth SELEXes: 30N7, counter-selected 30N7, 40N7 and counter-selected 40N7. The affinities ofeach ofthe 12th round pools ranged from 60 to 400 pM. Ligands were cloned from these pools.

B. Sequences of2'-F RNA Ligands to L-Selectin

In Table 16, ligand sequences are shown in standard single letter code (Cornish-Bowden, 1985 NAR 13: 3021-3030). Fixed region sequence is shown in lower case letters. By definition, each clone includes both the evolved sequence and the associated fixed region, unless specifically stated otherwise. A unique sequence is operationally defined as one that differs from all others by three or more nucleotides. Sequences that were isolated more than once are indicated by the parenthetical number, (n), following the ligand isolate number.

The 30N7 and 40N7 SELEX final pools shared acommon major sequence family, even though identical sequences from the two SELEXes are rare (Table 16). Most ligands (72 ofthe 92 unique sequences) from the 30N7 and 40N7 SELEXes contain one oftwo related sequence motifs, RYGYGUUUUCRAGY or

RYGYGUUWWUCRAGY. These motifs define family 1. Within the family there are three subfamilies. Subfamily la ligands (53/66) contain an additional sequence motif, CUYARRY, one nucleotide 5' to the family 1 consensus motifs. Subfamily lb (9/66 unique sequences) lacks the CUYARRY motif. Subfamily lc (5/66) is also missing the CUYARRY motif, has an A inserted between the Y and G ofconsensus YGUU and lacks the consensus GA base pair. The significance ofthe sequence subfamilies is reflected in the postulated secondary structure ofthe ligands (Example 25).

A second family, composed of5 sequences, has arelatively well defined consensus: UACUAN_0-1UGURCG...UYCACUAAGN_1-2CCC (Table 16). Family 3 has a short, unreliable consensus motif(Table 16). In addition, there are approximately 12 orphans or apparently unrelated sequences. Three ofthe orphan sequences were recovered at least twice (Table 16).

C. Affinities

The dissociation constants ofrepresentative ligands from Table 16 are shown in Table 17. These calculations assume two ligandbinding sites perchimera. The affinity ofrandom 2'-F RNA cannotbe reliably determinedbut is estimated to be approximately 10 μM.

The dissociation constants range from 34 pM to 315 nM at 37 °C. Binding affinity is not expected to be temperature sensitive since selection was at 37°C and 2'-F RNA forms thermal stable structures, but binding has not been tested at lower temperatures. For the mostpart, the extreme differences in affinity may be related to predicted secondary structure (Example 25).

The observed affinities ofthe evolved2'-F RNA ligands reaffirm our propositionthat itis possible to isolate oligonucleotide ligands with affinities that are several orders ofmagnitude greaterthan that ofcarbohydrate ligands.

Example 24

CellBinding Studies

The abiHty offull length 2'-F Ugands to bind to L-selectin presented in the context ofacell surface was testedby competition-flow cytometry experiments with humanperipheral blood lymphocytes. Lymphocytes were stained with 10 nM

FITC-conjugated DNA ligandFTTC-LD201T1 (SEQ ED NO: 185) in the presence of increasing concentrations ofunlabeled 2'-F ligands as describedinExample 22, paragraphE. Ligands LF1513 (SEQ ID NO: 321), LF1514 (SEQID NO: 297), LF1613 (SEQEDNO:331)andLF1618 (SEQ ID NO: 351) inhibitedthe binding ofFITC-LD201T1 in a concentration dependent manner, withcomplete inhibition observed at competitor concentrations of 10 to 300 nM. These results demonstrate that the 2'-F ligands are capable ofbinding cell surface L-selectin and suggest that the 2'-F ligands and LD201T1 bind the same or overlapping sites. The affinities of the fluoro ligands, calculated from the competition curves, range from 0.2 to 25 nM. The affinity oftwo ofthe ligands forL-selectin on human lymphocytes, LF1613 (Kd = 0.2 nM) andLF1514 (Kd = 0.8 nM), is significantly better than that ofthe DNA ligandLD201T1 (Kd = 3 nM). The reasonable agreementbetween the affinities forpurified protein and lymphocyte L-selectin suggests thatbinding to lymphocytes is specific forL-selectin. These data validate the feasibility ofusing immobiUzed, purified protein to isolate ligands against a cell surface protein.

Example 25

Secondary Structure ofHigh Affinity 2'-F RNA Ligands to L-Selectin

In favorable instances, comparative analysis ofaligned sequences allows deduction ofsecondary structure andstructure-function relationships. Ifthe nucleotides attwo positions in a sequence covary according toWatson-Crickbase pairing rules, then the nucleotides at these positions are apt to be paired.

Nonconserved sequences, especially those that vary in length are not apt to be directly involved in function, while highly conserved sequence are likely tobe directly involved.

The deduced secondary structure offamily la ligands fromcomparative analysis of21 unique sequences is a hairpin motif (Figure 15) consisting ofa4 to 7 nucleotide terminal loop, a 6 base upper stem and a lower stem of4 or more base pairs. The consensus terminal loops are either aUUUU tetraloop or a UUWWU pentaloop. Hexa- and heptaloops are relatively rare. The upperandlower stems are delineatedby a7 nucleotide bulge in the 5'-halfofthe stem. Four ofthe six base pairs in the upper stem and all base pairs in the lowerstem are supportedbyWatson- Crickcovariation. Ofthe two invariantbase pairs in the upper stem, one is the loop closing GC, while the otheris a non-standard GA. The lower stem is most often 4 or 5 base pairs longbut canbe extended. While the sequence oftheupper stemis strongly conserved, that ofthe lower stemis not, withthe possible exception ofthe YR' basepair adjacent to the internal bulge. This base pair appears to covary with the 3' position ofthe 7 nucleotidebulge in amannerwhichminimizes the HkeHhood ofextending the upper stem. Both the sequence (CUYARRY) andlength (7 nt) of the bulge are highly conserved.

In terms ofcomparative analysis, the 7 nucleotide bulge, the upper stem and the 5' and 3' positions ofthe terminal loop are most apt to be directly involved in L- selectin binding. Specifically, the 5' U and 3' U ofthe terminal loop, the invariant GC and GAbase pairs ofthe upper stem and the conserved C, U and A ofthe bulge are the mostly likely candidates. The lower stem, because ofits variability in length and sequence, is less likely to be directly involved. The importance ofthe bulge for binding is supported by the poor affinity ofligand LF1512 (SEQ ID NO: 357; Kd= 315 nM); the simplest structure for this ligand is a UUUU tetraloop and a ten base pair, nearly perfect, consensus stem which is missing only the 7 nucleotide bulge.

The deduced secondary structure offamily lb is similar to that offamily la, except thatthe upper stem is usually 7 base pairs in length and that the single strandedbulge which does not have ahighly conserved consensus is only 4 nucleotide long. This structure may be an acceptable variation ofthe 1a secondary structure with the upper stem's increased length allowing a shorterbulge; the affinity ofligandLF1511 (SEQ ID NO: 332) is 300 pM.

Although family lc has a consensus sequence, GUUUUCNR that is related to la and lb, aconvincing consensus secondary structure is not evident, perhaps due to insufficient data. The most highly structured member ofthe family, LF1618 (SEQ ID NO: 351), permits aUUUU tetraloop and "upper" stemof7 base pairs but has neither a lower stemnorthe consensus 7 nucleotide bulge sequence of la. The upper stem differs fromthose of la and lb in that it has an unpaired A adjacent to the loop closing G and does not have the invariant GAbase pair of 1a and 1b. The affinity ofLF1618 is amodest 10 nM which suggests that family lc forms aless successful structure.

Predictions ofminimal high affinity sequences forfarmly 1 ligands canbe made and serve as apartial test ofthe postulated secondary structure. Truncates which include only the upper stem and terminal loop, LF1514T1 (SEQEDNO: 385) orthese two elementsplus the 7 nucleotide bulge sequence, LF1514T2 (SEQED NO: 386), axe not expectedto bindwithhigh affinity. On the otherhand, there is a reasonable, but notrigorous, expectation that ligands truncatedatthebase ofthe lowerconsensus stem, LF1514T4 (SEQ ID NO: 387) andLF1807T4 (SEQ ED NO: 388), will bind with high affinity. In side by side comparisons, the affinities of LF1514T1 andLF1514T2 forLS-Rg were reduced atleast 100-fold in comparison to full lengthLD1514 (SEQ ID NO: 297), while the affinity ofLF1514T4 was reducedless than two fold and that ofLF1807T4 approximately three-fold. The correspondence between the predicted and observed truncate affinities supports the postulated secondary structure.

Since the ssDNA ligand LD201T1 (SEQ ID NO: 185) and the adhesion blocking anti-human L-selectin antibody DREG56 are known to bind to the lectin domain ofL-selectin, competition between radio-labeled LF1807 (SEQ ID NO: 309) and eitherunlabeled DREG56 or unlabeled LD201T1 can serve to determine ifthe 2'-F ligands also bind the lectin domain ofpurified LS-Rg. In these experiments, both DREG56 andLD201T1 gave concentration dependent inhibition ofLF1807 binding. Complete inhibition was attained with 300 nM Mab and 1 μM LD201T1. The competitors' affinities ofLS-Rg, calculated from the competition curves, were in good agreement with their known affinities. These results are consistent with the premise that LF1807, NX280 and DREG56 have the same or overlapping binding sites and consequently it is expected that 2'-F ligands will be antagonists ofL- selectin mediated adhesion. These results also reaffirm the proposition that the

SELEX protocol, with 5 mM elution ofbound oligonucleotides, preferentially elutes ligands bound at or near the lectin domain's bound calcium. Example 26

ssDNA Ligands to Human P-Selectin

PS-Rg is a chimeric protein in which the lectin, EGF, and the first two CRD domains ofhuman P-selectin arejoined to the Fc domain ofahuman G1

immunoglobulin (R.M. Nelson et al., 1993, supra). Purified chimera is provided by A.Varki. Soluble P-selectin is purchased fromR&D Systems. Unless otherwise indicated, all materials used in the ssDNA SELEX against the P- selectin/IgG, chimera, PS-Rg, are identical to those ofExamples 7 and 13.

The SELEX procedure is described in detail in United States Patent

5,270,163. The specific strategies andprocedures for evolving high affinity ssDNA antagonists to P-selectin are described in Examples 7 and 13.

Example 27

2'-F RNA Ligands to Human P-Selectin

The Experimental procedures outlined in this Example were usedto identify 2'-F RNA ligands to human P-selectin as described in Examples 28-34.

ExperimentalProcedures

A)Materials

PS-Rg is a chimeric protein in which the extracellular domain ofhuman P- selectin isjoined to the Fc domain ofa human G2 immunoglobulin (Norgard et al., 1993, PNAS 90:1068-1072). ES-Rg and CD22β-Rg are analogous constructs ofE- selectin and CD22βjoined to a human Gl immunoglobulinFc domain (R.M.

Nelson et al., 1993, supra; I. Stamenkovic et al., 1991, Cell 66, 1133-1144) while LS-Rg has L-selectinjoined to an IgG2 Fc domain. Purified chimerawere provided by A.Varki. Soluble P-selectin was purchased from R&D Systems. Protein A Sepharose 4 Fast Flow beads were purchased from Pharmacia Biotech. Anti-P- selectin monoclonal antibodies: G1 was obtained from Centocor. The 2'- F modified CTP and UTP were prepared according to Pieken et. al. (1991, Science 253:314-317). DNA oligonucleotides were synthesized by Operon. All other reagents and chemicals were purchased from commercial sources. Unless otherwise indicated, experiments utilizedHSMCbuffer (1 mM CaCl2, 1 mM MgCl₂, 150 mM NaCl, 20.0 mM HEPES, pH 7.4).

B) SELEX

The SELEX procedure is described in detail in United States Patent

5,270,163 and elsewhere. The nucleotide sequence ofthe synthetic DNA template for the PS-Rg SELEX was randomized at 50 positions. This variable region was flanked by N85' and 3' fixed regions. The transcript 50N8 has the sequence 5' gggagacaagaauaaacgcucaa-50N-uucgacaggaggcucacaacaggc 3' (SEQ ED NO: 390). All C and U have 2'-F substituted for 2'-OH on the ribose. The primers for the PCR were the following:

N85' Primer 5' taatacgactcactatagggagacaagaataaacgctcaa 3' (SEQ ED NO:

197)

N83' Primer 5' gcctgttgtgagcctcctgtcgaa 3' (SEQ ID NO: 198)

The fixedregions includeprimerannealing sites forPCR and cDNA synthesis as well as aconsensus T7 promoter to allow in vitro transcription. The initial RNA pool was made by first Klenow extending 1 nmol ofsynthetic single stranded DNA and then transcribing the resulting double stranded molecules with T7 RNA polymerase. Klenow extension conditions: 3.5 nmols primer 5N8, 1.4 nmols 40N8, IX Klenow Buffer, 0.4 mM each ofdATP, dCTP, dGTP and dTTP in a reaction volume of 1 ml.

For subsequent rounds, eluted RNA was the template for AMV reverse transcriptase mediated synthesis ofsingle stranded cDNA. These single-stranded DNA molecules were converted into double-stranded transcription templates by PCR amplification. PCR conditions were 50 mM KCl, 10 mM Tris-Cl, pH 8.3, 7.5 mM MgCl2, 1 mM ofeach dATP, dCTP, dGTP, and dTTP, and 25 U/ml ofTaq DNA polymerase. Transcription reactions contained 0.5 mM DNA template, 200 nM T7 RNA polymerase, 40 mM Tris-HCl (pH 8.0), 12 mM MgCl2, 5 mM DTT, 1 mM spermidine, 4% PEG 8000, 1 mM each of2'-OH ATP and 2'-OH GTP, 3.3 mM each of 2'-F CTP and 2'-F UTP, and 250 nM α-³²P 2'-OH ATP.

The strategy for partitioning PS-Rg/RNA complexes from unbound RNA is essentially identical to the strategy detailed in Example 7 for ligands to L-selectin (Table 18).

In the initial SELEX rounds, which were performed at 37 °C, the density of immobilized PS-Rg was 20 pmols/μl ofProtein A Sepharose 4 Fast Flow beads. In later rounds, the density ofPS-Rg was reduced (Table 18), as needed, to increase the stringency ofselection. Beginning with the second round, SELEX was often done at more than one PS-Rg density. At each round, the eluted material from only one PS-Rg density was carried forward.

Before each round, RNA was batch adsorbed to 100 μl ofprotein A sepharose beads for 1 hourin a 2 ml siliconized column. Unbound RNA and RNA eluted with minimal washing (two volumes) were combined and used for SELEX input material. For SELEX, extensively washed, immobilized PS-Rg was batch incubated with pre-adsorbed RNA for 0.5 to 1 hours in a 2 ml siliconized column with frequent mixing. Unbound RNA was removed by extensive batch washing (500 μl HSMC/wash). Bound RNA was eluted as two fractions; first, bound RNA was eluted by incubating and washing columns with 5 mM EDTA in HSMC without divalent cations; second, the remaining elutable RNA was removedby incubating and/or washing with 50 mM EDTA in HSMC without divalents. The percentage ofinput RNA that was eluted is recorded in Table 18. In every round, an equal volume of protein A sepharose beads without PS-Rg was treated identically to the SELEX beads to determine background binding. All unadsorbed, wash and eluted fractions were counted in a Beckman LS6500 scintillation counter in order to monitor each round ofSELEX.

The eluted fractions were processed for use in the foUowing round (Table 18). Afterprecipitating with 300 mM Sodium Acetate pH 7 in ethanol (2.5 volumes), the RNA was resuspended in 80 μl ofH₂O and 40 μl were reverse transcribed into cDNAby AMV reverse transcriptase at48°C for 30 minutes, in 50 mM Tris-Cl pH (8.3), 60 mM NaCl, 6 mM Mg(OAc)₂, 10 mM DTT, 200 pmol DNA primer, 0.4 mM each ofdNTPs, and 0.4 unit/μl AMV RT. Transcripts ofthe PCRproduct were used to initiate the next round ofSELEX. C) Nitrocellulose FilterBinding Assay

As described in SELEX Patent Applications, a nitrocellulose filter partitioning method was used to determine the affinity ofRNA ligands for PS-Rg and forotherproteins. FUter discs (nitrocellulose/cellulose acetate mixed matrix, 0.45 μm pore size, Millipore) were placed on a vacuum manifold and washed with 2 ml ofHSMC buffer under vacuum. Reaction mixtures, containing ³²P labeled

RNA pools and unlabeled PS-Rg, were incubated in HSMC for 10 - 20 min at 4°C, roomtemperature or 37 °C, filtered, and then immediately washed with 4 ml HSMC at the same temperature. The filters were air-dried and counted in aBeckman LS6500 liquid scintillation counterwithout fluor.

PS-Rg is a dimeric protein that is the expression product ofa recombinant gene constructed by fusing the DNA sequence that encodes the extracellular domains ofhuman P-selectin to the DNA that encodes a human IgG₁ Fc region. For affinity calculations, one ligand binding site per PS-Rg monomer (two per dimer) were assumed. The monomer concentration is defined as 2 times the PS-Rg dimer concentration. The equilibrium dissociation constant, Kd, for an RNA pool or specific ligand is calculated as described in Example 7, paragraph C.

D) Cloning and Sequencing

Twelfth round PCRproducts were re-amplified with primers which contain either aBamHl or aHinDIII restriction endonuclease recognition site. Using these restriction sites, the DNA sequences were inserted directionally into the pUC9 vector. These recombinant plasmids were transformed into E. coli strain JM109 (Life Technologies, Gaithersburg, MD). Plasmid DNA was prepared according to the alkaline hydrolysis method (PERFECTprep, 5'-3', Boulder, CO).

Approximately 50 clones were sequencedusing the Sequenase protocol (Amersham, Arlington Heights, IL). The resulting ligand sequences are shown in Table 19.

E) Boundary Experiments

The minimal high affinity sequence ofindividual ligands was determinedby boundary experiments (Tuerk et. al.1990, J. Mol. Biol.213: 749). Individual

RNA ligands, ³²P-labeled at the 5'-end for the 3' boundary and ³²P-labeled at the 3'-end for the 5' boundary, are hydrolyzed in 50 mM Na2CO₃ pH 9 for 8 minutes at 95°C. The resulting partial hydrolysate contains apopulation ofend-labeled molecules whose hydrolyzedends correspond to each ofthe purine positions in the full length molecule. The hydrolysate is incubated with PS-Rg (atconcentrations 5- fold above, below and atthe measured Kd for the ligand). The RNA concentration is significantly lower than the Kd. The reaction is incubated atroom temperature for 30 minutes, filtered, and then immediately washed with 5 ml HSMC at the same temperature. The bound RNA is extracted from the filter and then electrophoresed on an 8% denaturing gel adjacent to hydrolyzed RNA whichhas not been incubated with PS-Rg. Analysis is as described in Tuerk et. al.1990, J. Mol. Biol.213: 749.

F) 2'-O-Methyl Substitution Experiments

In orderto decrease the susceptibility ofthe 2'-F pyrimidine RNA ligands to nuclease digestion, post-SELEX modification experiments were performedto identify 2'-OH purines that are replaceable with 2'-OMe purines without loss of affinity as described in Green et. al. (1995, J. Mol. Biol.247: 60-68). Briefly, seven oligonucleotides were synthesized, each with three mixed positions. A mixed position is defined as a 2'-OH purine nucleotide within the RNA which has been synthesized with 2:1 ratio of2'-OH:2'-OMe. Since the coupling efficiency of2'- OH phosphoramidites is lower than that of2'-OMes, the resulting RNA has 25-50

% 2'-OH at each mixed position.³²P end-labeled RNA ligands are then incubated with concentrations ofPS-Rg 2-fold above and 2.5-foldbelow the Kd ofthe unmodified ligandat roomtemperature for 30 minutes, filtered, and then immediately washed with 5 ml HSMC at the same temperature. The bound RNA (Selected RNA) is extracted from the filter and then hydrolyzed with 50 mM

Na2CO3 pH 9 for 8 minutes at 95°C in paraUel with RNA which has not been exposed to binding and filtration (Unselected RNA). The Selected RNA is then electrophoresed on a20% denaturing gel adjacent to Unselected RNA.

To determine the affect on binding affinity of2'-OMe substitution at a particularposition, the ratio ofintensities ofthe Unselected:Selectedbands that correspondto the position in question are calculated. The Unselected:Selectedratio when the position is mixed is compared to the mean ratio forthatposition from experiments inwhich,the position is notmixed. Ifthe Unselected:Selectedratio of the mixedposition is significantly greater than that when theposition is not mixed, 2'-OMe may increase affinity. Conversely, ifthe ratio is significantly less, 2'-OMe may decrease affinity. Ifthe ratios are not significantly different, 2'-OMe substitution has no affect.

G) Cell Binding Studies

The ability ofevolved ligandpools and cloned ligands tobind to P-selectin presented in the context ofa cell surface was tested in experiments with human platelet suspensions. Whole blood from normal volunteers was collected in

Vacutainer 6457 tubes. Within 5 minutes ofcollection, 485 μl ofblood was stimulated with 15 μl Bio/DataTHROMBINEX for 5 minutes at room temperature. A 100 μl aliquot ofstimulatedblood was transferred to 1 ml ofBB- (140 mM NaCl,

20 mM HEPES pH 7.35, 5 mM KCl, 0.01% NaN₃) at 4°C and spun at 735 x g for 5 minutes. This step was repeated and the resulting pellet was re-suspended in 1 ml ofBB+ (140 mM NaCl, 20 mM HEPES pH 7.35, 5 mM KCl, 0.01% NaNs, 1 mM

CaCl₂, 1 mM MgCl₂) at 4°C.

To detect antigen expression, 15 μl BB+ containing FTTC conjugated anti-

CD61 orPE conjugated anti-CD62 antibody (Becton Dickinson) was incubated for

20-30 minutes at4°C with 10 μl ofplatelet suspension. This was diluted to 200 μl with 4°C BB+ and analyzed on aBecton Dickinson FACSCaHberusing 488 nm excitation and FL1 (530 nm emission) or FL2 (580 nm emission) with the machine live gated on platelets. Between 1000 and 5000 events in this gate were recorded.

To detect oligonucleotide ligand binding, 15 μl BB+ containing ligand conjugated to eitherFTEC orbiotin was incubated 20-30 minutes at4°C with 10 μl platelet suspension. The FTTC-ligand incubations were diluted to 200 μl withBB+ and analyzed on a FACSCaUberflow cytometer. The biotinylated-ligandreactions were incubated with streptavidin-phycoerythrin (SA-PE) (Becton Dickinson) for 20 minutes at4°C, before dilution and analysis. Wash steps with 500 μl BB+ and 700 x g spins have been used without compromising the quality ofthe results.

The specificity ofbinding to P-selectin (CD62P) expressed on platelets was testedby competition withthe P-selectin specific blocking monoclonal antibody, Gl. Saturability ofbinding was testedby self-competition withunlabeled RNA.

H) Inhibition ofSelectin Binding to sialyl-Lewis^x

The ability ofevolved RNA pools or cloned ligands to inhibit the binding of PS-Rg to sialyl-Lewis^x was tested in competitive ELISA assays (C. Foxall et al., 1992, supra). Forthese assays, the wells ofCorning (25801) 96 well microtiter plates were coatedwith 100 ng ofa sialyl-Lewis^x/BSA conjugate, airdried overnight, washed with 300 μl ofPBS(-) and then blocked with 1% BSA in HSMC for 60 min at room temperature. RNA ligands were incubated with PS-Rg in HSMC/1% BSA at room temperature for 15 min. Afterremoval ofthe blocking solution, 50 μl ofPS-Rg (lOnM) or a PS-Rg (10nM)/RNA ligand mix was addedto the coated, blocked wells and incubated at room temperature for 60 minutes. The binding solution was removed, wells were washed with 300 μl ofPBS(-) and then probed with HRP conjugated anti-human IgG, at room temperature to quantitate PS- Rg binding. After a 30 minute incubation atroomtemperature in the dark with OPD peroxidase substrate (Sigma P9187), the extent ofPS-Rg binding and percent inhibition was determined from the OD450.

Example 28

2'-F RNA Ligands to Human P-selectin

A. SELEX

The starting RNA pool for SELEX, randomized 50N8 (SEQ ID NO: 390), contained approximately 10¹⁵ molecules (1 nmol RNA). The SELEXprotocol is outlinedin Table 18. The dissociation constant ofrandomized RNA to PS-Rg is estimated tobe approximately 2.5 μM. An eight-fold difference was observed in the RNA elution profiles with 5 mM EDTA from SELEX and background beads for rounds 1 and 2, while the 50 mM elution produced a 30-40 fold excess over background Table 18. For rounds 1 through 3, the 5 mM and 50 mM eluted RNAs were pooled and processed for the next round. Beginning with round 4, only the 5 mM eluate was processed for the following round. To increase the stringency of selection, the density ofimmobilized PS-Rg was reduced five fold in round 2 and again in round three withoutgreatly reducing the fraction eluted fromthe column. The density ofimmobilized PS-Rg was further reduced 1.6-fold in round 4 and remained atthis density until round 8, with furtherreductions in protein density at laterrounds. The affinity ofthe selected pools rapidly increased and the pools gradually evolvedbiphasic binding characteristics.

Binding experiments with 12th roundRNA revealed that the affinity ofthe evolving pool for P-selectin was not temperature sensitive. Bulk sequencing of2nd, 6th, 11th and 12th RNA pools revealed noticeable non-randomnessby round twelve. The 6th round RNA bound monophasicaUy at 37°C with adissociation constant ofapproximately 85 nM, while the 11th and 12throundRNAs bound biphasicaUy with high affinity Kds ofapproximately 100 and 20 pM, respectively. The binding ofall tested pools required divalent cations. In the absence ofdivalent cations, the Kds ofthe 12th round pools increased to > 10 nM. (HSMC, minus Ca⁺⁺ /Mg⁺⁺ , plus 2 mM EDTA). The 12th round pool showed high specificity for PS-Rg with measured Kd's of 1.2 μM and 4.9 μM for ES-Rg and LS-Rg, respectively.

B. RNA Sequences

In Table 19, ligand sequences are shown in standard single letter code (Cornish-Bowden, 1985 NAR 13: 3021-3030). Fixed region sequence is shown in lower case letters. By definition, each clone includes both the evolved sequence and the associated fixed region, unless specifically stated otherwise. Fromthe twelfth round, 21 of44 sequenced ligands were unique. A unique sequence is

operationally defined as one that differs from all others by three ormore nucleotides. Sequences that were isolated more than once, are indicated by the parenthetical number, (n), following the ligand isolate number. These clones fall into five sequence families (1-5) and a group oftwo unrelated sequences (Orphans)(SEQ ED NOs: 199-219).

Family 1 is defined by 23 ligands from 13 independent lineages. The consensus sequence is composed oftwo variably spaced sequences, CUCAACGAMC and CGCGAG (Table 19). In 11 of 13 ligands the CUCAA of the consensus is from 5' fixed sequence which consequently minimizes variability and in turn reduces confidence in interpreting the importance ofCUCAA orthe paired GAG (see Example 27).

Families 2-5 are each represented by multiple isolates ofa single sequence which precludes determination ofconsensus sequences.

D. Affinities

The dissociation constants forrepresentative ligands, including all orphans, were determined by nitrocellulose filterbinding experiments and are Hsted in Table 20. These calculations assume two binding sites per chimera. The affinity of randomRNA is estimated to be approximately 2.5 μM.

In general, ligands bind monophasically with dissociation constants ranging from 15 pM to 450 pM at 37 °C. Some ofthe highest affinity ligands bind biphasicaUy. FuU length ligands offamiHes 1-4 show no temperature dependence. The observed affinities substantiate the proposition that it is possible to isolate oHgonucleotide ligands witii affinities that are several orders ofmagnitude greater than that ofcarbohydrate ligands.

Example 29

Specificity of2'-F RNA Ligands

The affinity ofP-selectin ligands to ES-Rg, LS-Rg and CD22β-Rg were determined by nitrocellulose partitioning. As indicated in Table 20, the ligands are highly specific for P-selectin. In general, a ligand's affinity for ES-Rg and LS-Rg is atleast 10⁴-fold lower than for PS-Rg. Binding above background is not observed for CD22β-Rg at the highest protein concentration tested (660 nM), indicating that ligands do notbind the Fc domain ofthe chimeric constructs nor do they have affinity forthe sialic acidbinding site ofthis unrelated lectin. The specificity of oligonucleotideligand binding contrasts sharply with the binding ofcognate carbohydrates by the selectins and confirms the propositionthat SELEX ligands will have greater specificity than carbohydrate ligands.

Example 30

Inhibition ofBinding to sialyl-Lewis^x

OHgonucleotide ligands, eluted by 2-5 mM EDTA, are expected to derive part oftheirbinding energy from contacts with the lectin domain's bound Ca⁺⁺ and consequently, are expected to compete with sialyl-Lewis^x forbinding. In competition assays, the selected oligonucleotide ligands competitively inhibit PS-Rg binding to immobilized sialyl-Lewis^x withIC50s ranging from 1 to 4 nM (Table 20). Specifically, ligandPF377 (SEQ ID NO: 206) has an IC50 ofapproximately 2 nM. Complete inhibition is attained at 10 nM ligand. This result is typical ofhigh affinity ligands and is reasonable under the experimental conditions. The IC50s of ligands whose Kds are much lower than the PS-Rg concentration (10 nM) are Umitedbythe proteinconcentration and are expected tobe approximately one half the PS-Rg concentration. The specificity ofcompetition is demonstratedby the inability ofround2 RNA (Kd~ 1 μM) to inhibit PS-Rg binding to immobilized sialyl-Lewis^x. These data verify that 2'-F RNA ligands are functional antagonists of PS-Rg.

Example 31

Secondary Structure ofHigh AffinityLigands

In favorable instances, comparative analysis ofaHgned sequences allows deduction ofsecondary structure and structure-function relationships. Ifthe nucleotides at two positions in a sequence covary according to Watson-Crickbase pairing rules, then the nucleotides at these positions are apt to be paired.

Nonconserved sequences, especially those that vary in length are not apt to be directly involved in function, while highly conserved sequences are likely to be directly involved.

Comparative analysis ofthe family 1 alignment suggests ahairpin motif, the stem ofwhich contains three asymmetrical internal loops (Figure 16). In the figure, consensus positions are specified, with invariant nucleotides in bold type. To the right ofthe stem is a matrix showing the number ofoccurrences ofparticularbase pairs forthe positions in the stemthat are on the same line. The matrix shows that 6 ofthe stem's 9 base pairs are supportedby Watson-Crickcovariation. Portions of the two consensus motifs, CUC and GAG, form the terminus ofthe stem.

Conclusions regarding a direct role ofthe terminus in binding are tempered by the use offixed sequence (11 of 13 ligands) which limits variability. The variability of the loop's sequence and length suggests that it is not directly involved in binding. This conclusion is reenforced by ligand PF422 (SEQ ID NO: 202) which is a circularpermutation ofthe consensus motif. Although the loop that connects the stem's two halves is at the opposite end relative to other ligands, PF422binds with high (Kd = 172 pM; Table 21) affinity. Example 32

Boundary Experiments

Boundary experiments were performed on anumber ofP-selectin ligands as described in Example 27 and the results are shown in Table 21. The results for family 1 ligands are consistent with their proposed secondary structure. The composite boundary species vary in size from 38-90 nucleotides, but are 40-45 nucleotides in family 1. Affinities ofthese truncated ligands are shown in Table 22.

In general, the truncates lose no more than 10-fold in affinity in comparison to the full length, effectively inhibit the binding ofPS-Rg to sialyl-Lewis^x and maintain binding specificity for PS-Rg (Table 22). These data validate the boundary method foridentifying the minimalhigh affinity binding element ofthe RNAligands.

Example 33

Binding of2'-F RNA Ligands to Human Platelets

Since the P-selectin ligands were isolated againstpurifiedprotein, their abiUty to bind P-selectin presented in the context ofa cell surface was determined in flow cytometry experiments with activated human platelets. Platelets were gatedby side scatter and CD61 expression. CD61 is a constitutively expressed antigen on the surface ofboth resting and activatedplatelets. The expression ofP-selectin was monitored with anti-CD62P monoclonal antibody (Becton Dickinson). The mean fluorescence intensity ofactivated platelets, stained with biotintylated-PF377sl (SEQ ID NO: 223)/SA-PE (Example 27, paragraph G), is 5 times greater than that ofsimilarly stainedresting platelets. In titration experiments, halfmaximal fluorescence occurs at approximately 50 pM PF377sl (EC50) which is consistent with its equUibrium dissociation constant, 60 pM, for PS-Rg. Binding to platelets is specific by the criterion that it is saturable. SaturabiUty has been demonstrated not only by titration but also by competition with unlabeled PF377sl.

Binding to platelets is P-selectin specific by the criteria that 1)

oligonucleotides that do not bind PS-Rg do not bind platelets; 2) that binding of PF377sl to platelets is divalent cation dependent; and most importantly 3) that binding is inhibitedby the anti-P-selectin adhesionblocking monoclonal antibody Gl, but notby an isotype control antibody. These data validate the feasibility of using immobiUzed, purified protein to isolate highly specific ligands against a cell surface P-selectin. Example 34

2'-O-Methyl Substitution Experiments

2'-OMe purine substitutions were performed on ligand PF377sl (SEQ ED

NO: 223) as described in Example 27 paragraph F and the results are shown in

Table 23. The data indicate that 2'-OMe purines at positions 7-9, 15, 27, 28 and 31 enhance binding while substitutions at positions 13, 14, 16, 18, 2122, 24, and 30 have little or no affect on affinity. Thus it appears that up to 15 positions may be substituted with only slight losses in affinity. In partial confirmation ofthis expectation, the affinity of377s1 simultaneously substituted with 2'-OMe purines at 11 positions (PF377M6, SEQ ID NO: 235) is 250 pM (Table 22).

Example 35

2'-NH₂ RNA Ligands to Human P-Selectin

The experimental procedures described in this Example are used in Examples

36-38 to isolate and characterize 2'-NH2 RNA ligands to human P-selectin.

Experimental Procedures

A)Materials

Unless otherwise indicated, all materials used in the 2'-NH₂ RNA SELEX against the P-selectin/IgG, chimera, PS-Rg, were identical to those ofExample 27. The 2'-NH₂ modified CTP and UTP were prepared according to Pieken et. al. (1991, Science 253:314-317). The buffer for SELEX experiments was 1 mM CaCl₂, 1 mM MgCl₂, 150 mM NaCl, 10.0 mM HEPES, pH 7.4.

B) SELEX

The SELEX procedure is described in detail in US patent 5,270,163 and elsewhere. The nucleotide sequence ofthe synthetic DNA template forthe PS-Rg SELEX was randomized at 50 positions. This variable region was flanked by N85' and 3' fixed regions. The transcript 50N8 has the sequence 5' gggagacaagaauaaac gcucaa-50N-uucgacaggaggcucacaacaggc 3' (SEQ ED NO: 248). All C and U have 2'-NH2 substituted for 2'-OH on the ribose. The primers for the PCR were the following:

N85' Primer 5' taatacgactcactatagggagacaagaataaacgctcaa 3' (SEQ ID NO: 249)

N83' Primer 5' gcctgttgtgagcctcctgtcgaa 3' (SEQ ED NO: 250). The procedures used to isolate 2'-NH₂ oligonucleotide ligands to P-selectin are identical to those described 2'-F ligands in Example 27, except that transcription reactions utilized 1 mM each, 2'-NH₂-CTP and 2'-NH₂-UTP, in place of3.3 mM each 2'-F- CTP and 2'-F-UTP.

C) Nitrocellulose Filter Binding Assay

As described in SELEX Patent Applications andin Example 27, paragraph C, a nitrocellulose filterpartitioning method was used to determine the affinity of RNA ligands for PS-Rg and for other proteins. Either a Gibco BRL 96 well manifold, as described inExample 23 or a 12 well MilHpore manifold (Example 7C) was used forthese experiments. Binding data were analyzed as described in Example 7, paragraph C.

D) Cloning and Sequencing

Twelfth round PCR products were re-amplified with primers which contain eitheraBamHl oraHinDEIIIrestriction endonuclease recognition site.

Approximately 75 ligands were cloned and sequenced using the procedures described in Example 7, paragraph D. The resulting sequences are shown in Table 25.

E) Cell Binding Studies

The abiUty ofevolved ligandpools tobind to P-selectin presented inthe context ofa cell surface was tested in flow cytometry experiments with human platelet suspensions as described in Example 7, paragraph E.

Example 36

2'-NΗ₂ RNA Ligands to Human P-Selectin

A. SELEX

The starting 2'-NH₂ RNA pool for SELEX, randomized 50N8 (SEQ ID

NO: 248), contained approximately 10¹⁵ molecules (1 nmol 2'-NH₂ RNA). The dissociation constant ofrandomized RNA to PS-Rg is estimated to be approximately 6.4 μM. The SELEX protocol is outlined in Table 24.

The initial round ofSELEX was performed at 37°C with an PS-Rg density of20 pmol/μl ofprotein A sepharose beads. Subsequent rounds were all at 37°C. In the firstround there was no signal above background for the 5 mM EDTA elution, whereas the 50 mM EDTA elution had a signal 7 fold above background, consequently, the two elutions were combined and processed for the next round. This scheme was continued through round 6. Starting with round seven only the 5 mM eluate was processed for the next round. To increase the stringency ofselection, the density ofimmobilized PS-Rg was reduced ten fold in round 6 with further reductions in protein density at later rounds. Under these conditions a rapid increase in the affinity ofthe selected pools was observed.

Binding experiments with 12th round RNA revealed that the affinity ofthe evolving pool for P-selectin was temperature sensitive despite performing the selection at 37°C, (Kds: 13 pM, 91 pM and 390 pM at 4°C, room temperature and 37 °C, respectively). Bulk sequencing ofRNA pools indicated dramatic non- randomness at round 10 with not many visible changes in round 12. Ligands were cloned and sequenced from round 12.

B.2'-NH2 RNA Sequences

In Table 25, the 2'-NH₂ RNA ligand sequences are shown in standard single letter code (Comish-Bowden, 1985 NAR 13: 3021-3030)(SEQ ID NOS: 251-290). The evolved randomregion is shown in upper case letters in Table 25. Any portion ofthe fixed region is shown in lowercase letters. By definition, each clone includes both the evolved sequence and the associated fixed region, unless specifically stated otherwise. From the twelfth round, 40/61 sequenced ligands were unique. A unique sequence is operationally defined as one that differs from all others by three or more nucleotides. Sequences that were isolated more than once are indicated by the parenthetical number, (n), following the ligand isolate number. Ligands from family 1 dominate the final pool containing 16/61 sequences, which are derived frommultiple lineages. Families 2 and 3 are represented by slight mutational variations ofa single sequence. Sequences labeled as "others" do not have any obvious similarities. Family 1 is characterized by the consensus sequence GGGAAGAAGAC (SEQ ID NO: 291).

C. Affinities

The dissociation constants ofrepresentative ligands are shown in Table 26. These calculations assume two RNA ligand binding sites per chimera. The affinity ofrandom 2-NH₂ RNA is estimated to be approximately 10 μM.

At 37°C, the dissociation constants range from 60 pM to 50 nM which is at least a 1x10³ to 1x10⁵ fold improvement over randomized 2'-NH₂ RNA (Table 26). There is amarked temperature sensitivity for Clone PA350 (SEQ ID NO:

252) with an increase in affinity of6 fold at 4°C (Table 26). The observed affinities ofthe evolved 2'-NH2 ligand pools reaffirm our proposition that it is possible to isolate oligonucleotide ligands with affinities that are several orders ofmagnitude greater than that ofcarbohydrate ligands.

Example 37

Specificity of2'-NH₂ RNA Ligands to P-Selectin The affinity ofclone PA350 (SEQ ID NO: 252) for LS-Rg and ES-Rg was determined by nitrocellulose partitioning and the results shown in Table 26. The ligands are highly specific for P-selectin. The affinity for ES-Rg is about 600-fold lower and that for LS-Rg is about 5x10⁵-fold less than for PS-Rg. Binding above background is not observed for CD22β-Rg indicating that ligands neitherbind the Fc domain ofthe chimeric constructs nor have affinity forunrelated siaUc acid binding sites.

The specificity ofoligonucleotide ligand binding contrasts sharply with the binding ofcognate carbohydrates by the selectins and reconfirms the proposition that SELEX ligands will have greater specificity than carbohydrate ligands.

Example 38

Cell Binding Studies

FITC-labeled ligand PA350 (FITC-350) (SEQ ID NO: 252) was tested for its abUity to bind to P-selectin presented in the context ofaplatelet cell surface by flow cytometry experiments as described in Example 23, paragraph G.

The specificity ofFITC-PA350 for binding to P-selectin was tested by competition experiments in which FTTC-PA350 and unlabeled blocking monoclonal antibody Gl were simultaneously added to stimulated platelets. Gl effectively competes with FTTC-PA350 for binding to platelets, while an isotype matched control has little orno effect which demonstrates that FTTC-PA350 specifically binds to P-selectin. The specificity ofbinding is further verified by the observation that oligonucleotide binding is saturable; binding of 10 nM FTTC-PA350 is inhibited by 200 nM unlabeled PA350. In addition, the binding of FTTC-PA350 is dependent on divalent cations; at 10 nM FTTC-PA350 activated platelets are not stained in excess ofautofluorescence in the presence of5 mM EDTA.

These data validate the feasibility ofusing immobilized, purifiedprotein to isolate ligands against a cell surface protein and the binding specificity of2'-NH2 ligands to P-selectin in the context ofa cell surface. Example 39

Inhibition ofP-selectin Binding to Sialyl Lewis^x

In competition assays, ligands PA341 (SEQ ID NO: 251) and PA350 (SEQ ID NO: 252) competitively inhibit PS-Rg binding to immobUized sialyl-Lewis^x with IC50s ranging from 2 to 5 nM (Table 26). This result is typical ofhigh affinity ligands andis reasonable underthe experimental conditions. The IC50s ofligands whose Kds are much lower than the PS-Rg concentration (10 nM) are limited by the protein concentration and are expected to be approximately one halfthe PS-Rg concentration. The specificity ofcompetitionis demonstrated by the inability of round 2 RNA (Kd- 1 μM) to inhibit PS-Rg binding to immobilized sialyl-Lewis^x. These dataverify that 2-NH₂ RNA ligands are functional antagonists ofP-selectin.

Example40

2'-NH₂ RNA Ligands to Human E-Selectin

ES-Rg is a chimeric protein in which the extracellular domain ofhuman E- selectin isjoined to the Fc domain ofahuman Gl immunoglobulin (R.M. Nelson et al., 1993, supra). Purified chimera were provided by A.Varki. Unless otherwise indicated, all materials used in this SELEX are similar to those ofExamples 7 and 13.

The SELEX procedure is described in detail in US patent 5,270,163 and elsewhere. The rationale and experimental procedures are the same as those described in Examples 7 and 13.

Claims

We Claim:

1. A method for identifying nucleic acid ligands and nucleic acid ligand sequences to a lectin comprising:

a) contacting a candidate mixture of nucleic acids with a lectin, wherein nucleic acids having an increased affinity to said lectin relative to the candidate mixture may be partitioned from the remainder of the candidate mixture;

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 said lectin, whereby nucleic acid ligands to said lectin may be identified.

2. The method of Claim 1 further comprising:

d) repeating steps a), b) and c).

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

4. The method of Claim 3 wherein said single-stranded nucleic acids are ribonucleic acids.

5. The method of Claim 4 wherein said nucleic acids comprise modified ribonucleic acids.

6. The method of Claim 5 wherein said nucleic acids comprise modified ribonucleic acids selected from the group consisting of 2'-amino (2'- NH2) modified ribonucleic acids and 2'-fluoro (2'-F) modified ribonucleic acids.

7. The method of Claim 3 wherein said single-stranded nucleic acids are deoxyribonucleic acids.

8. The method of Claim 2 further comprising

e) forming a multivalent Complex comprising two nucleic acid ligands identified in step c).

9. The method of Claim 5 further comprising

e) substituting 2'-O-methyl ribonucleic acids for 2'-OH ribonucleic acids in the nucleic acid ligands identified in step c).

10. The method of Claim 1 wherein said lectin is selected from the group consisting of a mammalian lectin, a plant lectin, a microbial lectin and a viral lectin.

11. The method of Claim 1 wherein said lectin is wheat germ agglutinin.

12. The method of Claim 1 wherein said lectin is a selectin.

13. The method of Claim 12 wherein said selectin is selected from the group consisting of L-selectin, E-selectin, and P-selectin.

14. The method of Claim 1 wherein said lectin is serum mannose binding protein.

15. A purified and isolated non-naturally occurring nucleic acid ligand to a lectin.

16. The nucleic acid ligand of Claim 15 which is a non-naturally occurring nucleic acid ligand having a specific binding affinity for said lectin, such lectin being a three dimensional chemical structure other than a polynucleotide that binds to said nucleic acid ligand through a mechanism which predominantly depends on Watson/Crick base pairing or triple helix binding, wherein said nucleic acid ligand is not a nucleic acid having the known physiological function of being bound by said lectin.

17. The nucleic acid ligand of Claim 15 wherein said lectin is selected from the group consisting of a mammalian lectin, a plant lectin, a microbial lectin and a viral lectin.

18. The nucleic acid hgand of Claim 15 wherein said lectin is selected from the group consisting of wheat germ agglutinin, L-selectin, E-selectin and P- selectin.

19. The nucleic acid ligand of Claim 15 wherein said lectin is wheat germ agglutinin.

20. The nucleic acid ligand to wheat germ agglutinin of Claim 19 wherein said nucleic acid ligand is a ribonucleic acid (RNA) ligand.

21. The nucleic acid ligand of Claim 20 which comprises a modified ribonucleic acid.

22. The nucleic acid ligand of Claim 21 wherein said modified ribonucleic acid is a 2'-amino (NH₂) modified ribonucleic acid.

23. The nucleic acid ligand to wheat germ agglutinin of Claim 22 wherein said ligand is an RNA ligand selected from the group consisting of the nucleotide sequences set forth in Table 2.

24. The nucleic acid ligand of Claim 23 wherein said ligand is selected from the group consisting of SEQ ID NOS: 4-55.

25. The nucleic acid ligand of Claim 20 wherein said ligand comprises sequences selected from the group consisting of SEQ ID NOS: 56-63.

26. The nucleic acid ligand to wheat germ agglutinin of Claim 19 wherein said ligand is substantially homologous to and has substantially the same ability to bind said wheat germ agglutinin as a ligand selected from the group consisting of the sequences set forth in Table 2.

27. The nucleic acid ligand to wheat germ agglutinin of Claim 19 wherein said ligand has substantially the same structure and the same ability to bind said wheat germ agglutinin as a ligand selected from the group consisting of the sequences set forth in Table 2.

28. The nucleic acid ligand of Claim 15 wherein said lectin is a selectin.

29. The nucleic acid ligand of Claim 28 wherein said selectin is selected from the group consisting of L-selectin, E-selectin and P-selectin.

30. The nucleic acid ligand of Claim 29 wherein said selectin is L- selectin.

31. The nucleic acid Hgand to L-selectin of Claim 30 wherein said nucleic acidligand is ribonucleic acid (RNA) Hgand.

32. The nucleic acid ligand of Claim 31 which comprises a modified ribonucleic acid.

33. The nucleic acid ligand of Claim 32 wherein said modified ribonucleic acid is selected from the group consisting of a 2'-amino (2'-NH₂) modified ribonucleic acid and a 2'-fluoro (2'-F) modified ribonucleic acid.

34. The nucleic acid ligand to L-selectin of Claim 33 wherein said ligand is an RNA ligand selected from the group consisting of the nucleotide sequences set forth in Tables 8 and 16.

35. The nucleic acidligand of Claim 34 wherein said Hgand is selected from the group consisting of SEQ ID NOS: 67-117 and 293-388.

36. The nucleic acid ligand of Claim 31 wherein said ligand comprises sequences selected from the group consisting of SEQ ID NOS: 118-125.

37. The nucleic acid ligand to L-selectin of Claim 30 wherein said ligand is substantially homologous to and has substantially the same ability to bind said L- selectin as a ligand selected from the group consisting of the sequences set forth in Tables 8, 12 and 16.

38. The nucleic acid ligand to L-selectin of Claim 30 wherein said ligand has substantiaUy the same structure and the same ability to bind said L-selectin as a ligand selected from the group consisting of the sequences set forth in Tables 8, 12 and 16.

39. The nucleic acid ligand to L-selectin of Claim 30 wherein said nucleic acid ligand is deoxyribonucleic acid (DNA).

40. The nucleic acid ligand to L-selectin of Claim 39 wherein said ligand is an DNA Hgand selected from the group consisting of the nucleotide sequences set forth in Table 12.

41. The nucleic acid ligand of Claim 40 wherein said Hgand is selected from the group consisting of SEQ ID NOS: 129-180 and 185-196.

42. The nucleic acid ligand of Claim 39 wherein said ligand comprises sequences selected from the group consisting of SEQ ID NOS : 181 - 184.

43. The nucleic acid Hgand of Claim 29 wherein said selectin is P- selectin.

44. The nucleic acid ligand to P-selectin of Claim 43 wherein said nucleic acid ligand is ribonucleic acid (RNA) ligand.

45. The nucleic acid ligand of Claim 44 which comprises a modified ribonucleic acid.

46. The nucleic acid ligand of Claim 45 wherein said modified ribonucleic acid is selected from the group consisting of a 2'-amino (2'-NH₂) modified ribonucleic acid, a 2'-fluoro (2'-F) modified ribonucleic acid, and a 2'-O- Methyl (2'-O-Me) modified ribonucleic acid.

47. The nucleic acid ligand to P-selectin of Claim 46 wherein said Hgand is an RNA ligand selected from the group consisting of the nucleotide sequences set forth in Tables 19, 21 and 25.

48. The nucleic acid ligand of Claim 47 wherein said ligand is selected from the group consisting of SEQ ID NOS: 199-219 and 236-290.

49. The nucleic acid ligand of Claim 44 wherein said ligand comprises sequences selected from the group consisting of SEQ ID NO: 291.

50. The nucleic acid ligand to P-selectin of Claim 43 wherein said ligand is substantially homologous to and has substantially the same ability to bind said P- selectin as a Hgand selected from the group consisting of the sequences set forth in Tables 19, 21 and 25.

51. The nucleic acid ligand to P-selectin of Claim 43 wherein said ligand has substantially the same structure and the same ability to bind said P-selectin as a ligand selected from the group consisting of the sequences set forth in Tables 19, 21 and 25.

52. The nucleic acid ligand to P-selectin of Claim 46 wherein said nucleic acid ligand is deoxyribonucleic acid (DNA).

53. The nucleic acid ligand of Claim 15 wherein said ligand is a ribonucleic acid ligand.

54. The nucleic acid ligand of Claim 53 which comprises a modified ribonucleic acid.

55. The nucleic acid ligand of Claim 54 wherein said modified ribonucleic acid is selected from the group consisting of 2'-amino (2'-NH₂) modified ribonucleic acids, 2'-fluoro (2'-F) modified ribonucleic acids and 2'-O- Methyl (2'-O-Me) modified ribonucleic acids.

56. The nucleic acid ligand of Claim 15 wherein said ligand is a deoxyribonucleic acid.

57. The nucleic acid ligand of Claim 15 wherein said ligand has been further chemically modified at the sugar and/or phosphate and/or base.

58. A multivalent Complex comprising a plurality ofligands of Claim 15.

59. A nucleic acid ligand to a lectin identified according to the method comprising:

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 said lectin, whereby nucleic acid ligands of said lectin may be identified.

60. A method for treating a lectin-mediated disease comprising administering a pharmaceutically effective amount of a nucleic acid ligand to a lectin.

61. The method of Claim 60 wherein said nucleic acid ligand to a lectin is identified according to the method of Claim 1.

62. The method of Claim 60 wherein said lectin is a selectin.

63. The method of Claim 62 wherein said selectin is L-selectin.

64. The method of Claim 62 wherein said selectin is P-selectin.