WO1995000668A1 - Method for selection of targets and nucleic acids which modulate target activity - Google Patents

Abstract

Methods for (i) identifying target molecules potentially susceptible to activity modulation upon binding with a nucleic acid sequence and (ii) isolating nucleic acid sequences selected for optimal target binding and target activity modulation potential.

Description

METHOD FOR SELECTION OF TARGETS AND NUCLEIC ACIDS WHICH MODULATE TARGET ACTIVITY

Bac σround of the Invention

Numerous proteins are known to bind nucleic acids with high affinity and, in some cases, high specificity. For example, enzymes which are involved in nucleic acid metabolism, such as polymerases, transcriptases, ligases and nucleases, interact with nucleic acids. With these enzymes limited degree of specificity is demonstrated by their ^ε' ability to discriminate between RNA and DNA. For example, RNases generally degrade RNA but not DNA. Additional specificity is present in the function among enzymes of a given class: e.g. endo-RNases cut RNAs internally, whereas exo-RNases remove• terminal ribonucleotides from an RNA; certain RNases will cut at single-stranded but not double- stranded sites whereas other RNases will have the converse specificity. In some cases, specificity of action, at least in vitro, can be compromised by assay components: mitochondrial DNA polymerase- , for example, usually prefer DNA to RNA as a template but can under certain assay conditions use RNA as a template and act as a reverse transcriptase. Many viral structural proteins exhibit specificity for the nucleic acids they will envelop. Not only will these proteins discriminate between RNA and DNA, but often a particular nucleic acid sequence is recognized to which the structural proteins will bind in the process of generating an intact virus particle.

There are large numbers of proteins which can influence selected gene expression qualitatively or quantitatively. These regulatory proteins appear to bind specifically to their cognate nucleic acid sequences. Often there is a family of related sequences to which a regulatory protein can bind. For example, the tumor suppressor protein p53 binds sequences comprised of two copies of a ten-base element, (RRRCATGYYY)₂, where R indicates a purine base, either A or G, and Y symbolizes a pyrimidine base, either T or C. The sequence requirement for p53 binding to DNA is not absolute since occasional substitutions can be tolerated.

Historically, specific nucleic acid sequences that are recognized by proteins have been determined by cataloguing the sequences known to interact with the protein and identifying their common features, or alternatively, by taking a sequence known to bind to a protein, mutating it in a variety of ways and determining the consequences on binding. This approach has been quite successful in identifying a variety of important sequence elements recognized by individual proteins or intracellular assemblies of components.

A conceptual and practical advance in the methodology to identify nucleic acid elements recognized by proteins came from the work of Oliphant and Struhl, Nucleic Acid Res., 16.:7673-7683 (1988). These authors devised a method that could be used for identifying nucleic acid sequence elements interacting with an intracellular target in the absence of prior information about the sequences to which the target molecules bound. Building on earlier work relating to a method for cloning random sequence oligonucleotides, (Oliphant, Nussbau and Struhl, .Gene, _ 4:177-183, 1986), they identified two important sequence elements required for gene transcription in prokaryotes. Their method involved an in vivo selection of cloned random sequence oligonucleotides based on their ability to promote gene transcription (i.e. interact with RNA polymerase complex) . Because this work involved a selection in vivo, there were practical limitations on the amount of sequence complexity (i.e. the total number of different sequences) that could be examined. However, by fixing the sequences resulting from the initial selection and then repeating the selection with a new randomized region, the authors in fact sampled a much wider universe of sequence complexity and actually identified consensus sequences that bound to both the original site on the transcription complex and one that was removed from the first site.

Subsequent to the studies described above, two reports described experiments that extended the technology for identifying nucleic acids that bind proteins to in vitro methods with even greater power and flexibility. Kinzler and Vogelstein, Nucleic Acids Res., .17:3645-3653 (1989), described a technique known as "whole genome PCR" that could be used to isolate specific nucleic acid fragments binding to a target protein. Briefly, these authors sheared genomic DNA to generate a population of fragments of nearly random sequence, modified the ends of the fragments so that they could be amplified by polymerase chain reaction or cloned, and then performed a reiterative process of: a) binding to target protein; b) separating bound fragments from unbound fragments by immunoprecipitation; c) amplifying the bound fragments by PCR; and d) repeating the binding with the newly prepared, enriched population of fragments as the source of binding sequences. After multiple reiterations, the surviving (i.e. selected by binding ability) fragments were cloned and their sequences determined. Analysis of the common elements found in the selected sequences in much the same way as was done with the in vivo identified sequences catalogued by common function, allowed the authors to identify the features needed for binding to target protein. This study demonstrated that amplification of the product from one selective binding overcame the potential limitation of recovering only very small amounts of bound nucleic acid and allowed for additional binding selections. The reiterative nature of the process allows for enrichment far in excess of what could be achieved in any single binding procedure and in fact enables the isolation of a population where substantially all of the sequences have the ability to bind target specifically. The authors asserted that their methods were suitable for finding sequences that bind specific proteins. They also stated that their general strategy could be used to identify and clone DNA sequences selectable by other methods.

Oliphant, Brandl and Struhl, Mol. and Cell Biol., 9.:2944-2949 (1989) described using synthetic random sequence oligonucleotides to select for sequences that bind to the yeast transcriptional activator protein GCN4. A synthetic oligonucleotide pool containing ²³ sequences (i.e.l0¹³-10¹⁴) was selected by repeated binding to and elution from a protein affinity column to derive a highly enriched population of oligonucleotides with strong binding to the target. After winnowing the large starting population of molecules to a relatively small fraction by repeated passes over the column, they cloned and sequenced the remaining oligonucleotides and discovered that the population of molecules shared a tightly-conserved common element of 7 bases and an extended, less stringently specified, element of at least 11 bases in length. In proposing that their methods were of general utility for finding oligonucleotides that could bind to proteins, they suggested that a variety of techniques (not just affinity chromatography, but filter binding, immunoprecipitation, or gel mobility shifting to name a few) could be used to isolate the protein-nucleic acid complexes away from the unbound nucleic acids, that PCR could be used to increase the amount of nucleic acid available for additional selection or for cloning, and that the stringency of the selection could be adjusted to obtain both weaker and stronger binding sequences.

The first example of using random sequence selection methods to identify the sequences binding to a protein of unknown function was reported by Kinzler and Vogelstein, Mol. Cell Biol. 10(2) :634-642 1990, as a follow-up of their earlier feasibility study. They had identified a protein amplified in certain cancers and suspected that it might bind to specific DNA sequences. Using whole genome PCR, they identified a specific nucleic acid sequence binding to this protein, GLI. Although they used double-stranded DNA in their study, the authors suggested that the technique could also be used to find single-stranded nucleic acids (generated easily by PCR) that bind specifically to proteins. Theisen and Bach, Nucleic Acids Res., 18:3202-3209 (1990), describe a combination of the techniques of Kinzler and Vogelstein with those of Oliphant, Brandl and Struhl by using PCR amplification of synthetic random sequence oligonucleotides following rounds of protein binding for selection, and in addition, demonstrate the use of nitrocellulose filter binding or electrophoretic mobility shifting to isolate the protein-bound DNA at each selection round. Shortly thereafter yet another suggestion in the earlier reports was implemented with the publication of two papers using single-stranded nucleic acids to identify specific sequences binding to a viral protein (Tuerk and Gold, Science, 249:505-510 (1990) or to non-protein target molecules (Ellington and Szostak, Nature, 3_46_:818-822, 1990). In both of these reports the investigators used single- stranded RNA as the source of random sequences, generating the starting material by in vitro transcription of sythesized DNA and then using PCR via reverse transcriptase (RT-PCR) to amplify the selected RNA before reiterating the cycle. It is of interest to note that both of these reports describe the isolation of specific nucleic acid sequences embedded in potential secondary structures. This is not unexpected since it is often the case that protein interactions with single- stranded nucleic acids are dependent on discrete secondary structures, but the Ellington and Szostak report represents the first demonstration of selection procedures applied to simpler, non-protein (organic dyes) target molecules. In a subsequent report from these same authors (Ellington and Szostak, Nature, 355:850-852. 1992), specific single-stranded DNA oligonucleotides were also selected for binding to these same dye molecule targets. The selected DNA molecules were also purported to have specific secondary structures although the structures and sequences found for the selected DNA oligomers were different from those originally found for RNA oligonucleotides. These findings show that random selection procedures may enable finding specific nucleic acid sequences which bind to non-protein targets.

Bock et. al. Nature, 355:564-566 (1992) describe the use of random single-stranded DNA to identify a specific nucleic acid sequence binding to thrombin, a protein whose major function is in the blood coagulation cascade. A thorough treatment of the considerations important in optimizing a selection protocol is presented in Irvine et al., J. Mol. Biol. 222:739-761 (1991). These considerations notwithstanding, a wide variety of conditions have been used to successfully select for oligonucleotides binding to specific targets.

As mentioned above, Kinzler and Vogelstein carried out their studies with GLI protein to determine whether the protein had any DNA binding activity as well as to identify a consensus sequence if in fact binding activity was present. In similar studies. Bock et al. (1992), supra. and Tuerk and Gold, PCT Publ. No. WO 91/19813 (26 December 1991) have determined that proteins such as thrombin and nerve growth factor, respectively, could bind oligonucleotides of a consensus sequence class. In a study carried out by Tuerk and Gold supra. a number of peptides and proteins were evaluated for ability to bind RNA randomers; some proteins appeared to bind the randomers readily whereas binding was not observed with other proteins.

There are methods in the art to determine whether a population of oligonucleotides containing a very large number of different sequences will bind to a target molecule (e.g., Kinzler & Vogelstein (1989), Oliphant, et al. (1989), Ellington & Szostak (1990), Tuerk & Gold (1990)); however, if the primary function of a molecule is not to bind nucleic acids these methods do not distinguish whether the binding to a target will modulate target function.

The present invention provides a method for determining the suitability of a protein for targeting by random sequence selection with oligonucleotides.

The most suitable candidate molecules against which to target oligonucleotides are those which are capable of binding an oligonucleotide with reasonable affinity at a site which modulates the activity of the target molecule. The present invention describes a screening procedure which evaluates these properties of a target molecule.

This method for identifying target molecules potentially susceptible to activity modulation upon binding with a nucleic acid sequence comprises, (i) contacting the target with nucleic acid in the presence and absence of a molecule known to modulate the activity of the target upon binding therewith (a "cognate molecule" or "cognate") under conditions that distinguish between binding of nucleic acid to target and binding of nucleic acid to cognate; (ii) separating target-bound nucleic acid from unbound nucleic acid in the presence and absence of cognate molecule under such conditions; and (iii) observing any detectable reduction of binding of nucleic acid with target in the presence of cognate as compared to nucleic acid binding with target in the absence of cognate. Generally, the cognate is contacted with the target before addition of nucleic acid or. alternatively, the cognate and nucleic acid are contacted with the target simultaneously.

Further, the invention also provides a method for isolating nucleic acid sequence candidates for target activity modulation upon binding therewith. This comprises (a) contacting a sample containing the target with a nucleic acid mixture under conditions that distinguish between binding of nucleic acid sequences to target from binding of nucleic acid sequences to cognate for a time sufficient to establish an association between the target and those sequences to which it binds; (b) adding target cognate thereto in an amount and for a time sufficient to displace a portion of the target-bound nucleic acid sequences; and (c) recovering nucleic acid sequences which remain bound to the target.

Preferably, in step (a), the sample contains the target in a concentration approximately that of the dissociation equilibrium constant (K_d) of the target for the bulk population of nucleic acid sequences in the mixture. This contact is preferably maintained for a time sufficient to establish an association equilibrium between the target and those nucleic acid sequences with which it binds before cognate is added. In step (b) cognate is preferably added at a concentration in excess of its Kd for the target and in molar excess over the target, for a time sufficient for all nucleic acid sequences except those strongly bound with target molecules to be replaced by cognate. Target molecules are then separated from unbound nucleic acids in the mixture. Nucleic acids which remain bound to target are then extracted and, optionally, the copy number of the extracted nucleic acid sequences can be amplified. Further purification and/or selection of the optimal nucleic acid sequence(ε) is achieved by repeating steps (a) through (c) on the isolated nucleic acid sequence(s) of the preceding iteration. The nucleic acids can be amplified between iteration by PCR or some other amplification procedure.

The target is preferably a protein. Further, the target is preferably a molecule not previously known to interact with nucleic acids. Also, the nucleic acids can be sequences not previously known to modulate protein activity. The nucleic acid sequences can be unmodified or modified oligonucleotides or polynucleotides and usually include at least 6 random nucleotides, preferably 6-600 random nucleotides and most preferably 20-60 random nucleotides. The nucleic acid sequences can be provided with flanking sequences on at least one of the 3' and 5' ends to facilitate amplification. The cognate is preferably an inhibitor or activator of the target, and it is also contemplated that the target can likewise be a molecule which acts on or affects the function of the cognate. The distinguishing conditions that can be varied are, for example, ionic strength (e.g., salt concentration), pH or temperature.

Figure 1. This histogram illustrates the effect of prebinding pentosan polysulfate (PPS) with gp30 on the interaction of labeled nucleic acid rando er with gp30 in the experiments described in Example 1.

Figure 2. This dose-response graph illustrates the displacement of gp30 bound oligonucleotides by PPS as reported in Example 2. Figure 3. This histogram illustrates the displacement of gp30 bound nucleic acid randomer by 100 nM and 1 mM PPS reported in Example 2.

Figure 4. This graph illustrates the specific binding over time of nucleic acid sequences with anti-DNA SLE- associated antibodies as reported in Example 3.

Figure 5. This histogram illustrates the results of the experiments reported in Example 4.

Figure 6. This graph illustrates the comparison of binding affinity of the unselected starting randomer population (RO) with the oligonucleotide recovered (RIC) from the dissociation reaction reported in Example 5.

Figure 7. This histogram illustrates the binding of bFGF and heparin with resultant total loss of oligonucleotide binding as reported in Example 6.

The ability to identify nucleic acids which bind with high affinity to proteins involved in disease has therapeutic relevance, but only if modulation of the target molecule activity results from the binding event. The mere fact that an oligonucleotide binds to a target molecule with high affinity does not ensure that the oligonucleotide will be located in proximity to an active site of the target molecule. For example, there might be multiple sites on a target molecule to which an oligonucleotide can bind and if the site with highest affinity is distal from a functional site on the target molecule, it will be difficult to select for oligonucleotides which bind closer to an active site on the target molecule but with less affinity. There are several scenarios in which an oligonucleotide can bind to a target molecule and productively interfere with activity. For example, an oligonucleotide that binds a ligand with high affinity in such a way that it blocks the interaction of that ligand with its receptor could be useful therapeutically, or as a lead compound for generating a chemically-modified analog that would be more pharmacologically stable, active or otherwise acceptable. Similarly, an oligonucleotide could bind to the ligand- binding site of a receptor in a therapeutically useful fashion. Or, an oligonucleotide could have therapeutic value if it bound to a catalytic site, cofactor-binding site or regulatory site of an enzyme. Alternatively, an oligonucleotide which bound to a low molecular weight ligand or enzyme substrate in such a way as to block its function or metabolism could have therapeutic relevance.

In general_/ "natural" oligonucleotides (i.e., those with natural bases, ribose or deoxyriboεe as the sugar and a phosphodiester backbone) are not pharmacologically suitable for use as drugs since they are very susceptible to nucleaseε in the circulation and in cells. Furthermore, if the target molecule is intracellular, this would be a disadvantage for natural oligonucleotides since these molecules do not efficiently penetrate cells. Accordingly, there is a need to chemically modify oligonucleotide leads to make them more effective as therapeutics. Even under the best of circumstances, for some oligonucleotide leads such chemical modification may result in a loss of specific activity. Accordingly, it is desirable to identify natural oligonucleotide lead compounds that bind to their target with a K_d in the micromolar range or lower. If, for example, nucleic acids in bulk bind to one target in the micromolar range and to another in the millimolar range, it is likely that the former target will select for a suitable lead sequence more often that the latter.

It is reasonable to assert that for a protein with a known physiological function that involves interaction with a nucleic acid (e.g., an activator or repressor), the strongest binding site for an oligonucleotide on that protein would be at the functional site. On the other hand, for molecules which are not known to function via.interaction with a nucleic acid, there is no a priori reason to expect that interaction with an oligonucleotide will affect function of that target molecule. If nucleic acid was found to bind to a functional site on a target molecule, this would increase the likelihood that the binding of an individual selected nucleic acid sequence would modulate activity of the target molecule. One way to make this determination would be to bind nucleic acid competitively with a "cognate molecule" (a molecule which is known to bind to a functional site on the target molecule). However, since the nucleic acid prior to selection of a specific sequence characteristically binds to target less avidly than cognate molecule binds to target, it is likely that unselected nucleic acid will not appreciably prevent interaction between a target molecule and a cognate molecule. The present invention takes advantage of this expected difference in affinity by evaluating for the converse effect, i.e., by determining whether cognate molecule can competitively inhibit nucleic acid binding to the target molecule.

Furthermore, the present invention describes dissociation rate-dependent selection procedures to effectively and rapidly sample high complexity random

- 13 - oligonucleotide mixtures to select those sequences which bind to the target molecule with high affinity and which effectively interfere with function of the target molecule by virtue of binding at or near an active site. The screening and selection procedures of the invention are of significant value in the identification and generation of oligonucleotide-based therapeutics.

In accordance with a principal aspect of the invention, a target molecule is incubated with-nucleic acids, preferably oligonucleotides (nucleic acids containing relatively few nucleotide residues, i.e., less than a few hundred residues, are commonly referred to as oligonucleotides) . After incubation the target molecule is separated from unbound nucleic acid and the amount of bound nucleic acid is measured. If measurable levels of nucleic acid binding to the target molecule are detected, binding of nucleic acid to a cognate molecule is carried out to determine conditions under which nucleic acid binding to target and cognate molecules can be distinguished (usually by selecting conditions under which cognate molecule binds nucleic acid very poorly if at all). Finally, the target molecule is again incubated with nucleic acid in the absence and presence of cognate molecules under conditions which discriminate between nucleic acid binding to target and cognate molecules. If the presence of cognate molecule decreases the total amount of nucleic acid bound to target, it is likely that the nucleic acid and the cognate molecule bind at or near the same site on the target molecule and this target molecule is considered to be suitable for further analysis. If the cognate molecule does not measurably reduce the amount of nucleic acid bound to target molecule, then this target molecule is deemed to be a less suitable candidate for targeting with oligonucleotides.

In one embodiment of the instant invention the nucleic acid is of a mixture of oligonucleotides in which at least a portion of each molecule contains randomly assorted bases or a mixture of bases and other chemical entities (this type of oligonucleotide is hereafter referred to as a "randomer" even if part of the molecule contains an invariant sequence). The use of randomer in place of a single nucleic acid sequence can be advantageous if the target has unknown sequence preferences.

The target molecule can be any macromolecule or small molecule, including but not limited to a protein, a fragment of a protein, a peptide, a sugar, a complex carbohydrate, a lipid, etc.

The cognate molecule can be a ligand if the target is a receptor; a receptor if the target is a ligand; a substrate, inhibitor, coenzyme, allosteric effector, etc. if the target is an enzyme; an enzyme if the target is a substrate, inhibitor, coenzyme, allosteric effector , etc. An antibody which neutralizes activity of the target molecule can also be used as a cognate molecule. Similarly a specific antigen can be the cognate molecule if the target is an antibody. The previous list is included as examples of cognate molecules but cognate molecules are not limited to those noted.

The nucleic acid can be constituted of single-stranded DNA, double-stranded DNA, single-stranded RNA, double- stranded RNA or mixtures of DNA and RNA. If the nucleic acid is a randomer, the randomer can contain a fixed sequence at the 5' end and a fixed sequence at the 3' end flanking a random sequence in the middle, but other structures containing a random region sequence are contemplated. The random region can contain approximately equal numbers of A,C,G and T (or U) at each base but cases in which there is a preponderance of one or another base(s), or in which other types of bases are substituted are also contemplated. The flanking sequences can be chosen to allow for PCR priming and/or restriction enzyme cutting for cloning and sequencing. For the method of the invention, the randomer population is labeled, preferably with a radioisotope such as ³²P, although other modes of labeling, such as with biotin, dyes, or other radionuclides can be used. ³²P end-labeling can conveniently be used after the randomers have been synthesized.

Single-stranded DNA is prepared with a nucleic acid synthesizer (for synthesis of DNA randomers, specific nucleotide monomers are added at each cycle to generate the defined 5' and 3' flanking sequences if such flanking sequences are desired, whereas a suitable mixture of nucleotide monomers is added at each cycle to generate the random core of the molecule). Mixtures containing four nucleotide monomers that allow approximately equal proportions of dpA, dpC, dpG, and dpT to be added randomly to the growing DNA chains are available commercially but others may be prepared from individual nucleotide monomers as desired. Double-stranded DNA is prepared by first synthesizing single-stranded DNA as above and then copying the synthesized single strands to double-stranded DNA by primer-dependent synthesis using DNA polymerase or another suitable enzyme. Single-stranded RNA can be generated with a nucleic acid synthesizer or from the double-stranded DNA described above by transcribing the DNA with RNA polymerase (provided the defined sequence at the 5' end of the DNA contains an RNA polymerase recognition site) and then removing the DNA template with DNase. Double- stranded RNA can be generated from single-stranded RNA using Qβ Replicase.

The random region of the randomer can be of variable length, but preferred lengths are from about six bases to about 100 bases or more. Most preferably, random regions between about 20 and 60 bases are used. Similarly, flanking invariant oligonucleotide sequences, when present, can contain as few as 6 bases or as many as 100 bases or more can be used. Flanking sequences can be present at the 5' end of the randomer, at the 3' end or at both ends. One or more invariant regions might be present in an internal region(ε) of the randomer or the randomer might lack an invariant region altogether.

To implement the instant method, the target molecule with any bound nucleic acid must be separated from unbound nucleic acid (as well as cognate molecule if the latter binds significant amounts of nucleic acid) . This can be achieved by any of a number of techniques known in the art such as immunoprecipitation, affinity column chromatography, filter binding and gel shifting on polyacrylamide gels. The latter two techniques are particularly convenient since reagent needs are minimal.

Incubation of nucleic acid with target can be carried out in solution (with subsequent separation by filter binding or gel shifting on polyacrylamide gels or by immunoprecipitation), with target in solid phase (with separation by affinity column chromatography as one example). Incubation buffers can vary depending upon those conditions required for maintaining target molecule integrity and permitting interaction between target and cognate molecules. However, salt concentration is important and should preferably be in the range of 25-250 mM. An intermediate salt concentration (e.g., 140 mM) can be used initially, followed by retesting at much lower salt concentration (e.g., 25mM); the higher salt concentration is likely to minimize non-specific electrostatic interactions between the target molecule and the phosphate backbone of the nucleic acid. The capacity to maintain binding in the.presence of a higher salt concentration may be predictive of a greater likelihood of finding a specific sequence interaction. Furthermore, if the selected molecule itself is to be a therapeutic, the ability to bind target at physiological concentrations of salt is an important consideration in evaluating target suitability. A typical incubation buffer is 25 mM Tris-HCl pH 7.5, 140 mM NaCl, 3mM MgCl₂.

In order to determine the appropriate conditions for testing, binding analysis for the target protein and cognate molecule is performed using each form of nucleic acid described above (i.e., single stranded or double stranded-DNA or RNA) . The binding analysis employs varying concentrations of target protein generally spanning the range from about 10" *M to lO^M and trace amounts of a labelled nucleic acid population and is well known in the art. Binding incubations can be at any temperature between about 4°C and 37°C that is consistent with maintaining the integrity of the target molecule. Binding incubation times can range from one or a few minutes to several hours. Times of 20-30 minutes are usually adequate. After incubation, the target molecule is recovered by any of the procedures described above that might be suitable. For proteins, filter binding or gel shift experiments are particularly desirable, although for non-protein targets, particularly small molecules, affinity chromatography or some other kind of column chromatography (e.g., ion exchange or size exclusion) might be more appropriate. Recovery procedures include but are not limited to those described above.

When filter binding is used to trap proteins or other targets, nitrocellulose filters (Millipore type HA) are commonly used. Although it is generally assumed that all proteins will be retained by such filters, retention is not universal, particularly for smaller proteins or peptideε. It is therefore important to ascertain that the target can be retained by the filter. One technique for doing this for protein targets is to stain filters with a non-specific protein stain such as amido black. The amount of protein adherent to the filter can be estimated by comparison with control filters in which various amounts of target protein are dried onto filters prior to staining.

If the target can be trapped by the filter, the incubation mix is filtered and washed with several volumes of incubation buffer. The amount of nucleic acid retained by the filter is then determined by appropriate means, e.g., scintillation spectrometry or Cerenkov determinations. A sample is scored as positive if there are significant levels of nucleic acid above background (e.g., filter and nucleic acid without target) . For gel shift studies, target molecule and radiolabeled nucleic acid are incubated, preferably at a number of target concentrations as described above. The mixture is then applied to a polyacrylamide gel (5% to 8%, depending upon the target) and electrophoresis performed. The gel is dried and subjected to autoradiography. If the target binds nucleic acid, there will be one or more bands of radioactivity closer to the position of the origin relative to the migration position of free labeled nucleic acid. The proportion of the nucleic acid bound can be determined by densitometric scan of the gel lane or by excising the bands from the gel and determining the radioactive content. An estimate of K_d can be made by plotting target concentration vs. amount of nucleic acid bound.

To determine whether a target binds nucleic acid in a region of the molecule that can interfere with function, the target molecule is preincubated with an excess amount of cognate molecule at a concentration of cognate molecule in excess of its K_d for the target. As a control, the target molecule is incubated with a similar concentration of irrelevant, non-cognate molecule that does not interact with the target. After preincubation (a few minutes to several hours), labeled nucleic acid is added to the mixture and its binding is measured as described above. If the amount of nucleic acid binding to the target in the presence of a cognate molecule is reduced relative to that observed with target alone or with target and non-cognate molecule, this could be interpreted as indicating that the nucleic acid has been competed effectively from binding to a site where cognate molecule binds and thus that at least some of the nucleic acid is binding at or near a site on the target that could have pharmacological relevance. Although it is preferred that the cognate molecule be incubated with target molecule before addition of nucleic acid, other procedures can be contemplated in which the cognate molecule is added to the target molecule simultaneously with, or after, the addition of nucleic acid to the target molecule.

This procedure works best when the target molecule binds much more nucleic acid than the cognate molecule under the test condition. Thus, if for example, both target and cognate molecules bind DNA at high levels, but only target molecule binds RNA at easily detectable levels, the latter type of nucleic acid should be used in the test. If it is not possible to identify a nucleic acid type which binds effectively to the target molecule but not to the cognate, the test is still possible. For example, gel-shift studies could be used for the test provided the cognate and target molecules shift the nucleic acid to different positions in the gel. Other procedures which can be used to isolate the target molecule free of cognate molecule and any unbound nucleic acid can be substituted for any of the procedures described above.

This procedure is used with a target molecule with which a cognate molecule competes for binding with nucleic acid. If the method indicates that such competition occurs, a time course experiment is carried out in which randomer is preincubated with target molecule as described below and then cognate molecule is added for varying periods of time prior to sampling to determine the amount of randomer which remains bound to target molecule. The length of time is determined at which only a fraction of the originally bound randomer (e.g., 10% or less) remains bound to the target molecule. The expectation is that the residual bound randomer molecules will have the slowest rate of dissociation from the target and thus will have the highest affinity (assuming a more or less constant association rate for all randomers).

In general, dissociation rate selection is initiated by preincubating target molecule with a randomer mixture at a concentration of target at or near the K_d of the target for the bulk population of random sequences, preferably such that there is a molar excess of target molecule. The preincubation time is preferably about 20-30 minutes, though periods as short as 1 minute or as long as several hours can be used. The temperature during the preincubation procedure is preferably between 4°C and 37°C although higher or lower temperatures can be contemplated. The temperature used during the incubation period with cognate molecule can be as high as or as low as any temperature at which the target molecule will retain activity and the ability to interact with cognate molecule. For example, if the target molecule is an enzyme and the cognate molecule its substrate, the cognate molecule could be modified during incubation and this could diminish its ability to bind to the target molecule. In this circumstance, the incubation should be carried out at 4°C provided that the cognate and target molecules interact at that temperature.

Following preincubation, cognate molecule is added, preferably at a concentration in excess of its K_d for the target and in molar excess over the target. A time course for the dissociation of target and randomer is determined. When a large percentage (more than 90%) of the bound randomer has been lost from the target, the remaining target-bound oligonucleotide is collected, extracted, amplified and the process is repeated. With reiterations, the rate of dissociation from the target will diminish as the level of high affinity oligonucleotide in the selected population increases. This will require lengthening the dissociation incubation time to continue to select the highest affinity sequences.

The invention further provides useful diagnostic methods in which oligonucleotides serve as specific binding partners which interact with active regions of target proteins or other target molecules. Thus, what .is provided is not analogous or readily extrapolated from doctrines of classical antibody immunoassay or hybridization assay in which only nucleic acid targets are evaluated. All the more, these diagnostic methods of the invention take advantage of the unique specificity between target and selected, sequence identifiable oligonucleotides. There are numerous suitable targets for diagnostic identification including but not limited to oncoproteins and the proteins that directly, e.g. transcription factors, and indirectly, e.g. signal tranεduction kinases for transcription factors, affect cellular or viral activity at the level of transcription/translation regulation.

The nucleic acids selected by the method of the present invention may be employed for binding to target molecules, such as, for example, proteins including, but not limited to, ligands, receptors, and/or enzymes, whereby such nucleic acids inhibit or stimulate the activity of the target molecules.

The identified nucleic acids are also applicable to the inhibition of any pathological activity, including, for example, viral replication, as well as the interference with the expression of genes which may contribute to cancer development. The nucleic acids selected by the method of the present invention are administered in an effective binding amount to the intended target. Preferably, they are administered to a host, such as a human or non-human animal host, so as to obtain a concentration of such nucleic acids in the blood of from about 0.1.to about 100 μmole/1. It is also contemplated, however, that they can be administered in vitro or ex vivo as well as in vivo.

The nucleic acids can be administered in conjunction with an acceptable pharmaceutical carrier as a pharmaceutical composition. Such pharmaceutical compositions may contain suitable excipients and are manufactured in a manner which is itself well known in the art.

Example 1 Analysis of αp30 Target Protein

Overexpression of the oncogene erb B2, which encodes a 185 kDa transmembrane growth factor receptor (plβS"*⁸²) with tyrosine kinase activity, correlates with poor prognosis in breast, ovarian, gastric, and endothelial cancers, and non- small cell lung adenocarcinoma. A 30 kDa glycoprotein (gp30) has been identified as an endogenous ligand for pl85^eAB2 by its direct binding to the receptor and induction of tyrosine kinase activity. Additionally, gp30 induces cell proliferation and soft agar colony formation, demonstrating its growth-stimulating ability. The ability to block the interaction of gp30 with its receptor could have important therapeutic consequences. Particularly, identification of nucleic acids which inhibits gp 30/pl85^eΛB2 ligand/receptor interaction has significant therapeutic potential with respect to arresting metastasis and treating breast cancer. It is of interest therefore to determine whether gp30 is a promising target for nucleic acid targeting. Pentosan polysulfate (PPS) binds to gp30 so as to inhibit its biological activity. Competition experiments were performed using PPS as a cognate molecule for gp30.

PPS (800 nM) was incubated with 200 nM gp30 for 15 minutes at room temperature in IX binding buffer (25 mM Tris, pH 7.5, 150 mM NaCl, 3 mM MgCl₂) . ³²Pr-labeled DNA oligonucleotides with a 60-baεe central random region were added to give a final concentration of 400 nM PPS, 100 nM gp30, and 485 pM DNA in IX binding buffer. After 15 minutes at room temperature the reaction was filtered through Millipore type HA nitrocellulose filters under vacuum, and the filters were immediately washed with IX binding buffer. The filters were dried and counted in a liquid scintillation counter without scintillant. Control reactions of labeled randomer mixed with either gp30 or PPS alone, or with water were performed. The input of ³²P-labelled randomer was determined in a mock reaction with water that was dried onto a nitrocellulose filter and counted without being filtered.

Preincubation of PPS with gp30 almost completely prevented randomer binding to gp30 (Figure 1). Since pentosan binding to gp30 is inhibitory to gp30 activity, it is likely that selected oligonucleotides will be pharmacologically active.

Example 2

Dissociation Rate Selection of Randomer Binding to σp30 The ability of PPS to displace randomer bound to gp30 was evaluated by incubating 1 nM ³²P-labeled randomer with 10 nM gp30 for 15 minutes at room temperature in IX binding buffer. PPS was added to give the following approximate final concentrations: 100 μM gp30, 1 μM PPS, 1.0 nM randomer in IX binding buffer. Residual binding of randomer to gp30 was measured by filter binding at various time points. In the absence of PPS addition, there was no change in the amount of randomer bound to gp30 over 30 minutes (Figure 2). Within the first minute of PPS addition, 87.6% of the gp30 bound randomer was displaced and by 30 minutes 92.7% was displaced.

Similar time-dependent displacement of randomer binding was observed at different PPS concentrations (Figure 3). Conditions for binding were as described with reference to Figure 2 except that upon addition of PPS the following concentrations were obtained: gp30 = 10 nM, randomer = 1.2 nM, PPS = 100 nM or 1 mM. With the above information, a selection involving prebinding of randomer to gp30 followed by a PPS challenge can be designed to enrich for high affinity sequences which bind to an the active site of gp30, such sequences being the ones that dissociate from gp30 more slowly than other oligonucleotides.

Ryam le fl

Selection of Nucleic Acids Binding to SLE Pathogenic Antibodies in Patients

There are many classes of anti-DNA antibodies in the sera of Systemic Lupus Erythematosis patients, but the high affinity ones appear to be most clearly involved in the pathogenesis of the disease. Therefore, a method to selectively identify and isolate nucleic acid sequences that bind specifically to these antibodies is highly desirable. Identification and use of small nucleic acid sequences that bind to SLE antibody will prevent or diminish the aggregation that results from binding with larger nucleic acid sequences and thereby diminish clogging of kidney clearance caused by the aggregates formed. The dissociation rate selection assay of the invention confers the needed specificity for efficient selection.

The method for this assay is to preincubate lupus serum- derived IgGs containing target antibodies as well as other antibodies together with PCR-amplifiable randomers and then to add an excess of extraneous DNA that will not be specifically amplifiable with the primer set used to amplify the randomer. The excess extraneous DNA will compete for binding to the lower affinity antibodies that are undergoing rapid dissociation/reasεociation. Therefore, those sequences binding to the high-affinity antibodies will be retained for longer times, so that when the nucleic acid-antibody complexes are isolated, the randomer that will be specifically amplifiable will be enriched for those higher affinity sequences that remained bound to the target antibodies.

Standard 20 μl binding reactions were set up to contain 0.5 μM IgG, and approximately 20,000 cpm end-labeled single or double-stranded randomer in 25 mM Tris-HCl pH 7.5, 150 mM NaCl and 3mM MgCl₂. The reaction mixtures were incubated at room temperature for 30 minutes. Immediately following, 2.5 μM single-stranded DNA or double-stranded DNA in a volume of about 1 μl were added to the samples and incubated for various times before the reaction mixture was filtered (see Figure 4). This established the 30-minute time point as useful for selection of the tightly-bound sequences. At this point, the DNA-IgG complexes are isolated and the bound DNA extracted and amplified in preparation for reiteration of the selection cycle.

Example 4

Analysis of -IFN Target Protein

Analysis by the method of the invention was performed to establish whether -IFN was a suitable candidate for nucleic acid targeting. -IFN was found to bind to all forms of nucleic acid examined. Binding of single stranded-DNA randomer with -IFN (1.0 μM) was used for the analysiε. Cognate molecule, -IFN receptor protein, bound little or no randomer above background at concentrations of protein in excess of those giving extensive binding to the -IFN (i.e. > 2μM). To determine if the binding of nucleic acid to -IFN was likely to be of pharmacological significance, binding inhibition by cognate molecule was determined.

Target protein, -IFN, at a concentration of 1 μM was incubated with trace amounts of ³²P-labeled DNA randomer for 15 minutes at room temperature and -IFN bound about 35% of randomer input (Figure 5). When the same -IFN (1 μM) preparation was preincubated for 10 minutes with a 2-fold molar excess of its cognate receptor prior to the addition of ³²P-labelled DNA randomer, and then incubated for a further 15 minutes in the presence of the randomer, the binding of the randomer was reduced to about 1.7% of input. When preincubation was performed in the presence of non-cognate receptor protein (e.g. TNF receptor) at a comparable concentration to the -IFN receptor, the binding of randomer to the -IFN was essentially unaffected. As with the cognate receptor, the non-cognate TNF receptor alone gave no binding to ³²P-labelled randomer under the conditions of the assay. This established that the major point of interaction of nucleic acid and target protein is at a position blocked by the interaction of ligand and receptor. Thus, the binding of

-IFN to its receptor can be blocked to inhibit or prevent hyperimmune responsiveness such as occurs in inflammatory disorders.

Exam le 5

Dissociation Rate Selection for Nucleic Acid

Binding to -IFN

In order to enrich for nucleic acid sequences capable of binding to -IFN with higher affinity, a dissociation rate selection approach was employed. -IFN protein at 0.5 μM in binding buffer (25 mM Triε-HCl pH 7.5, 140 mM NaCl, 3 mM MgCl₂) was incubated with DNA randomer at 185 nM for 20 minutes at room temperature. Following this incubation, IFN receptor protein was added to a final concentration of 1.0 μM (2-fold molar excess over target ligand). Immediately following the addition of receptor and thorough mixing of the solution, the sample was filtered to isolate protein-bound nucleic acid. Randomer was recovered from the sample, amplified by primer-dependent PCR and subjected to binding analysis to determine the change in affinity for the target protein after a single round of dissociation selection. Figure 6 compares the binding affinity of the unselected starting randomer population (RO) with the nucleic acid sample recovered from the dissociation reaction (RIC). The affinity of the new population for the target is approximately 3-fold greater than that of the starting sample.

Example 6

Analysis of bFGF Target Protein

An analysis was performed to establish whether basic Fibroblast Growth Factor (bFGF) was a suitable target for nucleic acid targeting. Binding analysis indicated that bFGF interacts with all forms of nucleic acid tested (i.e. ss-DNA and ds-DNA or RNA) . Conversely, the cognate FGF receptor did not bind ssDNA randomer up to a protein concentration of at least 1.7 μM. Single-stranded DNA randomer was chosen for use in the analysis. Incubation of 1 μM bFGF with ³²P- labelled tracer DNA randomer in binding buffer resulted in approximately 80% binding to the target. Preincubation of 1 μM bFGF with a 1.7-fold molar excess of cognate receptor prior to the addition of labelled randomer resulted in more than 50% reduction in the amount of randomer bound to the target (Figure 7). This indicates that a significant proportion of the nucleic acid binding to bFGF occurs at a site affected by receptor-ligand interaction and that specific nucleic acid sequences selected for binding to bFGF may have pharmacological activity.

Basic FGF also interacts with heparin in a reaction that affects the binding of ligand to receptor. It is believed that the sites of interaction on bFGF with receptor and heparin are proximal to one another. The effects of heparin on bFGF receptor binding suggest that it can serve as an appropriate cognate molecule for analysis by the method of the invention. Accordingly, heparin was evaluated for binding to DNA randomer and found to be negative above background up to 10 μM. Preincubation of 1 μM bFGF with this concentration of heparin prior to the addition of labelled randomer resulted in a total loss of randomer binding (Figure 7). Thus, the randomer and heparin seem to share a common site for binding. These data further support the finding that specific nucleic acid sequences selected for binding to bFGF are of pharmacological interest and indicate that bFGF is a suitable candidate for nucleic acid selection. This is significant in that bFGF has tumorgenic angiogenesis effects which are inhibited by prevention of its interaction with its receptor. Prevention of tumor growth stimulation is clearly a desirable and useful property, and the prevention of angiogensiε, particularly on a localized basis is useful in starving solid tumors or other neoplastic tissue of vascularization and resultant blood supply.

Claims

What Is Claimed Is:

1. A method for identifying target molecules potentially susceptible to activity modulation upon binding with a nucleic acid sequence, which comprises: contacting the target with a nucleic acid in the presence and absence of a cognate molecule known to modulate the activity of the target upon binding therewith under conditions that distinguish between binding of nucleic acid with target and binding of nucleic acid with cognate; separating target-bound nucleic acid from unbound nucleic acid in the presence and absence of cognate molecule under such conditions; and observing any detectable reduction of binding of nucleic acid with target in the presence of cognate as compared to nucleic acid binding to target in the absence of cognate.

2. The method of claim 1 wherein the target is a protein.

3. The method of claim 1 wherein the target is a molecule not previously known to interact with nucleic acidε.

4. The method of claim 1 wherein the nucleic acid sequences are not previously known to modulate protein activity.

5. The method of claim 1 wherein the nucleic acid sequences are modified or unmodified oligonucleotides or polynucleotides.

6. The method of claim 1 wherein the nucleic acid is a randomer.

7. The method of claim 1 wherein the nucleic acid sequences include at least 6 nucleotides.

8. The method of claim 1 wherein the nucleic acid includes flanking sequenceε at at leaεt one of the 3' and 5' end.

9. The method of claim 1 wherein the cognate is an inhibitor or activator of the target.

10. The method of claim 1 wherein the distinguishing conditions that can be varied are selected from ionic strength, pH and temperature.

11. A method for isolating nucleic acid sequence candidates for modulating target activity upon binding with target which comprises:

(a) contacting a sample containing the target with a nucleic acid mixture under conditions that distinguish between binding of nucleic acid sequences with target from binding of nucleic acid sequences with cognate for a time sufficient to establish an association between the target and those sequences with which it binds;

(b) adding target cognate thereto for a time sufficient to displace a portion of the target-bound nucleic acid sequences, and

(c) recovering nucleic acid sequences which remain bound to the target.

12. The method of claim 11 wherein the target is a protein.

13. The method of claim 11 wherein the target is a molecule not previously known to interact with nucleic acids.

14. The method of claim 11 wherein the nucleic acid sequences were not previously known to modulate protein activity.

15. The method of claim 11 wherein the nucleic acid sequences are modified or unmodified oligonucleotides or polynucleotides.

16. The method of claim 11 wherein the nucleic acid sequences include at least 6 nucleotides.

17. The method of claim 11 wherein the nucleic acid sequences include a flanking sequence at at least one of the 3' and 5' end.

18. The method of claim 11 wherein the cognate is an inhibitor or activator of the target.

19.. The method of claim 10 wherein the distinguishing conditions that can be varied are selected from ionic strength, pH and temperature.

20. An oligonucleotide, polynucleotide or analog thereof that binds to and affects the activity of gp30.

21. An oligonucleotide, polynucleotide or analog thereof that binds to and affects the activity of serum lupus IgG.

22. An oligonucleotide, polynucleotide or analog thereof that binds to and affects the activity of -IFN.

23. An oligonucleotide, polynucleotide or analog thereof that binds to and affects the activity of bFGF.