Method of Ordering Sequence Binding Preferences of a DNA-binding Molecule

The present invention relates to methods, systems, and kits useful for the identification of molecules that specifically bind to defined nucleic acid sequences. Also described are methods for designing molecules having the ability to bind defined nucleic acid sequences and compositions thereof.

Ambinder, R. F., et al., J. Virol. 65:1466-1478 (1991).

Ausubel, F. M., et al., Current Protocols in Molecular Biology, John Wiley and Sons, Inc., Media Pa.

Baguley, B. C., Mol. Cell. Bioch. 43:167-181 (1982).

Banerji, S. S., et al., Mol. Cell Biol. 11:4074-4087 (1991).

Beal, P. A., et al., Science 251:1360-1363 (1991).

Becker, Y., et al., Isr. J. Med. Sci. 8:1225 (1972).

Bialer, M., et al., J. Med. Chem. 23:1144 (1980).

Bialer, M., et al., J. Pharm. Sci. 70:822 (1981).

Birg, F., et al., Nucl. Acids Res. 18:2901-2908 (1990).

Chaiet, L., et al., Arch. Biochem. Biophys. 106:1 (1964).

Chaires, J. B., et al., Biochemistry 29:6145-6153 (1990).

Chang, H.-K,, et al., Mol. Cell. Biol. November:5189-5197 (1989).

Chen, K-X., et al., J. Biomol. Struct. Dyn. 3:445-466 (1985).

Chin, M. T., et al., J. Virol. 63:2967-2976 (1989).

Comai, L., et al., Cell 68:965-976 (1992).

Cooney, M., et al., Science 241:456-459 (1988).

Courtois, G., et al., Proc. Natl. Acad. Sci. USA 85:7937-7941 (1988).

Cullinane, C., et al., FEBS Lett. 293:195-198 (1991).

Debart, F., et al., J. Med. Chem. 32:1074-1083 (1989).

Dervan, P. B., Science 232:464-471 (1986).

Edwards, C. A. et al., J. Mol. Biol. 180:73-90 (1984).

Edwards, C. A., et al., in: Advances in Regulation of Cell Growth, Volume I: Regulation of Cell Growth and Activation, edited by Mond, J. J., et al., New York: Raven Press, p. 91-118 (1989).

Elias, P., et al., Proc. Natl. Acad. Sci. USA 85:2959-2963 (1988).

Fox, K. R., et al., Biochim. Biophys. Acta 840:383-392 (1985).

Fox, K. R., et al., Nucl. Acids Res. 16:2489-2507 (1988).

Fox, K. R., et al., Nucl. Acids Res. 18:1957-1963 (1990).

Fox, K. R., et al., Biochem J. 269:217-221.

Fried, M. G., et al., Nuc. Acid. Res. 9:6505 (1981).

Galas, D., et al., Nuc. Acid Res. 5:3157-3170 (1981).

Garner, M. M., et al., Nuc. Acid. Res. 9:3047 (1981).

Gaugain, B., et al., Biochemistry 17:5071 (1978).

Gessner, R. V., et al., Biochemistry 24:237-240 (1985).

Gilbert, D. F., et al., Proc. Natl. Acad. Sci. USA 86:3006 (1988).

Gilman, A. G., et al., eds., The Pharmacological Basis of Therapeutics, Eighth Edition, Pergamon Press (1990).

Goldin, A. L., et al., J. Virol. 38:5-58 (1981).

Goodisman, J., et al., Biochemistry 31:1046-1058 (1992).

Green, N. M., Adv. Protein Chem. 29:85 (1975).

Greenblatt, J., Cell 66:1067-1070 (1991).

Greene, W. C., Annu. Rev. Immunol. 8:453-475 (1990).

Griffin, L. C., et al., Science 245:967-971 (1989).

Griffin, J. H., et al., J. Am. Chem. Soc. (1992).

Gross, D. S., et al., Annu. Rev. Biochem. 57:159-197 (1988).

Gurskii, G. V., et al., Mol. Biol. 19:177 (1985).

Harlow, E., et al., Antibodies: A Laboratory Manual, Cold Spring Harbor Laboratory Press (1988).

Hausheer, F. H., et al., Anti-Cancer Drug Design 5:159-167 (1990).

Hawley, R. C., et al., Proc. Natl. Acad. Sci. USA 86:1105-1109 (1989).

Helene, C., et al., Biochim. Biophys. Acta 1049:99-125 (1990).

Helene, C., et al., Genome 31:413-420 (1989).

Hoogsteen, Acta Cryst. 12 822 (1959).

Jain, S. C., et al., J. Mol. Biol. 68:1-20 (1972).

Jeppesen, C., et al., Eur. J. Biochem. 182:437-444 (1989).

Kadonaga, J. T., PNAS 83:5889-5893 (1986).

Kissinger, K., et al., Biochemistry 26:5590-5595 (1987).

Koff, A., et al., J. Virol. 62:4096-4103 (1988).

Kotler, M., et al., FEBS. Lett. 21:222 (1972).

Krowicki, K., et al., J. Org. Chem. 52:3493 (1987).

Kuhlmann, K. F., et al., Nucl. Acids Res. 5:2629 (1978).

Laugaa, P., et al., Biochemistry 23:1336 (1985).

Le Pecq, J. B., et al., Proc. Natl. Acad. Sci. U.S.A. 72:2915-2919 (1975).

Lee, D. K., et al., Cell 67:1241-1250 (1991).

Lown, J. W., et al., J. Org. Chem. 50:3774 (1985).

Lown, J. W., et al., J. Med. Chem. 29:1210-1214 (1986).

Luck, G., et al., Nucl. Acids Res. 1:503 (1974).

Luckow, V. A., et al., Virology 170:31 (1989).

Maher III, L. J., et al., Science 245:725-730 (1989).

Maher, L. J., et al., Biochemistry 31(1):70-81 (1992).

Maniatis, T., et al. Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory (1982).

McGeoch, D. J., et al., J. Virol. 62:444-453 (1988).

Miller, et al., U.S. Pat. No. 4,757,055, issued 19 Jul. 1988.

Montenay-Garestier, T., et al., CIBA Found. Symp. 158:147-157.

Nakamura, S., et al., J. Antibiot., Ser. A. 17:220 (1964).

Olivo, P. D., et al., Proc. Natl. Acad. Sci. USA 85:5414-5418 (1988).

Olivo, P. D., et al., J. Virology 3:196-204 (1989).

Pelaprat, D., et al., J. Med. Chem. 23:1336-1343 (1980).

Perouault, L., et al., Nature 344:358-360 (1990).

Pitha, Biochem Biophys Acta 204:39 (1970a).

Pitha, Biopolymers 9:965 (1970b).

Pollsky, B., et al., Proc. Natl. Acad. Sci. U.S.A. 72:3310-3314 (1975).

Portugal, J., et al., FEBS Lett. 225:195-200 (1987).

Quigley, G. J., et al., Science 232:1255-1258 (1986).

Reisman, D., et al., Mol. Cell. Biol. 5:1822-1832 (1985).

Remers, W. A., Antineoplastic Agents, New York: John Wiley and Sons, Inc., 1992.

Rice, J. A., et al., Proc. Natl. Acad. Sci. USA 85:4158-4161 (1988).

Salas, X., et al., FEBS Lett. 292:223-228 (1991).

Sambrook, J., et al., In Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory Press, Vol. 2 (1989).

Schmidt, A., et al., J. Virol. 64:4037-4041 (1990).

Schuhmann, E., et al., Allg. Microbiol. 14:321 (1974).

Schultz, P. G., et al., Proc. Natl. Acad. Sci. USA 80:6834-6837 (1983).

Sherman, S. E., et al., Chem. Rev. 87:1153 (1987).

Siebenlist, U., et al., Proc. Natl. Acad. Sci. USA 122-126 (1980).

Skorobogaty, A., et al., Anti-Cancer Drug Design 3:41-56 (1988).

Smith, D. B., et al., Gene 67:31 (1988).

Sobell, H. M., et al., J. Mol. Biol. 68:21-34 (1972).

Sobell, H. M., Prof. Nucl. Acid. Res. Mol. Biol. 13:153-190 (1973).

Stow, N. D., et al., Virology 130:427-438 (1983).

Stow, N. D., et al., J. Gen. Virol. 67:1613-1623 (1986).

Strobel, S. A., et al., Science 249:73-75 (1990).

Summers, M. D., et al., A Manual of Methods for Baculovirus Vectors and Insect Cell Culture Procedures, Texas Agricultural Experimental Station Bulletin, No. 1555 (1987).

Summerton, J., et al., PCT International Application, Publication No. WO 86/05518, Published 25 Sep. 1986.

Summerton, J., et al., U.S. Pat. No. 5,034,506, issued 23 Jul. 1991.

Thrum, H., et al., Antimicrobial and Antineoplastic Chemotherapy, Prague: Czech. Med. Press, 1972. pp. 819-822 (1972).

Tullius, T. D., Ann. Rev. Biophys. Biochem. 18:213-237 (1989).

Wang, A. H.-J., et al., Science 225:1115-1121 (1984).

Wartel, R. M., et al., J. Biol. Chem. 15:285-318 (1975).

Weir, H. M., et al., Nucl. Acids Res. 17:1409-1425 (1989).

Werner, G. H., et al., Actual. Pharmaceut. Fr. 21:133 (1963).

White, R. J., et al., Biochemistry 28:6259-6269 (1989).

Woodbury, C. P., et al., Biochemistry 22(20):4730-4737 (1983).

Wu, C. A., et al., J. Virol. 62:435-443 (1988).

Young, S. L., et al., Proc. Natl. Acad. Sci. U.S.A. 88:10023-10026 (1991).

Zein, N., et al., Science 240:1198 (1988).

Zimmer, C., Pros. Nucl. Acid Res. Mol. Biol. 15:285-318 (1975).

Several classes of small molecules that interact with double-stranded DNA have been identified. Many of these small molecules have profound biological effects. For example, many aminoacridines and polycyclic hydrocarbons bind DNA and are mutagenic, teratogenic, or carcinogenic. Other small molecules that bind DNA include: biological metabolites, some of which have applications as antibiotics and antitumor agents including actinomycin D, echinomycin, distamycin, and calicheamicin; planar dyes, such as ethidium and acridine orange; and molecules that contain heavy metals, such as cisplatin, a potent antitumor drug.

The sequence binding preferences of most known DNA binding molecules have not, to date, been identified. However, several small DNA-binding molecules have been shown to preferentially recognize specific nucleotide sequences, for example: echinomycin has been shown to preferentially bind the sequence ›(A/T)CGT!/›ACG(A/T)! (Gilbert et al.); cisplatin has been shown to covalently cross-link a platinum molecule between the N7 atoms of two adjacent deoxyguanosines (Sherman et al.); and calicheamicin has been shown to preferentially bind and cleave the sequence TCCT/AGGA (Zein et al.).

Many therapeutic DNA-binding molecules (such as distamycin) that were initially identified based on their therapeutic activity in a biological screen have been later determined to bind DNA. There are several examples in the literature referring to synthetic or naturally-occurring polymers of DNA-binding drugs. Netropsin, for example, is a naturally-occurring oligopeptide that binds to the minor groove of double-stranded DNA. Netropsin contains two 4-amino-1-methylpyrrole-2-carboxylate residues and belongs to a family of similar biological metabolites from Streptomyces spp. This family includes distamycin, anthelvencin (both of which contain three N-methylpyrrole residues), noformycin, amidomycin (both of which contain one N-methylpyrrole residue) and kikumycin (which contains two N-methylpyrrole residues, like netropsin) (Debart, et al.). Synthetic molecules of this family have also been described, including the above-mentioned molecules (Lown, et al. 1985) well as dimeric derivatives (Griffin et al., Gurskii, et al.) and certain analogues (Bialer, et al. 1980, Bialer, et al. 1981, Krowicki, et al.).

Molecules in this family, particularly netropsin and distamycin, have been of interest because of their biological activity as antibacterial (Thrum et al., Schuhmann, et al.), antiparasitic (Nakamura et al.), and antiviral drugs (Becker, et al., Lown, et al. 1986, Werner, et al.).

Among the synthetic analogs of netropsin and distamycin are oligopeptides that have been designed to have sequence preferences different from their parent molecules. Such oligopeptides include the "lexitropsin" series of analogues. The N-methlypyrrole groups of the netropsin series were systematically replaced with N-methylimidazole residues, resulting in lexitropsins with increased and altered sequence specificities from the parent compounds (Kissinger, et al.). Further, a number of poly(N-methylpyrrolyl)netropsin analogues have been designed and synthesized which extend the number of residues in the oligopeptides to increase the size of the binding site (Dervan, 1986).

Co-pending, co-owned U.S. application Ser. No. 07/723,618 (herein incorporated by reference), filed 27 Jun. 1991, now U.S. Pat. No. 5,306,619, describes an in vitro assay for determining sequence preferences and relative binding affinities for DNA-binding molecules to specific DNA sequences. The present application describes the use of this assay to (i) determine the relative affinities of such a DNA-binding molecule for different DNA sequences, and (ii) use this information to design a binding agent capable of sequence-specific binding to a target duplex DNA target region. Once the relative sequence preferences of a DNA-binding molecule are defined, the information can be used to design DNA-binding molecules with sequence specificity and differential affinity that is predictably greater than the sequence specificity and differential affinity of the parent molecule(s).

In one aspect, the invention includes a method of constructing a DNA-binding agent capable of sequence-specific binding to a duplex DNA target region. The method includes identifying in the duplex DNA, a target region containing a series of at least two non-overlapping base-pair sequences of four base-pairs each, where the four base-pair sequences are adjacent, and each sequence is characterized by sequence-preferential binding to a duplex DNA-binding small molecule. The small molecules are coupled to form a DNA-binding agent capable of sequence-specific binding to said target region.

In one embodiment, the duplex-binding small molecules are identified as molecules capable of binding to a selected test sequence in a duplex DNA by first adding a molecule to be screened to a test system composed of (a) a DNA-binding protein that is effective to bind to a screening sequence in a duplex DNA, with a binding affinity that is substantially independent of the test sequence adjacent the screening sequence, but that is sensitive to binding of molecules to such test sequence, when the test sequence is adjacent the screening sequence, and (b) a duplex DNA having said screening and test sequences adjacent one another, where the binding protein is present in an amount that saturates the screening sequence in the duplex DNA.

The test molecule is incubated in the test system for a period sufficient to permit binding of the molecule being tested to the test sequence in the duplex DNA. The degree of binding protein bound to the duplex DNA before adding the test molecule is compared with that after adding the molecule. The screening sequence may be from the HSV origin of replication, and the binding protein may be UL9. Exemplary screening sequences are identified as SEQ ID NO: 1, SEQ ID NO: 2, and SEQ ID NO: 15.

Specific examples of tetrameric basepair sequences include TTTC, TTTG, TTAC, TTAG, TTGC, TTGG, TTCC, TTCG, TATC, TATG, TAAC, TAAG, TABC, TABB, TACC, TAGC sequences. A specific example of a binding small molecule capable of binding to these sequences is distamycin.

In another aspect, the invention includes a method of blocking transcriptional activity from a duplex DNA template. The method includes identifying in the duplex DNA, a binding site for a transcription factor and, adjacent the binding site, a target region having a series of at least two non-overlapping tetrameric base-pair sequences, where the four (tetrameric) base-pair sequences are adjacent and each sequence is characterized by sequence-preferential binding to a duplex DNA-binding small molecule. The sequences are contacted with a binding agent composed of the small molecules coupled to form a DNA-binding agent capable of sequence-specific binding to said target region.

The target may be selected, for example, from DNA sequences adjacent a binding site for a eucaryotic transcription factor, such as transcription factor TFIID, or a procaryotic transcription factor, such as transcription sigma factor.

Also disclosed is a DNA-binding agent capable of binding with base-sequence specificity to a target region in duplex DNA, where the target region contains at least two adjacent four base-pair sequences. The agent includes at least two subunits, where each subunit is a small molecule which has a sequence-preferential binding affinity for a sequence of four base-pairs in the target region. The subunits are coupled to form a DNA-binding agent capable of sequence-specific binding to said target region.

In one general embodiment, the agent is designed for binding to a sequence in which the two tetrameric basepair sequences are separated by at least 3 basepairs, and the small molecules in the agent are coupled to each other by a spacer molecule.

Also forming part of the invention is a method of constructing a binding agent capable of sequence-specific binding to a duplex DNA target region. The method includes identifying in the duplex DNA, a target region containing (i) a series of at least two adjacent non-overlapping base-pair sequences of four base-pairs each, where each four base-pair sequence is characterized by sequence-preferential binding to a duplex DNA-binding small molecule, and (ii) adjacent to (i) a DNA duplex region capable of forming a triplex with a third-strand oligonucleotide. The two small molecules are coupled to form a DNA-binding agent capable of sequence-specific binding to said target region, and the DNA-binding agent is attached to a third-strand oligonucleotide.

The binding of the DNA-binding agent to duplex DNA causes a shift from B form to A form DNA, allowing triplex binding between the third-strand polynucleotide and a portion of the target sequence.

Also disclosed is a triple-strand forming agent for use in practicing the method.

In still another aspect, the invention includes a method of ordering the sequence binding preferences of a DNA-binding molecule. The method includes adding a molecule to be screened to a test system composed of (a) a DNA-binding protein that is effective to bind to a screening sequence in a duplex DNA with a binding affinity that is substantially independent of such test sequence adjacent the screening sequence, but that is sensitive to binding of molecules to such test sequence, and (b) a duplex DNA having said screening and test sequences adjacent one another, where the binding protein is present in an amount that saturates the screening sequence in the duplex DNA.

The molecule in the test system is incubated for a period sufficient to permit binding of the molecule being tested to the test sequence in the duplex DNA, and the amount of binding protein bound to the duplex DNA before and after addition of the test molecule is compared. These steps are repeated using all test sequences of interest, and the sequences are then ordered on the basis of relative amounts of protein bound in the presence of the molecule for each test sequence.

The test sequences are selected, for example, from the group of 256 possible four base sequences composed of A, G, C and T. The DNA screening sequence is preferably from the HSV origin of replication, and the binding protein is preferably UL9.

FIG. 1A illustrates a DNA-binding protein binding to a screening sequence. FIGS. 1B and 1C illustrate how a DNA-binding protein may be displaced or hindered in binding by a small molecule by two different mechanisms: because of steric hinderance (1B) or because of conformational (allosteric) changes induced in the DNA by a small molecule (1C).

FIG. 2 illustrates an assay for detecting inhibitory molecules based on their ability to preferentially hinder the binding of a DNA-binding protein to its binding site. Protein (O) is displaced from DNA (/) in the presence of inhibitor (X). Two alternative capture/detection systems are illustrated, the capture and detection of unbound DNA or the capture and detection of DNA:protein complexes.

FIG. 3 shows a DNA-binding protein that is able to protect a biotin moiety, covalently attached to the oligonucleotide sequence, from being recognized by the streptavidin when the protein is bound to the DNA.

FIG. 4 shows the incorporation of biotin and digoxigenin into a typical oligonucleotide molecule for use in the assay of the present invention. The oligonucleotide contains the binding sequence (i.e., the screening sequence) of the UL9 protein, which is underlined, and test sequences flanking the screening sequence. The figure shows the preparation of double-stranded oligonucleotides end-labeled with either digoxigenin.

FIG. 5 shows a series of sequences that have been tested in the assay of the present invention for the binding of sequence-specific small molecules.

FIG. 6 outlines the cloning of a truncated form of the UL9 protein, which retains its sequence-specific DNA-binding ability (UL9-COOH), into an expression vector.

FIG. 7 shows the pVL1393 baculovirus vector containing the full length UL9 protein coding sequence.

FIG. 8 is a photograph of a SDS-polyacrylamide gel showing (i) the purified UL9-COOH/glutathione-S-transferase fusion protein and (ii) the UL9-COOH polypeptide.

FIGS. 9A and 9B shows the effect on UL9-COOH binding of alterations in the test sequences that flank the UL9 screening sequent. The data are displayed on band shift gels.

FIG. 10A shows the effect of the addition of several concentrations of Distamycin A to DNA:protein assay reactions utilizing different test sequences. FIG. 10B shows the effect of the addition of Actinomycin D to DNA:protein assay reactions utilizing different test sequences. FIG. 10C shows the effect of the addition of Doxorubicin to DNA:protein assay reactions utilizing different test sequences.

FIG. 11A illustrates a DNA capture system of the present invention utilizing biotin and streptavidin coated magnetic beads. The presence of the DNA is detected using an alkaline-phosphatase substrate that yields a chemiluminescent product. FIG. 11B shows a similar reaction using biotin coated agarose beads that are conjugated to streptavidin, that in turn is conjugated to the captured DNA.

FIG. 12 demonstrates a test matrix based on DNA:protein-binding data.

FIGS. 13A and 13B lists the top strands (5'-3') of all the possible four base pair sequences that could be used as a defined set of ordered test sequences in the assay.

FIG. 14A lists the top strands (5'-3') of all the possible four base pair sequences that have the same base composition as the sequence 5'-GATC-3'. This is another example of a defined, ordered set of sequences that could be tested in the assay. FIG. 14B presents the general sequence of a test oligonucleotide (SEQ ID NO:17), where XXXX is the test sequence and N=A,G,C, or T.

FIGS. 15A through 15F shows the results of 4 duplicate experiments in which the binding activity of distamycin was tested with all possible (256) four base pair sequences. The oligonucleotides are ranked from 1 to 256 (column 1, "rank") based on their average rank from the four experiments (column 13, "ave. rank"). (rank is shown in the first column of the chart).

FIG. 16 shows the average ranks (FIG. 15) plotted against the ideal ranks 1 to 256.

FIG. 17 shows the average r% scores (FIG. 15) plotted against the rank of 1 to 256.

FIGS. 18A to 18F shows the results of one experiment with actinomycin D. The n% scores and rank are shown for each of the 256 oligonucleotides.

FIG. 19 shows the individual n% scores (FIG. 18) plotted against the rank of 1 to 256.

FIG. 20 shows the top strands of 16 possible duplex DNA target sites for binding bis-distamycins.

FIG. 21 shows examples of bis-distamycin target sequences for bis-distamycins with internal flexible and/or variable length linkers targeted to sites comprised of two TTCC sequences, where N is any base.

FIGS. 22A to 22H show sample oligonucleotides for competition binding studies using the assay of the present invention.

FIG. 23 shows the DNA sequences of the HIV pro-viral promoter region. Several transcription factor binding sites are marked.

Definitions

Adjacent is used to describe the distance relationship between two neighboring sites. Adjacent sites are 20 or less bp apart, and can be separated any fewer number of bases including the situation where the sites are immediately abutting one another. "Flanking" is a synonym for adjacent.

Bound DNA, as used in this disclosure, refers to the DNA that is bound by the protein used in the assay (i.e., in the examples of this disclosure, the UL9 protein).

Coding sequences or coding regions are DNA sequences that code for RNA transcripts, unless specified otherwise.

Dissociation is the process by which two molecules cease to interact: the process occurs at a fixed average rate under specific physical conditions.

Functional binding is the noncovalent association of a protein or small molecule to the DNA molecule. In the assay of the present invention the functional binding of the protein to the screening sequence (i.e., its cognate DNA binding site) has been evaluated using filter binding or gel band-shift experiments.

Half-life is herein defined as the time required for one-half of the associated complexes, e.g., DNA:protein complexes, to dissociate.

Heteropolymers are molecules comprised of at least two different subunits, each representing a different type or class of molecule. The covalent coupling of different subunits, such as, DNA-binding molecules or portions of DNA-binding molecules, results in the formation of a heteropolymer: for example, the coupling of a non-intercalating homopolymeric DNA-binding molecule, such as distamycin, to an intercalating drug, such as daunomycin. Likewise, the coupling of netropsin, which is essentially a molecular subunit of distamycin, to daunomycin would also be a heteropolymer. As a further example, the coupling of distamycin, netropsin, or daunomycin to a DNA-binding homopolymer, such as a triplex-forming oligonucleotide, would be a heteropolymer.

Homopolymers are molecules that are comprised of a repeating subunit of the same type or class. Two examples of duplex DNA-binding homopolymers are as follows: (i) triplex-forming oligonucleotides or oligonucleotide analogs, which are composed of repeating subunits of nucleotides or nucleotide analogs, and (ii) oligopeptides, which are composed of repeating subunits linked by peptide bonds (e.g., distamycin, netropsin).

Sequence-preferential binding refers to DNA binding molecules that generally bind DNA but that show preference for binding to some DNA sequences over others. Sequence-preferential binding is typified by several of the small molecules tested in the present disclosure, e.g., distamycin. Sequence-preferential and sequence-specific binding can be evaluated using a test matrix such as is presented in FIG. 12. For a given DNA-binding molecule, there are a spectrum of differential affinities for different DNA sequences ranging from non-sequence-specific (no detectable preference) to sequence preferential to absolute sequence specificity (i.e., the recognition of only a single sequence among all possible sequences, as is the case with many restriction endonucleases).

Sequence-specific binding refers to DNA binding molecules which have a strong DNA sequence-preferential binding preference. For example, the following demonstrate typical sequence-specific DNA-binding: (i) multimers (heteropolymers and homopolymers) of the present invention (II i.e. 1, Multimerization; Example 13), and (ii) restriction enzymes and the proteins listed in Table V.

Screening sequence is the DNA sequence that defines the cognate binding site for the DNA binding protein: in the case of UL9 the screening sequence can, for example, be SEQ ID NO:1.

Small molecules are desirable as therapeutics for several reasons related to drug delivery: (i) they are commonly less than 10K molecular weight; (ii) they are more likely to be permeable to cells; (iii) unlike peptides or oligonucleotides, they are less susceptible to degradation by many cellular mechanisms; and, (iv) they are not as apt to elicit an immune response. Many pharmaceutical companies have extensive libraries of chemical and/or biological mixtures, often fungal, bacterial, or algal extracts, that would be desirable to screen with the assay of the present invention. Small molecules may be either biological or synthetic organic compounds, or even inorganic compounds (i.e., cisplatin).

Test sequence is a DNA sequence adjacent the screening sequence. The assay of the present invention screens for molecules that, when bound to the test sequence, affect the interaction of the DNA-binding protein with its cognate binding site (i.e., the screening sequence). Test sequences can be placed adjacent either or both ends of the screening sequence. Typically, binding of molecules to the test sequence interferes with the binding of the DNA-binding protein to the screening sequence. However, some molecules binding to these sequences may have the reverse effect, causing an increased binding affinity of the DNA-binding protein to the screening sequence. Some molecules, even while binding in a sequence specific or sequence preferential manner, might have no effect in the assay. These molecules would not be detected in the assay.

Unbound DNA, as used in this disclosure, refers to the DNA that is not bound by the protein used in the assay (i.e., in the examples of this disclosure, the UL9 protein).

I. The Assay

One feature of the present invention is that it provides an assay to identify small molecules that will bind in a sequence-specific manner to medically significant DNA target sites. The assay facilitates the development of a new field of pharmaceuticals that operate by interfering with specific DNA functions, such as crucial DNA:protein interactions. A sensitive, well-controlled assay to detect DNA-binding molecules and to determine their sequence-specificity and affinity has been developed. The assay can be used to screen large biological and chemical libraries; for example, the assay will be used to detect sequence-specific DNA-binding molecules in fermentation broths or extracts from various microorganisms. Furthermore, another application for the assay is to determine the sequence specificity and relative affinities of known DNA-binding drugs (and other DNA-binding molecules) for different DNA sequences. The drugs, which are currently used primarily as antibiotics or anticancer drugs, may have previously unidentified activities that make them strong candidates for therapeutics or therapeutic precursors in entirely different areas of medicine. The use of the assay to determine the sequence-binding preference of these known DNA-binding molecules enables the rational design of novel DNA-binding molecules with enhanced sequence-binding preference. The methods for designing and testing these novel DNA-binding molecules is described below.

The screening assay of the present invention is basically a competition assay that is designed to test the ability of a test molecule to compete with a DNA-binding protein for binding to a short, synthetic, double-stranded oligodeoxynucleotide that contains the recognition sequence for the DNA-binding protein flanked on either or both sides by a variable test site. The variable test site may contain any DNA sequence that provides a reasonable recognition sequence for a DNA-binding test molecule. Molecules that bind to the test site alter the binding characteristics of the protein in a manner that can be readily detected. The extent to which such molecules are able to alter the binding characteristics of the protein is likely to be directly proportional to the affinity of the test molecule for the DNA test site. The relative affinity of a given molecule for different oligonucleotide sequences at the test site (i.e., test sequences) can be established by examining the molecule's effect on the DNA:protein interaction using each of the test sequences.

The assay can be used to test specific target sequences to identify novel DNA-binding molecules. Also, the assay provides a means for the determination of the high affinity DNA binding sites for a given DNA-binding molecule, thus facilitating the identification of specific target sequences.

A. General Considerations

The assay of the present invention has been designed for detecting test molecules or compounds that affect the rate of transfer of a specific DNA molecule from one protein molecule to another identical protein in solution.

A mixture of DNA and protein is prepared in solution. The concentration of protein is in excess to the concentration of the DNA so that virtually all of the DNA is found in DNA:protein complexes. The DNA is a double-stranded oligonucleotide that contains the recognition sequence for a specific DNA-binding protein (i.e., the screening sequence). The protein used in the assay contains a DNA-binding domain that is specific for binding to the sequence within the oligonucleotide. The physical conditions of the solution (e.g., pH, salt concentration, temperature) are adjusted such that the half-life of the complex is amenable to performing the assay (optimally a half-life of 5-120 minutes), preferably in a range that is close to normal physiological conditions.

As one DNA:protein complex dissociates, the released DNA rapidly reforms a complex with another protein in solution. Since the protein is in excess to the DNA, dissociations of one complex always result in the rapid reassociation of the DNA into another DNA:protein complex. At equilibrium, very few DNA molecules will be unbound. If the unbound DNA is the component of the system that is measured, the minimum background of the assay is the amount of unbound DNA observed during any given measurable time period. If the capture/detection system used for capturing the unbound DNA is irreversible, the brevity of the observation period (the length of time used to capture the unbound DNA) and the sensitivity of the detection system define the lower limits of background DNA.

FIG. 1 illustrates how such a protein can be displaced from its cognate binding site or how a protein can be prevented from binding its cognate binding site, or how the kinetics of the DNA:protein interaction can be altered. In each case, the binding site for the test molecule is located at a site flanking the recognition sequence for the DNA-binding protein (FIG. 1A). One mechanism is steric hinderance of protein binding by a small molecule (competitive inhibition; FIG. 1B). Alternatively, a molecule may interfere with a DNA:protein binding interaction by inducing a conformational change in the DNA (allosteric interference, noncompetitive inhibition; FIG. 1C). In either event, if a test molecule that binds the oligonucleotide hinders binding of the protein, even transiently, the rate of transfer of DNA from one protein to another will be decreased. This will result in a net increase in the amount of unbound DNA and a net decrease in the amount of protein-bound DNA. In other words, an increase in the amount of unbound DNA or a decrease in the amount of bound DNA indicates the presence of an inhibitor, regardless of the mechanism of inhibition (competitive or noncompetitive).

Alternatively, molecules may be isolated that, when bound to the DNA, cause an increased affinity of the DNA-binding protein for its cognate binding site. In this case, the assay control samples (no drug added) are adjusted to less than 100% DNA:protein complex so that the increase in binding can be detected. The amount of unbound DNA (observed during a given measurable time period after the addition of the molecule) will decrease and the amount of bound DNA will increase in the reaction mixture as detected by the capture/detection system described in Section II.

B. Other Methods

There are several approaches that could be taken to look for small molecules that specifically inhibit the interaction of a given DNA-binding protein with its binding sequence (cognate site). One approach would be to test biological or chemical compounds for their ability to preferentially block the binding of one specific DNA:protein interaction but not the others. Such an assay would depend on the development of at least two, preferably three, DNA:protein interaction systems in order to establish controls for distinguishing between general DNA-binding molecules (polycations like heparin or intercalating agents like ethidium) and DNA-binding molecules having sequence binding preferences that would affect protein/cognate binding site interactions in one system but not the other(s).

One illustration of how this system could be used is as follows. Each cognate site could be placed 5' to a reporter gene (such as genes encoding .beta.-galactoside or luciferase) such that binding of the protein to the cognate site would enhance transcription of the reporter gene. The presence of a sequence-specific DNA-binding drug that blocked the DNA:protein interaction would decrease the enhancement of the reporter gene expression. Several DNA enhancers could be coupled to reporter genes, then each construct compared to one another in the presence or absence of small DNA-binding test molecules. In the case where multiple protein/cognate binding sites are used for screening, a competitive inhibitor that blocks one interaction but not the others could be identified by the lack of transcription of a reporter gene in a transfected cell line or in an in vitro assay. Only one such DNA-binding sequence, specific for the protein of interest, could be screened with each assay system. This approach has a number of limitations including limited testing capability and the need to construct the appropriate reporter system for each different protein/cognate site of interest.

Another example of a system to detect sequence-specific DNA-binding molecules would involve cloning a DNA-binding protein of interest, expressing the protein in an expression system (e.g., bacterial, baculovirus, or mammalian expression systems), preparing a purified or partially purified sample of protein, then using the protein in an in vitro competition assay to detect molecules that blocked the DNA:protein interaction. These types of systems are analogous to many receptor:ligand or enzyme:substrate screening assays developed in the past, but have the same limitations as outlined above in that a new system must be developed for every different protein/cognate site of interest and the capacity for different sequences is therefore limited.

Another example of a system designed to detect sequence-specific DNA-binding drugs would be the use of DNA footprinting procedures as described in the literature. These methods include DNase I or other nuclease footprinting (Chaires, et al.), hydroxy radical footprinting (Portugal, et al.), methidiumpropyl EDTA(iron) complex footprinting (Schultz, et al.), photofootprinting (Jeppesen, et al.), and bidirectional transcription footprinting (White, et al.). These procedures are likely to be accurate within the limits of their sequence testing capability but are seriously limited by (i) the number of different DNA sequences that can be used in one experiment (typically one test sequence that represents the binding site of the DNA-binding protein under study), and (ii) the difficulty of developing high throughput screening systems.

C. Choosing and Testing an Appropriate DNA-Binding Protein

Experiments performed in support of the present invention have defined a second approach for identifying molecules having sequence-preferential DNA-binding. In this approach small molecules binding to sequences adjacent the cognate binding sequence can inhibit the protein/cognate DNA interaction. This assay has been designed to use a single DNA:protein interaction to screen for sequence-specific or sequence-preferential DNA-binding molecules that recognize virtually any sequence.

While DNA-binding recognition sites are usually quite small (4-17 bp), the sequence that is protected by the binding protein is larger (usually 5 bp or more on either side of the recognition sequence--as detected by DNAase I protection (Galas et al.) or methylation interference (Siebenlist et al.). Experiments performed in support of the present invention demonstrated that a single protein and its cognate DNA-binding sequence can be used to assay virtually any DNA sequence by placing a sequence of interest adjacent to the cognate site: a small molecule bound to the adjacent site can be detected by alterations in the binding characteristics of the protein to its cognate site. Such alterations might occur by either steric hindrance, which would cause the dissociation of the protein, or induced conformational changes in the recognition sequence for the protein, which may cause either enhanced binding or more likely, decreased binding of the protein to its cognate site.

1) Criteria for choosing an appropriate DNA-binding protein. There are several considerations involved in choosing DNA:protein complexes that can be employed in the assay of the present invention including:

a) The half-life of the DNA:protein complex should be short enough to accomplish the assay in a reasonable amount of time. The interactions of some proteins with cognate sites in DNA can be measured in days not minutes: such tightly bound complexes would inconveniently lengthen the period of time it takes to perform the assay.

b) The half-life of the complex should be long enough to allow the measurement of unbound DNA in a reasonable amount of time. For example, the level of free DNA is dictated by the ratio between the time needed to measure free DNA and the amount of free DNA that occurs naturally due to the dissociation of the complex during the measurement time period.

In view of the above two considerations, practical useful DNA:protein half-lives fall in the range of approximately two minutes to several days, although shorter half-lives may be accommodated by faster equipment and longer half-lives may be accommodated by destabilizing the binding conditions for the assay.

c) A further consideration is that the kinetic interactions of the DNA:protein complex is relatively insensitive to the nucleotide sequences flanking the recognition sequence. The affinity of many DNA-binding proteins is affected by differences in the sequences adjacent to the recognition sequence. The most obvious example of this phenomenon is the preferential binding and cleavage of restriction enzymes given a choice of several identical recognition sequences with different flanking sequences (Polinsky et al.). If the half-life of the complex is affected by the flanking sequence, the analysis of comparative binding data between different flanking oligonucleotide sequences becomes difficult but is not impossible.

2) Testing DNA:Protein Interactions for Use in the Assay

Experiments performed in support of the present invention have identified a DNA:protein interaction that is particularly useful for the above described assay: the Herpes Simplex Virus (HSV) UL9 protein that binds the HSV origin of replication (oriS). The UL9 protein has fairly stringent sequence specificity. There appear to be three binding sites for UL9 in oriS, SEQ ID NO:1, SEQ ID NO:2 and SEQ ID NO:17 (Elias, P. et al., Stow et al.). One sequence (SEQ ID NO:1) binds with at least 10-fold higher affinity than the second sequence (SEQ ID NO:2): the embodiments described below use the higher affinity binding site (SEQ ID NO:1).

DNA:protein association reactions are performed in solution. The DNA:protein complexes can be separated from free DNA by any of several methods. One particularly useful method for the initial study of DNA:protein interactions has been visualization of binding results using band shift gels (Example 3A). In this method DNA:protein binding reactions are applied to polyacrylamide/TBE gels and the labelled complexes and free labeled DNA are separated electrophoretically. These gels are fixed, dried, and exposed to X-ray film. The resulting autoradiograms are examined for the amount of free probe that is migrating separately from the DNA:protein complex. These assays include (i) a lane containing only free labeled probe, and (ii) a lane where the sample is labeled probe in the presence of a large excess of binding protein. The band shift assays allow visualization of the ratios between DNA:protein complexes and free probe. However, they are less accurate than filter binding assays for rate-determining experiments due to the lag time between loading the gel and electrophoretic separation of the components.

The filter binding method is particularly useful in determining the half-life for protein:oligonucleotide complexes (Example 3B). In the filter binding assay, DNA:protein complexes are retained on a filter while free DNA passes through the filter. This assay method is more accurate for half-life determinations because the separation of DNA:protein complexes from free probe is very rapid. The disadvantage of filter binding is that the nature of the DNA:protein complex cannot be directly visualized. So if, for example, the competing molecule was also a protein competing for the binding of a site on the DNA molecule, filter binding assays cannot differentiate between the binding of the two proteins nor yield information about whether one or both proteins are binding.

There are many known DNA:protein interactions that may be useful in the practice of the present invention, including (i) the DNA protein interactions listed in Table V, (ii) bacterial, yeast, and phage systems such as lambda o.sub.L -o.sub.R /cro, and (iii) modified restriction enzyme systems (e.g., protein binding in the absence of divalent cations). Any protein that binds to a specific recognition sequence may be useful in the present invention. One constraining factor is the effect of the immediately adjacent sequences (the test sequences) on the affinity of the protein for its recognition sequence. DNA:protein interactions in which there is little or no effect of the test sequences on the affinity of the protein for its cognate site are preferable for use in the described assay; however, DNA:protein interactions that exhibit (test sequence-dependent) differential binding may still be useful if algorithms are applied to the analysis of data that compensate for the differential affinity. In general, the effect of flanking sequence composition on the binding of the protein is likely to be correlated to the length of the recognition sequence for the DNA-binding protein. In short, the kinetics of binding for proteins with shorter recognition sequences are more likely to suffer from flanking sequence effects, while the kinetics of binding for proteins with longer recognition sequences are more likely to not be affected by flanking sequence composition. The present disclosure provides methods and guidance for testing the usefulness of such DNA:protein interactions, i.e., other than the UL9 oriS binding site interaction, in the screening assay.

D. Preparation of Full Length UL9 and UL9-COOH Polypeptides

UL9 protein has been prepared by a number of recombinant techniques (Example 2). The full length UL9 protein has been prepared from baculovirus infected insect cultures (Example 3A, B, and C). Further, a portion of the UL9 protein that contains the DNA-binding domain (UL9-COOH) has been cloned into a bacterial expression vector and produced by bacterial cells (Example 3D and E). The DNA-binding domain of UL9 is contained within the C-terminal 317 amino acids of the protein (Weir et al.). The UL9-COOH polypeptide was inserted into the expression vector in-frame with the glutathione-S-transferase (gst) protein. The gst/UL9 fusion protein was purified using affinity chromatography (Example 3E). The vector also contained a thrombin cleavage site at the junction of the two polypeptides. Therefore, once the fusion protein was isolated (FIG. 8, lane 2) it was treated with thrombin, cleaving the UL9-COOH/gst fusion protein from the gst polypeptide (FIG. 8, lane 3). The UL9-COOH-gst fusion polypeptide was obtained at a protein purity of greater than 95% as determined using Coomaisie staining.

Other hybrid proteins can be utilized to prepare DNA-binding proteins of interest. For example, fusing a DNA-binding protein coding sequence in-frame with a sequence encoding the thrombin site and also in-frame with the .beta.-galactoside coding sequence. Such hybrid proteins can be isolated by affinity or immunoaffinity columns (Maniatis et al.; Pierce, Rockford Ill.). Further, DNA-binding proteins can be isolated by affinity chromatography based on their ability to interact with their cognate DNA binding site. For example, the UL9 DNA-binding site (SEQ ID NO:1) can be covalently linked to a solid support (e.g., CnBr-activated Sepharose 4B beads, Pharmacia, Piscataway N.J.), extracts passed over the support, the support washed, and the DNA-binding then isolated from the support with a salt gradient (Kadonaga). Alternatively, other expression systems in bacteria, yeast, insect cells or mammalian cells can be used to express adequate levels of a DNA-binding protein for use in this assay.

The results presented below in regard to the DNA-binding ability of the truncated UL9 protein suggest that full length DNA-binding proteins are not required for the DNA:protein assay of the present invention: only a portion of the protein containing the cognate site recognition function may be required. The portion of a DNA-binding protein required for DNA-binding can be evaluated using a functional binding assay (Example 4A). The rate of dissociation can be evaluated (Example 4B) and compared to that of the full length DNA-binding protein. However, any DNA-binding peptide, truncated or full length, may be used in the assay if it meets the criteria outlined in part I.C.1, "Criteria for choosing an appropriate DNA-binding protein". This remains true whether or not the truncated form of the DNA-binding protein has the same affinity as the full length DNA-binding protein.

E. Functional Binding and Rate of Dissociation

The full length UL9 and purified UL9-COOH proteins were tested for functional activity in "band shift" assays (see Example 4A). The buffer conditions were optimized for DNA:protein-binding (Example 4C) using the UL9-COOH polypeptide. These DNA-binding conditions also worked well for the full-length UL9 protein. Radiolabeled oligonucleotides (SEQ ID NO:14) that contained the 11 bp UL9 DNA-binding recognition sequence (SEQ ID NO:1) were mixed with each UL9 protein in appropriate binding buffer. The reactions were incubated at room temperature for 10 minutes (binding occurs in less than 2 minutes) and the products were separated electrophoretically on non-denaturing polyacrylamide gels (Example 4A). The degree of DNA:protein-binding could be determined from the ratio of labeled probe present in DNA:protein complexes versus that present as free probe. This ratio was typically determined by optical scanning of autoradiograms and comparison of band intensities. Other standard methods may be used as well for this determination, such as scintillation counting of excised bands. The UL9-COOH polypeptide and the full length UL9 polypeptide, in their respective buffer conditions, bound the target oligonucleotide equally well.

The rate of dissociation was determined using competition assays. An excess of unlabelled oligonucleotide that contained the UL9 binding site was added to each reaction. This unlabelled oligonucleotide acts as a specific inhibitor, capturing the UL9 protein as it dissociates from the labelled oligonucleotide (Example 4B). The dissociation rate, as determined by a band-shift assay, for both full length UL9 and UL9-COOH was approximately 4 hours at 4.degree. C. or approximately 10 minutes at room temperature. Neither non-specific oligonucleotides (a 10,000-fold excess) nor sheared herring sperm DNA (a 100,000-fold excess) competed for binding with the oligonucleotide containing the UL9 binding site.

F. oriS Flanking Sequence Variation

As mentioned above, one feature of a DNA:protein-binding system for use in the assay of the present invention is that the DNA:protein interaction is not affected by the nucleotide sequence of the regions adjacent the DNA-binding site. The sensitivity of any DNA:protein-binding reaction to the composition of the flanking sequences can be evaluated by the functional binding assay and dissociation assay described above.

To test the effect of flanking sequence variation on UL9 binding to the otiS SEQ ID NO:1 sequences oligonucleotides were constructed with 20-30 different sequences (i.e., the test sequences) flanking the 5' and 3' sides of the UL9 binding site. Further, oligonucleotides were constructed with point mutations at several positions within the UL9 binding site. Most point mutations within the binding site destroyed recognition. Several changes did not destroy recognition and these include variations at sites that differ between the three UL9 binding sites (SEQ ID NO:1 and SEQ ID NO:2 and SEQ ID NO:17): the second UL9 binding site (SEQ ID NO:2) shows a ten-fold decrease in UL9:DNA binding affinity (Elias et al.) relative to the first (SEQ ID NO:1). On the other hand, sequence variation at the test site (also called the test sequence), adjacent to the screening site (FIG. 5, Example 5), had virtually no effect on binding or the rate of dissociation.

The results demonstrating that the nucleotide sequence in the test site, which flanks the screening site, has no effect on the kinetics of UL9 binding in any of the oligonucleotides tested is a striking result. This allows the direct comparison of the effect of a DNA-binding molecule on test oligonucleotides that contain different test sequences. Since the only difference between test oligonucleotides is the difference in nucleotide sequence at the test site(s), and since the nucleotide sequence at the test site has no effect on UL9 binding, any differential effect observed between the two test oligonucleotides in response to a DNA-binding molecule must be due solely to the differential interaction of the DNA-binding molecule with the test sequence(s). In this manner, the insensitivity of UL9 to the test sequences flanking the UL9 binding site greatly facilitates the interpretation of results. Each test oligonucleotide acts as a control sample for all other test oligonucleotides. This is particularly true when ordered sets of test sequences are tested (e.g., testing all 256 four base pair sequences (FIG. 13) for binding to a single drug).

Taken together the above experiments support that the UL9-COOH polypeptide binds the SEQ ID NO:1 sequence with (i) appropriate strength, (ii) an acceptable dissociation time, and (iii) indifference to the nucleotide sequences flanking the assay (binding) site. These features suggested that the UL9/oriS system could provide a versatile assay for detection of small molecule/DNA-binding involving any number of specific nucleotide sequences.

The above-described experiment can be used to screen other DNA:protein interactions to determine their usefulness in the present assay.

G. Small Molecules as Sequence-Specific Competitive Inhibitors

To test the utility of the present assay system several small molecules that have sequence preferences (i.e., a preference for AT-rich versus GC-rich sequences) have been tested.

Distamycin A binds relatively weakly to DNA (K.sub.A =2.times.10.sup.5 M.sup.-1) with a preference for non-alternating AT-rich sequences (Jain et al.; Sobell; Sobell et al.). Actinomycin D binds DNA more strongly (K.sub.A =7.6.times.10.sup.-7 M.sup.-1) than Distamycin A and has been reported to have a relatively strong preference for the dinucleotide sequence dGdC (Luck et al.; Zimmer; Wartel). Each of these molecules poses a stringent test for the assay. Distamycin A tests the sensitivity of the assay because of its relatively weak binding. Actinomycin D challenges the ability to utilize flanking sequences since the UL9 recognition sequence contains a dGdC dinucleotide: therefore, it might be anticipated that all of the oligonucleotides, regardless of the test sequence flanking the assay site, might be equally affected by actinomycin D.

In addition, Doxorubicin, a known anti-cancer agent that binds DNA in a sequence-preferential manner (Chen, K-X, et al.), has been tested for preferential DNA sequence binding using the assay of the present invention.

Actinomycin D, Distamycin A, and Doxorubicin have been tested for their ability to preferentially inhibit the binding of UL9 to oligonucleotides containing different sequences flanking the UL9 binding site (Example 6, FIG. 5). Furthermore, distamycin A and actinomycin D have been screened against all possible 256 4 bp DNA sequences. Binding assays were performed as described in Example 5. These studies were completed under conditions in which UL9 is in excess of the DNA (i.e., most of the DNA is in complex).

In the preliminary studies, distamycin A was tested with 5 different test sequences flanking the UL9 screening sequence: SEQ ID NO:5 to SEQ ID NO:9. The results shown in FIG. 10A demonstrate that distamycin A preferentially disrupts binding to the test sequences UL9 polyT, UL9 polyA and, to a lesser extent, UL9 ATAT. FIG. 10A also shows the concentration dependence of the inhibitory effect of distamycin A: at 1 .mu.M distamycin A most of the DNA:protein complexes are intact (top band) with free probe appearing in the UL9 polyT and UL9 polyA lanes, and some free probe appearing in the UL9 ATAT lane; at 4 .mu.M free probe can be seen in the UL9 polyT and UL9 polyA lanes; at 16 .mu.M free probe can be seen in the UL9 polyT and UL9 polyA lanes; and at 40 .mu.M the DNA:protein in the polyT, UL9 polyA and UL9 ATAT lanes are near completely disrupted while some DNA:protein complexes in the other lanes persist. These results were consistent with the reported preference of Distamycin A for non-alternating AT-rich sequences.

Actinomycin D was tested with 8 different test sequences flanking the UL9 screening sequence: SEQ ID NO:5 to SEQ ID NO:9, and SEQ ID NO:11 to SEQ ID NO:13. The results shown in FIG. 10B demonstrate that actinomycin D preferentially disrupts the binding of UL9-COOH to the oligonucleotides UL9 CCCG (SEQ ID NO:5) and UL9 GGGC (SEQ ID NO:6). These oligonucleotides contain, respectively, three or five dGdC dinucleotides in addition to the dGdC dinucleotide within the UL9 recognition sequence. This result is consistent with the results described in the literature for Actinomycin D binding to the dinucleotide sequence dGdC. Apparently the presence of a potential preferred target site within the screening sequence (oriS, SEQ ID NO:1), as mentioned above, does not interfere with the function of the assay.

Doxorubicin was tested with 8 different test sequences flanking the UL9 screening sequence: SEQ ID NO:5 to SEQ ID NO:9, and SEQ ID NO:11 to SEQ ID NO:13. The results shown in FIG. 10C demonstrate that Doxorubicin preferentially disrupts binding to oriEco3, the test sequence of which differs from oriEco2 by only one base (compare SEQ ID NO:12 and SEQ ID NO:13). FIG. 10C also shows the concentration dependence of the inhibitory effect of Doxorubicin: at 15 .mu.M Doxorubicin, the UL9 binding to the screening sequence is strongly affected when oriEco3 is the test sequence, and more mildly affected when polyT, UL9 GGGC, or oriEco2 was the test sequence; and at 35 .mu.M Doxorubicin most DNA:protein complexes are nearly completely disrupted, with UL9 polyT and UL9ATAT showing some DNA still complexed with protein. Also, effects similar to those observed at 15 .mu.M were also observed using Doxorubicin at 150 nM, but at a later time point.

The feasibility studies performed with the limited set of test sequences, described above, provided evidence that the results of the assay are not inconsistent with the results reported in the literature. However, the screening of all possible 256 four base-pair sequences, using the assay of the present invention, provides a much more extensive overview of the sequence preferences of distamycin A and actinomycin D.

First, the results obtained in the feasibility studies with both distamycin A and actinomycin D were corroborated by the results obtained in the screen of all 256 sequences. In other words, the rank of the oligonucleotides remained the same in the larger screen. Second, the screens of distamycin A and actinomycin D both support the general hypotheses described in the literature: that is, distamycin A has a preference for binding AT-rich sequences while actinomycin D has a preference for binding GC-rich sequences. However, both drug screens of all possible 4 bp sequences revealed additional characteristics that have not been described in the literature.

Based on the data from 4 separate experiments (Examples 10 and 11; FIGS. 15, 16 and 17), consensus sequences can be derived for distamycin binding. One consensus sequence (Example 11) is relatively AT-rich, although the preference in the 4th base position is distinctly G or C. The other consensus sequence (Example 11) is relatively GC-rich, with some of the sequences having a 75% GC-content. The assay data is consistent with distamycin binding data shown in the literature, but the present assay provides additional sequence preference data not previously uncovered.

The actinomycin D data (Example 12; FIGS. 18 and 19) is also consistent with much of the published data. Further, the data from the assay of the present invention demonstrate that 8 of the top-ranking 35 test sequences in the actinomycin D screen are self-complementary sequences. There are 16 self-complementary sequences among the 256 4 bp test sequences. The rank of these sequences as determined by the assay are presented in Table I.

The assay data shown here would suggest a distinct preference of actinomycin D for self-complementary 4 bp sequences that begin with G, or next best, T, with a distinct repulsion from self-complementary sequences that begin with A or C. This data is supported by data in the literature that found actinomycin to bind with high affinity to DNA duplex containing the sequence, 5'-TCGA-3'. Furthermore, the assay suggests that self-complementary may be important for recognition and, at least in the case of self-complementary sequences, that the 1st and 4th bases may be more important in the recognition process than the two internal base pairs. These observations underscore the potential usefulness of the assay system in uncovering the mechanisms of macromolecular recognition sequences.

The ability of the assay to distinguish sequence binding preference using weak DNA-binding molecules with relatively poor sequence-specificity (such as distamycin A) is a stringent test of the assay. Accordingly, the present assay seems well-suited for the identification of molecules having better sequence specificity and/or higher sequence binding affinity. Further, the results demonstrate sequence preferential binding with the known anti-cancer drug Doxorubicin. This result indicates the assay may be useful for screening mixtures for molecules displaying similar characteristics that could be subsequently tested for anti-cancer activities as well as sequence-specific binding.

Other compounds that may be suitable for testing in the present DNA:protein system or for defining alternate DNA:protein systems include the following categories of DNA-binding molecules.

A first category of DNA-binding molecules includes non-intercalating major and minor groove DNA-binding molecules. For example, two major classes of major groove binding molecules are DNA-binding proteins (or peptides) and nucleic acids (or nucleic acid analogs such as those with peptide or morpholino backbones) capable of forming triplex DNA. There are a number of non-intercalating minor groove DNA-binding molecules including, but not limited to the following: distamycin A, netropsin, mithramycin, chromomycin and oligomycin, which are used as antitumor agents and antibiotics; and synthetic antitumor agents such as berenil, phthalanilides, aromatic bisguanylhydrazones and bisquaternary ammonium heterocycles (for review, see Baguley, 1982). Non-intercalating DNA-binding molecules vary greatly in structure: for example, the netropsin-distamycin series are oligopeptides compared to the diarylamidines berenil and stilbamidine.

A second category of DNA-binding molecules includes intercalating DNA-binding molecules. Intercalating agents are an entirely different class of DNA-binding molecules that have been identified as antitumor therapeutics and include molecules such as daunomycin (Chaires, et al.) and nogalomycin (Fox, et al., 1988)(see Remers, 1984).

A third category of DNA-binding molecules includes molecules that have both groove-binding and intercalating properties. DNA-binding molecules that have both intercalating and minor groove binding properties include actinomycin D (Goodisman, et al.), echinomycin (Fox, et al. 1990), triostin A (Wang, et al.), and luzopeptin (Fox, 1988). In general, these molecules have one or two planar polycyclic moieties and one or two cyclic oligopeptides. Luzopeptins, for instance, contain two substituted quinoline chromophores linked by a cyclic decadepsipeptide. They are closely related to the quinoxaline family, which includes echinomycin and triostin A, although they luzopeptins have ten amino acids in the cyclic peptide, while the quinoxaline family members have eight amino acids.

In addition to the major classes of DNA-binding molecules, there are also some small inorganic molecules, such as cobalt hexamine, which is known to induce Z-DNA formation in regions that contain repetitive GC sequences (Gessner et al.). Another example is cisplatin, cis-di-amminedichloroplatinum(II), which is a widely used anticancer therapeutic. Cisplatin forms a covalent intrastrand crosslink between the N7 atoms of adjacent guanosines (Rice, et al.).

Furthermore, there are a few molecules, such as calichemicin, that have unusual biochemical structures that do not fall in any of the major categories. Calichemicin is an antitumor antibiotic that cleaves DNA and is thought to recognize DNA sequences through carbohydrate moieties (Hawley, et al.). Several DNA-binding molecules, such as daunomycin, A447C, and cosmomycin B have sugar group, which may play a role in the recognition process.

Limited sequence preferences for some of the above drugs have been suggested: for example, echinomycin is thought to preferentially bind to the sequence (A/T) CGT (Fox, et al.). However, the absolute sequence preferences of the known DNA-binding drugs have never been demonstrated. Despite the large number of publications in this field, prior to the development of the assay described herein, no methods were available for determining sequence preferences among all possible binding sequences.

H. Theoretical Considerations on the Concentration of Assay Components

There are only two components in the assay, the test oligonucleotide (i.e., the test sequence) and the DNA-binding domain of UL9, which is described below. A number of theoretical considerations have been employed in establishing the assay system. In one embodiment of the invention, the assay is used as a mass-screening assay: in this embodiment the smallest volumes and concentrations possible were desirable. Each assay typically uses about 0.1-0.5 ng DNA in a 15-20 .mu.l reaction volume (approximately 0.3-1.5 nM). The protein concentration is in excess and can be varied to increase or decrease the sensitivity of the assay. In the simplest scenario (steric hindrance), where the small molecule is acting as a competitive inhibitor and the ratio of protein:DNA and DNA-binding test molecule:DNA is 1:1, the system kinetics can be described by the following equations:

and

D=DNA, P=protein, X=DNA-binding molecule, k.sub.fp and k.sub.fx are the rates of the forward reaction for the DNA:protein interaction and DNA:drug interaction, respectively, and k.sub.bp and K.sub.bx are the rates of the backwards reactions for the respective interactions. Brackets, ›!, indicate molar concentration of the components.

In the assay, both the protein, P, and the DNA-binding molecule or drug, X, are competing for the DNA. If steric hindrance is the mechanism of inhibition, the assumption can be made that the two molecules are competing for the same site. When the concentration of DNA equals the concentration of the DNA:drug or DNA:protein complex, the equilibrium binding constant, K.sub.eq, is equal to the reciprocal of the protein concentration (1/›P!). When all three components are mixed together, the relationship between the drug and the protein can be described as:

where "z" defines the difference in affinity for the DNA between P and X. For example, if z=4, then the affinity of the drug is 4-fold lower than the affinity of the protein for the DNA molecule. The concentration of X, therefore, must be 4-fold greater than the concentration of P, to compete equally for the DNA molecule. Thus, the equilibrium affinity constant of UL9 will define the minimum level of detection with respect to the concentration and/or affinity of the drug. Low affinity DNA-binding molecules will be detected only at high concentrations; likewise, high affinity molecules can be detected at relatively low concentrations. With certain test sequences, complete inhibition of UL9 binding at markedly lower concentrations than indicated by these analyses have been observed, probably indicating that certain sites among those chosen for feasibility studies have affinities higher than previously published. Note that relatively high concentrations of known drugs can be utilized for testing sequence specificity. In addition, the binding constant of UL9 can be readily lowered by altering the pH or salt concentration in the assay if it ever becomes desirable to screen for molecules that are found at low concentration (e.g., in a fermentation broth or extract).

The system kinetic analysis becomes more complex if more than one protein or drug molecule is bound by each DNA molecule. As an example, if UL9 binds as a dimer,

then the affinity constant becomes dependent on the square of the protein concentration:

The same reasoning holds true for the DNA-binding test molecule, X; if,

then the affinity constant becomes dependent on the square of the protein concentration:

Similarly, if the molar ratio of DNA to DNA-binding test molecule was 1:3, the affinity constant would be dependent on the cube of the drug concentration.

Experimentally, the ratio of molar components can be determined. Given the chemical equation:

the affinity constant may be described as

where ›! indicates concentration, D=DNA, P=protein, x=number of DNA molecules per DNA:protein complex, and y=number of protein molecules per DNA:protein complex. By determining the ratio of DNA:protein complex to free DNA, one can solve for x and y:

if a=the fraction of DNA that is free, then the fraction of DNA that is bound can be described as 1-a; and if x.sub.bound :x.sub.free (the ratio of DNA:protein complex to free DNA) is known for more than one DNA concentration. This is because the affinity constant should not vary at different DNA concentrations. Therefore,

Substituting the right side of the equation above,

Because the concentration of components in the assay can be varied and are known, the molar ratio of the components can be determined. Therefore, ›D1.sub.x P.sub.y ! and ›D2.sub.x P.sub.y ! can be described as (1-a.sub.1) ›x.sub.1 ! and (1-a.sub.2) ›x.sub.2 !, respectively, and ›D1! and ›D2! can be described as (a.sub.1) ›x.sub.1 ! and (a.sub.2) ›x.sub.2 !, respectively. ›P! remains constant and is described as (y)-(y/x) (1-a) (x) , where y is the total protein concentration and (y/x)(1-a)(x) is the protein complexed with DNA.

Experiments performed in support of the present invention give the result that x.about.1 and y.about.2. In other words, UL9 binds the DNA as a dimer.

The system kinetic analyses become more complex if the inhibition is allosteric (non-competitive inhibition) rather than competition by steric hindrance. Nonetheless, the probability that the relative effect of an inhibitor on different test sequences is due to its relative and differential affinity to the different test sequences is fairly high. This is particularly true in the assays in which all sequences within an ordered set (e.g., possible sequences of a given length or all possible variations of a certain base composition and defined length) are tested. In short, if the effect of inhibition in the assay is particularly strong for a single sequence, then it is likely that the inhibitor binds that particular sequence with higher affinity than any of the other sequences. Furthermore, while it may be difficult to determine the absolute affinity of the inhibitor, the relative affinities have a high probability of being reasonably accurate. This information will be most useful in facilitating, for instance, the refinement of molecular modeling systems.

I. The Use of the Assay Under Conditions of High Protein Concentration

When the screening protein is added to the assay system at very high concentrations, the protein binds to non-specific sites on the oligonucleotide as well as the screening sequence. This has been demonstrated using band shift gels: when serial dilutions are made of the protein and mixed with a fixed concentration of oligonucleotide, no binding (as seen by a band shift) is observed at very low dilutions (e.g., 1:100,000), a single band shift is observed at moderate dilutions (e.g., 1:100) and a smear, migrating higher than the single band observed at moderate dilutions, is observed at high concentrations of protein (e.g., 1:10). The observation of a smear is indicative of a mixed population of complexes, all of which presumably have the screening protein binding to the screening sequence with high affinity, but in addition have a larger number of proteins bound with markedly lower affinity.

Some of the low affinity binding proteins are bound to the test sequence. For example, when using the UL9-based system, the low affinity binding proteins are likely UL9 or less likely glutathione-S-transferase: these are the only proteins in the assay mixture. These proteins are significantly more sensitive to interference by a molecule binding to the test sequence for two reasons. First, the interference is likely to be by direct steric hinderance and does not rely on induced conformational changes in the DNA; secondly, the protein is a low affinity binding protein because the test site is not a cognate-binding sequence. In the case of UL9, the difference in affinity between the low affinity binding and the high affinity binding appears to be at least two orders of magnitude.

The filter binding assays capture more DNA:protein complexes when more protein is bound to the DNA. The relative results are accurate, but under moderate protein concentrations, not all of the bound DNA (as demonstrated by band shift assays) will bind to the filter unless there is more than one DNA:protein complex per oligonucleotide (e.g., in the case of UL9, more than one UL9:DNA complex). This makes the assay exquisitely sensitive under conditions of high protein concentration. For instance, when actinomycin binds DNA at a test site under conditions where there is one DNA:UL9 complex per oligonucleotide, a preference for binding GC-rich oligonucleotides has been observed; under conditions of high protein concentration, where more than one DNA:UL9 complex is found per oligonucleotide, this binding preference is even more apparent. These results suggest that the effect of actinomycin D on a test site that is weakly bound by protein may be more readily detected than the effect of actinomycin D on the adjacent screening sequence. Therefore, employing high protein concentrations may increase the sensitivity of the assay.

II. Capture/Detection Systems

As an alternative to the above described band shift gels and filter binding assays, the measurement of inhibitors can be monitored by measuring either the level of unbound DNA in the presence of test molecules or mixtures or the level of DNA:protein complex remaining in the presence of test molecules or mixtures. Measurements may be made either at equilibrium or in a kinetic assay, prior to the time at which equilibrium is reached. The type of measurement is likely to be dictated by practical factors, such as the length of time to equilibrium, which will be determined by both the kinetics of the DNA:protein interaction as well as the kinetics of the DNA:drug interaction. The results (i.e., the detection of DNA-binding molecules and/or the determination of their sequence preferences) should not vary with the type of measurement taken (kinetic or equilibrium).

FIG. 2 illustrates an assay for detecting inhibitory molecules based on their ability to preferentially hinder the binding of a DNA-binding protein. In the presence of an inhibitory molecule (X) the equilibrium between the DNA-binding protein and its binding site (screening sequence) is disrupted. The DNA-binding protein (O) is displaced from DNA (/) in the presence of inhibitor (X), the DNA free of protein or, alternatively, the DNA:protein complexes, can then be captured and detected.

For maximum sensitivity, unbound DNA and DNA:protein complexes should be sequestered from each other in an efficient and rapid manner. The method of DNA capture should allow for the rapid removal of the unbound DNA from he protein-rich mixture containing the DNA:protein complexes.

Even if the test molecules are specific in their interaction with DNA they may have relatively low affinity and they may also be weak binders of non-specific DNA or have non-specific interactions with DNA at low concentrations. In either case, their binding to DNA may only be transient, much like the transient binding of the protein in solution. Accordingly, one feature of the assay is to take a molecular snapshot of the equilibrium state of a solution comprised of the target/assay DNA, the protein, and the inhibitory test molecule. In the presence of an inhibitor, the amount of DNA that is not bound to protein will be greater than in the absence of an inhibitor. Likewise, in the presence of an inhibitor, the amount of DNA that is bound to protein will be lesser than in the absence of an inhibitor. Any method used to separate the DNA:protein complexes from unbound DNA, should be rapid, because when the capture system is applied to the solution (if the capture system is irreversible), the ratio of unbound DNA to DNA:protein complex will change at a predetermined rate, based purely on the off-rate of the DNA:protein complex. This step, therefore, determines the limits of background. Unlike the protein and inhibitor, the capture system should bind rapidly and tightly to the DNA or DNA:protein complex. The longer the capture system is left in contact with the entire mixture of unbound DNA and DNA:protein complexes in solution, the higher the background, regardless of the presence or absence of inhibitor.

Two exemplary capture systems are described below for use in the assay of the present invention. One capture system has been devised to capture unbound DNA (part II.A). The other has been devised to capture DNA:protein complexes (part II.B). Both systems are amenable to high throughput screening assays. The same detection methods can be applied to molecules captured using either capture system (part II.C)

A. Capture of Unbound DNA

One capture system that has been developed in the course of experiments performed in support of the present invention utilizes a streptavidin/biotin interaction for the rapid capture of unbound DNA from the protein-rich mixture, which includes unbound DNA, DNA:protein complexes, excess protein and the test molecules or test mixtures. Streptavidin binds with extremely high affinity to biotin (K.sub.d =10.sup.-15 M) (Chaiet et al.; Green), thus two advantages of the streptavidin/biotin system are that binding between the two molecules can be rapid and the interaction is the strongest known non-covalent interaction.

In this detection system a biotin molecule is covalently attached in the oligonucleotide screening sequence (i.e., the DNA-binding protein's binding site). This attachment is accomplished in such a manner that the binding of the DNA-binding protein to the DNA is not destroyed. Further, when the protein is bound to the biotinylated sequence, the protein prevents the binding of streptavidin to the biotin. In other words, the DNA-binding protein is able to protect the biotin from being recognized by the streptavidin. This DNA:protein interaction is illustrated in FIG. 3.

The capture system is described herein for use with the UL9/oriS system described above. The following general testing principles can, however, be applied to analysis of other DNA:protein interactions. The usefulness of this system depends on the biophysical characteristics of the particular DNA:protein interaction.

1) Modification of the Protein Recognition Sequence with Biotin

The recognition sequence for the binding of the UL9 (Koff et al.) protein is underlined in FIG. 4. Oligonucleotides were synthesized that contain the UL9 binding site and site-specifically biotinylated a number of locations throughout the binding sequence (SEQ ID NO:14; Example 1, FIG. 4). These biotinylated oligonucleotides were then used in band shift assays to determine the ability of the UL9 protein to bind to the oligonucleotide. These experiments using the biotinylated probe and a non-biotinylated probe as a control demonstrate that the presence of a biotin at the #8-T (biotinylated deoxyuridine) position of the bottom strand meets the requirements listed above: the presence of a biotin moiety at the #8 position of the bottom strand does not markedly affect the specificity of UL9 for the recognition site; further, in the presence of bound UL9, streptavidin does not recognize the presence of the biotin moiety in the oligonucleotide. Biotinylation at other A or T positions did not have the two necessary characteristics (i.e., UL9 binding and protection from streptavidin): biotinylation at the adenosine in position #8, of the top strand, prevented the binding of UL9; biotinylation of either adenosines or thymidines (top or bottom strand) at positions #3, #4, #10, or #11 all allowed binding of UL9, but in each case, streptavidin also was able to recognize the presence of the biotin moiety and thereby bind the oligonucleotide in the presence of UL9.

The above result (the ability of UL9 to bind to an oligonucleotide containing a biotin within the recognition sequence and to protect the biotin from streptavidin) was unexpected in that methylation interference data (Koff et al.) suggest that methylation of the deoxyguanosine residues at positions #7 and #9 of the recognition sequence (on either side of the biotinylated deoxyuridine) blocks UL9 binding. In these methylation interference experiments, guanosines are methylated by dimethyl sulfate at the N.sup.7 position, which corresponds structurally to the 5-position of the pyrimidine ring at which the deoxyuridine is biotinylated. These moieties all protrude into the major groove of the DNA. The methylation interference data suggest that the #7 and #9 position deoxyguanosines are contact points for UL9, it was therefore unexpected that the presence of a biotin moiety between them would not interfere with binding.

The binding of the full length protein was relatively unaffected by the presence of a biotin at position #8 within the UL9 binding site. The rate of dissociation was similar for full length UL9 with both biotinylated and un-biotinylated oligonucleotides. However, the rate of dissociation of the truncated UL9-COOH polypeptide was faster with the biotinylated oligonucleotides than with non-biotinylated oligonucleotides, which is a rate comparable to that of the full length protein with either DNA.

The binding conditions were optimized for UL9-COOH so that the half-life of the truncated UL9 from the biotinylated oligonucleotide was 5-10 minutes (optimized conditions are given in Example 4), a rate compatible with a mass screening assay. The use of multi-well plates to conduct the DNA:protein assay of the present invention is one approach to mass screening.

2) Capture of Site-Specific Biotinylated Oligonucleotides

The streptavidin:biotin interaction can be employed in several different ways to remove unbound DNA from the solution containing the DNA, protein, and test molecule or mixture. Magnetic polystyrene or agarose beads, to which streptavidin is covalently attached or attached through a covalently attached biotin, can be exposed to the solution for a brief period, then removed by use, respectively, of magnets or a filter mesh. Magnetic streptavidinated beads are currently the method of choice. Streptavidin has been used in many of these experiments, but avidin is equally useful.

An example of a second method for the removal of unbound DNA is to attach streptavidin to a filter by first linking biotin to the filter, binding streptavidin, then blocking nonspecific protein binding sites on the filter with a nonspecific protein such as albumin. The mixture is then passed through the filter, unbound DNA is captured and the bound DNA passes through the filter. This method can give high background due to partial retention of the DNA:protein complex on the filter.

One convenient method to sequester captured DNA is the use of streptavidin-conjugated superparamagnetic polystyrene beads as described in Example 7. These beads are added to the assay mixture to capture the unbound DNA. After capture of DNA, the beads can be retrieved by placing the reaction tubes in a magnetic rack, which sequesters the beads on the reaction chamber wall while the assay mixture is removed and the beads are washed. The captured DNA is then detected using one of several DNA detection systems, as described below.

Alternatively, avidin-coated agarose beads can be used. Biotinylated agarose beads (immobilized D-biotin, Pierce) are bound to avidin. Avidin, like streptavidin, has four binding sites for biotin. One of these binding sites is used to bind the avidin to the biotin that is coupled to the agarose beads via a 16 atom spacer arm: the other biotin binding sites remain available. The beads are mixed with binding mixtures to capture biotinylated DNA (Example 7). Alternative methods (Harlow et al.) to the bead capture methods just described include the following streptavidinated or avidinated supports: low-protein binding filters, or 96-well plates.

B) Capture of DNA:protein complexes. The amount of DNA:protein complex remaining in the assay mixture in the presence of an inhibitory molecule can also be determined as a measure of the relative effect of the inhibitory molecule. A net decrease in the amount of DNA:protein complex in response to a test molecule is an indication of the presence of an inhibitor. DNA molecules that are bound to protein can be captured on nitrocellulose filters. Under low salt conditions, DNA that is not bound to protein freely passes through the filter. Thus, by passing the assay mixture rapidly through a nitrocellulose filter, the DNA:protein complexes and unbound DNA molecules can be rapidly separated. This has been accomplished on nitrocellulose discs using a vacuum filter apparatus or on slot blot or dot blot apparatuses (all of which are available from Schleicher and Schuell, Keene, N.H.). The assay mixture is applied to and rapidly passes through the wetted nitrocellulose under vacuum conditions. Any apparatus employing nitrocellulose filters or other filters capable of retaining protein while allowing free DNA to pass through the filter would be suitable for this system.

C) Detection systems. For either of the above capture methods, the amount of DNA that has been captured is quantitated. The method of quantitation depends on how the DNA has been prepared. If the DNA is radioactively labelled, beads can be counted in a scintillation counter, or autoradiographs can be taken of dried gels or nitrocellulose filters. The amount of DNA has been quantitated in the latter case by a densitometer (Molecular Dynamics, Sunnyvale, Calif.); alternatively, filters or gels containing radiolabeled samples can be quantitated using a phosphoimager (Molecular Dynamics). Alternatively, the captured DNA may be detecting using a chemiluminescent or colorimetric detection system.

Radiolabelling and chemiluminescence (i) are very sensitive, allowing the detection of sub-femtomole quantities of oligonucleotide, and (ii) use well-established techniques. In the case of chemiluminescent detection, protocols have been devised to accommodate the requirements of a mass-screening assay. Non-isotopic DNA detection techniques have principally incorporated alkaline phosphatase as the detectable label given the ability of the enzyme to give a high turnover of substrate to product and the availability of substrates that yield chemiluminescent or colored products.

1) Radioactive Labeling

Many of the experiments described above for UL9 DNA:protein-binding studies have made use of radio-labelled oligonucleotides. The techniques involved in radiolabelling of oligonucleotides have been discussed above. A specific activity of 10.sup.8 -10.sup.9 dpm per .mu.g DNA is routinely achieved using standard methods (e.g., end-labeling the oligonucleotide with adenosine .gamma.-›.sup.32 P!-5' triphosphate and T4 polynucleotide kinase). This level of specific activity allows small amounts of DNA to be measured either by autoradiography of gels or filters exposed to film or by direct counting of samples in scintillation fluid.

2) Chemiluminescent Detection

For chemiluminescent detection, digoxigenin-labelled oligonucleotides (Example 1) can be detected using the chemiluminescent detection system "SOUTHERN LIGHTS," developed by Tropix, Inc. The detection system is diagrammed in FIGS. 11A and 11B. The technique can be applied to detect DNA that has been captured on either beads, filters, or in solution.

Alkaline phosphatase is coupled to the captured DNA without interfering with the capture system. To do this several methods, derived from commonly used ELISA (Harlow et al.; Pierce, Rockford Ill.) techniques, can be employed. For example, an antigenic moiety is incorporated into the DNA at sites that will not interfere with (i) the DNA:protein interaction, (ii) the DNA:drug interaction, or (iii) the capture system. In the UL9 DNA:protein/biotin system the DNA has been end-labelled with digoxigenin-11-dUTP (dig-dUTP) and terminal transferase (Example 1, FIG. 4). After the DNA was captured and removed from the DNA:protein mixture, an anti-digoxigenin-alkaline phosphatase conjugated antibody was then reacted (Boehringer Mannheim, Indianapolis Ind.) with the digoxigenin-containing oligonucleotide. The antigenic digoxigenin moiety was recognized by the antibody-enzyme conjugate. The presence of dig-dUTP altered neither the ability of UL9-COOH protein to bind the oriS SEQ ID NO:1-containing DNA nor the ability of streptavidin to bind the incorporated biotin.

Captured DNA was detected using the alkaline phosphatase-conjugated antibodies to digoxigenin as follows. One chemiluminescent substrate for alkaline phosphatase is 3-(2'-spiroadamantane)-4-methoxy-4-(3"-phosphoryloxy) phenyl-1,2-dioxetane disodium salt (AMPPD) (Example 7). Dephosphorylation of AMPPD results in an unstable compound, which decomposes, releasing a prolonged, steady emission of light at 477 nm. Light measurement is very sensitive and can detect minute quantities of DNA (e.g., 10.sup.2 -10.sup.3 attomoles) (Example 7).

Colorimetric substrates for the alkaline phosphatase system have also been tested. While the colorimetric substrates are useable in the present assay system, use of the light emission system is more sensitive.

An alternative to the above biotin capture system is to use digoxigenin in place of biotin to modify the oligonucleotide at a site protected by the DNA-binding protein at the assay site: biotin is then used to replace the digoxigenin moieties in the above described detection system. In this arrangement the anti-digoxigenin antibody is used to capture the oligonucleotide probe when it is free of bound protein. Streptavidin conjugated to alkaline phosphatase is then used to detect the presence of captured oligonucleotides.

D) Alternative methods for detecting molecules that increase the affinity of the DNA-binding protein for its cognate site. In addition to identifying molecules or compounds that cause a decreased affinity of the DNA-binding protein for the screening sequence, molecules may be identified that increase the affinity of the protein for its cognate binding site. In this case, leaving the capture system for unbound DNA in contact with the assay for increasing amounts of time allows the establishment of a fixed half-life for the DNA:protein complex (for example SEQ ID NO:1/UL9). In the presence of a stabilizing molecule, the half-life, as detected by the capture system time points, will be shortened.

Using the capture system for DNA:protein complexes to detect molecules that increase the affinity of the DNA-binding protein for the screening sequence requires that an excess of unlabeled oligonucleotide containing the UL9 binding site (but not the test sequences) is added to the assay mixture. This is, in effect, an off-rate experiment. In this case, the control sample (no test molecules or mixtures added) will show a fixed off-rate (i.e., samples would be taken at fixed intervals after the addition of the unlabeled competition DNA molecule, applied to nitrocellulose, and a decreasing amount of radiolabeled DNA:protein complex would be observed). In the presence of a DNA-binding test molecule that enhanced the binding of UL9, the off-rate would be decreased (i.e., the amount of radiolabeled DNA:protein complexes observed would not decrease as rapidly at the fixed time points as in the control sample).

III. Utility

A. The Usefulness of Sequence-Specific DNA-Binding Molecules

The present invention defines a high through-put in vitro screening assay to test large libraries of biological or chemical mixtures for the presence of DNA-binding molecules having sequence binding preference. The assay is also capable of determining the sequence-specificity and relative affinity of known DNA-binding molecules or purified unknown DNA-binding molecules. Sequence-specific DNA-binding molecules are of particular interest for several reasons, which are listed here. These reasons, in part, outline the rationale for determining the usefulness of DNA-binding molecules as therapeutic agents:

First, generally, for a given DNA:protein interaction, there are several thousands fewer target DNA-binding sequences per cell than protein molecules that bind to the DNA. Accordingly, even fairly toxic molecules might be delivered in sufficiently low concentration to exert a biological effect by binding to the target DNA sequences.

Second, DNA has a relatively more well-defined structure compared to RNA or protein. Since the general structure of DNA has less tertiary structural variation, identifying or designing specific binding molecules should be easier for DNA than for either RNA or protein. Double-stranded DNA is a repeating structure of deoxyribonucleotides that stack atop one another to form a linear helical structure. In this manner, DNA has a regularly repeating "lattice" structure that makes it particularly amenable to molecular modeling refinements and hence, drug design and development.

Third, since many single genes, of which there are only 1 or 2 copies in the cell, are transcribed into more than one, potentially as many as thousands of RNA molecules, each of which may be translated into many proteins, targeting any DNA site, whether it is a regulatory sequence, non-coding sequence or a coding sequence, may require a much lower drug dose than targeting RNAs or proteins. Proteins (eg, enzymes, receptors, or structural proteins) are currently the targets of most therapeutic agents. More recently, RNA molecules have become the targets for antisense or ribozyme therapeutic molecules.

Fourth, blocking the function of a RNA that encodes a protein or of the protein itself when that protein regulates several cellular genes may have detrimental effects, particularly if some of the regulated genes are important for the survival of the cell. However, blocking a DNA-binding site that is specific to a single gene regulated by such a protein results in reduced toxicity.

An example situation (4) is HNF-1 binding to Hepatitis B virus (HBV): HNF-1 binds an HBV enhancer sequence and stimulates transcription of HBV genes (Chang et al.). In a normal cell HNF-1 is a nuclear protein that appears to be important for the regulation of many genes, particularly liver-specific genes (Courtois et al.). If molecules were isolated that specifically bound to the DNA-binding domain of HNF-1, all of the genes regulated by HNF-1 would be down-regulated, including both vital and cellular genes. Such a drug could be lethal since many of the genes regulated by HNF-1 may be necessary for liver function. However, the assay of the present invention presents the ability to screen for a molecule that could distinguish the HNF-1 binding region of the Hepatitis B virus DNA from cellular HNF-1 sites by, for example, including divergent flanking sequences when screening for the molecule. Such a molecule would specifically block HBV expression without effecting cellular gene expression.

B. General Applications of the Assay

General applications of the assay include but are not limited to: screening libraries of unknown chemicals, either biological or synthetic compounds, for sequence-specific DNA-binding molecules (part III.B.1), determining the sequence-specificity or preference and/or relative affinities of DNA-binding molecules (part III.B.2), testing of modified derivatives of DNA-binding molecules for altered specificity or affinity (part III.B.3), using the assay in secondary confirmatory or mechanistic experiments, (part III.B.4), using the data generated from the above applications to refine the predictive capabilities of molecular modeling systems (III.B.5), and using the refined molecular modeling systems to generate a new "alphabet" of DNA-binding subunits that can be polymerized to make novel heteropolymers designed de novo to bind specific DNA target sites (part III.B.6).

1) Mass-Screening of Libraries for the Presence of Sequence-Specific DNA-Binding Molecules

Many organizations (e.g., the National Institutes of Health, pharmaceutical and chemical corporations) have large libraries of chemical or biological compounds from synthetic processes or fermentation broths or extracts that may contain as yet unidentified DNA-binding molecules. One utility of the assay is to apply the assay system to the mass-screening of these libraries of different broths, extracts, or mixtures to detect the specific samples that contain the DNA-binding molecules. Once the specific mixtures that contain the DNA-binding molecules have been identified, the assay has a further usefulness in aiding in the purification of the DNA-binding molecule from the crude mixture. As purification schemes are applied to the mixture, the assay can be used to test the fractions for DNA-binding activity. The assay is amenable to high throughput (e.g., a 96-well plate fermat automated on robotics equipment such as a Beckman Biomek workstation ›Beckman, Palo Alto, Calif.! with detection using semiautomated plate-reading densitometers, luminometers, or phosphoimagers).

The concentration of protein used in mass-screening is determined by the sensitivity desired. The screening of known compounds, as described in III.B.2, is typically performed in protein excess at a protein concentration high enough to produce 90-95% of the DNA bound in DNA:protein complex. The assay is very sensitive to discriminatory inhibition at this protein concentration. For some mass-screening, it may be desirable to operate the assay under higher protein concentration, thus decreasing the sensitivity of the assay so that only fairly high affinity molecules will be detected: for example, when screening fermentation broths with the intent of identifying high affinity binding molecules. The range of sensitivities in the assay will be determined by the absolute concentration of protein used.

One utility of the method of the present invention, under conditions using a relatively insensitive system (high ›P!:›D! ratio), is as a screening system for novel restriction enzymes. In this case, an ability to discriminate between slight differences in affinity to different sequences may not be necessary or desirable. Restriction enzymes have highly discriminatory recognition properties--the affinity constant of a restriction endonuclease for its specific recognition sequence versus non-specific sequences are orders of magnitude different from one another. The assay may be used to screen bacterial extracts for the presence of novel restriction endonucleases. The 256 test oligonucleotides described in Example 10, for example, may be used to screen for novel restriction endonucleases with 4 bp recognition sequences. The advantages of the system are that all possible 4 bp sequences are screened simultaneously, that is, it is not limited to self-complementary sequences. Further, any lack of specificity (such as, more than one binding site) is uncovered during the primary screening assay.

2 ) Directed Screening

The assay of the present invention is also useful for screening molecules that are currently described in the literature as DNA-binding molecules but with uncertain DNA-binding sequence specificity (i.e., having either no well-defined preference for binding to specific DNA sequences or having certain higher affinity binding sites but without defining the relative preference for all possible DNA binding sequences). The assay can be used to determine the specific binding sites for DNA-binding molecules, among all possible choices of sequence that bind with high, low, or moderate affinity to the DNA-binding molecule. Actinomycin D, Distamycin A, and Doxorubicin (Example 6) all provide examples of molecules with these modes of binding. Many anti-cancer drugs, such as Doxorubicin (see Example 6) show binding preference for certain identified DNA sequences, although the absolute highest and lowest specificity sequences have yet to be determined, because, until the invention described herein, the methods (Salas, X. and Portugal, J.; Cullinane, C. and Phillips, D. R.; Phillips, D. R.,; and Phillips, D. R. et al.) for detecting differential affinity DNA-binding sites for any drug were limited. Doxorubicin is one of the most widely used anti-cancer drugs currently available. As shown in Example 6, Doxorubicin is known to bind some sequences preferentially. Another example of such sequence binding preference is Daunorubicin (Chen et al.) that differs slightly in structure from Doxorubicin (Goodman et al.). Both Daunorubicin and Doxorubicin are members of the anthracycline antibiotic family: antibiotics in this family, and their derivatives, are among the most important newer antitumor agents (Goodman et al.).

The assay of the present invention allows the sequence preferences or specificities of DNA-binding molecules to be determined. The DNA-binding molecules for which sequence preference or specificity can be determined may include small molecules such as aminoacridines and polycyclic hydrocarbons, planar dyes, various DNA-binding antibiotics and anticancer drugs, as well as DNA-binding macromolecules such as peptides and polymers that bind to nucleic acids (eg, DNA and the derivatized homologs of DNA that bind to the DNA helix).

The molecules that can be tested in the assay for sequence preference/specificity and relative affinity to different DNA sites include both major and minor groove binding molecules as well as intercalating and non-intercalating DNA binding molecules.

3) Molecules Derived from Known DNA-Binding Molecules

The assay of the present invention facilitates the identification of different binding activities by molecules derived from known DNA-binding molecules. An example would be to identify and test derivatives of anti-cancer drugs with DNA-binding activity and then test for anti-cancer activity through, for example, a battery of assays performed by the National Cancer Institute (Bethesda Md.). Further, the assay of the present invention can be used to test derivatives of known anti-cancer agents to examine the effect of the modifications on DNA-binding activity and specificity. In this manner, the assay may reveal activities of anticancer agents, and derivatives of these agents, that facilitates the design of DNA-binding molecules with therapeutic or diagnostic applications in different fields, such as antiviral or antimicrobial therapeutics. The binding-activity information for any DNA-binding molecule, obtained by application of the present assay, can lead to a better understanding of the mode of action of more effective therapeutics.

4 ) Secondary Assays

As described above, the assay of the present invention is used (i) as a screening assay to detect novel DNA-binding molecules, or (ii) to determine the relative specificity and affinity of known molecules (or their derivatives). The assay may also be used in confirmatory studies or studies to elucidate the binding characteristics of DNA-binding molecules. Using the assay as a tool for secondary studies can be of significant importance to the design of novel DNA-binding molecules with altered or enhanced binding specificities and affinities.

a) Confirmatory Studies

The assay of the present invention can be used in a competition study to confirm and refine the direct binding data obtained from the assay. For example, in the screening of distamycin with all possible 256 bp sequences (Example 10), the confirmatory assay can be used (i) to confirm the rankings observed in the assay, (ii) to refine the rankings among the 5-10 highest ranked binders (which show no statistical difference in rank with data from 4 experiments), and (iii) to resolve perceived discrepancies in the assay data.

All of these goals may be accomplished using a competition experiment which determines the relative ability of test sequences to compete for the binding of distamycin.

The perceived discrepancy in the distamycin experiment is as follows: test oligonucleotides scored poorly in the assay which were complementary to most of the top-ranking test sequence oligonucleotides (Examples 10 and 11). This result was unexpected since it is unlikely that the affinity of distamycin for binding a test site depends on the orientation of the screening site to the test site. More likely, the assay detects the binding of distamycin when the molecule is bound to the test oligonucleotide in one orientation, but fails to detect the binding of distamycin when the test sequence is in the other orientation. A competition study will resolve this question, since the binding of distamycin to a competitor sequence will be orientation-independent; the competition does not depend on the mechanism of the assay.

For the competition experiment, the assay may be performed under any conditions suitable for the detection of drug binding. When these conditions are established, different competitor DNAs are added to the assay system to determine their relative ability to compete for drug binding with the radiolabeled test oligonucleotide in the assay system.

The test oligonucleotide against which competitor DNAs are tested in the competition experiment may be any oligonucleotide of interest. Generally in competition assays using a test oligonucleotide with relatively higher binding affinity for the drug will be more useful than using a test oligonucleotide with lower binding activity.

The competitor DNAs may be any sequence of interest. Several classes of DNA may be tested as competitor molecules including, but not limited to, the following: genomic DNAs, synthetic DNAs (e.g., poly(dA), poly(dI-dC), and other DNA polymers), test oligonucleotides of varying sequences, or any molecule of interest that is thought to compete for distamycin binding.

When using the competition assay to verify the results of a 256 oligonucleotide panel screen (like Example 10), the following criteria are useful for selecting the competitor test oligonucleotides:

(i) sequences that rank high in the assay but which do not have relative binding affinities with differences that are statistically significant from each other, in order to determine their relative affinity with greater precision;

(ii) sequences that are purported by other techniques (e.g., footprinting or transcriptional block analysis) to be high affinity binding sites, in order to compare the results of those techniques with the screening assay results;

(iii) sequences that are complementary to test sequences that rank high in the assay, in order to determine whether these test sequences are false negatives; and

(iv) sequences of any rank in the assay, in order to confirm the assay results.

Several methods may be used to perform the competition study as long as the relative affinities of the competing DNA molecules are detectable. One such method is described in Example 14. In this example, the concentration of the assay components (drug, protein, and DNA) is held constant relative to those used in the original screening assay, but the molar ratio of the test oligonucleotide to the competitor oligonucleotides is varied.

Another method for performing a competition assay is to hold the concentrations of protein, drug and initial amount of test oligonucleotide constant, then add a variable concentration of competitor DNA. In this design, the protein and drug concentration must be sufficiently high to allow the addition of further DNA (i.e., the competitor DNA) without i) decreasing the amount of DNA:protein complex in the absence of drug to a level that is unsuitable for detection of DNA:protein complex, and ii) increasing the amount of DNA:protein complex in the presence of drug to a level that is unsuitable for the detection of drug binding. The window between detectable DNA:protein complex and detectable effect of the drug must be wide enough to determine differences among competitor DNAs.

In any competition method, it is important that the relative concentrations of the competing DNA molecules are accurately determined. One method for accomplishing accurate determination of the relative concentrations of the DNA molecules is to tracer-label competitor molecules to a low specific activity with a common radiolabeled primer (Example 14). In this manner, the competitor molecules have the same specific activity, but are not sufficiently radioactive (200-fold less than the test oligonucleotide) to contribute to the overall radioactivity in the assay.

b) Secondary Studies to Elucidate Binding Characteristics

The studies outlined in III.B.4.a describe methods of determining some of the binding processes of distamycin A. The assay of the present invention may also be used to explore mechanistic questions about distamycin binding.

For example, several of the complements of the putative high affinity binding sites for distamycin have low scores in the assay. As described above, this may imply directionality in binding. The results may also imply that the test sites are not equal with respect to the effect exerted on UL9-COOH binding. Oligonucleotides can be designed to test the hypothesis of directionality.

The basic test oligonucleotide has the structure presented in FIG. 22A (SEQ ID NO:21). In one scenario, the score in the binding assay is high, i.e., the greatest effect of distamycin, when the test sequences is XYZZ (FIG. 22A, with the base X complementary to the base Y and the base Q complementary to the base Z), and the complement (FIG. 22B; SEQ ID NO:22) scores low. These results imply that the test sites are not equivalent with respect to their effect on UL9, otherwise the right side would have the effect in one oligonucleotide and the left site would have the effect in the other. These results further suggest that the effect of distamycin is directional. The only assumption is that distamycin should bind with the same affinity to the XYZZ/QQXY sequence (FIGS. 22A and 22B) regardless of its position or orientation in the oligonucleotide. Since the scores are derived at equilibrium, this is likely to be the case.

To test the hypothesis that one site is effective in the assay, oligonucleotides may be designed that have the UL9 site inverted with respect to the test sites (FIGS. 22C and 22D; SEQ ID NO:23 and SEQ ID NO:24, respectively). If only one site is active with respect to UL9 and if the FIG. 22A oligo was most effective in binding distamycin, then the oligo C should be less active in the assay then oligo D; in other words, flipping the UL9 site will result in QQXY ranking high, XYZZ ranking low.

Finally, to determine the "direction" of distamycin binding, mix test sequences and invert the binding site as shown in the four oligonucleotides presented in FIGS. 22E, 22F, 22G and 22H. Alternatively, one test site or the other could be deleted from the test oligonucleotide.

This type of analysis provides an example of the usefulness in the assay in determining binding properties of DNA-binding drugs.

5) Generation of Binding Data and Refinement of Molecular Modeling Systems

The assay of the present invention generates data which can be applied to the refinement of molecular modeling systems that address DNA structural analysis: the data is also useful in the design and/or refinement of DNA-binding drugs. Traditionally, mass screening has been the only reasonable method for discovering new drugs. Modern rational drug design seeks to minimize laboratory screening. However, ab initio rational drug design is difficult at this time given (i) insufficiencies in the underlying theories used for de novo design, and (ii) the computational intensity which accompanies such design approaches.

The ab initio approach requires calculations from first principles by quantum mechanics: such an approach is expensive and time-consuming. The introduction of data concerning the relative binding affinities of one or more DNA-binding molecules to all 256 four base pair DNA sequences allows the development, via molecular modeling, of ad hoc protocols for DNA structural analysis and subsequent DNA-binding drug design. The accumulation of data for the DNA sequences to which small molecules bind is likely to result in more accurate, less expensive molecular modeling programs for the analysis of DNA.

The screening capacity of the assay of the present invention is much greater than screening each separate DNA sequence with an individual cognate DNA-binding protein. Direct competition assays involving individual receptor:ligand complexes (e.g., a specific DNA:protein complex) are most commonly used for mass screening efforts, each assay requires the identification, isolation, purification, and production of the assay components. In particular, a suitable DNA:protein interactions must be identified. Using the assay of the present invention, libraries of synthetic chemicals or biological molecules can be screened to detect molecules that have preferential binding to virtually any specified DNA sequence--all using a single assay system. When employing the assay of the present invention, secondary screens involving the specific DNA:protein interaction may not be necessary, since inhibitory molecules detected in the assay may be tested directly on a biological system: for example, the ability to disrupt viral replication in a tissue culture or animal model.

6) The Design of New DNA-Binding Heteropolymers Comprised of Subunits Directed to Different DNA Sequences

The assay of the present invention will facilitate the predictive abilities of molecular modeling systems in two ways. First, ad hoc methods of structural prediction will be improved. Second, by employing pattern matching schemes, the comparison of sequences having similar or different affinities for a given set of DNA-binding molecules should empirically reveal sets of sequences that have similar structures (see part III.D, Using a Test Matrix). Molecular modeling programs are "trained" using the information concerning DNA-binding molecules and their preferred binding sequences. With this information coupled to the predicative power of molecular modeling programs, the design of DNA-binding molecules (subunits) that could be covalently linked becomes feasible.

These molecular subunits would be directed at defined sections of DNA. For example, a subunit would be designed for each possible DNA unit. For example, if single bases were the binding target of the subunits, then four subunits would be required, one to correspond to each base pair. These subunits could then be linked together to form a DNA-binding polymer, where the DNA binding preference of the polymer corresponds to the sequence binding preferences of the subunits in the particular order in which the subunits are assembled.

Another example of such a polymer is using subunits whose binding was directed at two base sections of DNA. In this case, 4.sup.2 =16 subunits would be used, each subunit having a binding affinity for a specific two base pair sequence (e.g., AA, AC, AG, AT, CA, CC, CG, CT, GA, GC, GG, GT, TA, TC, TG, TT). If the polymers were to be comprised of subunits targeted to 3 base pair sections of DNA, then 4.sup.3 =64 subunits would be prepared. The design of such molecular subunits is dependent upon the establishment of a refined database using empirical data derived by the method of the present invention, as described in Section III.B.

C. Sequences Targeted by the Assay

The DNA:protein assay of the present invention has been designed to screen for compounds that bind a full range of DNA sequences that vary in length as well as complexity. Sequence-specific DNA-binding molecules discovered by the assay have potential usefulness as either molecular reagents, therapeutics, or therapeutic precursors. Sequence-specific DNA-binding molecules are potentially powerful therapeutics for essentially any disease or condition that in some way involves DNA. Examples of test sequences for the assay include: a) binding sequences of factors involved in the maintenance or propagation of infectious agents, especially viruses, bacteria, yeast and other fungi, b) sequences causing the inappropriate expression of certain cellular genes, and c) sequences involved in the replication of rapidly growing cells. Furthermore, gene expression or replication need not necessarily be disrupted by blocking the binding of specific proteins. Specific sequences within protein-coding regions of genes (e.g., oncogenes) are equally valid test sequences since the binding of small molecules to these sequences is likely to perturb the transcription and/or replication of the region. Finally, any molecules that bind DNA with some sequence specificity, that is, not just to one particular test sequence, may be still be useful as anti-cancer agents. Several small molecules with some sequence preference are already in use as anticancer therapeutics. Molecules identified by the present assay may be particularly valuable as lead compounds for the development of congeners having either different specificity or different affinity.

One advantage of the present invention is that the assay is capable of screening for binding activity directed against any DNA sequence. Such sequences can be medically significant target sequences (see part 1, Medically Significant Target Sites, in this section), scrambled or randomly generated DNA sequences, or well-defined, ordered sets of DNA sequences (see part 2, Ordered Sets of Test Sequences, in this section), which could be used for screening for molecules demonstrating sequence preferential binding (like Doxorubicin) to determine the sequences with highest binding affinity and/or to determine the relative affinities between a large number of different sequences. There is usefulness in taking either approach for detecting and/or designing new therapeutic agents. Part 3 of this section, Theoretical Considerations for Choosing Target Sequences, outlines the theoretical considerations for choosing DNA target sites in a biological system.

1) Medically Significant Target Sequences

Few effective viral therapeutics are currently available; yet several potential target sequences for antiviral DNA-binding drugs have been well-characterized. Furthermore, with the accumulation of sequence data on all biological systems, including vital genomes, cellular genomes, pathogen genomes (bacteria, fungi, eukaryotic parasites, etc.), the number of target sites for DNA-binding drugs will increase greatly in the future.

There are numerous methods for identifying medically significant target sequences for DNA-binding drugs, including, but not limited to, the following. First, medically significant target sequences are found in pathogens of the biological kingdoms, for example in genetic sequences that are key to biochemical pathways or physiological processes. Second, a target is identified, such as (i) a pathogen involved in an infectious disease, or (ii) a biochemical pathway or physiological process of a noninfectious disease, genetic condition, or other biological process. Then specific genes that are important for the survival of the pathogen or modulation of the endogenous pathway involved in the target system is identified. Third, specific target sequences that affect the expression or activity of a DNA molecule, such as genes or sites involved in replication.

There are numerous pathogens that are potential targets for DNA-binding drugs designed using the methods described in this application. Table II lists a number of potential target pathogens.

In addition to pathogens, many non-infectious diseases may be controlled at the level of DNA. These diseases are therefore potential candidates for treatment with DNA-binding therapeutics that are discovered or designed using the methods described in this application. Table III lists a number of potential non-infectious diseases that may be targeted for treatment using DNA-binding molecules.

Once a disease or condition is identified as a potential candidate for treatment with a DNA-binding therapeutic, specific genes or other DNA sequences that are crucial for the expression of the disease associated gene (or survival of a pathogen) are identified within the biochemical or physiological pathway (or the pathogen). In humans, many genes involved in important biological functions have been identified. Virtually any DNA sequence is a potential target site for a DNA-binding molecule, including mRNA coding sequences, promoter sequences, origins of replication, and structural sequences, such as telomeres and centromeres. One of the classes of sites that may be preferable are the recognition sequences for proteins that are involved in the regulation or expression of genetic material. For this reason, genes containing the promoter regions, which bind many regulatory proteins are listed in Table IV (see also Example 15).

Once the gene target or, in the case of small pathogens, the genome target has been identified, short sequences within the gene or genome target are identified as medically significant target sites. Medically significant target sites can be defined as short DNA sequences (approximately 4-30 base pairs) that are required for the expression or replication of genetic material. For example, sequences that bind regulatory factors, either transcriptional or replicatory factors, are ideal target sites for altering gene or viral expression.

Further, coding sequences may be adequate target sites for disrupting gene function, although the disruption of a polymerase complex that is moving along the DNA sequence may require a stronger binder than for the disruption of the initial binding of a regulatory protein.

Finally, even non-coding, non-regulatory sequences may be of interest as target sites (e.g., for disrupting replication processes or introducing an increased mutational frequency).

Some specific examples of medically significant target sites are shown in Table V. For example, origin of replication binding proteins have short, well-defined binding sites within the viral genome and are therefore excellent target sites for a competitive DNA-binding drug. Examples of such proteins include, Epstein Barr virus nuclear antigen 1 (EBNA-1) (Ambinder, R. F., et al.; Reisman, D. et al.), E2 (which is encoded by the human papilloma virus) (Chin, M. T., et al.), UL9 (which is encoded by herpes simplex virus type 1) (McGeoch, D. J., et al.), and the homologous protein in varicella zoster virus (VZV) (Stow, N. D. and Davison, A. J.).

Similarly, recognition sequences for DNA-binding proteins that act as transcriptional regulatory factors are also good target sites for antiviral DNA-binding drugs. Examples listed in Table V include (i) the binding site for hepatic nuclear factor (HNF-1), which is required for the expression of human hepatitis B virus (HBV) (Chang, H.-K.), and (ii) NF.kappa.B and NFAT-1 binding sites in the human immunodeficiency virus (HIV) long terminal repeat (LTR), one or both of which may be involved in the expression of the virus (Greene, W. C.).

Examples of non-viral DNA targets for DNA-binding drugs are also shown in Table V to illustrate the wide range of potential applications for sequence-specific DNA-binding molecules. For example, nuclear factor of activated T cells (NFAT-1) is a regulatory factor that is crucial to the inducible expression of the interleukin 2 gene in response to signals from the antigen receptor, which, in turn, is required for the cascade of molecular events during T cell activation (for review, see Edwards, C. A. and Crabtree, G. R.). The mechanism of action of two immunosuppressants, cyclosporin A and FK506, is thought to be to block the inducible expression of NFAT-1 (Schmidt, A. et al. and Banerji, S. S. et al.). However, the effects of these drugs are not specific to NFAT-1; therefore, a drug targeted specifically to the NFAT-1 binding site in the IL-2 enhancer would be desirable as an improved immunosuppressant.

Targeting the DNA site with a DNA-binding drug rather than targeting with a drug that affects the DNA-binding protein (presumably the target of the current immunosuppressants) is advantageous for at least two reasons: first, there are many fewer target sites for specific DNA sequences than specific proteins (e.g., in the case of glucocorticoid receptor, a handful of DNA-binding sites vs. about 50,000 protein molecules in each cell); and second, only the targeted gene need be affected by a DNA-binding drug, while a protein-binding drug would disable all the cellular functions of the protein. An example of the latter point is the binding site for HNF-1 in the human fibrinogen promoter. Fibrinogen level is one of the most highly correlated factor with cardiovascular disease. A drug targeted to either HNF-1 or the HNF-1 binding site in the fibrinogen promoter might be used to decrease fibrinogen expression in individuals at high risk for disease because of the over-expression of fibrinogen. However, since HNF-1 is required for the expression of a number of normal hepatic genes, blocking the HNF-1 protein would be toxic to liver function. In contrast, by blocking a DNA sequence that is composed in part of the HNF-1 binding site and in part by flanking sequences for divergence, the fibrinogen gene can be targeted with a high level of selectivity, without harm to normal cellular HNF-1 functions. Further examples of medically significant DNA-binding sequences are presented in Table V.

(Abbreviations: EBV, Epstein-Barr virus; EBNA, Epstein-Barr virus nuclear antigen; HSV, Herpes Simplex virus; VZV, Varicella zoster virus; HPV, human papilloma virus; HIV LTR, Human immunodeficiency virus long terminal repeat; NFAT, nuclear factor of activated T cells; NFkB, nuclear factor kappaB; AIDS, acquired immune deficiency syndrome; ARC, AIDS related complex; HBV, hepatitis B virus; HNF, hepatic nuclear factor).

The assay has been designed to screen virtually any DNA sequence. Test sequences of medical significance include viral or microbial pathogen genomic sequences and sequences within or regulating the expression of oncogenes or other inappropriately expressed cellular genes. In addition to the detection of potential antiviral drugs, the assay of the present invention is also applicable to the detection of potential drugs for (i) disrupting the metabolism of other infectious agents, (ii) blocking or reducing the transcription of inappropriately expressed cellular genes (such as oncogenes or genes associated with certain genetic disorders), and (iii) the enhancement or alteration of expression of certain cellular genes.

2) Defined Sets of Test Sequences

The approach described in the above section emphasizes screening large numbers of fermentation broths, extracts, or other mixtures of unknowns against specific medically significant DNA target sequences. The assay can also be utilized to screen a large number of DNA sequences against known DNA-binding drugs to determine the relative affinity of the single drug for every possible defined specific sequence. For example, there are 4.sup.n possible sequences, where n=the number of nucleotides in the sequence. Thus, there are 4.sup.3 =64 different three base pair sequences, 4.sup.4 =256 different four base pair sequences, 4.sup.5 =1024 different 5 base pair sequences, etc. If these sequences are placed in the test site, the site adjacent to the screening sequence (the example used in this invention is the UL9 binding site), then each of the different test sequences can be screened against many different DNA-binding molecules.

The test sequences may be placed on either or both sides of the screening sequence, and the sequences flanking the other side of the test sequences are fixed sequences to stabilize the duplex and, on the 3' end of the top strand, to act as an annealing site for the primer (see example 1). For example, oligonucleotide sequences could be constructed as shown in Example 15 (SEQ ID NO:18). In FIG. 15, the TEST and SCREENING sequences are indicated. The preparation of such double-stranded oligonucleotides is described in Example 1 and illustrated in FIGS. 4A and 4B. The test sequences, denoted in FIG. 15 as X:Y (where X=A,C,G, or T and Y=the complementary sequence, T,G,C, or A), may be any of the 256 different 4 base pair sequences shown in FIG. 13.

Once a set of test oligonucleotides containing all possible four base pair sequences has been synthesized (see Example 1), the set can be screened with any DNA-binding drug. The relative effect of the drug on each oligonucleotide assay system will reflect the relative affinity of the drug for the test sequence. The entire spectrum of affinities for each particular DNA sequence can therefore be defined for any particular DNA-binding drug. This data, generated using the assay of the present invention, can be used to facilitate molecular modeling programs and/or be used directly to design new DNA-binding molecules with increased affinity and specificity.

Another type of ordered set of oligonucleotides that may be useful for screening are sets comprised of scrambled sequences with fixed base composition. For example, if the recognition sequence for a protein is 5'-GATC-3' and libraries were to be screened for DNA-binding molecules that recognized this sequence, then it would be desirable to screen sequences of similar size and base composition as control sequences for the assay. The most precise experiment is one in which all possible 4 bp sequences are screened. In the case of a 4 base-pair sequence, this represents 4.sup.4 =256 different test sequences: a number of screening sequences that may not be practical in every situation. However, there are many fewer possible 4 bp sequences with the same base composition (1G, 1A, 1T, 1C) (n|=24 different 4 bp sequences with this particular base composition), such sequences provide excellent controls without having to screen large numbers of sequences.

3) Theoretical Considerations in Choosing Biological Target Sites: Specificity and Toxicity

One consideration in choosing sequences to screen using the assay of the present invention is test sequence accessibility, that is, the potential exposure of the sequence in vivo to binding molecules. Cellular DNA is packaged in chromatin, rendering most sequences relatively inaccessible. Sequences that are actively transcribed, particularly those sequences that are regulatory in nature, are less protected and more accessible to both proteins and small molecules. This observation is substantiated by a large literature on DNAase I sensitivity, footprinting studies with nucleases and small molecules, and general studies on chromatin structure (Tullius). The relative accessibility of a regulatory sequence, as determined by DNAase I hypersensitivity, is likely to be several orders of magnitude greater than an inactive portion of the cellular genome. For this reason the regulatory sequences of cellular genes, as well as viral regulatory or replication sequences, are useful regions to choose for selecting specific inhibitory small molecules using the assay of the present invention.

Another consideration in choosing sequences to be screened using the assay of the present invention is the uniqueness of the potential test sequence. As discussed above for the nuclear protein HNF-1, it is desirable that small inhibitory molecules are specific to their target with minimal cross reactivity. Both sequence composition and length effect sequence uniqueness. Further, certain sequences are found less frequently in the human genome than in the genomes of other organisms, for example, mammalian viruses. Because of base composition and codon utilization differences, viral sequences are distinctly different from mammalian sequences. As one example, the dinucleotide CG is found much less frequently in mammalian cells than the dinucleotide sequence GC: further, in SV40, a mammalian virus, the sequences AGCT and ACGT are represented 34 and 0 times, respectively. Specific viral regulatory sequences can be chosen as test sequences keeping this bias in mind. Small inhibitory molecules identified which bind to such test sequences will be less likely to interfere with cellular functions.

There are approximately 3.times.10.sup.9 base pairs (bp) in the human genome. Of the known DNA-binding drugs for which there is crystallographic data, most bind 2-5 bp sequences. There are 4.sup.4 =256 different 4 base sequences; therefore, on average, a single 4 bp site is found roughly 1.2.times.10.sup.7 times in the human genome. An individual 8 base site would be found, on average, about 50,000 times in the genome. On the surface, it might appear that drugs targeted at even an 8 bp site might be deleterious to the cell because there are so many binding sites; however, several other considerations must be recognized. First, most DNA is tightly wrapped in chromosomal proteins and is relatively inaccessible to incoming DNA-binding molecules as demonstrated by the nonspecific endonucleolytic digestion of chromatin in the nucleus (Edwards, C. A. and Firtel, R. A.).

Active transcription units are more accessible, but the most highly exposed regions of DNA in chromatin are the sites that bind regulatory factors. As demonstrated by DNAase I hypersensitivity (Gross, D. S. and Garrard, W. T.), regulatory sites may be 100-1000 times more sensitive to endonucleolytic attack than the bulk of chromatin. This is one reason for targeting regulatory sequences with DNA-binding drugs. Secondly, the argument that several anticancer drugs that bind 2, 3, or 4 bp sequences have sufficiently low toxicity that they can be used as drugs indicates that, if high affinity binding sites for known drugs can be matched with specific viral target sequences, it may be possible to use currently available drugs as antiviral agents at lower concentrations than they are currently used, with a concomitantly lower toxicity.

4) Further Considerations in Choosing Target Sites: Finding Eukaryotic Promoters

Eukaryotic organisms have three RNA polymerases (Pol I, II, and III) that transcribe genetic information from DNA to RNA. The correct regulation of this information flow is essential for the survival of the cell. These multisubunit enzymes need additional proteins to regulate transcription. Many of these additional proteins bind to DNA in a region 5' of the translation start site for a gene: this region is generally known as the promoter region of the gene. All three polymerases use a core set of general transcription proteins to bind to this region. A central component of this complex is the protein called TBP or TFIID. The site this protein binds to is known as the TATA-box because the sequence usually contains a sequence motif similar to TATA (e.g., TATAa/tAa/t). Originally it was thought that each of the three polymerases used a separate set of general transcription factors and that Pol II used TFIID exclusively. Recently it has been shown that all three classes of RNA polymerase need TFIID for transcriptional regulation (see Comai, et al. and Greenblatt).

A molecule that binds to a DNA sequence closely adjacent or slightly overlapping a TATA binding site will likely alter transcriptional regulation of the gene. If the molecule bound directly to the TATA-box itself, then this molecule is expected to be very toxic to cells since the transcription of most genes would be altered. The sequences adjacent to TATA boxes, however, are not conserved. Accordingly, if a particular sequence is selected adjacent a TATA box of a particular gene, a molecule that binds to this specific sequence would be expected to altering the transcriptional regulation of just that gene.

TATA-boxes were first identified by determining the sequence of the DNA located 5' of the RNA start sites of a number of genes. Examination of these sequences revealed that most genes had a TATA-box motif (consensus sequence) in the range of nucleotides 50 to 15 nucleotides 5' of the RNA start site. In vitro studies, typically DNA protection (footprinting) studies, lead to the conclusion that proteins were binding to these sites. Further in vitro DNA binding experiments demonstrated that some proteins could specifically bind to these sites. This lead to assays that allowed purification and subsequent sequencing of the binding proteins. This information facilitated the cloning and expression of genes encoding the binding proteins. A large number of transcription factors are now known. The protein designated TFIID has been demonstrated to bind to the TATA-box (Lee, et al.).

Molecules that interfere with the interaction of these transcription factors and their target DNA (i.e., DNA/Protein transcription complexes) are also expected to alter transcription initiated from the target DNA. A publicly available database of these factors and the sequences which they bind is available from the National Library of Medicine and is called "The Transcription Data Base, or TFD."

The binding sites of these transcription factors can be identified in the 5' non-coding region of genes having known sequences (Example 15). For a selected gene, non-conserved DNA surrounding the transcription factor binding site can then be chosen as a specific target sequence for small molecule binding. Small molecules that bind such specific target sequences can be identified and/or designed using the assay and methods of the present invention.

5) Further Considerations in Choosing Target Sites: Procaryotes and Viruses

Bacterial gene expression is regulated at several different levels, including transcription. General and specific transcription factors are needed along with the core RNA polymerase to accurately produce appropriate amounts of mRNA. Antibiotics that bind to the RNA polymerase and prevent mRNA production are potent bacterial poisons: molecules that could interfere with the initiation of transcription for specific essential genes are expected to have similar effects.

Many bacterial promoters have been sequenced and carefully examined. In general, the majority of bacterial promoters have two well characterized regions, the -35 region which has a consensus sequence similar to SEQ ID NO:25 and the -10 region with a consensus sequence of SEQ ID NO:26. The sequence of the start site for RNA polymerase, however, is not always the same. The start site is determined by a supplementary protein called the sigma factor, which confers specificity for binding the RNA polymerase core. Several sigma factors are present in any species of bacteria. Each sigma factor recognizes a different set of promoter sequences. Expression of sigma factors is regulated, typically, by the growth conditions the bacteria is encountering. These sigma factor promoter sequences represent excellent targets for sequence specific DNA binding molecules.

As an example of choosing target sequences for the purpose of designing a DNA-binding therapeutic for a bacterial disease, consider the example of tuberculosis. Tuberculosis is caused by Mycobacterium tuberculosis.

All bacteria need to make ribosomes for the purpose of protein synthesis. The -35 and -10 regions of M. tuberculosis ribosome RNA synthesis has been determined. In the EMBL locus MTRRNOP the -35 signal is located at coordinants 394 . . . 400 and the -10 signal is found at coordinants 419 . . . 422. These regions represent excellent targets for a DNA binding drug that would inhibit the growth of the bacteria by disrupting its ability to make ribosomes and synthesize protein. Multiple other essential genes could be targeted in a similar manner.

M. tuberculosis is a serious public health problem for several reasons, including the development of antibiotic resistant strains. Many antibiotics inhibit the growth of bacteria by binding to a specific protein and inhibiting its function. An example of this is the binding of rifampicin to the beta subunit of the bacterial RNA polymerase. Continued selection of bacteria with an agent of this kind can lead to the selection of mutants having an altered RNA polymerase so that the antibiotic can no longer bind it. Such mutants can arise from a single mutation.

However, binding a drug to a DNA regulatory region requires at least two mutations to escape the inhibitory effect of the drug: one mutation in the target DNA sequence so that the drug could not bind the target sequence, and one mutation in the regulatory binding protein so that it can recognize the new, mutated regulatory sequence. Such a double mutation event is much less frequent than the single mutation discussed above, for example, with rifampicin. Accordingly, it is expected that the development of drug resistant bacteria would be much less common for DNA-binding drugs that bind to promoter sequences.

The HIV viral promoter region (shown in FIG. 23) provides an example of choosing DNA target sequences for sequence-specific DNA binding drugs to inhibit viral replication.

Many eukaryotic viruses use promoter regions that have similar features to normal cellular genes. The replication of these viruses depends on the general transcription factors present in the host cell. As such, the promoter sequences in DNA viruses are similar to those found in cellular genes and have been well-studied. The binding factors Sp-1 and TFIID are important generalized factors that most viral promoters use.

In the HIV promoter sequence found in LOCUS HIVBH101 in version 32 of the EMBL databank, three tandem decanucleotide Sp1 binding sites are located between positions 377 and 409. Site III shows the strongest affinity for the cellular factor. The three cause up to a tenfold effect on transcriptional efficiency in vitro. The transcription start site is at position 455, with a TATA box at 427-431 in the sequence listed below. In addition to these sites, there are two NF-kappa-B sites in this region between nucleotides 350 and 373. These sites are annotated in FIG. 23.

Sequence-specific DNA binding molecules that specifically disrupted this binding would be expected to disrupt HIV replication. For example, the sequences adjacent to the TFIID binding site (SEQ ID NO:28 and/or SEQ ID NO:29), would be target sites for a DNA-binding molecule designed to disrupt TFIID binding. These sequences are found in HIV but are not likely to occur overlapping TFIID binding sites in the endogenous human genome. Multiple sites could be targeted to decrease the likelihood that a single mutation could prevent drug binding.

D. Using Test Matrices and Pattern Matching for the Analysis of Data

The assay described herein has been designed to use a single DNA:protein interaction to screen for sequence-specific and sequence-preferential DNA-binding molecules that can recognize virtually any specified sequence. By using sequences flanking the recognition site for a single DNA:protein interaction, a very large number of different sequences can be tested. The analysis of data yielded by such experiments displayed as matrices and analyzed by pattern matching techniques should yield information about the relatedness of DNA sequences.

The basic principle behind the DNA:protein assay of the present invention is that when molecules bind DNA sequences flanking the recognition sequence for a specific protein the binding of that protein is blocked. Interference with protein binding likely occurs by either (or both) of two mechanisms: (i) directly by steric hindrance, or (ii) indirectly by perturbations transmitted to the recognition sequence through the DNA molecule.

Both of these mechanisms will presumably exhibit distance effects. For inhibition by direct steric hindrance direct data for very small molecules is available from methylation and ethylation interference studies. These data suggest that for methyl and ethyl moieties, the steric effect is limited by distance effects to 4-5 base pairs. Even still the number of different sequences that can theoretically be tested for these very small molecules is still very large (i.e., 5 base pair combinations total 4.sup.5 (=1024) different sequences).

In practice, the size of sequences tested can be explored empirically for different sized test DNA-binding molecules. A wide array of sequences with increasing sequence complexity can be routinely investigated. This may be accomplished efficiently by synthesizing degenerate oligonucleotides and multiplexing oligonucleotides in the assay process (i.e., using a group of different oligonucleotides in a single assay) or by employing pooled sequences in test matrices.

In view of the above, assays employing a specific protein and oligonucleotides containing the specific recognition site for that protein flanked by different sequences on either side of the recognition site can be used to simultaneously screen for many different molecules, including small molecules, that have binding preferences for individual sequences or families of related sequences. FIG. 12 demonstrates how the analysis of a test matrix yields information about the nature of competitor sequence specificity. As an example, to screen for molecules that could preferentially recognize each of the 256 possible tetranucleotide sequences (FIG. 13), oligonucleotides could be constructed that contain these 256 sequences immediately adjacent to a 11 bp recognition sequence of UL9 oriS SEQ ID NO:(15), which is identical in each construct.

In FIG. 12 "+" indicates that the mixture retards or blocks the formation of DNA:protein complexes in solution and "-" indicates that the mixture had no marked effect on DNA:protein interactions. The results of this test are shown in Table VI.

These results demonstrate how such a matrix provides data on the presence of sequence specific binding activity is a test mixture and also provides inherent controls for non-specific binding. For example, the effect of test mix #8 on the different test assays reveals that the test mix preferentially affects the oligonucleotides that contain the sequence CCCT. Note that the sequence does not have to be within the test site for test mix #8 to exert an affect. By displaying the data in a matrix, the analysis of the sequences affected by the different test mixtures is facilitated.

Furthermore, defined, ordered sets of oligonucleotides can be screened with a chosen DNA-binding molecule. The results of these binding assays can then be examined using pattern matching techniques to determine the subsets of sequences that bind the molecule with similar binding characteristics. If the structural and biophysical properties (such as, geometric shape and electrostatic properties) of sequences are similar, then it is likely that they will bind the molecule with similar binding characteristics. If the structural and biophysical properties of sequences are different, then it is likely that they will not bind the molecule with similar binding characteristics. In this context, the assay might be used to group defined, ordered sequences into subsets based on their binding characteristics: for example, the subsets could be defined as high affinity binding sites, moderate affinity binding sites, and low affinity binding sites. Sequences in the subsets with positive attributes (e.g., high affinity binding) have a high probability of having similar structural and biophysical properties to one another.

By screening and analyzing the binding characteristics of a number of DNA-binding molecules against the same defined set of DNA sequences, data can be accumulated about the subsets of sequences that fall into the same or similar subsets. Using this pattern matching approach, which can be computer-assisted, the sequences with similar structural and biophysical properties can by grouped empirically. The database arising from pattern matching analysis of raw assay data will lead to the increased understanding of sequence structure and thereby lead to the design of novel DNA-binding molecules with related but different binding activities.

E) Applications for the Determination of the Sequence Specificity of DNA-Binding Drugs

Applications for the determination of the sequence specificity of DNA-binding drugs are described below. The applications are divided into drug homo- and heteromeric polymers (part 1) and sequence-specific DNA-binding molecules as facilitators of triple strand formation (part 2).

One utility of the assay of the invention is the identification of highest affinity binding sites among all possible sites of a certain length for a given DNA-binding molecule. This information may be valuable to the design of new DNA-binding therapeutics.

1) Multimerization of Sequence-Preferential or Sequence-Specific DNA-Binding Molecules Identified in the Assay

Any particular DNA-binding small molecule screened in the assay may only recognize a 2-4 base pair site, and even if the recognition is quite specific, the molecule may be toxic because there are so many target sites in the genome (3.times.10.sup.9 /4.sup.4 4 bp sites, for example). However, if drugs with differential affinity for different sites are identified, the toxicity of DNA-binding drugs may be drastically reduced by creating dimers, trimers, or multimers with these drugs (Example 13). From theoretical considerations of the free energy changes accompanying the binding of drugs to DNA, the intrinsic binding constant of a dimer should be the square of the binding constant of the monomer (LePecq, J. B.). Experimental data confirmed this expectation in 1978 with dimer analogs of ethidium bromide (Kuhlman, et al.). Dimerization of several intercalating molecules, in fact, yields compounds with DNA affinities raised from 10.sup.5 M.sup.-1 for the corresponding monomer to 10.sup.8 to 10.sup.9 M.sup.-1 for the dimers (Skorobogaty, A. et al.; Gauguin, B. et al. (1978a and b); LePecq, J. B. et al.; Pelaprat, D. et al.). Trimerization, which theoretically should yield binding affinities that are the cube of the affinity of the homomonomeric subunit or the product of affinities of the heteromonomeric subunits, has yielded compounds with affinities as high as 10.sup.12 M.sup.-1 (Laugaa, P. et al.). Such affinity is markedly better than the affinities seen for many DNA regulatory proteins.

As a hypothetical example, if a relatively weak DNA-binding drug, drug X, which binds a 4 bp site with an affinity of 2.times.10.sup.5 M.sup.-1 was dimerized, the bis-X drug would now recognize an 8 bp site with a theoretical affinity of 4.times.10.sup.10 M.sup.-1. The difference in affinity between the monomer X and the bis-X form is 200,000-fold. The number of 4 bp sites in the genome is approximately 1.2.times.10.sup.7 versus the number of 8 bp sites in the genome which is approximately 5.times.10.sup.4. Accordingly, there are 256-fold fewer 8 bp sites than 4 bp sites. Thus, the number of high affinity target sites is 256-fold fewer for the bis-X molecule than the number of low affinity target sites for the monomer X, with a 200,000-fold difference in affinity between the two types of sites.

Since the binding constant of a dimer is the product of the binding constants of the monomers, when monomers with higher initial binding constants are formed into dimers (or multimers) the differential effect is proportionately increased, creating a wider "window" of affinity versus the number of binding sites. The breadth of the window essentially reflects the margin of effective drug concentration compared to the relative toxicity.

There are two immediate ramifications of dimerization (or multimerization) of monomeric drugs with moderate toxicity and sequence preference. First, the concentration of drug needed is lowered because of the higher affinity, so that even relatively toxic molecules can be used as drugs. Second, since toxicity is likely linked to the average number of drug molecules bound to the genome, as specificity is increased by increasing the length of the binding site, toxicity is decreased.

Given the information already available on sequence-preferential binding of DNA-binding drugs, it is likely that each drug presented to the screening assay will have (i) a number of high affinity binding sites (e.g., 10 to 100-fold better affinity than the average site), (ii) a larger number of sites that are bound with moderate affinity (3 to 10-fold better affinity than average), (iii) the bulk of the binding sites having average affinity, and (iv) a number of sites having worse-than-average affinity. This range of binding affinities will likely resemble a bell-shaped curve. The shape of the curve will probably vary for each drug. To exemplify, assume that approximately five 4 bp sites will be high affinity binding sites, and twenty 4 bp sites will be moderately high affinity binding sites, then any given drug may recognize roughly 25, high or moderately high affinity binding sites. If 50 to 100 drugs are screened, this represents a "bank" of potentially 250-500 high affinity sites and 1000-2500 moderately high affinity sites. Thus, the probability of finding a number of high affinity drug binding sites that match medically significant target sites is good. Furthermore, heterodimeric drugs can be designed to match DNA target sites of 8 or more bp, lending specificity to the potential pharmaceuticals.

As discussed above, once the sequence preferences are known, the information may be used to design oligomeric molecules (homopolymers or heteropolymers) with substantially greater sequence specificity and substantially higher binding affinity. For example, if a DNA-binding molecule, X, binds a 4 bp sequence 5'-ACGT-3'/5'-ACGT-3' with an equilibrium affinity constant of 2.times.10.sup.5 M.sup.-1, then the dimer of X, X.sub.2, should bind the dimer of the sequence, 5'-ACGTACGT-3'/5'-ACGTACGT-3', with an equilibrium affinity constant of (2.times.10.sup.5 M.sup.-1).sup.2 =4.times.10.sup.10 M.sup.-2. The DNA-binding dimer molecule, X.sub.2, recognizes an 8 bp sequence, conferring higher sequence specificity, with a binding affinity that is theoretically 200,000-fold higher than the DNA-binding monomer, X.

The same argument can be extended to trimer molecules: the trimer of X, X.sub.3, would bind a 12 bp sequence, 5'-ACGTACGTACGT-3'/5'-ACGTACGTACGT-3', with a theoretical equilibrium affinity constant of 8.times.10.sup.15 M.sup.-2.

DNA-binding polymers constructed using the above-mentioned approach may be homo- or hetero-polymers of the parent compounds or oligomeric compounds composed of mixed subunits of the parent compounds. Homopolymers are molecules constructed using two or more subunits of the same monomeric DNA-binding molecule. Heteropolymers are molecules constructed using two or more subunits of different monomeric DNA-binding molecules. Oligomeric compounds are constructed of mixed pieces of parent compounds and may be hetero- or homomeric.

For example, distamycin is a member of a family of non-intercalating minor groove DNA-binding oligopeptides that are composed of repeating units of N-methylpyrrole groups. Distamycin has 3 N-methylpyrrole groups. Daunomycin is a member of an entirely different class of DNA-binding molecules, the anthracycline antibiotics, that bind to DNA via intercalation. Examples of homopolymers would be bis-distamycin, the dimer of distamycin, a molecule containing 6 N-methylpyrrole groups or tris-distamycin, the trimer of distamycin, a molecule containing 9 N-methylpyrrole groups. Heteropolymers are molecules composed of different types of DNA-binding subunits; for example, compounds composed of a distamycin molecule linked to a daunomycin molecule or a distamycin molecule linked to two daunomycin molecules. The term "oligomeric" is being used to describe molecules comprised of linked subunits each of which may be smaller than the parent compound.

An example of an homo-oligomeric compound would be a distamycin molecule linked to 1 or 2 additional N-methylpyrrole groups; the resulting molecule would not be as large as bis-distamycin, but would fundamentally be composed of the same component organic moieties that comprise the parent molecule. Examples of a hetero-oligomeric compounds would be daunomycin linked to one or two N-methylpyrrole groups.

The construction of these polymers will be directed by the information derived from the sequence preferences of the parent compounds tested in the assay. In one embodiment of the assay, a database of preferred sequences is constructed, providing a source of information about the 4 bp sequences that bind with relatively higher affinity to particular drugs that may be linked together to target any particular larger DNA sequence.

DNA-binding subunits can be chemically coupled to form heteropolymers or homopolymers. The subunits can be joined directly to each other, as in the family of distamycin molecules, or the subunits can be joined with a spacer molecule, such as carbon chains or peptide bonds. The coupling of subunits is dependent on the chemical nature of the subunits: appropriate coupling reactions can be determined for any two subunit molecules from the chemical literature. The choice of subunits will be directed by the sequence to be targeted and the data accumulated through the methods discussed in Section III.B of this application.

2) Sequence-Specific DNA-Binding Molecules Identified in the Assay as Facilitators of Triplex Formation

Several types of nucleic acid base-containing polymers have been described that will form complexes with nucleic acids (for reviews, see Helene, C. and Toulme, J.-J.). One type of such a polymer forms a triple-stranded complex by the insertion of a third strand into the major groove of the DNA helix. Several types of base-recognition specific interactions of third strand oligonucleotide-type polymers have been observed. One type of specificity is due to Hoogsteen bonding (Hoogsteen). This specificity arises from recognition between pyrimidine oligonucleotides and double-stranded DNA by pairing thymine and adenine:thymine base pairs and protonated cytosine and guanine:cytosine base pairs (Griffin, L. and Dervan, P. B.). Another type of specific interaction involves the use of purine oligonucleotides for triplex formation. In these triplexes, adenine pairs with adenine:thymine base pairs and guanine with guanine:cytosine base pairs (Cooney, M. et al.; Beal, P. A. and Dervan, P. B.).

Other motifs for triplex formation have been described, including the recognition of thymine:adenine base pairs by guanine (Griffin, L. and Dervan, P. B.), the incorporation of nucleic acid analogs (eg, methylphosphonates, phosphorothioates; Miller, et al.), and the invention of backbones other than the phosphoribose backbones normally found in nucleic acids (Pitha, et al.; Summerton, et al.). In several cases, the formation of triplex has been demonstrated to inhibit the binding of a DNA-binding protein (e.g., Young, S. L., et al.; Maher, L. J., et al.) or the expression of a cellular protein (Cooney, M., et al.).

Furthermore, several experiments have been reported in which a small DNA-binding molecule has been covalently attached to polymer capable of forming a triplex structure. An acridine:polypyrimidine molecule has been demonstrated to inhibit SV40 in CV-1 cells (Birg, F., et al.), cleavage at a single site in a yeast chromosome was achieved with an oligonucleotide:EDTA-Fe molecule (Strobel, S. A., et al.; Moser, H. E. and Dervan, P. B.), and a photoinducible endonuclease was created by similar strategy by attaching an ellipticine derivative to a homopyrimidine oligonucleotide (Perrouault, L., et al.). Several other small intercalating agents coupled to oligonucleotides have been described (for review, see Montenay-Garestier).

One utility of the assay of the present invention is to identify the sequence-specificity of DNA-binding molecules for use in designing and synthesizing heteromeric therapeutics consisting of a DNA-binding polymer (e.g., an oligonucleotide) attached to a sequence-preferential or sequence-specific DNA-binding molecule, yielding a heteropolymer. The attached small molecule may serve several functions. First, if the molecule has increased affinity for a specific site (such as, a particular 4 base pair sequence) over all other sites of the same size, then the local concentration of the hetero-molecule, including the oligonucleotide, will be increased at those sites. The amount of heteropolymer, containing a sequence-specific moiety attached to one end, needed for treatment purposes is reduced compared to a heteropolymer that has a non-specific DNA-binding moiety attached. This reduction in treatment amount is directly proportional to both the differential specificity and the relative affinities between the sequence-specific binder and the non-specific binder. For the simplest example, if a sequence-specific molecule with absolute specificity (i.e., it binds only one sequence) had equal affinity for a specific 4 base-pair target site (1/256 possible combinations) as a non-specific molecule, then the amount of drug needed to exert the same effective concentration at that site could potentially be as much as 256-fold less for the specific and non-specific drugs. Accordingly, attaching a sequence-specific DNA-binding molecule to a polymer designed to form triplex structures allows increased localized concentrations.

A second utility of the assay of the present invention is to identify small molecules that cause conformational changes in the DNA when they bind. The formation of triplex DNA requires a shift from B form to A form DNA. This is not be energetically favorable, necessitating the use of increased amounts of polymer for triplex formation to drive the conformational change. The insertion of a small DNA-binding molecule (such as, actinomycin D) which induces a conformational change in the DNA, thereby reduces the amount of polymer needed to stabilize triplex formation.

Accordingly, one embodiment of the invention is to use the assay to test known DNA-binding molecules with all 256 possible four base pair test sequences to determine the relative binding affinity to all possible 4 bp sequences. Then, once the sequence preferences are known, the information may be used to design heteropolymeric molecules comprised of a small DNA-binding molecule and a macromolecule, such as a triplex-forming oligonucleotide, to obtain a DNA-binding molecule with enhanced binding characteristics. The potential advantages of attaching a sequence-specific or sequence-preferential DNA-binding small molecule to a triplex forming molecule are to (i) target the triplex to a subset of specific DNA sequences and thereby (ii) anchor the triplex molecule in the vicinity of its target sequence and in doing so, (iii) increase the localized concentration of the triplex molecule, which allows (iv) lower concentrations of triplex to be used effectively. The presence of the small molecule may also (v) facilitate localized perturbations in DNA structure, destabilizing the B form of DNA, which is unsuitable for triplex formation, thereby facilitating the formation of other structures, such as those necessary for triplex formation (A form DNA); the net effect would be to decrease the amount of triplex needed for efficacious results.

E) Other Applications

The potential pharmaceutical applications for sequence-specific DNA-binding molecules are very broad, including antiviral, antifungal, antibacterial, antitumor agents, immunosuppressants, and cardiovascular drugs. Sequence-specific DNA-binding molecules can also be useful as molecular reagents as, for example, specific sequence probes.

As more DNA-binding molecules are detected, information about their DNA binding affinities, sequence recognition, and mechanisms of DNA-binding will be gathered, eventually facilitating the design and/or modification of new molecules with different or specialized activities.

Although the assay has been described in terms of the detection of sequence-specific DNA-binding molecules, the reverse assay could be achieved by adding DNA in excess to protein to look for peptide sequence specific protein-binding inhibitors.

The following examples illustrate, but in no way are intended to limit the present invention.

Synthetic oligonucleotides were prepared using commercially available automated oligonucleotide synthesizers. Alternatively, custom designed synthetic oligonucleotides may be purchased, for example, from Synthetic Genetics (San Diego, Calif.). Complementary strands were annealed to generate double-strand oligonucleotides.

Restriction enzymes were obtained from Boehringer Mannheim (Indianapolis Ind.) or New England Biolabs (Beverly Mass.) and were used as per the manufacturer's directions.

Distamycin A and Doxorubicin were obtained from Sigma (St. Louis, Mo.). Actinomycin D was obtained from Boehringer Mannheim or Sigma.

Standard cloning and molecular biology techniques are described in Ausubel, et al., and Sambrook, et al.

This example describes the preparation of (A) biotinylated/digoxigenin/radiolabeled, and (B) radiolabeled double-stranded oligonucleotides that contain the screening sequence and selected Test sequences.

A. Biotinylation

The oligonucleotides were prepared as described above. The wild-type control sequence for the UL9 binding site, as obtained from HSV, is shown in FIG. 4. The screening sequence, i.e. the UL9 binding sequence, is CGTTCGCACTT (SEQ ID NO:1) and is underlined in FIG. 4A. Typically, sequences 5' and/or 3' to the screening sequence were replaced by a selected Test sequence (FIG. 5).

One example of the preparation of a site-specifically biotinylated oligonucleotide is outlined in FIG. 4. An oligonucleotide primer complementary to the 3' sequences of the screening sequence-containing oligonucleotide was synthesized. This oligonucleotide terminated at the residue corresponding to the C in position 9 of the screening sequence. The primer oligonucleotide was hybridized to the oligonucleotide containing the screening sequence. Biotin-11-dUTP (Bethesda Research Laboratories (BRL), Gaithersburg Md.) and Klenow enzyme were added to this complex (FIG. 4) and the resulting partially double-stranded biotinylated complexes were separated from the unincorporated nucleotides using either pre-prepared "G-25 SEPHADEX" spin columns (Pharmacia, Piscataway N.J.) or "NENSORB" columns (New England Nuclear) as per manufacturer's instructions. The remaining single-strand region was converted to double-strands using DNA polymerase I Klenow fragment and dNTPs resulting in a fully double-stranded oligonucleotide. A second "G-25 SEPHADEX" column was used to purify the double-stranded oligonucleotide. Oligonucleotides were diluted or resuspended in 10 mM Tris-HCl, pH 7.5, 50 mM NaCl, and 1 mM EDTA and stored at -20.degree. C. For radiolabelling the complexes, .sup.32 P-alpha-dCTP (New England Nuclear, Wilmington, Del.) replaced dCTP for the double-strand completion step. Alternatively, the top strand, the primer, or the fully double-stranded oligonucleotide have been radiolabeled with .gamma.-.sup.32 P-ATP and polynucleotide kinase (NEB, Beverly, Mass.). Most of our preliminary studies have employed radiolabeled, double-stranded oligonucleotides. The oligonucleotides are prepared by radiolabeling the primer with T4 polynucleotide kinase and .gamma.-.sup.32 P-ATP, annealing the "top" strand full length oligonucleotide, and "filling-in" with Klenow fragment and deoxynucleotide triphosphates. After phosphorylation and second strand synthesis, oligonucleotides are separated from buffer and unincorporated triphosphates using "G-25 SEPHADEX" preformed spin columns (IBI or Biorad). This process is outlined in FIG. 4B. The reaction conditions for all of the above Klenow reactions were as follows: 10 mM Tris-HCl, pH 7.5, 10 mM MgCl.sub.2, 50 mM NaCl, 1 mM dithioerythritol, 0.33-100 .mu.M deoxytriphosphates, 2 units Klenow enzyme (Boehringer-Mannheim, Indianapolis Ind.). The Klenow reactions were incubated at 25.degree. C. for 15 minutes to 1 hour. The polynucleotide kinase reactions were incubated at 37.degree. C. for 30 minutes to 1 hour.

B) End-labeling with digoxigenin. The biotinylated, radiolabelled oligonucleotides or radiolabeled oligonucleotides were isolated as above and resuspended in 0.2M potassium cacodylate (pH=7.2), 4 mM MgCl.sub.2, 1 mM 2-mercaptoethanol, and 0.5 mg/ml bovine serum albumin. To this reaction mixture digoxigenin-11-dUTP (an analog of dTTP, 2'-deoxy-uridine-5'-triphosphate, coupled to digoxigenin via an 11-atom spacer arm, Boehringer Mannheim, Indianapolis Ind.) and terminal deoxynucleotidyl transferase (GIBCO BRL, Gaithersburg, Md.) were added. The number of Dig-11-dUTP moieties incorporated using this method appeared to be less than 5 (probably only 1 or 2) as judged by electrophoretic mobility on polyacrylamide gels of the treated fragment as compared to oligonucleotides of known length.

The biotinylated or non-biotinylated, digoxygenin-containing, radiolabelled oligonucleotides were isolated as above and resuspended in 10 mM Tris-HCl, 1 mM EDTA, 50 mM NaCl, pH 7.5 for use in the binding assays.

The above procedure can also be used to biotinylate the other strand by using an oligonucleotide containing the screening sequence complementary to the one shown in FIG. 4 and a primer complementary to the 3' end of that molecule. To accomplish the biotinylation Biotin-7-dATP was substituted for Biotin-11-dUTP. Biotinylation was also accomplished by chemical synthetic methods: for example, an activated nucleotide is incorporated into the oligonucleotide and the active group is subsequently reacted with NHS-LC-Biotin (Pierce). Other biotin derivatives can also be used.

C. Radiolabelling the Oligonucleotides

Generally, oligonucleotides were radiolabelled with gamma-.sup.32 P-ATP or alpha-.sup.32 P-deoxynucleotide triphosphates and T4 polynucleotide kinase or the Klenow fragment of DNA polymerase, respectively. Labelling reactions were performed in the buffers and by the methods recommended by the manufacturers (New England Biolabs, Beverly Mass.; Bethesda Research Laboratories, Gaithersburg Md.; or Boehringer/Mannheim, Indianapolis Ind.). Oligonucleotides were separated from buffer and unincorporated triphosphates using "G-25 SEPHADEX" preformed spin columns (IBI, New Haven, Conn.; or Biorad, Richmond, Calif.) or "NENSORB" preformed columns (New England Nuclear, Wilmington, Del.) as per the manufacturers instructions.

There are several reasons to enzymatically synthesize the second strand. The two main reasons are that by using an excess of primer, second strand synthesis can be driven to near completion so that nearly all top strands are annealed to bottom strands, which prevents the top strand single strands from folding back and creating additional and unrelated double-stranded structures, and secondly, since all of the oligonucleotides are primed with a common primer, the primer can bear the end-label so that all of the oligonucleotides will be labeled to exactly the same specific activity.

A. Cloning of the UL9 Protein-Coding Sequences into pAC373.

To express full length UL9 protein a baculovirus expression system has been used. The sequence of the UL9 coding region of Herpes Simplex Virus has been disclosed by McGeoch et al. and is available as an EMBL nucleic acid sequence. The recombinant baculovirus AcNPV/UL9A, which contained the UL9 protein-coding sequence, was obtained from Mark Challberg (National Institutes of Health, Bethesda, Md.). The construction of this vector has been previously described (Olivo et al. (1988, 1989)). Briefly, the NarI/EcoRV fragment was derived from pMC160 (Wu et al.). Blunt-ends were generated on this fragment by using all four dNTPs and the Klenow fragment of DNA polymerase I (Boehringer Mannheim, Indianapolis Ind.) to fill in the terminal overhangs. The resulting fragment was blunt-end ligated into the unique BamHI site of the baculoviral vector pAC3T3 (Summers et al.).

B. Cloning of the UL9 Sequence in pVL1393

The UL9 protein-coding region was cloned into a second baculovirus vector, pVL1393 (Luckow et al.). The 3077 bp NarI/EcoRV fragment containing the UL9 gene was excised from vector pEcoD (obtained from Dr. Bing Lan Rong, Eye Research Institute, Boston, Mass.): the plasmid pEcoD contains a 16.2 kb EcoRI fragment derived from HSV-I that bears the UL9 gene (Goldin et al.). Blunt-ends were generated on the UL9-containing fragment as described above. EcoRI linkers (10 mer) were blunt-end ligated (Ausubel et al.; Sambrook et al.) to the blunt-ended NarI/EcoRV fragment.

The vector pVL1393 (Luckow et al.) was digested with EcoRI and the linearized vector isolated. This vector contains 35 nucleotides of the 5' end of the coding region of the polyhedron gene upstream of the polylinker cloning site. The polyhedron gene ATG has been mutated to ATT to prevent translational initiation in recombinant clones that do not contain a coding sequence with a functional ATG. The EcoRI/UL9 fragment was ligated into the linearized vector, the ligation mixture transformed into E. coli and ampicillin resistant clones selected. Plasmids recovered from the clones were analyzed by restriction digestion and plasmids carrying the insert with the amino terminal UL9 protein-coding sequences oriented to the 5' end of the polyhedron gene were selected. This plasmid was designated pVL1393/UL9 (FIG. 7).

pVL1393/UL9 was cotransfected with wild-type baculoviral DNA (AcMNPV; Summers et al.) into SF9 (Spodoptera frugiperda) cells (Summers et al.). Recombinant baculovirus-infected Sf9 cells were identified and clonally purified (Summers et al.).

C. Expression of the UL9 Protein

Clonal isolates of recombinant baculovirus infected Sf9 cells were grown in Grace's medium as described by Summers et al. The cells were scraped from tissue culture plates and collected by centrifugation (2,000 rpm, for 5 minutes, 4.degree. C.). The cells were then washed once with phosphate buffered saline (PBS) (Maniatis et al.). Cell pellets were frozen at -70.degree. C. For lysis the cells were resuspended in 1.5 volumes 20 mM HEPES, pH 7.5, 10% glycerol, 1.7M NaCl, 0.5 mM EDTA, 1 mM dithiothreitol (DTT), and 0.5 mM phenyl methyl sulfonyl fluoride (PMSF). Cell lysates were cleared by ultracentrifugation (Beckman table top ultracentrifuge, TLS 55 rotor, 34 krpm, 1 hr, 4.degree. C.). The supernatant was dialyzed overnight at 4.degree. C. against 2 liters dialysis buffer (20 mM HEPES, pH 7.5, 10% glycerol, 50 mM NaCl, 0.5 mM EDTA, 1 mM dtt, and 0.1 mM PMSF).

These partially purified extracts were prepared and used in DNA:protein binding experiments. If necessary extracts were concentrated using a "CENTRICON 30" filtration device (Amicon, Danvers Mass.).

D. Cloning the Truncated UL9 Protein

The sequence encoding the C-terminal third of UL9 and the 3' flanking sequences, an approximately 1.2 kb fragment, was subcloned into the bacterial expression vector, pGEX-2T (FIG. 6). The pGEX-2T is a modification of the pGEX-1 vector of Smith et al. which involved the insertion of a thrombin cleavage sequence in-frame with the glutathione-S-transferase protein (gst).

A 1,194 bp BamHI/EcoRV fragment of pEcoD was isolated that contained a 951 bp region encoding the C-terminal 317 amino acids of UL9 and 243 bp of the 3' untranslated region.

This BamHI/EcoRV UL9 carboxy-terminal (UL9-COOH) containing fragment was blunt-ended and EcoRI linkers added as described above. The EcoRI linkers were designed to allow in-frame fusion of the UL9 protein-coding sequence to the gst-thrombin coding sequences. The linkered fragment was isolated and digested with EcoRI. The pGEX-2T vector was digested with EcoRI, treated with Calf Intestinal Alkaline Phosphatase (CIP) and the linear vector isolated. The EcoRI linkered UL9-COOH fragment was ligated to the linear vector (FIG. 6). The ligation mixture was transformed into E. coli and ampicillin resistant colonies were selected. Plasmids were isolated from the ampicillin resistant colonies and analyzed by restriction enzyme digestion. A plasmid which generated a gst/thrombin/UL9-COOH in frame fusion was identified (FIG. 6) and designated pGEX-2T/UL9-COOH.

E. Expression of the Truncated UL9 Protein

E. coli strain JM109 was transformed with pGEX-2T/C-UL9-COOH and was grown at 37.degree. C. to saturation density overnight. The overnight culture was diluted 1:10 with LB medium containing ampicillin and grown from one hour at 30.degree. C. IPTG (isopropyllthio-.beta.-galactoside) (GIBCO-BRL) was added to a final concentration of 0.1 mM and the incubation was continued for 2-5 hours. Bacterial cells containing the plasmid were subjected to the temperature shift and IPTG conditions, which induced transcription from the tac promoter.

Cells were harvested by centrifugation and resuspended in 1/100 culture volume of MTPBS (150 mM NaCl, 16 mM Na.sub.2 HPO.sub.4, 4 mM NaH.sub.2 PO.sub.4). Cells were lysed by sonication and lysates cleared of cellular debris by centrifugation.

The fusion protein was purified over a glutathione agarose affinity column as described in detail by Smith et al. The fusion protein was eluted from the affinity column with reduced glutathione, dialyzed against UL9 dialysis buffer (20 mM HEPES pH 7.5, 50 mM NaCl, 0.5 mM EDTA, 1 mM DTT, 0.1 mM PMSF) and cleaved with thrombin (2 ng/ug of fusion protein).

An aliquot of the supernatant obtained from IPTG-induced cultures of pGEX-2T/C-UL9-COOH-containing cells and an aliquot of the affinity-purified, thrombin-cleaved protein were analyzed by SDS-polyacrylamide gel electrophoresis. The result of this analysis is shown in FIG. 8. The 63 kilodalton GST/C-UL9 fusion protein is the largest band in the lane marked GST-UL9 (lane 2). The first lane contains protein size standards. The UL9-COOH protein band (lane GST-UL9+Thrombin, FIG. 8, lane 3) is the band located between 30 and 46 kD: the glutathione transferase protein is located just below the 30 kD size standard. In a separate experiment a similar analysis was performed using the uninduced culture: it showed no protein corresponding in size to the fusion protein.

Extracts are dialyzed before use. Also, if necessary, the extracts can be concentrated typically by filtration using a "CENTRICON 30" filter.

A. Band Shift Gels

DNA:protein binding reactions containing both labelled complexes and free DNA were separated electrophoretically on 4-10% polyacrylamide/Tris-Borate-EDTA (TBE) gels (Freid et al.; Garner et al.). The gels were then fixed, dried, and exposed to X-ray film. The autoradiograms of the gels were examined for band shift patterns.

B. Filter Binding Assays

A second method used particularly in determining the half-lives for protein:oligonucleotide complexes is filter binding (Woodbury et al.). Nitrocellulose disks (Schleicher and Schuell, BA85 filters) that have been soaked in binding buffer (see below) were placed on a vacuum filter apparatus. DNA:protein binding reactions (see below; typically 15-30 .mu.l) are diluted to 0.5 ml with binding buffer (this dilutes the concentration of components without dissociating complexes) and applied to the discs with vacuum applied. Under low salt conditions the DNA:protein complex sticks to the filter while free DNA passes through. The discs are placed in scintillation counting fluid (New England Nuclear), and the cpm determined using a scintillation counter.

This technique has been adapted to 96-well and 72-slot nitrocellulose filtration plates (Schleicher and Schuell) using the above protocol except (i) the reaction dilution and wash volumes are reduced and (ii) the flow rate through the filter is controlled by adjusting the vacuum pressure. This method greatly facilitates the number of assay samples that can be analyzed. Using radioactive oligonucleotides, the samples are applied to nitrocellulose filters, the filters are exposed to x-ray film, then analyzed using a Molecular Dynamics scanning densitometer. This system transfers data directly into analytical software programs (eg, Excel) for analysis and graphic display.

A. Functional DNA-Binding Activity Assay

Purified protein was tested for functional activity using band-shift assays. Radiolabelled oligonucleotides (prepared as in Example 1B) that contain the 11 bp recognition sequence were mixed with the UL9 protein in binding buffer (optimized reaction conditions: 0.1 ng .sup.32 P-DNA, 1 ul UL9 extract, 20 mM HEPES, pH 7.2, 50 mM KCl, and 1 mM DTT). The reactions were incubated at room temperature for 10 minutes (binding occurs in less than 2 minutes), then separated electrophoretically on 4-10% non-denaturing polyacrylamide gels. UL9-specific binding to the oligonucleotide is indicated by a shift in mobility of the oligonucleotide on the gel in the presence of the UL9 protein but not in its absence. Bacterial extracts containing (+) or without (-) UL9 protein and affinity purified UL9 protein were tested in the assay. Only bacterial extracts containing UL9 or affinity purified UL9 protein generate the gel band-shift indicating protein binding.

The degree of extract that needed to be added to the reaction mix, in order to obtain UL9 protein excess relative to the oligonucleotide, was empirically determined for each protein preparation/extract. Aliquots of the preparation were added to the reaction mix and treated as above. The quantity of extract at which the majority of the labelled oligonucleotide appears in the DNA:protein complex was evaluated by band-shift or filter binding assays. The assay is most sensitive under conditions in which the minimum amount of protein is added to bind most of the DNA. Excess protein can decrease the sensitivity of the assay.

B. Rate of Dissociation

The rate of dissociation is determined using a competition assay. An oligonucleotide having the sequence presented in FIG. 4, which contained the binding site for UL9 (SEQ ID NO:14), was radiolabelled with .sup.32 P-ATP and polynucleotide kinase (Bethesda Research Laboratories). The competitor DNA was a 17 base pair oligonucleotide (SEQ ID NO:16) containing the binding site for UL9.

In the competition assays, the binding reactions (Example 4A) were assembled with each of the oligonucleotides and placed on ice. Unlabelled oligonucleotide (1 .mu.g) was added 1, 2, 4, 6, or 21 hours before loading the reaction on an 8% polyacrylamide gel (run in TBE buffer (Maniatis et al.)) to separate the reaction components. The dissociation rates, under these conditions, for the truncated UL9 (UL9-COOH) and the full length UL9 is approximately 4 hours at 4.degree. C. In addition, random oligonucleotides (a 10,000-fold excess) that did not contain the UL9 binding sequence and sheared herring sperm DNA (a 100,000-fold excess) were tested: neither of these control DNAs competed for binding with the oligonucleotide containing the UL9 binding site.

C. Optimization of the UL9 Binding Assay

1. Truncated UL9 from the Bacterial Expression System

The effects of the following components on the binding and dissociation rates of UL9-COOH with its cognate binding site have been tested and optimized: buffering conditions (including the pH, type of buffer, and concentration of buffer); the type and concentration of monovalent cation; the presence of divalent cations and heavy metals; temperature; various polyvalent cations at different concentrations; and different redox reagents at different concentrations. The effect of a given component was evaluated starting with the reaction conditions given above and based on the dissociation reactions described in Example 4B.

The optimized conditions used for the binding of UL9-COOH contained in bacterial extracts (Example 2E) to oligonucleotides containing the HSV ori sequence (SEQ ID NO:1) were as follows: 20 mM HEPES, pH 7.2, 50 mM KCl, 1 mM DTT, 0.005-0.1 ng radiolabeled (specific activity, approximately 10.sup.8 cpm/.mu.g) or digoxiginated, biotinylated oligonucleotide probe, and 5-10 .mu.g crude UL9-COOH protein preparation (1 mM EDTA is optional in the reaction mix). Under optimized conditions, UL9-COOH binds very rapidly and has a dissociation rate of about 4 hours at 4.degree. C. with non-biotinylated oligonucleotide and 5-10 minutes with biotinylated oligonucleotides. The dissociation rate of UL9-COOH changes markedly under different physical conditions. Typically, the activity of a UL9 protein preparation was assessed using the gel band-shift assay and related to the total protein content of the extract as a method of standardization. The addition of herring sperm DNA depended on the purity of UL9 used in the experiment Binding assays were incubated at 25.degree. C. for 5-30 minutes.

2. Full Length UL9 Protein from the Baculovirus System

The binding reaction conditions for the full length baculovirus-produced UL9 polypeptide have also been optimized. The optimal conditions for the current assay were determined to be as follows: 20 mM Hepes; 100 mM NaCl; 0.5 mM dithiothreitol; 1 mM EDTA; 5% glycerol; from 0 to 10.sup.4 -fold excess of sheared herring sperm DNA; 0.005-0.1 ng radiolabeled (specific activity, approximately 10.sup.8 cpm/.mu.g) or digoxiginated, biotinylated oligonucleotide probe, and 5-10 .mu.g crude UL9 protein preparation. The full length protein also binds well under the optimized conditions established for the truncated UL9-COOH protein.

The oligonucleotides shown in FIG. 5 were radiolabelled as described above. The competition assays were performed as described in Example 4B using UL9-COOH. Radiolabelled oligonucleotides were mixed with the UL9-COOH protein in binding buffer (typical reaction: 0.1 ng oligonucleotide .sup.32 P-DNA, 1 .mu.l UL9-COOH extract, 20 mM HEPES, pH 7.2, 50 mM KCl, 1 mM EDTA, and 1 mM DTT). The reactions were incubated at room temperature for 10 minutes. A zero time point sample was then taken and loaded onto an 8% polyacrylamide gel (run use TBE). One .mu.g of the unlabelled 17 bp competitive DNA oligonucleotide (SEQ ID NO:16) (Example 4B) was added at 5, 10, 15, 20, or 60 minutes before loading the reaction sample on the gel. The results of this analysis are shown in FIG. 9: the screening sequences that flank the UL9 binding site (SEQ ID NO:5-SEQ ID NO:13) are very dissimilar but have little effect on the off-rate of UL9. Accordingly, these results show that the UL9 DNA binding protein is effective to bind to a screening sequence in duplex DNA with a binding affinity that is substantially independent of test sequences placed adjacent the screening sequence. Filter binding experiments gave the same result.

Different oligonucleotides, each of which contained the screening sequence (SEQ ID NO:1) flanked on the 5' and 3' sides by a test sequence (SEQ ID NO:5 to SEQ ID NO:13), were evaluated for the effects of distamycin A, actinomycin D, and doxorubicin on UL9-COOH binding.

Binding assays were performed as described in Example 5. The oligonucleotides used in the assays are shown in FIG. 5. The assay mixture was allowed to pre-equilibrate for 15 minutes at room temperature prior to the addition of drug.

A concentrated solution of Distamycin A was prepared in dH.sub.2 O and was added to the binding reactions at the following concentrations: 0, 1 .mu.M, 4 .mu.M, 16 .mu.M, and 40 .mu.M. The drug was added and incubated at room temperature for 1 hour. The reaction mixtures were then loaded on an 8% polyacrylamide gel (Example 5) and the components separated electrophoretically. Autoradiographs of these gels are shown in FIG. 10A. The test sequences tested were as follows: UL9 polyT, SEQ ID NO:9; UL9 CCCG, SEQ ID NO:5; UL9 GGGC, SEQ ID NO:6; UL9 polyA, SEQ ID NO:8; and UL9 ATAT, SEQ ID NO:7. These results demonstrate that Distamycin A preferentially disrupts binding to UL9 polyT, UL9 polyA and UL9 ATAT.

A concentrated solution of Actinomycin D was prepared in dH.sub.2 O and was added to the binding reactions at the following concentrations: 0 .mu.M and 50 .mu.M. The drug was added and incubated at room temperature for 1 hour. Equal volumes of dH.sub.2 O were added to the control samples. The reaction mixtures were then loaded on an 8% polyacrylamide gel (Example 5) and the components separated electrophoretically. Autoradiographs of these gels are shown in FIG. 10B. In addition to the test sequences tested above with Distamycin A, the following test sequences were also tested with Actinomycin D: AToril, SEQ ID NO:11; oriEco2, SEQ ID NO:12, and oriEco3, SEQ ID NO:13. These results demonstrate that actinomycin D preferentially disrupts the binding of UL9 to the oligonucleotides UL9 CCCG and UL9 GGGC.

A concentrated solution of Doxorubicin was prepared in dH.sub.2 O and was added to the binding reactions at the following concentrations: 0 .mu.M, 15 .mu.M and 35 .mu.M. The drug was added and incubated at room temperature for 1 hour. Equal volumes of dH.sub.2 O were added to the control samples. The reaction mixtures were then loaded on an 8% polyacrylamide gel (Example 5) and the components separated electrophoretically. Autoradiographs of these gels are shown in FIG. 10C. The same test sequences were tested as for Actinomycin D. These results demonstrate that Doxorubicin preferentially disrupts the binding of UL9 to the oligonucleotides UL9polyT, UL9 GGGC, oriEco2, and oriEco3. Doxorubicin appears to particularly disrupt the UL9:screening sequence interaction when the test sequence oriEco3 is used. The sequences of the test sequences for oriEco2 and oriEco3 differ by only one base: an additional T residue inserted at position 12, compare SEQ ID NO:12 and SEQ ID NO:13.

A. The Capture of Protein-Free DNA

Several methods have been employed to sequester unbound DNA from DNA:protein complexes.

1. Magnetic Beads

Streptavidin-conjugated superparamagnetic polystyrene beads (Dynabeads M-280 Streptavidin, Dynal AS, 6-7.times.10.sup.8 beads/ml) are washed in binding buffer then used to capture biotinylated oligonucleotides (Example 1). The beads are added to a 15 ul binding reaction mixture containing binding buffer and biotinylated oligonucleotide. The beads/oligonucleotide mixture is incubated for varying lengths of time with the binding mixture to determine the incubation period to maximize capture of protein-free biotinylated oligonucleotides. After capture of the biotinylated oligonucleotide, the beads can be retrieved by placing the reaction tubes in a magnetic rack (96-well plate magnets are available from Dynal). The beads are then washed.

2. Agarose Beads

Biotinylated agarose beads (immobilized D-biotin, Pierce, Rockford, Ill.) are bound to avidin by treating the beads with 50 .mu.g/.mu.l avidin in binding buffer overnight at 4.degree. C. The beads are washed in binding buffer and used to capture biotinylated DNA. The beads are mixed with binding mixtures to capture biotinylated DNA. The beads are removed by centrifugation or by collection on a non-binding filter disc.

For either of the above methods, quantification of the presence of the oligonucleotide depends on the method of labelling the oligonucleotide. If the oligonucleotide is radioactively labelled: (i) the beads and supernatant can be loaded onto polyacrylamide gels to separate protein:DNA complexes from the bead:DNA complexes by electrophoresis, and autoradiography performed; (ii) the beads can be placed in scintillation fluid and counted in a scintillation counter. Alternatively, presence of the oligonucleotide can be determined using a chemiluminescent or colorimetric detection system.

B. Detection of Protein-Free DNA

The DNA is end-labelled with digoxigenin-11-dUTP (Example 1). The antigenic digoxigenin moiety is recognized by an antibody-enzyme conjugate, anti-digoxigenin-alkaline phosphatase (Boehringer Mannheim Indianapolis, Ind.). The DNA/antibody-enzyme conjugate is then exposed to the substrate of choice. The presence of dig-dUTP does not alter the ability of protein to bind the DNA or the ability of streptavidin to bind biotin.

1. Chemiluminescent Detection

Digoxigenin-labelled oligonucleotides are detected using the chemiluminescent detection system "SOUTHERN LIGHTS" developed by Tropix, Inc. (Bedford, Mass.). Use of this detection system is illustrated in FIGS. 11A and 11B. The technique can be applied to detect DNA that has been captured on either beads or filters.

Biotinylated oligonucleotides, which have terminal digoxygenin-containing residues (Example 1), are captured on magnetic (FIG. 11A) or agarose beads (FIG. 11B) as described above. The beads are isolated and treated to block non-specific binding by incubation with I-Light blocking buffer (Tropix) for 30 minutes at room temperature. The presence of oligonucleotides is detected using alkaline phosphatase-conjugated antibodies to digoxygenin. Anti-digoxigenin-alkaline phosphatase (anti-dig-AP, 1:5000 dilution of 0.75 units/ul, Boehringer Mannheim) is incubated with the sample for 30 minutes, decanted, and the sample washed with 100 mM Tris-HCl, pH 7.5, 150 mM NaCl. The sample is pre-equilibrated with 2 washes of 50 mM sodium bicarbonate, pH 9.5, 1M MgCl.sub.2, then incubated in the same buffer containing 0.25 mM 3-(2'-spiroadamantane)-4-methoxy-4-(3'-phosphoryloxy) phenyl-1,2-dioxetane disodium salt (AMPPD) for 5 minutes at room temperature. AMPPD was developed (Tropix Inc.) as a chemiluminescent substrate for alkaline phosphatase. Upon dephosphorylation of AMPPD the resulting compound decomposes, releasing a prolonged, steady emission of light at 477 nm.

Excess liquid is removed from filters and the emission of light occurring as a result of the dephosphorylation of AMPPD by alkaline phosphatase can be measured by exposure to x-ray film or by detection in a luminometer.

In solution, the bead-DNA-anti-dig-AP is resuspended in "SOUTHERN LIGHT" assay buffer and AMPPD and measured directly in a luminometer. Large scale screening assays are performed using a 96-well plate-reading luminometer (Dynatech Laboratories, Chantilly, Va.). Subpicogram quantities of DNA (10.sup.2 to 10.sup.3 attomoles (an attomole is 10.sup.-18 moles)) can be detected using the Tropix system in conjunction with the plate-reading luminometer.

2. Colorimetric Detection

Standard alkaline phosphatase colorimetric substrates are also suitable for the above detection reactions. Typically substrates include 4-nitrophenyl phosphate (Boehringer Mannheim). Results of colorimetric assays can be evaluated in multiwell plates (as above) using a plate-reading spectrophotometer (Molecular Devices, Menlo Park Calif.). The use of the light emission system is more sensitive than the colorimetric systems.

The top strands of 256 oligonucleotides, containing all possible 4 bp sequences in the test sites flanking the UL9 recognition site, were synthesized. The oligonucleotides were composed of identical sequences except for the 4 bp sites flanking either side of the UL9 recognition sequence (SEQ ID No:1). The oligonucleotides had the general sequence presented in FIG. 14B (SEQ ID NO:17), where XXXX is the test sequence and N=A,G,C, or T. A 12 bp primer sequence, which is the complementary sequence to the 3'-end of the test oligonucleotide, was also synthesized: the primer was designated the HSV primer and is presented as SEQ ID NO:18.

The HSV primer was used to prime second strand synthesis and to facilitate labeling the oligonucleotides to the same specific activity. Oligonucleotide labelling was accomplished by labeling the 5' end of the HSV primer and then using the same primer to prime second strand synthesis of all 256 test oligonucleotides. The 5' end of the primer can be labeled with radioisotopes such as .sup.32 P, .sup.33 P, or .sup.35 S, or with non-radioactive detection systems such as digoxygenin or biotin as discussed in the Capture/Detection section.

Radioactive-labeling of the primer with .sup.32 P is accomplished by the enzymatic transfer of a radioactive phosphate from .gamma.-.sup.32 P-ATP to the 5' end of the primer oligonucleotide using T4 polynucleotide kinase (Ausubel, et al.). For labeling 256 oligonucleotides, approximately 60 .mu.g HSV primer was labeled as follows. The oligonucleotide was incubated for 1 hour at 37.degree. C. with 125 .mu.l .gamma.-.sup.32 P-ATP (20 mCi total, 7000 Ci/mmol) and 600 units of T4 polynucleotide kinase in a 3 ml reaction volume containing 50 mM Tris-HCL, pH 7.5, 10 mM MgCl.sub.2, 10 mM spermadine, and 1.5 mM dithiothreitol (freshly prepared). To stop the reaction, EDTA was added to a final concentration of 20 mM. Unincorporated nucleotides were removed using "G-25 SEPHADEX" chromatography in 10 mM Tris-HCL, pH 7.5, 50 mM NaCl, and 1 mM EDTA (TE+50).

The radioactive primer was individually annealed to the top strand of each of the 256 test oligonucleotides. The bottom strand is synthesized using deoxyribonucleotides and Klenow fragment or T4 polymerase (Ausubel, et al.). The annealing mixture typically contained 200 ng HSV primer mixed with 1 .mu.g top strand in 20 mM Tris-HCL, pH 7.5, 1 mM spermadine, and 0.1 mM EDTA (35 .mu.l reaction volume). The primer was annealed to the top strand by incubating the sample for 2 minutes at 70.degree. C., then placing the sample at room temperature or on ice. To the annealing mixture, 4.5 .mu.l 10.times. Klenow buffer (10.times.=200 mM Tris-HCL, 500 mM NaCl, 50 mM MgCl.sub.2, 10 mM dithiothreitol), 5 .mu.l 0.5 mM each dNTP (dATP, dCTP, dGTP, dTTP), and 1 .mu.l Klenow fragment were added. This reaction mixture was incubated 30-60 minutes at room temperature (or up to 37.degree. C.).

The volume of the reaction mixture was increased by adding 75 .mu.l a solution of 10 mM Tris-HCl, pH 7.5, 50 mM NaCl, and 10 mM EDTA. The reaction mixture was applied to a 1 ml "G-25 SEPHADEX" (in TE+50) spin column. The spin columns were prepared by plugging 1 cc tuberculin syringes with silanized glass wool and adding a slurry of "G-25 SEPHADEX." The columns were prespun at 2000 rpm in a tabletop centrifuge for 4 minutes. The samples (reaction mixtures) were passed through the column by centrifugation (2000 rpm, 4 minutes at room temperature) to remove unincorporated deoxyribonucleotides. The incorporation of .sup.32 P was measured by placing a small volume of the sample in scintillation fluor and determining the disintegrations per minute (dpms) in a scintillation counter.

The radiolabeled double-stranded oligonucleotides were then diluted to the same specific activity (equal dpms per volume). Typically, a concentration of 0.1 to 1 ng/.mu.l oligonucleotide was used in the assay.

The same procedure can be used for second strand synthesis and labeling to equal specific activity regardless of the type of label on the HSV primer.

Screening large numbers of test molecules or test sequences is most easily accomplished in an arrayed sample format, for example, a 96-well plate format. Such formats are readily amenable to automation using robotics systems. Several different types of disposable plastic plates are available for use in screening assays including the following: polyvinyl chloride (PVC), polypropylene (PP), polyethylene (PE), and polystyrene (PS) plates. Plates, or any testing vehicle in which the assay is performed, are tested for protein and DNA adsorption and coated with a blocking reagent if necessary.

One method for testing protein or DNA adsorption to plates is to place assay mixtures in the wells of the plates for varying lengths of time. Samples are then removed from the wells and a nitrocellulose dot blot capture system (Ausubel, et al.; Schleicher and Schuell) is used to measure the amount of DNA:protein complex remaining in the mixture over time.

When radiolabeled oligonucleotides are used for the test, signal can be measured using autoradiography and a scanning laser densitometer. A decrease in the amount of DNA:protein complex in the absence of competitor molecules is indicative of plate adsorption. If plate adsorption occurs, the plates are coated with a blocking agent prior to use in the assay.

None of the plates listed above showed marked adsorption at a 30 minute time point under the conditions of the assay. However, most plates, regardless of brand, showed significant adsorption at times greater than 2 hours.

Coating the plates with a blocking agent decreases variability in the assay. Several types of blocking reagents typically used to block the adsorption of macromolecules to plastic are known, primarily from immunoscreening procedures. For example, plates may be blocked with either 1% bovine serum albumin (BSA) in phosphate-buffered saline (PBS), or 0.1% gelatin, 0.05% "TWEEN29" in PBS.

To test for the effectiveness of using such blocking reagents, the plates were treated with the above reagents for 1 hour at room temperature, then washed three times with 0.05% "TWEEN20" in PBS and once with the assay buffer. Assay reaction mixtures were aliquoted to the plates and tested as described above using dot blot capture assays. Both of the blocking reagents (BSA or gelatin) were effective in blocking DNA and protein binding--except when polypropylene plates were used. Based on these experiments, PVC plates blocked with BSA were determined to work well in the assay of the present invention.

Plates were tested for inter- and intra-plate variability by aliquoting duplicate samples to all 96-wells of several plates, and determining the amount of DNA:protein complex recovered using the dot blot/nitrocellulose system. The coefficient of variation ›%CV=(the standard deviation/mean)*100! was calculated for intra-plate variability (i.e., between samples on the same plate) and inter-plate variability (i.e., between plates). Blocked PVC plates showed an intra-plate %CV of 5-20%; inter-plate variability was about 8%.

Using the assay method of the present invention, distamycin was tested for sequence selectivity and relative binding affinity to 256 different 4 bp sequences.

1) The Assay Mixture

Water, buffer and UL9 were mixed on ice and aliquoted to the wells of a 96-well plate. The addition of water/UL9/buffer mix was accomplished with an 8-channel repipettor, which holds a relatively large volume and allowed rapid, accurate pipetting to all 96 wells of a master experimental plate.

Radiolabeled double-stranded oligonucleotides were aliquoted from 96-well master stock plates (containing the array of all 256 oligonucleotides diluted to the same specific activity) to the wells of the master experimental plates.

Master assay mixtures in the master experimental plates were thoroughly mixed by pipetting up and down. The mixtures were aliquoted to the test plates. Each test plate typically included one sample as a control (no test molecules added) and as many test samples as were needed for different test molecules or test molecule concentrations. There were 3 master oligonucleotide stock plates, containing the array of 256 oligonucleotides. Accordingly, an experiment testing distamycin at different concentrations would require 256 control assays (one for each oligonucleotide) and 256 assays at each of the drug concentrations to be tested.

The following assay mixture was used for testing distamycin in the assay of the present invention: 1.5 nM radiolabeled DNA and 12.8 nM UL9-COOH protein (prepared as described above in the UL9 binding buffer; 20 mM Hepes, pH 7.2, 50 mM KCl, and 1 mM dithiothreitol). The concentration of the components in the assay mixture can be varied as described above in the Detailed Description.

Assay mixtures containing both UL9 and DNA were incubated at room temperature for at least 10 minutes to allow the DNA:protein complexes to form and for the system to come to equilibrium. At time=0, the assay was begun by adding water (control samples) or distamycin (5-15 .mu.M, test samples) to the assay mixtures using a 12-channel micropipettor. After incubation with drug for 5-120 minutes, samples were taken and applied to nitrocellulose on a 96-well dot blot apparatus (Schleicher and Schuell). The samples were held at 4.degree. C.

Tests were performed in duplicate. Typically, one set of 256 test oligonucleotides was scrambled with respect to location on the 96-well plate to eliminate any effects of plate location.

2) The Capture/Detection System

A 96-well dot blot apparatus was used to capture the DNA:protein complexes on a nitrocellulose filter. The filters used in the dot blot apparatus were pretreated as follows. The nitrocellulose filter was pre-wetted with water and soaked in UL9 binding buffer. The filter was then placed on 1 to 3 pieces of 3 MM filter paper, which were also presoaked in UL9 binding buffer. All filters were chilled to 4.degree. C. prior to placement in the apparatus.

Prior to the application of the assay sample to the wells of the dot-blot apparatus, the wells were filled with 375 .mu.l of UL9 binding buffer. Typically, 5-50 .mu.l of sample (usually 10-15 .mu.l) were pipetted into the wells containing binding buffer and a vacuum applied to the system to pull the sample through the nitrocellulose. Unbound DNA passes through the nitrocellulose, protein-bound DNA sticks to the nitrocellulose. The filters were dried and exposed to X-ray film to generate autoradiographs.

3) Quantitation of Data

The autoradiographs of the nitrocellulose filters were analyzed with a Molecular Dynamics (Sunnyvale, Calif.) scanning laser densitometer using an ImageQuant software package (Molecular Dynamics). Using this software, a 96-well grid was placed on the image of the autoradiograph and the densitometer calculated the "volume" of each dot ("volume" is equivalent to the density of each pixel in the grid square multiplied by the area of the grid square). The program automatically subtracts background. The background was determined by either the background of a line or object drawn outside the grid or by using the gridlines as background for each individual dot.

The data is exported to a spreadsheet program, such as "EXCEL" (Microsoft Corporation, Redmond, Wash.) for further analysis.

4) Analysis of Data

The data generate from the densitometry analysis was analyzed using the spreadsheet program "EXCEL."

For each test oligonucleotide, at each drug concentration and/or each time point, a raw % score was calculated. The raw % score (r%) can be described as

where T was the densitometry volume of the test sample and C was the densitometry volume of the control sample. The oligonucleotides were then ranked from 1 to 256 based on their r% score. Further calculations were based on the rank of each oligonucleotide with respect to all other oligonucleotides.

The rank of each oligonucleotide was averaged over several experiments (where one experiment is equivalent to testing all 256 test oligonucleotides by the assay of the present invention) in view of the variability in rank between any two experiments. The confidence level for the ranking of the oligonucleotides increased with repetition of the experiment.

FIG. 15 shows the results of 4 separate experiments with distamycin. The test samples were treated with 10 .mu.M distamycin for 30 minutes. The r% scores are shown for each of the 4 experiments (labeled 918A, 918B, 1022A, and 1022B) and the ranks of each oligonucleotide in each experiment are shown. The test oligonucleotides have been ranked from 1 to 256 based on their average rank. The average rank was the sum of the ranks in the individual experiments divided by the number of experiments. FIGS. 16 and 17 show the results presented in FIG. 15 in graphic form. FIG. 16 shows the average ranks plotted against the ideal ranks 1 to 256. FIG. 17 shows the average r% scores plotted against the rank of 1 to 256. These data demonstrate the reproducible ability of the assay to detect differential binding and effects of distamycin on different 4 bp sequences.

One method used to determine the sequence preferences for distamycin was to examine the sequences that rank highest in the assay for sequence similarities. This process may be accomplished visually or by designing computer programs to inspect the data.

Using the data shown in FIG. 15, consensus sequences can be constructed for distamycin in the following manner.

Sequences with rankings less than 50 (indicating a strong effect of distamycin on the test sequence) in all four experiments were:

Sequences with rankings less than 50 (indicating a strong effect of distamycin on the test sequence) in three of the four experiments were:

Sequences with rankings less than 50 (indicating a strong effect of distamycin on the test sequence) in two of the four experiments were:

The following assumptions allow prediction of a consensus sequence for a distamycin recognition sequence: (i) the most favored sequences are the test sequences that rank in the top 50 in all four experiments; (ii) the next favored sequences will be the test sequences that rank in the top 50 in 3 of 4 experiments; and (iii) the next favored sequences will be the test sequences that rank in the top 50 in 2 of 4 experiments.

The positions in the test sequence are represented by the numerals 1, 2, 3 and 4. One consensus sequence that predicted from the above binding data is:

The nucleotides at each position can also be ranked:

Furthermore, the importance of the position of the nucleotide can be ranked. Examination of this data would indicate that the importance of the positions is

These data can be tested for validity by deriving all possible consensus sequences and examining their scores in the assay. The consensus sequences derived from the above information, in order of rank as predicted by the consensus sequence, are:

Note that the actual rank numbers are out of a possible 256 and that only one number is greater than 50. The average rank of these 16 oligos is only 17. These data indicate that the consensus sequence has predictive value.

Using the same data, a second consensus sequence can be derived that has slightly worse average rank with respect to the relative effect of distamycin in the assay.

The test sequences predicted by this consensus sequence are as follows:

This consensus sequence also appears to be predictive of favored distamycin binding sites since the average rank of test oligonucleotides predicted by this sequence is 29, substantially below the median rank of 128. However, the sequences predicted by this consensus sequence do not appear to be affected as strongly by distamycin as the sequences in the first consensus sequence, described above.

Actinomycin D has been tested for sequence selectivity and relative binding affinity to the 256 different 4 bp sequences.

1) The Assay Mixture and Calibrator Samples

The assay mixture was prepared as described in Example 10.

One assay mixture useful for the testing of actinomycin D contained 1.5 nM radiolabeled DNA and 12.8 nM UL9-COOH protein prepared as described above in the UL9 binding buffer (20 mM Hepes, pH 7.2, 50 mM KCl, and 1 mM dithiothreitol). The concentration of the components can be varied as described in the Detailed Description.

The assay mixtures containing both UL9 and DNA were incubated at room temperature for at least 10 minutes to allow the DNA:protein complexes to form and for the system to come to equilibrium. At time=0, the assay was begun by adding water (control samples) or actinomycin D (1-5 .mu.M, test samples) to the assay mixtures using a 12-channel micropipettor. After incubation with drug for 5-120 minutes, samples were taken and applied to nitrocellulose filters using a 96-well dot blot apparatus (Schleicher and Schuell) held at 4.degree. C.

In this experiment, calibrator samples were used to normalize the results between plates, that is, to take plate-to-plate variability into account. Calibrator samples were prepared using 2-fold serial dilutions of DNA in the assay mixture and incubating duplicate samples in one column of the 96-well assay plate. The highest concentration of DNA used was the same concentration used in the screening samples. In general, calibrator samples were used in all experiments. However, use of calibrator samples appeared to be less important for experiments using blocked plates since the variability between blocked plates is lower than between unblocked plates.

The calibrator samples were used to normalize the values between plates as follows. The volume values (Example 10) for the calibrator samples were obtained from densitometry. Volume values were plotted against DNA concentration. The plots were examined to ensure linearity. The volume values for the points on the calibrator line were then averaged for each plate. A factor, designated the normalization factor, was then determined for each calibrator line. When the normalization factor was multiplied by the average of the points on each calibrator line, the product was the same number for all plates. Usually, the average of the line averages was used for determining the normalization factor, although in theory, any of the line average numbers could be used. The operating assumption in this analysis was that the differences in the calibrator samples reflected the differences in adsorption for each plate. By normalizing to the calibrator samples, these variations were minimized.

Once the normalizing factor was obtained, all of the raw volume values for each of the test assays on the plate was multiplied by the normalizing factor. For example, if the following data were obtained, the process of normalization would be as follows:

Plate I has a normalization factor of 1; Plate II has a normalization factor of 1875/1969=0.95; Plate III has a normalization factor of 1875/1781=1.05. The equation used to establish these numbers is as follows: "Average average"/line average=normalization factor.

The sample data on each plate were then multiplied by the normalization factor to obtain normalized volume values.

The following actinomycin D experiment did not employ blocked 96-well plates. Accordingly, the results of this experiment are likely to exhibit more variability than if blocked plates had been used. Since unblocked plates were used, the above described calibrator sample method was used to normalize the data. However, the adsorption observed on the plates used in these experiments (Falcon PVC 96-well plates) was minimal during the time period used for this experiment (30'), as described in Example 9. The calibrator samples used in this experiment confirmed the low variability.

2) The Capture/Detection System

A 96-well dot blot apparatus was used to capture the DNA:protein complexes on a nitrocellulose filter as described in Example 10.

3) Quantitation of Data

The autoradiographs of the nitrocellulose filters were analyzed as described in Example 10.

4) Analysis of Data

After densitometry, the data was analyzed using the spreadsheet program "EXCEL." For each plate, the calibrator samples were examined and used to determine the normalization value. Then, for each test oligonucleotide, at each drug concentration and/or each time point, a normalized % score was calculated. The normalized % score (n%) can be described as follows:

where (i) nT is the densitometry volume of the test sample multiplied by the normalization factor for the plate from which the sample was obtained, and (ii) nC is the densitometry volume of the control sample multiplied by the normalization factor for the plate from which the sample was obtained. The oligonucleotides were then ranked from 1 to 256 based on their n% scores.

FIG. 18 shows the results of one experiment with actinomycin D. The test samples were treated with 2 .mu.M distamycin for 30 minutes. The n% scores and rank are shown for each of the 256 oligonucleotides.

FIG. 19 shows the individual n% scores plotted against the rank of 1 to 256.

For samples in which the plates are blocked, the use of calibrator samples becomes a backup check for consistency, but "normalization" is not necessary.

Once the relative binding preferences of a distamycin have been determined, sequences are selected for target sites for DNA-binding molecules composed of two distamycin molecules, bis-distamycins, or three distamycin molecules, tris-distamycins.

1.) Selecting Sequences for Binding with Highest Affinity to Distamycin Oligomers

The top binding sites for distamycin, determined as described above, are defined by the consensus sequence, 5'-T:T/A:C/A:C-3': accordingly, the top sequences are TTCC, TTAC, TACC and TAAC. Using this information, 2.sup.4 =16 possible dimer sequences, i.e., combinations of the four top binding sequences, can be targeted by a bis-distamycin in which the distamycin molecules are immediately adjacent to one another.

The top strands of the 16 possible duplex DNA target sites for binding bis-distamycins are shown in FIG. 20. Similarly, trimers of distamycin, tris-distamycins, could be targeted toward selected 12 bp sequences, comprised of all possible combinations of the four 4 bp sequences. There are 3.sup.4 =81 possible highest affinity target trimer sequences.

There are several advantages to targeting longer sequences with bis- or tris-distamycin:

2.) As the Number of Potential Target Sites Decreases, Specificity Increases

All 8 bp combinatorial possibilities of the 4 top favored binding sites for distamycin are potential high affinity binding sites for bis-distamycin. The consensus sequence used in this example predicts four favored binding sites for distamycin. This represents (4/4.sup.4)*100=about 1.6% of the possible 4 bp sites in the genome. Since there are 4.sup.8 possible 8 bp sequences, this represents, on average, only (2.sup.4 /4.sup.8)*100=about 0.02% of the total genome. There are 4.sup.12 possible 12 bp sequences, this represents, on average, only (3.sup.4 /4.sup.12)*100=0.00000075% of the genome.

The following discussion provides perspective and illustrates the improvement in the actual number of target sites in the human genome for when using a dimer of distamycin versus a monomer of distamycin. The human genome is about 3.times.10.sup.9 bp. If the number of favored target sites for distamycin is four, and the number of possible 4 bp sequences is 4.sup.4 =256, then the number of favored target sites in the genome is (4/256)(3.times.10.sup.9)=4.7.times.10.sup.7, or about 50 million favored target sites.

Given that the number of possible 8 bp sites is 4.sup.8 =65,536, if all possible combinatorial 8 bp sites derived from the favored 4 bp sites (2.sup.4 =16; FIG. 20) are favored, then the number of favored 8 bp target sites is (16/65,536)(3.times.10.sup.9)=7.3.times.10.sup.5 or about 700,000 possible sites. This represents a 64-fold reduction in the number of highest affinity target sites between distamycin and bis-distamycin; alternatively, this result can be viewed as a 64-fold increase in specificity.

Likewise, given that the number of possible 12 bp sites is 4.sup.12 =1.7.times.10.sup.7, if all possible favored 12 bp sites (3.sup.4 =81) are favored, then the number of favored 12 bp target sites is (81/1.7.times.10.sup.7)(3.times.10.sup.9)=1.4.times.10.sup.4 : i.e., 14,000 possible highest affinity sites. This represents an approximately 3000-fold decrease in the number of highest affinity target sites between distamycin and tris-distamycin and a 500-fold decrease in the number of highest affinity target sites between bis-distamycin and tris-distamycin.

3) An Exponential Increase in Affinity

As the target site increases in size, (i) the number of target sites in a defined number of nucleotides decreases, and (ii) the specificity increases. Further, the affinity of binding is typically the product of the binding affinities of component parts (see Section IIi.e.1 above). As an example, the published binding constant for distamycin to bulk genomic DNA is about 2.times.10.sup.5 M.sup.-1. Dimers of distamycin will have a theoretical binding affinity of the square of the binding constant of distamycin:

Trimers of distamycin will have binding affinities of the cube of the binding affinity of distamycin:

Thus, if distamycin shows only a 10-fold higher affinity (2.times.10.sup.6 M.sup.-1) for the top favored binding sites than the average binding sites in DNA, then the affinity constant for bis-distamycin to an 8 bp site comprised of two favored binding sites is 100-fold higher than for an 8 bp sequence comprised of two average binding sites:

While this does not represent absolute sequence specificity in binding, the binding affinity is 100-fold greater for 0.02% (16/65, 536) of the total possible 8 bp target sequences.

The use of a trimer targeted sequence will afford an even higher increase in affinity to the most favored binding sites:

Thus, with only 10-fold differential activity in binding between favored sites and average sites, a 1000-fold difference in affinity can be achieved by designing trimer molecules to specific target sites. When considering the administration of DNA-binding molecules as drugs, a 1000-fold lower dose of tris-distamycin, versus the distamycin monomer, could be administered and an increase in relatively specific binding to selected target sites achieved.

In this example, the differential activity of distamycin is only 10-fold. Clearly, differential activities of larger magnitudes will greatly accentuate the increased affinity effect. For example, a 100-fold difference in activity of a 4 bp DNA-binding molecule toward high affinity and average affinity sequences would result in (i) a 10,000-fold difference in the binding affinity of a dimer of the molecule targeted to an 8 bp sequence, and (ii) a million-fold increase in the binding affinity of the trimer to a 12 bp sequence.

4) Selecting Target Sequences for Distamycin Oligomers with Flexible and/or Variable-Length Linkers in Between the Distamycin Moieties

The sequences that can be targeted with bis- or tris-distamycin molecules are not limited to sequences in which the two 4 bp favored binding sites are immediately adjacent to one another. Flexible linkers can be placed between the distamycin moieties and sequences can be targeted that are not immediately adjacent. The target sequences can have distances of 1 to several bases between them: this distance depends on the length of the chemical linker. Examples of bis-distamycin target sequences for bis-distamycins with internal flexible and/or variable length linkers targeted to sites comprised of two TTCC sequences are shown in FIG. 21, where N is any base.

For each particular bis-distamycin, the explanations of increased affinity and specificity remain the same as described above with the following exception. For the case in which the linker was sufficiently flexible to span different numbers of bases in between the two distamycin sites, the number of sites targeted with highest affinity would be multiplied by the number of bases spanned.

In respect to the ease of drug design and target selection, there are several advantages to the above described targeting strategies, including the following:

i) Any conformational changes induced by binding at the half-site would be minimized.

ii) The affinity, therefore, would be more likely to be the product of the affinities of the interactions observed for the monomeric sites.

iii) The half-molecule (e.g, 1 distamycin unit) would anchor the bis-molecule (e.g., bis-distamycin) thus increasing the localized concentration for the binding of the second half of the bis-molecule.

iv) If a simple linking chain is used, with a variable number of atoms, the number of sites that can be targeted by multimers of the monomer increases. This targeting method can be of value when, for example, there are no medically significant target sites with adjacent favored binding sites for distamycin. Therefore there are no good target sites for bis-distamycin. In this situation, the database can be screened for additional target sequences with N.sub.1 to n (where N is any base) between the two target binding sequences. For example, where n=4, the number of sequences to be searched becomes (4.sup.2)*4=64. The likelihood of finding such a sequence is reasonably high.

5) Selecting a Specific Target Site

Using the above approach, a sequence was identified from the medically significant target site database that contains SEQ ID NO:19, which is a subset of the group of sequences represented by SEQ ID NO:20. SEQ ID NO:19 occurs overlapping the binding site for a transcription factor, Nuclear Factor of Activated T Cells (NFAT-1), which is a major regulatory factor in the induction of interleukin 2 expression early in the T cell activation response. NFAT-1 is crucial in (i) the T cell response, and (ii) in blocking the expression of IL-2, which causes immunosuppression. The sequences TTCC and TTTC, the distamycin target binding sequences in SEQ ID NO:19, rank first and seventh in the assay.

The assay of the present invention measures the effect of the binding of a DNA-binding molecule to a test site by the release of a protein from an adjacent screening site. Accordingly, the assay is an indirect assay. Following here is the description of an application of the assay useful to provide confirmatory evidence of the data obtained in the initial screening processes.

The results of the distamycin screening assay described in Example 10 suggested that there were possible false negatives: specifically, test sequences that bind distamycin but fail to show an effect on the binding of the reporter protein. The data suggesting false negatives was as follows. If the assay detected strictly the affinity of binding of distamycin, then the scores of the test sequences complementary to the high-scoring test sequences should always be equally high. However, an examination of the highest ranking test sequences and the complementary test sequences reveals that this is not the case (see Table XIV).

All but one of the complementary sequences rank in the lower half, 4 of them in the lowest 20%, i.e., these was little effect on reporter protein binding in the presence of distamycin when using these sequences as test sequences in the assay.

This observation reflects the usefulness of a confirmatory assay that examines the relative affinity of a particular sequence for binding distamycin. A confirmatory assay may also be useful in revealing additional information about the physical characteristics of drug binding. For example, one can hypothesize that the reason for the apparent inverse relationship between test sequences with high activity in the assay and their complements is that the effect of distamycin is directional and only active at one test site. This hypothesis can be tested using the following competition experiment. Competitor oligonucleotides, containing test sequences of interest, are added to the assay mixture. This allows the determination of which test sequences compete most effectively with the radiolabeled test oligonucleotide for binding distamycin.

Assay mixtures are prepared as described in Example 10, using a high-ranking test oligonucleotide, e.g., TTCC (ranking=#1), as the radiolabelled oligonucleotide in the experiment. The test oligonucleotide TTCC is labelled to high specific activity with .gamma.-.sup.32 P-ATP as described in Example 8: in this example, the labeled TTCC oligonucleotide will be referred to as the "high specific activity test oligonucleotide".

The competitor oligonucleotides are labeled as described in Example 8, except that the ATP used for kinasing the primer is 1:200 radiolabeled:nonradiolabeled. In other words, the competitor oligonucleotides are tracer labeled with radioactive phosphorous to a 200-fold lower specific activity than the high specific activity test oligonucleotide. Since all of the competitor oligonucleotides are labeled with the same radiolabeled primer molecule, the relative concentrations of the competitor DNAs can be determined with high accuracy. Further, since the specific activity is the same, the concentrations can be adjusted to be the same. For the purposes of this example, the competitor DNAs are referred to as "low specific activity competitor oligonucleotides."

The use of competitor DNAs for which the concentration is known is important for the competition experiment. The accuracy of the competition assay may be further enhanced by separating any unincorporated radiolabeled primer from the double stranded competitor oligonucleotides. This separation can be achieved using, for example, a 6-20% polyacrylamide gel. The gel is then exposed to x-ray film and the amount of double-stranded oligonucleotide determined by use of a scanning laser densitometer, essentially as described in the Examples above.

The competition assay is performed as described in Example 10, except that competitor DNAs are added in increasing relative concentration to the high specific activity test oligonucleotide. The DNA concentration (›DNA!) is held constant and the UL9 concentration (›UL9!) and distamycin concentration (›distamycin!) are as described in Example 10. The components in the competition assay samples are as follows.

Controls:

UL9+TTCC,; UL9+TTCC*+Competitors; UL9+TTCC*+distamycin;

Test Samples:

UL9+TTCC*+distamycin+Competitors;

where UL9 is UL9-COOH, TTCC* is the high specific activity test oligonucleotide, and Competitors are the low specific activity competitor oligonucleotides.

TTCC-low (the tracer-labeled low specific activity competitor) competes with TTCC, on an equimolar basis for the binding of both protein and distamycin. A competitor molecule with lower affinity for distamycin than TTCC requires a higher molar ratio to TTCC, to compete for distamycin binding. The competition for protein between all competitors is constant. Only the competition for distamycin varies; the variability is due to the differential affinity of the competitor oligonucleotides for distamycin. The concentration of competitor used in these experiments ranges from 0.1-fold of the TTCC* concentration to 100-fold of the TTCC* concentration.

The competition assay described here facilitates the determination of actual rank between the test oligonucleotides that are detected as highly effective molecules in the original assay. The competition assay also facilitates the detection of false negatives. As described above, the results of the assay discussed in Example 10 imply "directional" binding of distamycin, in which the effect of binding is only detected when the molecule is bound in one direction with respect to the UL9 protein. Binding in the opposite direction (i.e., to the complementary test sequence) is not detected with the same activity in the assay.

The purpose of this competition experiment is to use the test oligonucleotides to compete for the binding of distamycin. If the sequences complementary to the "best binders" are false negatives in the assay, they should nonetheless be effective competitors in the competition assay.

The binding of a drug or other DNA-binding molecule to the recognition sequence for TFIID, or other selected transcription factors, is expected to alter the transcriptional activity of the associated gene. TATA-boxes, which are the recognition sequences for the transcriptional regulatory factor TFIID, are associated with most eukaryotic promoters and are critical for the expression of most eukaryotic genes. Targeting a DNA-binding drug to TATA boxes in general would be undesirable. However, sequences flanking TATA box sequences are typically unique between genes. By targeting such flanking sequences, perhaps with one base overlapping the TFIID recognition site, each gene can be targeted with specificity using the novel DNA-binding molecules designed from the data generated from the DNA-binding drug assay. One method for determining novel and specific target sequences for novel DNA-binding drugs is described here. The method may be applied to any known binding site for any specific transcription factor, regardless of whether the identity of the transcription factor itself is known. TATA-boxes have been determined for a large number of genes. Typically, the TATA-box consensus sequence has been identified by examining the DNA sequence 5' of the RNA start site of a selected gene. However, the most rigorous determinations of TATA boxes have also demonstrated the transcription factor binding site by DNA protection experiments and DNA:protein binding assays (using electrophoretic methods). Many of these sites are annotated in the public databases "EMBL" and "GENBANK", which both contain sequences of nucleic acids sequences. Unfortunately, the flat field listing of these databases do not consistently annotate these sites. It is possible, however, to automatically search a database, using a text parsing language called AWK, to extract most sequence information that relates to annotated promoter sequences.

The following is a description of how selected promoter sites were located in the public database from "EMBL." The flat field annotations from "EMBL" Version 32 as processed by "INTELLIGENTICS" (Mountain View, Calif.), were obtained with the set of UNIX programs call "IG-SUITE." These programs were executed on a "SUN IPX" workstation. An AWK script was used to parse all the primate annotation files listed in the "EMBL" database. The AWK interpreter is supplied as part of the system software that comes with the "SUN IPX" workstation.

The following is a description of how the AWK parses annotation files looking for and printing information relating to promoters and TATA-boxes. The system is asked to examine the input files for certain key words in the header lines or annotations to the sequence. The AWK interpreter reads input files line by line and executes functions based on patterns found in each line. In this case, the AWK system read the annotation files of EMBL. The following is a description of how the AWK script can be used to parse out sequences containing TATA-boxes.

The program first examines the files for all header lines containing the word "complete" but not "mRNA" or "pseudogene"; the output is printed. Complete genes sometimes contain the promoter sequences but complete mRNA genes do not contain the promoters. mRNA genes are not of interest for the purpose of detecting promoter elements. Next, the AWK system looks for the word "exon 1" and if it finds it prints the header and "DE" line. Then it looks for "5'" and prints the header line if it does not contain the word "mRNA". Next it looks for the word "transcription" and if it finds it prints the preceding and following line along with description line.

Next, the AWK system examines the files for the word "TATA" in the header lines or references. This results is printed. After this it looks for the word "promoter" and if it finds it prints that line and the line after it which contains the information about the promoter. Then the program looks for "protein.sub.-- bind" and prints that line along with the next one. The description of "protein.sub.-- bind" is usually used to mark potential binding sites of transcription factors in the "EMBL" database. AWK then scans for any annotated primary mRNA start sites. The promoter sequence is found in front of the start site. Finally, any exon 1 start sites that are annotated in the feature table are extracted. Exon 1 start sites should also be the primary transcription start site and the TATA boxes usually are found approximately 25-35 base pairs 5' to the transcriptional start site.

The actual AWK script is included here as an example of how to parse a database to extract promoter sites:

After the AWK script is run on the database the output is manually examined. Those sites that are clearly promoter sites are noted and nucleotide coordinates recorded. Other gene sequences are examined using the "FINDSEQ" program of "IG.sub.-- SUITE" to see if the promoter sites can be determined or if the references in the database describe the promoter sequences. If so, those nucleotide coordinates are noted. At the end of this examination "FINDSEQ" is used to extract any sequences containing promoter sequences by using an indirect file of "LOCUS" names constructed using a text editor.

A parsing program was also written to extract each of the annotated sites from the file that "FINDSEQ" extracted from "EMBL." This program extracts the following information: the promoter site name and four numbers representing the nucleotide coordinates of where the sequence is to start, what the coordinate of the first base of the site is, the coordinate of the last base of the site and the end of the sequence to be extracted. A large batch file was constructed to automatically extract each of the promoter sites. These sequences formed the basis of Table IV.

While the invention has been described with reference to specific methods and embodiments, it will be appreciated that various modifications and changes may be made without departing from the invention.