JP2003524364A - Identification of molecular targets - Google Patents

Abstract

(57) [Abstract] Identification of molecular targets of drugs or toxins is the first step in understanding how drugs or toxins work, learning how to improve drugs or assess the dangers of toxins. Is an important step forward. The primary action of an agent usually involves binding to a protein. The second effect manifests itself in the form of side effects, and in some cases, by binding to other proteins. Therefore, it is useful to identify all physiologically relevant sites of a drug or toxin. Simple methods for obtaining a list of possible targets of drugs, toxins or other biologically active substances (collectively called ligands) include multi-step methods. The first step is to screen a protein or peptide library to identify library members that exhibit high affinity for a particular ligand. The second step involves searching a sequence database for proteins containing sequences of library members that have been shown to have high affinity for the ligand. The proteins thus identified constitute a list of possible targets for the ligand. If a random peptide library is used, the location of the identified consensus sequence within the identified protein will identify potential ligand binding sites on the target.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION

The present invention is generally directed to methods for the identification of molecular targets for drugs or toxins in organisms or other biological systems.

Most drugs or toxins express their activity by binding to proteins. These proteins are called receptors, drug targets or molecular targets (Gies, 1996). A drug (pharmaceutical agent), toxin or other biologically active molecule is referred to herein as a ligand. The identification of ligand targets is a crucial first step in understanding how ligands affect biological systems. Today, this identification is generally a long and difficult process. However, identification of the target of the ligand is desirable because it provides essential information for drug modification or assessment of toxicity or side effects.

Many drugs are currently designed to specifically bind to specific target proteins, and their primary target is certain. However, these agents may have additional targets that they bind to and produce unexpected or undesired biological effects (toxicity or side effects). The origin of these side effects or toxicities is not always apparent from the main mode of action of the drug. The identification of the second step, whose interaction causes side effects, helps to monitor the initial toxicological assessment in humans by identifying possible biological systems, helps interpret the observed side effects, or Provide information that can be used to interfere with

In addition to engineered agents, natural or synthetic organic substances are often screened for specific biological activities (eg, suppression of human cancer cells in culture) and those exhibiting desirable activities are shown. Identified and developed without prior knowledge of the molecular target from which its activity is obtained. The first step in understanding the mode of action of these drugs is to establish their molecular targets. this is,
It is often a slow and expensive process. However, the identification of primary and secondary targets for these types of agents is crucial for their further development and toxicological evaluation.

Prior to assaying the new drug in humans, the drug is tested in animals to assess its toxicity. The success of these toxicological screens relies on the potency that mimics the human system in which animal models are to be performed. If the animal molecular target is essentially the same as the human molecular target, toxicological evaluation in animals can provide a good example of the toxicity of a drug in humans. However, this is not universally applicable. Many drugs and toxins are highly species-dependent in their action (eg aspirin is toxic to mice. Ohdo, et a.
l., 1995). More suitable test animals can be selected if a list of possible human molecular targets is available prior to testing in animals. For example, if the potential target of the drug is the enzyme hexokinase (the first enzyme in the glycolytic pathway), the sequences of human and mouse hexokinase can be compared. If these sequences are similar at putative drug binding sites, mice are an acceptable model for the evaluation of drug effects on glycolysis. If not, the use of another animal model is indicated. Thus, the ability to predict potential drug binding sites in advance of animal studies lends itself to the design and evaluation of pharmacological screens. Furthermore, during clinical trials, the list of possible targets simplifies the evaluation of drug side effects.

During clinical use of a drug, the unexpected benefit of identifying a drug being used to treat one medical condition as being effective against another medical condition may be observed. This is particularly beneficial because the drug in clinical use has already passed many regulatory hurdles and has undergone toxicological evaluation. The list of possible targets for drugs already in use may provide clues for new applications as well as a list of medical conditions for which the drug needs to be tested. This is especially beneficial for rare illnesses where there is little financial incentive for drug development.

In addition to defining the desired mode of drug interaction, the identification of molecular targets is also essential to understanding the action of environmental toxins. Man-made and natural toxins pose a constant risk to the human population. Assessment of the risk posed by these molecules depends on the determination of the mode of action of the toxin. Moreover, these molecules, in some cases, have several distinct physiologically important targets. A full characterization of the risks associated with exposure to these toxins involves the identification and characterization of all relevant molecular targets.

Current methods for identifying molecular targets for drugs or toxins in biological systems include:
It's troublesome. Generally, they involve culturing large numbers of mammalian or other organism cells in order to harvest sufficient protein extract to test binding to the ligand. Once these proteins have been extracted, they must be isolated in sufficient quantity to sequence the proteins by affinity for their ligands (a difficult task for low expressed proteins). next,
The purified protein must be partially sequenced by Edman degradation to confirm the putative peptide sequence. If this sequence is of sufficient quality,
In order to screen the genomic library of the original cell of interest, a set of synonymous DNA hybridization probes must be devised. If this step is successful, the gene for the protein is recovered, cloned into an expression vector and then sequenced. This method results in protein identity that is suspected to bind the ligand, but the steps of cell culture, purification, peptide sequencing, and probing for hybridization of the gene of interest are all expensive and time consuming.

Sparks et al. (1996) and Hoffmann et al. (1996) reported screening human and mouse protein libraries generated from cDNA to identify proteins with high affinity for a particular peptide. . They state that they only screened for peptides (eg, not small molecule drugs or toxins). In addition, a random peptide library (New England Biolabs Product Catalog, “Ph.D” prod) for screening against antibodies to identify an epitope that is another form of protein-protein interaction is identified.
ucts, 1998) are commercially available. Screening for proteins that show high affinity for peptide ligands is conceptually and practically different from using small molecules as ligands. Protein-protein interactions generally involve large-scale chemical interactions of large structures on each protein that generally include relatively large sites of interaction. Therefore, the binding energy is usually
It is much more powerful with respect to peptide ligands. However, the use of this technique is relatively limited because many such biologically active molecules are not proteinaceous.

[0010]

[Outline of the Invention]

Accordingly, one of the objects of the present invention is to provide a method for identifying potential molecular targets for drugs, toxins or other biologically active small molecules.

Thus, in summary, the present invention has a molecular weight of less than 5,000 Daltons,
Moreover, the method for identifying a protein that binds to a ligand other than a peptide or a protein is targeted. In this method, the ligand contains a gene package (ge) containing the corresponding base sequence to identify a member of the library that has an affinity for the ligand that is greater than the affinities of the other members of the library.
screened against a library of peptides or proteins, each displayed on the surface of netic packages). Each member of the peptide or protein library is physically linked by a genetic package to a nucleic acid polymer encoding that member, which also allows the peptide or protein to interact with a ligand. Members of the library that have an affinity for a ligand that is greater than the affinities of other members of the library are separated from the library, and the base sequences encoding these members are determined and translated into peptide or consensus peptide sequences. To be done. A protein sequence database is then searched to identify proteins that contain the peptide sequence or that correspond to the consensus peptide sequence.

Other objects and features of the invention will in part be obvious and will in part be described below.

Definitions “Ligand” as used herein means a small molecule (less than 5 kD) capable of binding to a protein, preferably other than a nucleic acid, peptide or protein.

“Peptide” as used herein means an unbranched amino acid polymer ranging in size from 3 amino acids (the lowest number for which consensus sequence information is useful) to any length. To do. Some very long peptides as defined herein are sometimes referred to by those of skill in the art as "proteins".

A “gene package” as used herein connects a protein fused with a peptide from a peptide library that carries the genetic information encoding the fused peptide, while interacting with the ligand of interest. It means any manner of presenting the peptide in such a way that it can act. A non-exhaustive list of genetic packages includes: phage peptide display systems, bacterial pili display systems, yeast surface protein display systems, plasmid DNA binding fusion protein systems and other similar modalities.

By “plating agent” as used herein is meant any molecule that can be used to immobilize a conjugated molecule on a solid support, including molecules including solid supports.

“Amplification,” as used herein, refers to the replication of a genetic package that displays a member of a peptide library and that includes DNA encoding that member of the library.

“Taxane,” as used herein, has the following A, B, and C rings (where the ring positions are numbered):

[Chemical 1] Means a compound having

[0019]

Detailed Description of the Preferred Embodiments

The action of the ligand on the molecular target depends on the energy of interaction or binding energy with the target. It is usually characterized as a binding or dissociation constant. In this specification, the concept of dissociation constant is used. Regarding dissociation of the protein-ligand complex PL into the ligand L and the protein P, PL ----> P + L (1) The dissociation constant K _D is defined as follows. _{K D = [P] [L} ] / [PL] (2) ( wherein, [P] and [L] is the concentration of each protein and a ligand, [PL] protein - is the concentration of the ligand complex ). The most intuitive characterization of K _D is that it corresponds to the concentration of ligand at which half the protein remains bound to the ligand. Physiologically relevant binding constants are generally 10-15
Smaller than micromolar, often in the nanomolar range.

Small molecules usually bind to the accessible surface of proteins in pockets or grooves where the contact area between the ligand and the protein is larger. The larger the contact area, the larger the number of bonds that can be made between the ligand and the protein, and the smaller the dissociation constant. If the ligand binding induces a substantial conformational change in the protein, the ligand can become deeply embedded within it. Proteins are, of course, polypeptide chains consisting of unbranched polymers of amino acids. To form the original conformation, protein chains fold into a complex structure in which amino acids from different parts of the amino acid chain interact with each other. Usually, but not always, the binding site consists of amino acids from several different regions of the protein chain. Occasionally, the presence of a ligand may stabilize the disordered portion of the protein chain as it wraps around the ligand. Thus, in most cases no more than a few contiguous amino acids make contact with the small ligand when it is bound to the protein in its original conformation.

It is useful to identify the amino acids that make up the binding site for the ligand and that form the interaction between the protein and the ligand from those that form the scaffold around the binding site. These amino acids help make that site, but have no direct interaction with the ligand. The binding site amino acids do not necessarily exhibit a particular affinity for the ligand if they are not supported by the appropriate scaffold. As a result, a proper scaffold is required to construct or mimic the binding site.

In a preferred embodiment of the invention, the first step in the identification of possible ligand targets is
Screening a peptide or protein library for sequences that exhibit a relatively high affinity for a particular ligand when compared to other sequences in the library. Most protein or peptide libraries are expressed on the surface of genetic packages that provide the basic proteins on which the library peptides are displayed.

The protein or foreign peptide may be introduced into a phage, a bacterium, a yeast cell or by the insertion of a nucleic acid corresponding to the peptide or protein sequence into the genome (DNA or RNA) of the genetic package or a vehicle of choice. It can be displayed on the surface of other gene packages. To screen a library of peptides for binding with the ligand of interest, it is desirable to make a physical or logical connection between each peptide and the nucleic acid encoding the peptide. Usually this is accomplished by fusing the DNA sequence corresponding to the peptide with the gene encoding the surface protein of the gene package. If the alterations to the surface protein do not prove harmful to the biological system in which it is displayed, the genetic package with the foreign protein or peptide on its surface can be propagated.
After several rounds of screening for affinity for the ligand and reculturing of the binding candidate, such ligation allows identification of the genetic material encoding the peptide. Once the proteins or peptides with the desired binding properties are obtained, their sequences can be obtained by sequencing the corresponding nucleic acids within the genetic package. The sequence thus identified may correspond to a sequence within the protein that binds the ligand. In order to optimize the probability of getting a positive result, it is necessary to use several libraries that exhibit many scaffolds.

Although display on a variety of different gene packages has been achieved, most studies have involved bacteriophage particles. The concepts involved in library construction and screening are introduced here using the example of phage as a mediator, but this does not exclude other gene packages. Other types of uses in the present invention include DNA binding fusion protein systems and yeast membrane fusion protein systems. The use of other protein or peptide library display systems is well understood in the art and the choice of a particular display system for use is largely a matter of convenience of the particular inventor.

Several phage-based systems for displaying peptides on the surface of phage have been described in the literature. The fusion phage approach of Parmley and Smith (1988, Gene 73: 305-318) has been used to display proteins. Others
A phage-based system in which a peptide is fused to the p3 coat protein of filamentous phage has been described (Scott and Smith, 1990, Science 249: 386-390; Devlin et.
al., 1990, Science 249: 404-406; and Cwirla et al., 1990, Proc. Natl. A.
cad. Sci. USA 897: 6378-6382 (each description is incorporated herein by reference). See also the discussion of phage fusion technology in US Pat. No. 5,498,530 (Schatz, PJ, etc.). These latter references describe the fusion of peptides at the amino terminus of the p3 protein). The fusion protein is part of the capsid encapsulating the phage genomic DNA, thus establishing a connection between the isolated peptide and the genetic material encoding the peptide. Phage encoding peptide ligands for the receptor of interest can be isolated from the peptide display library after several rounds of affinity enrichment followed by phage expansion.

By inserting different nucleotide sequences into each phage genome, a library of billions of different phage particles, each displaying a different peptide or protein, can be constructed (Scott and Smith, 1990). The diversity of the library is
Selection depends on the length of the inserted peptide or protein (ie, greater diversity can be achieved by 12 amino acid long inserts than 4 amino acid long inserts) and by the utility of handling very large numbers of phage particles. Limited by censoring sequences that are harmful to the living system. The latter restriction can be reduced by using many phage libraries generated separately in parallel experiments.

The library can be constructed in many ways known in the art, but two methods are described here in some detail. The first method is a random peptide library similar to that used in Example 1. In this case, all or part of the nucleic acid that is inserted into the phage genome is randomized (eg, by chemically synthesizing a partial or fully random oligonucleotide and then inserting it into the genome). This technique is typically used to generate 5-12 random amino acids on the surface of phage particles.

Several methods exist for generating DNA sequences encoding random peptide libraries. The most common consists of the production of DNA oligomers of synthetic trinucleotide codons. These nucleotide trimers can be easily synthesized using the solid support method (McBride and Caruthers, 1983, Tetr. Let.
ters 22: 245). Next, the trimers are mixed in a desired molar ratio to give a solid support D
It is used as a structural unit in NA synthesis. The ratios are usually approximately equimolar, but can be controlled in unequal ratios to obtain over- or under-representation of certain amino acids encoded by the oligonucleotide population. Condensation of trimers to generate oligocodons is performed essentially as described for conventional syntheses using activated mononucleosides as building blocks (generally Atkinson and Smith, 1984). , Oligonucleotide synthesis (JJ Gait
, ed.), pp.35-82). This technique is equal (or controlled unequal)
To produce a population of oligonucleotides for cloning that could encode the possible peptide sequences of and to minimize accidental synthesis of stop codons (US Pat.
5,498,530 (Schatz, etc.)).

In order to identify consensus sequences that occur in some clones,
Screening a random library requires the sequencing of many clones. Multiple occurrences of a sequence or similar sequences in clones isolated from a random library may result in relatively low affinity that was accidentally isolated due to non-specific binding during affinity selection. It is necessary to distinguish those sequences that have a high affinity for those in the rest of the library from those that have. Identification of commonly occurring sequences that are the result of desirable growth properties as opposed to desirable binding properties can be accomplished by statistical analysis of 100 randomly selected members of the library. Computer software that screens these 100 random sequences on a scale of 20 or more amino acid signatures (including properties such as hydrophobicity, flexibility, etc.) detects unique trends in each library and identifies useful growth characteristics. To identify the library members that predominate in number.

The consensus binding sequences identified by screening a random peptide library usually do not make up all random inserts. The rest of the randomized inserts provide scaffold variation around the consensus sequence and / or completion of binding sites. For example, a 5 amino acid consensus buried in a 12 amino acid insert is bound to as many as 8x20 ⁷ (1.024x10 ¹⁰ ) possible scaffolds in the library. This suggests that the use of extensive libraries with peptide presentation, including linear and cyclized configurations, allows pentapeptides to be displayed in unusually many conformations. Some of these may correspond to the conformation it adopts at the surface of the protein that constitutes the natural target for the ligand.

The second method for constructing the library is DN from cDNA library.
Insert A. A cDNA library is constructed from messenger RNAs within a particular tissue and may be constructed by PCR from single cells. m
RNA was isolated from tissue and, using reverse transcription, D corresponding to the sequence of the mRNA.
Synthesize NA. See US Pat. No. 5,629,179 (Mierdorf et al.) for a cDNA insert library generation protocol. When inserted into the phage gene package, this results in a library of phage with a surface displaying peptides having the same sequence as the protein or fragment of the protein from the tissue used to isolate the mRNA. The cDNA library is censored by the same mechanism as the random peptide library. Some inserts, especially large ones (eg, Rodi and Makowski, 1997), can be lethal to the biological system being used. Some inserts do not fold into the conformation of the original protein, rendering their binding properties inappropriate, and others constitute fragments of the protein that may or may not exhibit their original binding properties. You can

A list of target proteins that could be considered successful in screening a cDNA library is provided, all corresponding to fully expressed proteins. If the sequence of the identified protein is in the sequence database, information about binding adds to the information already known about the protein. If the protein sequence is not in the database, the sequence means the identification of a new protein. False positives can result from misfolded proteins that bind the ligand in a non-physiological manner and proteins that bind non-specifically to the ligand. False negatives are m from proteins that are misfolded and do not bind in their normal physiologically relevant manner.
It can arise from proteins that are not displayed in the library due to non-expression at the time the RNA is reverse transcribed, and from proteins that are censored from the library because they are lethal to the systems that express them.

Successful screening of random peptide libraries results in the identification of one or more consensus sequences that exhibit affinity for the ligand. Proteins that have been previously characterized and contain this sequence are described in GenBank and SWISS.
-Can be identified by searching a sequence data bank such as Prot. Some of the identified proteins contain consensus sequences in a conformation that does not bind ligand. These are false positives. Other proteins may display sequences in a manner that does not bind the ligand. These ligand binding proteins may exhibit binding constants that may or may not be within physiological relevance. False negatives occur for all proteins that have not yet been sequenced and placed in the database. However, the number of false negatives is substantially reduced as the Human Genome Project research continues. Other false negatives occur when the library being screened does not properly mimic the environment of the ligand binding site in the protein.

The use of cDNA libraries has both advantages and disadvantages over random libraries. All clones isolated on the screen correspond to proteins and therefore a small number of extraneous sequences are obtained. And usually cd
The NA library can represent the entire sequence of the protein target. However, only proteins that are actively expressed in cells have the opportunity to be detected on the screen. An unknown number of proteins fold incorrectly on the surface and go undetected (or provide a false positive). Membrane proteins or proteins that form large macromolecular assemblies do not appear to fold properly on the phage surface. Proteins from other species, early-onset proteins or rarely expressed proteins each need to be screened for different libraries. Selection of the appropriate cDNA library for drug or toxin screening can be guided by knowledge of the physiological effects of the drug or toxin and / or known target organs. Due to the smaller peptides displayed in random peptide libraries, precise folding of their short sequences is not required for isolation of the consensus sequence.
Amino acids involved in ligand binding are displayed in a wide range of molecular backbone environments, either of which may correspond to the correct molecular context for binding. A further advantage of screening a random peptide library over a cDNA library is that it provides information about the binding site as well as the proteins involved.

The peptide library displayed in the gene package is biopanning (
(As used in Example 1) (Kay et al., 1996; Yu, Y., et al., 1996, Method
s Enzymol. 267: 3-27; Petrenko, VA, et al., 1996, Protein Engineering,
9: 797-801), column chromatography, Southern and Western blotting, and electrophoresis, and can be screened for peptides that exhibit a relatively high affinity for a particular ligand. The advantage of physically connecting the displayed peptide to the nucleic acid that encodes it is, at least in principle, to isolate even a single particle that binds the ligand because the genetic package in which it is displayed is characterized. It is suitable for detection because it is used to grow a large amount of the same particles. Thus, the isolation of clones with a particular affinity for a selected ligand provides the opportunity to determine the sequence of peptides or proteins exhibiting that affinity. Once the gene package displaying peptides with the desired binding properties has been isolated, their sequences can be obtained by sequencing the corresponding nucleic acids within the gene package. The sequence thus identified may correspond to a sequence within the protein that binds the ligand. Several libraries need to be used to optimize the probability of obtaining useful results.

The number of positives identified by a particular sequence can be estimated. A given pentapeptide sequence occurs approximately once every 3.2 million amino acids. Given the 500 amino acids in the human genome and an average protein size of 75-100,000 proteins, a particular pentapeptide may be expected to occur approximately 16 times. Some sequences are far more common than others, so the protein is in a random fashion, but contains consensus sequences if it is not developed like repeats and modifications of existing genetic units or elements. The number of proteins can easily be doubled or substantially less. Furthermore, even if the entire human genome is sequenced, screening with random peptide libraries rarely yields an impractical list of human proteins containing that sequence.

Kabsch and Sander, Proc. Natl. Acad. Sci. USA 1984, Feb; 81 (4): 1075-
1078 searched for 62 related proteins with 10,000 residues and identified 25 cases in which the same pentapeptide appeared in two unrelated proteins. In 6 out of 25 identical 5 residues were present in the α-helix of one protein and in the β-sheet of another. Nevertheless, in many cases the structural similarity between pentapeptides of identical sequence was significant. Minor and Kim (1996) can be engineered so that the 11 amino acid sequence folds into an α-helical conformation when inserted at one position in the protein and a β-sheet conformation when inserted at another position. Proved that. This suggests that the same sequence may be displayed in very different conformations in different proteins, and in order to have a good opportunity to identify drug targets, many libraries should be identified to identify binding motifs. Indicates that screening is needed.

A limitation in most previous studies involving peptide display libraries was that they were only screened for affinity for a given protein. The present invention relates to screening for affinity for small molecule ligands (drugs, toxins). This difference is operationally very large because the dissociation constants used are very large (low affinity) and therefore more difficult to detect. The differences are physiologically very large, as the properties of the drug and toxin binding sites appear to render this approach infeasible.

However, our search of protein databases for three-dimensional structures identified hundreds of unique and known three-dimensional structures with bound ligands.
Analysis of these structures shows that for almost all ligands with molecular weights above ˜300 daltons, at least one stretch of 5 or more consecutive amino acids was involved in direct binding to the ligand (FIG. 1). reference). The graph shows the number of consecutive amino acids that "contact" with the binding ligand based on the three-dimensional structural data collected for some proteins. "Contact" was defined to be less than 4 or 5 angstroms from the ligand. As expected, many amino acids “contact” the ligand when using the 5 Å criterion. However, the maximum average number of consecutive amino acids does not exceed 8 and 5
It may even be lower than the Angstrom standard. For particularly large ligands,
The number fills 7 to 10, with larger binding sites most commonly creating more loops than larger loops of discontinuous amino acids. It is unusual for the ligand binding site to be made up of only a single stretch of amino acids. Thus, any of the identified stretches of contiguous amino acids involved in the ligand binding site will constitute only a partial binding site. As a result, the part of the binding site that constitutes the longest stretch of contiguous amino acids has a relatively weak affinity for the ligand in the absence of the remaining binding site residues. However, it was demonstrated in Example 1 that the ratio of specific binding to non-specific binding was strong enough to be detectable for the other peptides in the library.

When analyzing peptide sequences selected to utilize the claimed method,
Strong consensus sequences can emerge. In this case, the investigator preferably uses one of many public domain search engine devices (eg, GenBank's BLAST search).
Can be used to find sequences for proteins that contain consensus sequences or the like. Examples 1 (SEQ ID NO: 3) and 3 (SEQ ID NO :)
This technique was applied when analyzing the consensus sequence of 5).

However, in the absence of a clear consensus sequence, the selected peptide still contains information useful in identifying possible binding sites in the protein. Utilization of this information preferably involves computer programming to perform subsequent algorithms to identify selected peptides and proteins that are believed to be involved (possibly toxic or therapeutic) in the particular molecular activity. This is done by scoring the homology between the sequences. For example, taxol-selective peptides were investigated against all proteins known to be involved in apoptosis. Elsewhere in other proteins where it is known to be involved in apoptosis, even when removing the two peptides containing HTPHP (SEQ ID NO: 3) from the taxol-selected peptide set from Example 1. This algorithm can identify the flexible loop of Bcl-2 as the site of taxol binding because it has a higher homology with the more selected peptides.

The algorithm is a simple brute force comparison of a protein's sequence with a control set of all selected and random (non-selective) peptides from the same peptide library. Randomly selected peptide controls are needed because most libraries have preference for amino acids and triplets that produce false positives in the absence of control peptides. For example, the NEB12mer library used in Example 1 has a preference for the proline pair (PP). The proline pairs found in taxol-selected peptides are therefore not necessarily due to taxol selection, but rather due to their preference in the library. If, by some agreement, the ligand had a high affinity for the sequence PP, the selected peptide has a higher PP frequency than that of the random (control) peptide.

An important consideration when programming algorithms for comparison is the determination of means for comparing peptide and protein sequences. this is,
(I) the length of the segment where the comparison occurs (the number of amino acids considered at one time), (i
i) a means of calculating homology scores, which usually involves the selection of a matrix of amino acid similarities, and (iii) a lower homology of the peptide or peptide fragment considered to be unrelated to the protein sequence being compared. Requires selection.

With regard to (i), the comparison is preferably made on the peptide length associated with the binding of the ligand. Examination of the binding of peptides to small ligands appears to involve only 4-8 amino acids in binding. With less than 4 amino acids, too many false positives (
Noisy output), and using more than 8 amino acids produces too many false negatives. In most cases, 6 amino acids were used at a time, but 5 or 7 are preferred for other applications.

Regarding (ii), choosing the correct homology matrix or amino acid similarity matrix helps to score the similarity between two amino acid sequences (Dayhof
f, MO, etc.). Applicants have used several different homology matrices for these calculations and have found that they all show similar results, but produce results with different signal to noise ratios. Different matrices are preferred in different environments. A typical matrix is 20x20, one row and one column for each amino acid. Additional columns and columns may be added for special purposes or to enhance certain types of homology searches. The diagonal of the matrix marks the identities (eg when alanine is matched with alanine). Diagonals usually have different scores for different amino acids (e.g. tryptophan is very rarely replaced by anything else, so its slanted score may, for example, severely alter the structure or function of a protein. Higher than alanine, which can be frequently replaced without. Elements that deviate from the slope of the matrix show a similarity score. For example, alanine and serine have low positive scores, but tryptophan and serine have significant negative scores. Applicants used a homology matrix called Blossum 62 (Henikoff, S. et al., 1992) for their comparison.

With respect to (iii), the selected noise level was fully empirically determined using Bcl-2 as an example and other apoptosis-related proteins as controls. The absolute number depends on the similarity matrix used.

The algorithm is characterized by the following steps: (a) Selecting the first block of protein sequences (eg the first 6 amino acids). (B) The first block of peptidic sequence (eg, the first 6 of the first peptide).
Individual amino acids). (C) Calculate homology scores between protein blocks and peptide blocks using a similarity matrix. This is done by summing the matrix elements corresponding to the amino acid pairs made from the protein and the corresponding amino acids in the peptide. (D) Advance the comparison data set to the second location in the first peptide and repeat (c). (E) Repeat (b)-(d) for each peptide, proceeding through all possible positions in all peptides selected by the affinity screen. (F) Once this comparison is complete with all selected peptides, proceed to position 2 of the protein and repeat steps (b)-(e). (G) Repeat steps (b)-(f) for all positions of the protein while preserving the cumulative homology score for each position. (H) Perform steps (b)-(g) on a control (random) set of peptides selected from the library. (I) Subtract the homology score for the random peptide from the homology score for the selected peptide for each position in the protein. (J) Output the calculation result.

The maximum homology score obtained by this algorithm for any given protein is a measure of its potential involvement in the binding of related ligands. The distribution of homology scores in proteins provides information about where binding may occur. In the case of Bcl2, this is the SEQ ID NO: containing the peptide in the data set:
It was shown that the binding of taxol, even without the inclusion of 3, included the extended region of the flexible loop of the protein.

Discovery of human Bcl-2 as a possible molecular target for taxol, and μ
After confirmation of the binding of Bcl-2 to taxol with molar affinity, Applicants
Based on their binding to Bcl-2, a method was devised to determine the apoptotic activity of taxanes. The method involves simply measuring the dissociation curve of the taxane / Bcl-2 complex using a standard ELISA assay. Taxanes, which bind Bcl-2 more strongly than taxol, have more potency than taxol to induce apoptosis in cancer cells and may therefore produce better formulations.

[0050]   The following examples illustrate the principles and advantages of this invention.

Example 1 Identification of Taxol-Binding Peptides Taxol is an anti-cancer agent with proven efficacy against a wide range of malignancies,
Most clinically interesting are human and ovarian cancers (Rowinsky and Donehow
er, 1991). Its only known molecular target is tubulin, which induces it to polymerize, disrupts the dynamic instability of microtubules in the mitotic machinery, and pauses mitosis during the metaphase / late transition ( Jordan et al., 1993). Programmed cell death or apoptosis, many intracellular messengers and growth factors,
For example, increased levels of p53 (Roth, W. 1998), and phosphorylation and subsequent inactivation of the anti-apoptotic protein Bcl-2 (Haldar et al., 1995; 199).
A wide range of other responses to taxol have been reported, including 6). These activities are considered to be indirectly induced by the action of taxol on microtubules.

To identify possible binding sites for taxol on β-tubulin or other proteins, a random phage-displayed peptide library was searched for peptides with high affinity for taxol. It was inferred that peptides that have a high affinity for taxol when exposed to the surface of phage particles may also have a high affinity for taxol when present on the surface of cellular proteins.
Bacteriophage M13 (Ph.D.-12 library, New England Biol
(Abs, Beverly, Massachusetts)), a random dodecameric library displayed at the N-terminus of p3 was screened for members with high affinity for taxol. The taxol derivative biotinylated at the C ₇ position was synthesized as described in Example 2. The taxol derivative was immobilized on the surface of streptavidin coated plates and selected for members with high affinity for biotinylated taxol using standard techniques of biopanning, followed by modification. A 10-fold higher molar amount of coordination (ie, biotinylated taxol) than was normally used was added to streptavidin-coated culture dishes. Two separate screens yielded enhancement values of 1.7 × 10 ⁵ and 2.8 × 10 ⁵ fold, respectively (where enhancement = input phage / output phage). These taxol-binding phages were named Round I. Twenty individual phage particles were randomly selected from the Round I pool and sequenced. The rest of the Round I phage was non-competitively amplified so that taxol-binding peptides with low dissociation constants and low growth properties had a good chance of survival. Less than 90,000 plaque forming units (pfus) were fixed in agar on 150 mm culture plates. A solution of Tris buffered saline + 0.1% Tween 20 (nonionic detergent) at 4 ° C.
The viral particles were eluted from the agar by gently swirling for about 2 hours.
The virus suspension was purified and concentrated by performing two consecutive polyethylene glycol precipitations.

2 for amplification Round I clones with 40-fold amount of input phage (enhanced for low K ^D or “tight” bound phage particles versus non-specific bound phage particles) on culture dishes containing complex taxol. After the first screening, 5.1x1
0 ^four times resulted in a strengthening. 2.3x by amplification of these round II phage particles followed by a third round of screening with taxol (round III)
A 10 ⁴ times enhancement value was obtained. 70 members of Round II bacteriophage and 23 members of Round III bacteriophage were randomly selected and sequenced in the same manner.

For amino acid sequence characteristics, each group of peptides was analyzed as follows: The original live by generating a Poisson distribution from the incidence of individual amino acids at each of the 12 positions for the random insert. Characterization of the sequence characteristics of the rally (ie before screening with taxol) has been accomplished. The information content of each clone is
It was defined as the negative natural logarithm of its probability of occurrence within the library. This information content value is a convenient measure of probability, the greater the information content, the more sparse the related sequences. The amino acid sequences of peptides isolated by affinity selection have a relatively high content of related information as compared to the sequences from the parent library. The average information content is
33.1 ± 1.6 for non-selected phage (n = 101) to 35.4 ± 2.5 for round I phage (n = 20), round II phage (n = 70)
Increased to 35.4 ± 2.1 for Round III phage (n = 23) and 34.8 ± 2.4 for Round III phage (n = 23). These figures show that after round I, affinity selection resulted in no further enhancement for taxol-binding phages, and that round II
Later, the growth characteristics became a measurable factor in the composition of the phage population. This is a scenario assumed by another group (Folgori et al., EMBO J. 1994 May 1; 13 (9).
: 2236-2243).

Round I, II and III sequence analysis was performed to identify any of the bacteriophages sharing a minimum of 3 contiguous amino acids out of 12. All three rounds, as well as randomly selected non-screened phage harbored pairs with contiguous trimers in common. A control search with 102 randomly deselected sequences yielded 5 tetramers, 0 pentamers and 1 hexamer, the latter 30.2 each.
And had an information content of only 31.0. In addition, a synthetic sequence database OWL using a control hexamer (Bleasby et al. (1994) Nucleic Acids Research,
A DELPHOS sequence search within 22 (17), 3574-3577) derived only two entities, the hypothetical human cytomegalovirus protein (ULB7 HCMVA) and the hypothetical yeast protein (S61023). Of the putative taxol-binding sequences, only Round II sequences contained clone pairs sharing consensus sequences of 4 and 5 residues in length. Two tetramers, SEQ ID NO: 1 and SEQ ID NO:
2 was identified in the Round II clone. However, tetramers are not currently sufficiently specific to generate a manageable list of target proteins that are most often suspect. For example, a sequence search using SEQ ID NO: 1 yielded 873 matches, of which 152 were human sequences.

In addition to the two tetramers in Round II, two pairs of pentamers were identified. Clone T
AX2.6 and TAX12.29 share the pentapeptide SEQ ID NO: 3 at the same position within the insert peptide, clones TAX2.8 and TAX2.74 share SEQ ID NO: 4 at different positions along the insert sequence. . The localization within these four clones was determined by TAX2.6 and TAX2.29 at 35.9 and 36.
It has an information content of 3, while TAX2.8 and TAX2.74 each have 31
． It is shown to carry information content values of 6 and 31.2. Since the information values are expressed as logarithms, these numerical values are about
A factor of 0 indicates that the statistical appearance rate is low. After two rounds of biopanning, a single pentapeptide consensus sequence, SEQ ID NO: 3, was identified as the one that best corresponds to the taxol binding motif.

A search of the OWL database revealed that SEQ ID NO: 3 does not occur in any known tubulin and only two known human proteins, Bcl-2 (residues 55-5).
9) and ataxin-2 (SCA2).
Pulst, et al., Nat. Genet. 1996 Nov.; 14 (3): 269-276; Imbert et al., Na
t. Genet 1996 Nov; 14 (3): 285-291). In Bcl-2, it is obtained both central to the appearance (X-ray crystallography and NMR structure of 60 amino acids loop high flexibility which is identified by comparison with the Bcl-2 homologue, Bcl-X _L (Mu
chmore et al., 1996). This loop acts as a regulatory domain for both Bcl X _L and Bcl-2 (Chang et al. , 1997), not required for its anti-apoptotic activity (Muchmore et al., 1996) . Since Bcl-2 function is known to be regulated by taxol, it represents a very likely molecular target for taxol.

An ELISA binding assay was used to determine if the Bcl-2 / GST fusion protein bound directly to immobilized biotinylated taxol. Bcl-2 binding was detected using anti-Bcl-2 antibody. These results indicate that Bcl-2 binds to the biotinylated taxol derivative with a K _d of approximately 0.4 μM (see FIG. 2). The graph shows taxol (Δ), biotin (of Bcl-2 / GST fusion protein).
▲) and binding of biotinylated dioxin (■), as well as Bcl-X _L, indicating the ELISA data for binding to taxol (○) and biotin (●). The x-axis is the log nM concentration of protein in solution (Bcl-2 or Bcl
A -X _L), y-axis is the uncorrected optical density at 490 nm. The insert is the chemical structure of the biotinylated taxol used for peptide selection. In a similar experiment, B
cl-2 is a biotinylated taxotere with an affinity of approximately 10-50 times.
Was also shown to bind.

By circular dichroism spectroscopy, using the CD results demonstrating that the Bcl-2 / GST construct undergoes a substantial conformational change in the binding of taxol, Bcl-2
The binding of taxol to the Bcl-2 portion of the / GST fusion protein was confirmed. The spectrum of the fusion protein is significantly different with and without added taxol (see Figure 3). The graph shows circular dichroism spectra of human Bcl / GST fusion protein with and without taxol (solid line) and dotted line. Each spectrum is the average of the results of 5 spectra and the standard deviation was used to subtract the baseline as shown. These differences are much larger than the error bars (standard deviation) in the measured values, and the shape of the curve (220/2
10 nm peak ratio) and peak position changes are included. The secondary structure produced for the Bcl-2 fusion protein with and without taxol differed by about 4% and appears to involve changes in the region of the molecule found in the beta-rotational conformation. This corresponds to the type of secondary structure predicted for the HTPHP containing portion of the Bcl-2 loop region (SEQ ID NO: 3). The magnitude of the difference (with error level in mind) suggests that about 10-12 amino acids are involved in the change. TA
The homology between X2.29 and human Bcl-2 is between positions 1-12 and residues 55-66, respectively, meaning a minimal drug / protein interaction of at least 12 amino acids.

In a control experiment, taxol was added to purified GST from S. japonicum. Bcl
In contrast to the -2 / GST fusion construct, the GST spectra are almost identical with and without taxol. The difference is much smaller than the error bar, no change in 220/210 nm ratio is observed and no peak shift is observed.
This means that there is any net structural change within the GST molecule alone, which comprises less than 0.5% of its overall secondary structure. These results indicate that taxol bound to the Bcl-2 / GST fusion protein resulted in Bcl-2
It reveals that the observed spectral difference for the / GST fusion protein must be obtained from the Bcl-2 portion of the molecule. Furthermore, this bond is B
It has significant binding structural consequences for the cl-2 protein and includes approximately 15% -20% loop residues.

Taxol or taxotere was shown to induce phosphorylation of Bcl-2 and apoptosis of 697 human pre-acute leukemia B cells (Haldar et al., 199).
5; 1997). The concentrations of these two agents required to induce apoptosis in these cells are similar to those at which taxol and taxotere initiate in vitro interactions with Bcl-2. This suggests that it may be a direct interaction of these agents with Bcl-2 that causes the observed phosphorylation and subsequent apoptosis. Furthermore, human Bcl-2 differs significantly from the murine version of the protein in the region of the SEQ ID NO: 3 motif. This difference may explain the poor response of mouse tumors to taxol treatment. Currently, a mouse model with human tissue tumors has to be used in the taxol test.

The technique used here identifies Bcl-2 as a possible target for taxol and taxotere. The method used is quite simple. It is possible that the members of the p3 library used in this study and containing the consensus SEQ ID NO: 3 mimic the taxol binding site on Bcl-2 with the same amino acids present on Bcl-2. .

Example 2 Preparation of Biotinylated Taxol The taxol derivative 7-biotinamide carbamate 6a-c was synthesized as shown in Scheme 1. The reaction of TEXCI with taxol in pyridine at 0 ° C is 2
'-TES-Taxol 2 was produced in a yield of 84%. Compound 2 is CH ₂ 2 at room temperature
Reacts with the carbonyl Jimi indazole in Cl _2, and produced 2'-TES-7- imidazo Ride taxol 3 in 94% yield. 3 was treated with the amine H ₂ N (CH ₂ ) _n NH ₂ (50 equivalents, n = 4,6,12) in excess in CH ₂ Cl ₂ at room temperature to give the amide carbonate 4 (a, n = 4; b, n = 12) was obtained. 4 without further purification
ac at room temperature in DMF biotin N-hydroxyl succinimide ester 7
Treatment with 2'-TES'7-biotinamide-carbamate 5a-c. 5a using HF / Py / MeC patterned stitch (1:10:10 v / v / v) at room temperature
After removal of the 2'-TES group from -c and chromatography, the desired 7-
Biotinamide carbamate 6a-c was obtained.

Scheme 1

Scheme 1 (continued)

Scheme 1 (continued)

Scheme 1 (continued)

Scheme 1 (continued)

2′-TES-Taxol 2.0 To a solution of taxol (247 mg, 0.289 mmol) in pyridine (2.9 mL) at 0 ° C., pyridine (0.87 mmol).
0.87 mL of a 1.0 M solution of TESCI in was added. This mixture is stirred at 0 ° C for 7
Stir for hours then room temperature for 8 hours and dilute with EtOAc. The resulting mixture of saturated aqueous NaHCO _3, water, 10% CuSO _4, was washed with water and brine (brine), dried over Na ₂ SO _4. The solvent was removed in vacuo to give a pale yellow oil (345 m
g) was obtained. After removing the solvent, chromatography (3: 2 v / v hexane / Et)
OAc) treatment gave 2 as a white solid (236 mg, 84%).

2'-TES-7-imidazolide taxol 3. CH ₂ Cl ₂ (10mL) solution of 2'-TES-taxol (2.251mg, 0.259mmol) to a solution of was added carbonyl diimidazole (422 mg, 2.60 mmol). The mixture was stirred at room temperature for 19 hours, diluted with EtOAc (50 mL), saturated Na
It was washed with aqueous HCO ₃ solution, water, brine and dried over Na ₂ SO ₄ . The solvent was removed in vacuo to give 3 as a white solid (249mg, 0.234mmol, 9
1%).

Aminocarbonate 4a. 3 (148 mg, 0 in CH ₂ Cl ₂ (10 mL)
． 139 mmol) in H ₂ N (CH ₂ ) ₄ NH ₂ (663 mg, 7.52 m)
mol) was added. The mixture was stirred at room temperature for 16 hours, EtOAc (50 mL)
Diluted with and washed with saturated aqueous NaHCO ₃ , water and brine, dried over Na ₂ SO ₄ . Removal of solvent in vacuo gave 4a as a white solid (141 mg,
0.130 mmol, 94%).

[0067]   Aminocarbonate 4b. CH₂Cl₂3 (101 mg, 0 in 10 mL)
． 095 mmol) in H 2₂N (CH₂)₆NH₂(558 mg, 4.80
mmol) was added. The mixture was stirred at room temperature for 19 hours and washed with EtOAc (50
mL) and saturated aqueous NaHCO.₃Washed with water, brine, Na₂SO _Four Dried on. Removal of solvent in vacuo gave 4b as a white solid (99m
g, 0.089 mmol, 94%).

Aminocarbonate 4c. Compound 4c was similarly treated with 3 and H ₂ N (CH ₂ ) ₁₂
Synthesized in good yield from NH ₂ .

2'-TES-7-Biotinamide carbamate 5a. To a solution of 4a (141 mg, 0.130 mmol) in DMF (3 mL) was added biotin N-hydroxyl succinimide ester (7.50 mg, 0.15 mmol).
The mixture was stirred at room temperature for 11 hours, diluted with EtOAc (50 mL), washed with saturated aqueous NaHCO ₃ , water and brine and dried over Na ₂ SO ₄ . The solvent was removed in vacuo and chromatographed to give a white solid (132m).
g, 0.101 mmol, 78%).

2'-TES-7-Biotinamide carbamate 5b. To a solution of 4a (99 mg, 0.089 mmol) in DMF (3 mL) was added biotin N-hydroxyl succinimide ester (7.36 mg, 0.11 mmol). The mixture was stirred at room temperature for 11 hours, diluted with EtOAc (50 mL), washed with saturated aqueous NaHCO ₃ , water and brine and dried over Na ₂ SO ₄ . After removing the solvent in vacuo, 5b was obtained as a white solid (117 mg, 0.088 mmol,
99%).

2'-TES-7-Biotinamide carbamate 5c. A solution of 5c was similarly synthesized from 4c and 7 in good yield.

7-Biotinamide carbamate 6a. HF / Pyridine / MeCN (1: 1
0:10 v / v / v) A solution of 5a (132 mg, 0.101 mmol) in 2 mL was stirred at room temperature for 30 minutes, diluted with EtOAc (50 mL) and saturated aqueous NaH.
CO _3, water, CuSO _4, was washed with water and brine, dried over Na ₂ SO _4.
Chromatography (1: 5 v / v MeOH / CHCl ₃ ) gave 6a as a white solid (112 mg, 0.094 mmol, 93%). C ₆₂ H ₇₅ N ₅
Analysis for O ₁₇ S. (1/2) CHCl ₃ . Theoretical C: 59.90; H: 6.
07. Found C: 60.42; H: 6.06.

7-Biotinamide carbamate 6b. HF / Pyridine / MeCN (1: 1
0:10 v / v / v) A solution of 5b (117 mg, 0.088 mmol) in 2 mL was stirred at room temperature for 30 minutes, diluted with EtOAc (50 mL) and saturated aqueous NaH.
CO _3, water, CuSO _4, was washed with water and brine, dried over Na ₂ SO _4.
Chromatography (1: 5 v / v MeOH / CHCl ₃ ) gave 6b as a white solid (97 mg, 0.079 mmol, 90%). Analysis of _{C 64 H 79 N 5 O 1} 7 S · 3H 2. Theoretical C: 60.22; H: 6.66. Measured value C: 6
0.47; H: 6.68.

7-Biotinamide carbamate 6c. 5c to 6 with similar reaction and operation
Obtained c as a white solid (110 mg). C ₇₀ H ₉₁ N ₅ O ₁₇ S ・ (1/2) SH
Analysis for Cl ₃ . Theoretical C: 61.98; H: 6.75. Measured value C: 61.
77; H: 6.98.

Example 3 Identification of Dioxin-Binding Peptides Dioxins are pervasive environmental toxins that have toxic effects at very low doses. For the most part, toxicologists believe that the action is due to the interaction of dioxins with the single hydrocarbon target, the aryl hydrocarbon receptor (AhR) (
Birnbaum, 1994). However, the action of dioxins is so pleiotropic that the possibility of other molecular targets cannot be ruled out. Valery Petrenko
Dr. U. Missouri, Columbia screened a random peptide library for affinity for biotinylated dioxin (Petrenko et al.
., 1996). This screening of a random peptide library fused to the p8 coat protein of M13 phage resulted in the single tetrapeptide consensus sequence EPFP (SEQ ID NO: 5). The p8 library is significantly different from the p3 library used by Applicants in Example 1 in that there are ~ 3000 copies of p8 on the virion but only 5 p3 proteins. Therefore, the strength of the interaction becomes more important, and weaker interactions can result in target-phage binding. Another weakness of the p8 library is the fact that the large number of inserts per virion (virus particle) results in observable effects of insert synthesis on host metabolism (Rodi et al. (1997) Proceedings of the 22 ^nd Tanaguchi
International Symposium, (Nov.18-21, 1996)). Because of this, there is considerable censership of the library due to the wide selection of select factors, ranging from protein synthesis to virion assemblies. Certain sequences, especially proline (
Sequences containing P) and phenylalanine (F) have been shown to be highly preferred in p8 libraries (Cesareni, et al., 1997). Therefore, it is difficult to know whether the putative dioxin-binding motif observed by Petrenko is a dioxin-binding motif, or simply the result of the need for growth of his library. These data demonstrate the need to consider growth and selection factors when selecting peptide libraries. Nevertheless, Petrenko showed that the EPFP (SEQ ID NO: 5) motif did indeed bind dioxin when displayed on p8.

Applicants searched the peptide databank for a protein containing SEQ ID NO: 5. The details of random peptides in the library are relatively small (8
Amino acids), and the resulting consensus was only 4 amino acids in length, so there were many human and mammalian proteins in which it was found. After searching the literature, one of them, a leukemia inhibitory factor, was overexpressed in mice, similar to that observed with very high doses of dioxin (Hilton et al., 1991). The applicant has determined that it will cause some toxic effects. By this correlation, applicants determined that the affinity of dioxin for LIF was the first of the proteins containing the putative dioxin binding motif tested for binding to biotinylated dioxin.

The fact that biotinylated dioxin actually binds to LIF was confirmed by ELIS.
Experiment A showed. Applicants have determined that the EPFP homology domain with LIF (residues 50-5
3) with several mutants of the LIF protein altered, with anti-dioxin antibody as a positive control and as a negative control ciliary neurotrophic factor (CNTF) (structurally similar to LIF The same biotinylated dioxins as Petrenko were used in an ELISA assay for binding with (but lacking the SEQ ID NO: 5 motif). The results of these experiments are shown in Figure 4. LIF binds to human and murine LIF with a K _D of about 10-100 μM, which shows an affinity of about 10,000-fold less than the anti-dioxin antibody used as a positive control. The negative control CNTF showed no binding to LIF. When tested at several mutation points in LIF, binding was reduced to undetectable levels in the Pro-53-Ala mutant, with the Phe-156-Ala mutant at about 3 minutes of the dioxin binding motif. 1 shows reduced binding (in the three-dimensional structure of LIL, Phe15
6 was flanked by EPFP (SEQ ID NO: 5)), and the Pro-51-Ala mutant was found to have no altered binding. These results indicate that (i) dioxins bind LIF with detectable affinity and greater affinity than other proteins of similar sequence and structure but lacking the EPFP motif, and (ii) alteration of the EPFP motif. Results in a reduction in the observed dioxin levels, and (iii) other residues in the immediate vicinity of the EPFP motif (
That is, it shows that Phe156) is also involved in the dioxin bond.

LI of the putative binding site for dioxin and the LIF receptor (LIFR)
Correlation with residues known to be involved in F binding is that dioxin
We show direct binding to regulatory regions that regulate the F-LIFR interaction (Robinson
et al., 1994). Enhancement of the LIF-LIFR interaction can cause debilitating and possibly toxic effects associated with dioxin exposure and LIF overexpression such as endometriosis.

Petrenko (Petrenko et al., 1996) used the peptide consensus sequence EPFP (
glu-pro-phe-pro) demonstrated an affinity for biotinylated dioxins. However, it did not yield any further results. Applicants have followed this result by screening GenBank for EPFP-containing proteins and found that the aryl hydrocarbon receptor (AhR), which is known to bind dioxin, has a similar sequence. Found to have. Applicants have also shown that the sequence EPFP allows the identification of leukemia inhibitors (LIFs) as possible targets for dioxin, and that dioxin actually binds LIF with at least micromolar affinity. Have demonstrated.

Applicants have repeated Petrenko binding experiments in the p3M13 phage coat protein expression system using a random 12 amino acid peptide library (Ph.D-12). In the p3 system, about 3 to 5 peptides are displayed per phage for Example 1. Peptide motifs have different conformational skeletons (12 to 8 amino acids)
Applicants did not necessarily obtain the same consensus data as Petrenko, as they were present under very non-competitive conditions with very high affinity stringency (3 to 3000 copies of peptide per phage). I wasn't expecting it. In fact, Applicants did not see the same EPFP (SEQ ID NO: 5) consensus sequence appearance.

Rather, Applicants did not recognize a particular statistical tendency for preference for glutamate and for serine residues among peptides showing improved dioxin binding. The results are shown in the table below. Column A contains data on amino acid frequencies from 101 peptides selected from the peptide library. Column B shows frequency from 69 peptides selected for binding to taxol (2 times selection), column C shows frequency from 89 peptides selected for binding to dioxin (2 and 3 times selection). Selection data is merged). The frequencies for all these amino acids are very similar, reflecting the fact that many (70-80%) of the peptides isolated after selection for binding are, in fact, due to non-specific binding.

[0082]

[Table 1]

From these data, it was found that significant additional information was gathered about the selection of certain peptides for dioxin, and the data were as statistically significant as those obtained for taxol. Conceivable. However, this example may provide different information that the use of different peptide libraries and gene packages alters the utility for a particular study, and several different libraries and gene packages are preferably present in the method. It is also noted that one will try to get the best results for the particular application.

In view of the detailed description of the invention and the examples provided above, it is believed that some of the objects of the invention are achieved. The description and illustrations presented herein are intended to acquaint others skilled in the art with the invention, its principles, and its practical application. One of ordinary skill in the art can adapt and apply the invention in many forms to best suit a particular use requirement. Therefore, the particular embodiments of the present invention described above are not an exhaustive examination or limitation of the present invention.

Sequence Identification SEQ ID NO: 1 lists the sequences of the tetrameric consensus peptide sequences identified from the second round of affinity selection enhancement in Example 1.

SEQ ID NO: 2 lists the sequence of the tetrameric consensus peptide sequence identified from the second round of affinity selection enhancement in Example 1.

SEQ ID NO: 3 lists the sequence of the pentameric consensus peptide sequence identified from the second round of affinity selection enhancement in Example 1.

SEQ ID NO: 4 lists the sequence of the pentameric consensus peptide sequence identified from the second round of affinity selection enhancement in Example 1.

[0089]

【Reference diagram】

【Reference diagram】

【Reference diagram】

[Brief description of drawings]

FIG. 1 is a “contact” of maximum separation of 4 Å and 5 Å.
The criteria are used to indicate the number of amino acids that make contact with the binding ligand for the selected protein when the analysis of their three-dimensional structure is confirmed.

Figure 2 shows ELISA data showing binding of taxol and various control Bcl-2 / GST and Bcl-X _L. Bcl-X _L lacks homology to SEQ ID NO: 3 in the putative binding site.

FIG. 3 shows circular dichroism spectra of Bcl-2 / GST with and without taxol. This spectrum shows that the Bcl-2 / GST fusion protein undergoes a substantial conformational change in the presence of taxol, unlike GST alone.

[Figure 4]   FIG. 4 shows the results of the dioxin binding ELISA experiment of Example 3.

[Sequence list]

─────────────────────────────────────────────────── ─── Continued front page    (81) Designated countries EP (AT, BE, CH, CY, DE, DK, ES, FI, FR, GB, GR, IE, I T, LU, MC, NL, PT, SE), CN, JP, K R (72) Inventor Sangany, High Tech Jay.             32306, Florida, United States             Lahassee West Pensacolas             Treat 1701 F-term (reference) 4B024 AA01 AA11 AA20 CA04 DA05                       DA12 DA20 HA11                 4B063 QA13 QA18 QQ79 QR41 QR51                       QS02 QS17 [Continued summary] It When using a random peptide library, the same Of identified consensus sequences within defined proteins The location will identify the possible ligand binding sites on the target. It

Claims (15)

[Claims] 1. A method for identifying a protein having a molecular weight of less than 5,000 Da and binding to a ligand other than a nucleic acid, a peptide or a protein, each peptide or protein of a peptide or protein library. Screening said ligand against a peptide or protein library comprising a set of gene packages physically linking a member of said to a nucleic acid polymer encoding said member, and having an affinity for said ligand. Separating the members of the library in the library, the affinity of which is greater than the affinities of other members of the library for the ligand, and a nucleotide sequence encoding the member separated from the library. And base these sequences. Translating into a peptide sequence and comprising a portion of said translated peptide sequence or corresponding to a consensus peptide sequence obtained from a statistical analysis of the peptide sequences of said translated library members, a singular or Identifying a plurality of proteins. 2. The method of claim 1, wherein the ligand is an organic molecule conjugated to a plating agent. 3. The method of claim 2, wherein the peptide or protein member of the library is selected from the group consisting of expression products of cDNA libraries obtained from cells and fragments of the expression products. 4. The method of claim 2, wherein the plating agent is biotin. 5. The method of claim 3, wherein the ligand is biotinylated dioxin.
The method described. 6. The method of claim 1, wherein the peptide or protein members of said library are selected from the group consisting of expression products of cDNA libraries obtained from cells and fragments of said expression products. 7. The method of claim 6, wherein the peptide or protein members of said library are selected by a method selected from the group consisting of random selection and engineered random selection. 8. The method of claim 6, wherein the peptide or protein members of said library are selected by a method selected from the group consisting of random selection and engineered random selection. 9. The method of claim 1, wherein the peptide or protein members of said library are selected by a method selected from the group consisting of random selection and engineered random selection. 10. The method of claim 1, wherein the library of peptides or proteins comprises, on the surface of a gene package, a gene package representing individual peptide or protein members of the library. 11. The method of claim 10, wherein the peptide or protein member of the library is genetically fused to a protein that integrates to become the surface of the gene package. 12. The gene package comprises a phage, a virus, a yeast,
11. The method of claim 10, selected from the group consisting of and bacteria. 13. A member of the library having a higher affinity for the ligand is labeled with the ligand by a method of discrimination selected from the group consisting of biopanning, chromatography, Western or Southern blotting and electrophoresis. The method of claim 1, wherein the method is separated from members having a lower affinity. 14. A member of said library having a higher affinity for said ligand is biopanned with amplification of a gene package having a member of said library having a higher affinity for said ligand. The method of claim 1, wherein a member having a lower affinity for the ligand is separated by a distinguishing method selected from the group consisting of chromatography, Western or Southern blotting and electrophoresis. 15. The method of claim 14, wherein said isolating is performed by alternating said method of discrimination and said amplification of said gene package about 2-3 times.