US 20050255548 A1
Disclosed herein are proteins that include a fibronectin type III domain having at least one randomized loop. Also disclosed herein are nucleic acids encoding such proteins and the use of such proteins in diagnostic methods and in methods for evolving novel compound-binding species and their ligands.
68. A protein comprising:
(a) a first fibronectin type III (Fn3) domain wherein said first Fn3 domain has an amino acid sequence that has been altered relative to the amino acid sequence of a first naturally occurring Fn3 domain such that the first Fn3 domain binds to a first compound that is not bound by the first naturally occurring Fn3 domain; and
(b) a second Fn3 domain, wherein said second Fn3 domain has an amino acid sequence that has been altered relative to the amino acid sequence of a second naturally occurring Fn3 domain such that the second Fn3 domain binds to a compound that is not bound by the second naturally occurring Fn3 domain.
69. The protein of
70. The protein of
71. The protein of
72. The protein of
73. The protein of
74. The protein of
75. The protein of
76. The protein of
77. The protein of
78. The protein of
79. The protein of
80. The protein of
81. The protein of
82. The protein of
83. The protein of
84. The protein of
85. The protein of
86. The protein of
87. The protein of
88. The protein of
(a) a first Fn3 domain wherein said first Fn3 domain has an amino acid sequence that has been altered relative to the amino acid sequence of the eighth Fn3 module of human fibronectin (8Fn3) such that the first Fn3 domain binds to a first compound that is not bound by human 8Fn3;
(b) a second Fn3 domain wherein said first Fn3 domain has an amino acid sequence that has been altered relative to the amino acid sequence of the ninth Fn3 module of human fibronectin (9Fn3) such that the second Fn3 domain binds to a second compound that is not bound by human 9Fn3; and
(c) a third Fn3 domain wherein said third Fn3 domain has an amino acid sequence that has been altered relative to the amino acid sequence of the tenth Fn3 module of human fibronectin (10Fn3) such that the third Fn3 domain binds to a first compound that is not bound by human 10Fn3.
This application claims the benefit of the filing date of provisional application, U.S. Ser. No. 60/111,737, filed Dec. 10, 1998, and utility application, U.S. Ser. No. 09/456,693, filed Dec. 9, 1999.
This invention relates to protein scaffolds useful, for example, for the generation of products having novel binding characteristics.
Proteins having relatively defined three-dimensional structures, commonly referred to as protein scaffolds, may be used as reagents for the design of engineered products. These scaffolds typically contain one or more regions which are amenable to specific or random sequence variation, and such sequence randomization is often carried out to produce libraries of proteins from which desired products may be selected. One particular area in which such scaffolds are useful is the field of antibody design.
A number of previous approaches to the manipulation of the mammalian immune system to obtain reagents or drugs have been attempted. These have included injecting animals with antigens of interest to obtain mixtures of polyclonal antibodies reactive against specific antigens, production of monoclonal antibodies in hybridoma cell culture (Koehler and Milstein, Nature 256:495, 1975), modification of existing monoclonal antibodies to obtain new or optimized recognition properties, creation of novel antibody fragments with desirable binding characteristics, and randomization of single chain antibodies (created by connecting the variable regions of the heavy and light chains of antibody molecules with a flexible peptide linker) followed by selection for antigen binding by phage display (Clackson et al., Nature 352:624, 1991).
In addition, several non-immunoglobulin protein scaffolds have been proposed for obtaining proteins with novel binding properties. For example, a “minibody” scaffold, which is related to the immunoglobulin fold, has been designed by deleting three beta strands from a heavy chain variable domain of a monoclonal antibody (Tramontano et al., J. Mol. Recognit. 7:9, 1994). This protein includes 61 residues and can be used to present two hypervariable loops. These two loops have been randomized and products selected for antigen binding, but thus far the framework appears to have somewhat limited utility due to solubility problems. Another framework used to display loops has been tendamistat, a 74 residue, six-strand beta sheet sandwich held together by two disulfide bonds (McConnell and Hoess, J. Mol. Biol. 250:460, 1995). This scaffold includes three loops, but, to date, only two of these loops have been examined for randomization potential.
Other proteins have been tested as frameworks and have been used to display randomized residues on alpha helical surfaces (Nord et al., Nat. Biotechnol. 15:772, 1997; Nord et al., Protein Eng. 8:601, 1995), loops between alpha helices in alpha helix bundles (Ku and Schultz, Proc. Natl. Acad. Sci. USA 92:6552, 1995), and loops constrained by disulfide bridges, such as those of the small protease inhibitors (Markland et al., Biochemistry 35:8045, 1996; Markland et al., Biochemistry 35:8058, 1996; Rottgen and Collins, Gene 164:243, 1995; Wang et al., J. Biol. Chem. 270:12250, 1995).
The present invention provides a new family of proteins capable of evolving to bind any compound of interest. These proteins, which make use of a fibronectin or fibronectin-like scaffold, function in a manner characteristic of natural or engineered antibodies (that is, polyclonal, monoclonal, or single-chain antibodies) and, in addition, possess structural advantages. Specifically, the structure of these antibody mimics has been designed for optimal folding, stability, and solubility, even under conditions which normally lead to the loss of structure and function in antibodies.
These antibody mimics may be utilized for the purpose of designing proteins which are capable of binding to virtually any compound (for example, any protein) of interest. In particular, the fibronectin-based molecules described herein may be used as scaffolds which are subjected to directed evolution designed to randomize one or more of the three fibronectin loops which are analogous to the complementarity-determining regions (CDRs) of an antibody variable region. Such a directed evolution approach results in the production of antibody-like molecules with high affinities for antigens of interest. In addition, the scaffolds described herein may be used to display defined exposed loops (for example, loops previously randomized and selected on the basis of antigen binding) in order to direct the evolution of molecules that bind to such introduced loops. A selection of this type may be carried out to identify recognition molecules for any individual CDR-like loop or, alternatively, for the recognition of two or all three CDR-like loops combined into a non-linear epitope.
Accordingly, the present invention features a protein that includes a fibronectin type III domain having at least one randomized loop, the protein being characterized by its ability to bind to a compound that is not bound by the corresponding naturally-occurring fibronectin.
In preferred embodiments, the fibronectin type III domain is a mammalian (for example, a human) fibronectin type III domain; and the protein includes the tenth module of the fibronectin type III (10Fn3) domain. In such proteins, compound binding is preferably mediated by either one, two, or three 10Fn3 loops. In other preferred embodiments, the second loop of 10Fn3 may be extended in length relative to the naturally-occurring module, or the 10Fn3 may lack an integrin-binding motif. In these molecules, the integrin-binding motif may be replaced by an amino acid sequence in which a basic amino acid-neutral amino acid-acidic amino acid sequence (in the N-terminal to C-terminal direction) replaces the integrin-binding motif; one preferred sequence is serine-glycine-glutamate. In another preferred embodiment, the fibronectin type III domain-containing proteins of the invention lack disulfide bonds.
Any of the fibronectin type III domain-containing proteins described herein may be formulated as part of a fusion protein (for example, a fusion protein which further includes an immunoglobulin Fc domain, a complement protein, a toxin protein, or an albumin protein). In addition, any of the fibronectin type III domain proteins may be covalently bound to a nucleic acid (for example, an RNA), and the nucleic acid may encode the protein. Moreover, the protein may be a multimer, or, particularly if it lacks an integrin-binding motif, it may be formulated in a physiologically-acceptable carrier.
The present invention also features proteins that include a fibronectin type III domain having at least one mutation in a β-sheet sequence which changes the scaffold structure. Again, these proteins are characterized by their ability to bind to compounds that are not bound by the corresponding naturally-occurring fibronectin.
In addition, any of the fibronectin scaffolds of the invention may be immobilized on a solid support (for example, a bead or chip), and these scaffolds may be arranged in any configuration on the solid support, including an array.
In a related aspect, the invention further features nucleic acids encoding any of the proteins of the invention. In preferred embodiments, the nucleic acid is DNA or RNA.
In another related aspect, the invention also features a method for generating a protein which includes a fibronectin type III domain and which is pharmaceutically acceptable to a mammal, involving removing the integrin-binding domain of said fibronectin type III domain. This method may be applied to any of the fibronectin type III domain-containing proteins described above and is particularly useful for generating proteins for human therapeutic applications. The invention also features such fibronectin type III domain-containing proteins which lack integrin-binding domains.
In yet other related aspects, the invention features screening methods which may be used to obtain or evolve randomized fibronectin type III proteins capable of binding to compounds of interest, or to obtain or evolve compounds (for example, proteins) capable of binding to a particular protein containing a randomized fibronectin type III motif. In addition, the invention features screening procedures which combine these two methods, in any order, to obtain either compounds or proteins of interest.
In particular, the first screening method, useful for the isolation or identification of randomized proteins of interest, involves: (a) contacting the compound with a candidate protein, the candidate protein including a fibronectin type III domain having at least one randomized loop, the contacting being carried out under conditions that allow compound-protein complex formation; and (b) obtaining, from the complex, the protein which binds to the compound.
The second screening method, for isolating or identifying a compound which binds to a protein having a randomized fibronectin type III domain, involves: (a) contacting the protein with a candidate compound, the contacting being carried out under conditions that allow compound-protein complex formation; and (b) obtaining, from the complex, the compound which binds to the protein.
In preferred embodiments, the methods further involve either randomizing at least one loop of the fibronectin type III domain of the protein obtained in step (b) and repeating steps (a) and (b) using the further randomized protein, or modifying the compound obtained in step (b) and repeating steps (a) and (b) using the further modified compound. In addition, the compound is preferably a protein, and the fibronectin type III domain is preferably a mammalian (for example, a human) fibronectin type III domain. In other preferred embodiments, the protein includes the tenth module of the fibronectin type III domain (10Fn3), and binding is mediated by one, two, or three 10Fn3 loops. In addition, the second loop of 10Fn3 may be extended in length relative to the naturally-occurring module, or 10Fn3 may lack an integrin-binding motif. Again, as described above, the integrin-binding motif may be replaced by an amino acid sequence in which a basic amino acid-neutral amino acid-acidic amino acid sequence (in the N-terminal to C-terminal direction) replaces the integrin-binding motif; one preferred sequence is serine-glycine-glutamate.
The selection methods described herein may be carried out using any fibronectin type III domain-containing protein. For example, the fibronectin type III domain-containing protein may lack disulfide bonds, or may be formulated as part of a fusion protein (for example, a fusion protein which further includes an immunoglobulin Fc domain, a complement protein, a toxin protein, or an albumin protein). In addition, selections may be carried out using the fibronectin type III domain proteins covalently bound to nucleic acids (for example, RNAs or any nucleic acid which encodes the protein). Moreover, the selections may be carried out using fibronectin domain-containing protein multimers.
Preferably, the selections involve the immobilization of the binding target on a solid support. Preferred solid supports include columns (for example, affinity columns, such as agarose columns) or microchips.
In addition, the invention features diagnostic methods which employ the fibronectin scaffold proteins of the invention. Such diagnostic methods may be carried out on a sample (for example, a biological sample) to detect one analyte or to simultaneously detect many different analytes in the sample. The method may employ any of the scaffold molecules described herein. Preferably, the method involves (a) contacting the sample with a protein which binds to the compound analyte and which includes a fibronectin type III domain having at least one randomized loop, the contacting being carried out under conditions that allow compound-protein complex formation; and (b) detecting the complex, and therefore the compound in the sample.
In preferred embodiments, the protein is immobilized on a solid support (for example, a chip or bead) and may be immobilized as part of an array. The protein may be covalently bound to a nucleic acid, preferably, a nucleic acid, such as RNA, that encodes the protein. In addition, the compound is often a protein, but may also be any other analyte in a sample. Detection may be accomplished by any standard technique including, without limitation, radiography, fluorescence detection, mass spectroscopy, or surface plasmon resonance.
As used herein, by “fibronectin type III domain” is meant a domain having 7 or 8 beta strands which are distributed between two beta sheets, which themselves pack against each other to form the core of the protein, and further containing loops which connect the beta strands to each other and are solvent exposed. There are at least three such loops at each edge of the beta sheet sandwich, where the edge is the boundary of the protein perpendicular to the direction of the beta strands. Preferably, a fibronectin type III domain includes a sequence which exhibits at least 30% amino acid identity, and preferably at least 50% amino acid identity, to the sequence encoding the structure of the 10Fn3 domain referred to as “1ttg” (ID=“1ttg” (one ttg)) available from the Protein Data Base. Sequence identity referred to in this definition is determined by the Homology program, available from Molecular Simulation (San Diego, Calif.). The invention further includes polymers of 10Fn3-related molecules, which are an extension of the use of the monomer structure, whether or not the subunits of the polyprotein are identical or different in sequence.
By “naturally occurring fibronectin” is meant any fibronectin protein that is encoded by a living organism.
By “randomized” is meant including one or more amino acid alterations relative to a template sequence.
By a “protein” is meant any sequence of two or more amino acids, regardless of length, post-translation modification, or function. “Protein” and “peptide” are used interchangeably herein.
By “RNA” is meant a sequence of two or more covalently bonded, naturally occurring or modified ribonucleotides. One example of a modified RNA included within this term is phosphorothioate RNA.
By “DNA” is meant a sequence of two or more covalently bonded, naturally occurring or modified deoxyribonucleotides.
By a “nucleic acid” is meant any two or more covalently bonded nucleotides or nucleotide analogs or derivatives. As used herein, this term includes, without limitation, DNA, RNA, and PNA.
By “pharmaceutically acceptable” is meant a compound or protein that may be administered to an animal (for example, a mammal) without significant adverse medical consequences.
By “physiologically acceptable carrier” is meant a carrier which does not have a significant detrimental impact on the treated host and which retains the therapeutic properties of the compound with which it is administered. One exemplary physiologically acceptable carrier is physiological saline. Other physiologically acceptable carriers and their formulations are known to one skilled in the art and are described, for example, in Remington's Pharmaceutical Sciences, (18th edition), ed. A. Gennaro, 1990, Mack Publishing Company, Easton, Pa., incorporated herein by reference.
By “selecting” is meant substantially partitioning a molecule from other molecules in a population. As used herein, a “selecting” step provides at least a 2-fold, preferably, a 30-fold, more preferably, a 100-fold, and, most preferably, a 1000-fold enrichment of a desired molecule relative to undesired molecules in a population following the selection step. A selection step may be repeated any number of times, and different types of selection steps may be combined in a given approach.
By “binding partner,” as used herein, is meant any molecule which has a specific, covalent or non-covalent affinity for a portion of a desired compound (for example, protein) of interest. Examples of binding partners include, without limitation, members of antigen/antibody pairs, protein/inhibitor pairs, receptor/ligand pairs (for example cell surface receptor/ligand pairs, such as hormone receptor/peptide hormone pairs), enzyme/substrate pairs (for example, kinase/substrate pairs), lectin/carbohydrate pairs, oligomeric or heterooligomeric protein aggregates, DNA binding protein/DNA binding site pairs, RNA/protein pairs, and nucleic acid duplexes, heteroduplexes, or ligated strands, as well as any molecule which is capable of forming one or more covalent or non-covalent bonds (for example, disulfide bonds) with any portion of another molecule (for example, a compound or protein).
By a “solid support” is meant, without limitation, any column (or column material), bead, test tube, microtiter dish, solid particle (for example, agarose or sepharose), microchip (for example, silicon, silicon-glass, or gold chip), or membrane (for example, the membrane of a liposome or vesicle) to which a fibronectin scaffold or an affinity complex may be bound, either directly or indirectly (for example, through other binding partner intermediates such as other antibodies or Protein A), or in which a fibronectin scaffold or an affinity complex may be embedded (for example, through a receptor or channel).
The present invention provides a number of advantages. For example, as described in more detail below, the present antibody mimics exhibit improved biophysical properties, such as stability under reducing conditions and solubility at high concentrations. In addition, these molecules may be readily expressed and folded in prokaryotic systems, such as E. coli, in eukaryotic systems, such as yeast, and in in vitro translation systems, such as the rabbit reticulocyte lysate system. Moreover, these molecules are extremely amenable to affinity maturation techniques involving multiple cycles of selection, including in vitro selection using RNA-protein fusion technology (Roberts and Szostak, Proc. Natl. Acad. Sci USA 94:12297, 1997; Szostak et al., U.S. Ser. No. 09/007,005 and U.S. Ser. No. 09/247,190; Szostak et al. WO98/31700), phage display (see, for example, Smith and Petrenko, Chem. Rev. 97:317, 1997), and yeast display systems (see, for example, Boder and Wittrup, Nature Biotech. 15:553, 1997).
Other features and advantages of the present invention will be apparent from the following detailed description thereof, and from the claims.
The novel antibody mimics described herein have been designed to be superior both to antibody-derived fragments and to non-antibody frameworks, for example, those frameworks described above.
The major advantage of these antibody mimics over antibody fragments is structural. These scaffolds are derived from whole, stable, and soluble structural modules found in human body fluid proteins. Consequently, they exhibit better folding and thermostability properties than antibody fragments, whose creation involves the removal of parts of the antibody native fold, often exposing amino acid residues that, in an intact antibody, would be buried in a hydrophobic environment, such as an interface between variable and constant domains. Exposure of such hydrophobic residues to solvent increases the likelihood of aggregation.
In addition, the antibody mimics described herein have no disulfide bonds, which have been reported to retard or prevent proper folding of antibody fragments under certain conditions. Since the present scaffolds do not rely on disulfides for native fold stability, they are stable under reducing conditions, unlike antibodies and their fragments which unravel upon disulfide bond breakdown.
Moreover, these fibronectin-based scaffolds provide the functional advantages of antibody molecules. In particular, despite the fact that the 10Fn3 module is not an immunoglobulin, its overall fold is close to that of the variable region of the IgG heavy chain (
There are now described below exemplary fibronectin-based scaffolds and their use for identifying, selecting, and evolving novel binding proteins as well as their target ligands. These examples are provided for the purpose of illustrating, and not limiting, the invention.
10Fn3 Structural Motif
The antibody mimics of the present invention are based on the structure of a fibronectin module of type III (Fn3), a common domain found in mammalian blood and structural proteins. This domain occurs more than 400 times in the protein sequence database and has been estimated to occur in 2% of the proteins sequenced to date, including fibronectins, tenscin, intracellular cytoskeletal proteins, and prokaryotic enzymes (Bork and Doolittle, Proc. Natl. Acad. Sci. USA 89:8990, 1992; Bork et al., Nature Biotech. 15:553, 1997; Meinke et al., J. Bacteriol. 175:1910, 1993; Watanabe et al., J. Biol. Chem. 265:15659, 1990). In particular, these scaffolds include, as templates, the tenth module of human Fn3 (10Fn3), which comprises 94 amino acid residues. The overall fold of this domain is closely related to that of the smallest functional antibody fragment, the variable region of the heavy chain, which comprises the entire antigen recognition unit in camel and llama IgG (
The three loops of 10Fn3 corresponding to the antigen-binding loops of the IgG heavy chain run between amino acid residues 21-31, 51-56, and 76-88 (
In contrast, the second loop of 10Fn3 is only 6 residues long, whereas the corresponding loop in antibody heavy chains ranges from 16-19 residues. To optimize antigen binding, therefore, the second loop of 10Fn3 is preferably extended by 10-13 residues (in addition to being randomized) to obtain the greatest possible flexibility and affinity in antigen binding. Indeed, in general, the lengths as well as the sequences of the CDR-like loops of the antibody mimics may be randomized during in vitro or in vivo affinity maturation (as described in more detail below).
The tenth human fibronectin type III domain, 10Fn3, refolds rapidly even at low temperature; its backbone conformation has been recovered within 1 second at 5° C. Thermodynamic stability of 10Fn3 is high (ΔGu=24 kJ/mol=5.7 kcal/mol), correlating with its high melting temperature of 110° C.
One of the physiological roles of 10Fn3 is as a subunit of fibronectin, a glycoprotein that exists in a soluble form in body fluids and in an insoluble form in the extracellular matrix (Dickinson et al., J. Mol. Biol. 236:1079, 1994). A fibronectin monomer of 220-250 kD contains 12 type I modules, two type II modules, and 17 fibronectin type III modules (Potts and Campbell, Curr. Opin. Cell Biol. 6:648, 1994). Different type III modules are involved in the binding of fibronectin to integrins, heparin, and chondroitin sulfate. 10Fn3 was found to mediate cell adhesion through an integrin-binding Arg-Gly-Asp (RGD) motif on one of its exposed loops. Similar RGD motifs have been shown to be involved in integrin binding by other proteins, such as fibrinogen, von Wellebrand factor, and vitronectin (Hynes et al., Cell 69:11, 1992). No other matrix- or cell-binding roles have been described for 10Fn3.
The observation that 10Fn3 has only slightly more adhesive activity than a short peptide containing RGD is consistent with the conclusion that the cell-binding activity of 10Fn3 is localized in the RGD peptide rather than distributed throughout the 10Fn3 structure (Baron et al., Biochemistry 31:2068, 1992). The fact that 10Fn3 without the RGD motif is unlikely to bind to other plasma proteins or extracellular matrix makes 10Fn3 a useful scaffold to replace antibodies. In addition, the presence of 10Fn3 in natural fibrinogen in the bloodstream suggests that 10Fn3 itself is unlikely to be immunogenic in the organism of origin.
In addition, we have determined that the 10Fn3 framework possesses exposed loop sequences tolerant of randomization, facilitating the generation of diverse pools of antibody mimics. This determination was made by examining the flexibility of the 10Fn3 sequence. In particular, the human 10Fn3 sequence was aligned with the sequences of fibronectins from other sources as well as sequences of related proteins (
For the human 10Fn3 sequence, this analysis indicates that, at a minimum, amino acids 1-9, 44-50, 61-54, 82-94 (edges of beta sheets); 19, 21, 30-46 (even), 79-65 (odd) (solvent-accessible faces of both beta sheets); 21-31, 51-56, 76-88 (CDR-like solvent-accessible loops); and 14-16 and 36-45 (other solvent-accessible loops and beta turns) may be randomized to evolve new or improved compound-binding proteins. In addition, as discussed above, alterations in the lengths of one or more solvent exposed loops may also be included in such directed evolution methods. Alternatively, changes in the β-sheet sequences may also be used to evolve new proteins. These mutations change the scaffold and thereby indirectly alter loop structure(s). If this approach is taken, mutations should not saturate the sequence, but rather few mutations should be introduced. Preferably, no more than 10 amino acid changes, and, more preferably, no more than 3 amino acid changes should be introduced to the β-sheet sequences by this approach.
The antibody mimics described herein may be fused to other protein domains. For example, these mimics may be integrated with the human immune response by fusing the constant region of an IgG (Fc) with a 10Fn3 module, preferably through the C-terminus of 10Fn3. The Fc in such a 10Fn3-Fc fusion molecule activates the complement component of the immune response and increases the therapeutic value of the antibody mimic. Similarly, a fusion between 10Fn3 and a complement protein, such as C1q, may be used to target cells, and a fusion between 10Fn3 and a toxin may be used to specifically destroy cells that carry a particular antigen. In addition, 10Fn3 in any form may be fused with albumin to increase its half-life in the bloodstream and its tissue penetration. Any of these fusions may be generated by standard techniques, for example, by expression of the fusion protein from a recombinant fusion gene constructed using publically available gene sequences.
Fibronectin Scaffold Multimers
In addition to fibronectin monomers, any of the fibronectin constructs described herein may be generated as dimers or multimers of 10Fn3-based antibody mimics as a means to increase the valency and thus the avidity of antigen binding. Such multimers may be generated through covalent binding between individual 10Fn3 modules, for example, by imitating the natural 8Fn3−9Fn3-10Fn3 C-to-N-terminus binding or by imitating antibody dimers that are held together through their constant regions. A 10Fn3-Fc construct may be exploited to design dimers of the general scheme of 10Fn3-Fc::Fc-10Fn3. The bonds engineered into the Fc::Fc interface may be covalent or non-covalent. In addition, dimerizing or multimerizing partners other than Fc can be used in 10Fn3 hybrids to create such higher order structures.
In particular examples, covalently bonded multimers may be generated by constructing fusion genes that encode the multimer or, alternatively, by engineering codons for cysteine residues into monomer sequences and allowing disulfide bond formation to occur between the expression products. Non-covalently bonded multimers may also be generated by a variety of techniques. These include the introduction, into monomer sequences, of codons corresponding to positively and/or negatively charged residues and allowing interactions between these residues in the expression products (and therefore between the monomers) to occur. This approach may be simplified by taking advantage of charged residues naturally present in a monomer subunit, for example, the negatively charged residues of fibronectin. Another means for generating non-covalently bonded antibody mimics is to introduce, into the monomer gene (for example, at the amino- or carboxy-termini), the coding sequences for proteins or protein domains known to interact. Such proteins or protein domains include coil-coil motifs, leucine zipper motifs, and any of the numerous protein subunits (or fragments thereof) known to direct formation of dimers or higher order multimers.
Although 10Fn3 represents a preferred scaffold for the generation of antibody mimics, other molecules may be substituted for 10Fn3 in the molecules described herein. These include, without limitation, human fibronectin modules 1Fn3-9Fn3 and 11Fn3-17Fn3 as well as related Fn3 modules from non-human animals and prokaryotes. In addition, Fn3 modules from other proteins with sequence homology to 10Fn3, such as tenascins and undulins, may also be used. Modules from different organisms and parent proteins may be most appropriate for different applications; for example, in designing an antibody mimic, it may be most desirable to generate that protein from a fibronectin or fibronectin-like molecule native to the organism for which a therapeutic or diagnostic molecule is intended.
Directed Evolution of Scaffold-Based Binding Proteins
The antibody mimics described herein may be used in any technique for evolving new or improved binding proteins. In one particular example, the target of binding is immobilized on a solid support, such as a column resin or microtiter plate well, and the target contacted with a library of candidate scaffold-based binding proteins. Such a library may consist of 10Fn3 clones constructed from the wild type 10Fn3 scaffold through randomization of the sequence and/or the length of the 10Fn3 CDR-like loops. If desired, this library may be an RNA-protein fusion library generated, for example, by the techniques described in Szostak et al., U.S. Ser. Nos. 09/007,005 and 09/247,190; Szostak et al., WO98/31700; and Roberts & Szostak, Proc. Natl. Acad. Sci. USA (1997) vol. 94, p. 12297-12302. Alternatively, it may be a DNA-protein library (for example, as described in Lohse, DNA-Protein Fusions and Uses Thereof, U.S. Ser. No. 60/110,549, U.S. Ser. No. 09/459,190, and US 99/28472). The fusion library is incubated with the immobilized target, the support is washed to remove non-specific binders, and the tightest binders are eluted under very stringent conditions and subjected to PCR to recover the sequence information or to create a new library of binders which may be used to repeat the selection process, with or without further mutagenesis of the sequence. A number of rounds of selection may be performed until binders of sufficient affinity for the antigen are obtained.
In one particular example, the 10Fn3 scaffold may be used as the selection target. For example, if a protein is required that binds a specific peptide sequence presented in a ten residue loop, a single 10Fn3 clone is constructed in which one of its loops has been set to the length of ten and to the desired sequence. The new clone is expressed in vivo and purified, and then immobilized on a solid support. An RNA-protein fusion library based on an appropriate scaffold is then allowed to interact with the support, which is then washed, and desired molecules eluted and re-selected as described above.
Similarly, the 10Fn3 scaffold may be used to find natural proteins that interact with the peptide sequence displayed in a 10Fn3 loop. The 10Fn3 protein is immobilized as described above, and an RNA-protein fusion library is screened for binders to the displayed loop. The binders are enriched through multiple rounds of selection and identified by DNA sequencing.
In addition, in the above approaches, although RNA-protein libraries represent exemplary libraries for directed evolution, any type of scaffold-based library may be used in the selection methods of the invention.
The antibody mimics described herein may be evolved to bind any antigen of interest. These proteins have thermodynamic properties superior to those of natural antibodies and can be evolved rapidly in vitro. Accordingly, these antibody mimics may be employed in place of antibodies in all areas in which antibodies are used, including in the research, therapeutic, and diagnostic fields. In addition, because these scaffolds possess solubility and stability properties superior to antibodies, the antibody mimics described herein may also be used under conditions which would destroy or inactivate antibody molecules. Finally, because the scaffolds of the present invention may be evolved to bind virtually any compound, these molecules provide completely novel binding proteins which also find use in the research, diagnostic, and therapeutic areas.
Exemplary scaffold molecules described above were generated and tested, for example, in selection protocols, as follows.
A complex library was constructed from three fragments, each of which contained one randomized area corresponding to a CDR-like loop. The fragments were named BC, DE, and FG, based on the names of the CDR-H-like loops contained within them; in addition to 10Fn3 and a randomized sequence, each of the fragments contained stretches encoding an N-terminal His6 domain or a C-terminal FLAG peptide tag. At each junction between two fragments (i.e., between the BC and DE fragments or between the DE and FG fragments), each DNA fragment contained recognition sequences for the EarI Type IIS restriction endonuclease. This restriction enzyme allowed the splicing together of adjacent fragments while removing all foreign, non-10Fn3, sequences. It also allows for a recombination-like mixing of the three 10Fn3 fragments between cycles of mutagenesis and selection.
Each fragment was assembled from two overlapping oligonucleotides, which were first annealed, then extended to form the double-stranded DNA form of the fragment. The oligonucleotides that were used to construct and process the three fragments are listed below; the “Top” and “Bottom” species for each fragment are the oligonucleotides that contained the entire 10Fn3 encoding sequence. In these oligonucleotides designations, “N” indicates A, T, C, or G; and “S” indicates C or G.
The pairs of oligonucleotides (500 pmol of each) were annealed in 100 μL of 10 mM Tris 7.5, 50 mM NaCl for 10 minutes at 85° C., followed by a slow (0.5-1 hour) cooling to room temperature. The annealed fragments with single-stranded overhangs were then extended using 100 U Klenow (New England Biolabs, Beverly, Mass.) for each 100 μL aliquot of annealed oligos, and the buffer made of 838.5 μl H2O, 9 μl 1 M Tris 7.5, 5 μl 1 M MgCl2, 20 μl 10 mM dNTPs, and 7.5 μl 1M DTT. The extension reactions proceeded for 1 hour at 25° C.
Next, each of the double-stranded fragments was transformed into a RNA-protein fusion (PROfusion™) using the technique developed by Szostak et al., U.S. Ser. No. 09/007,005 and U.S. Ser. No. 09/247,190; Szostak et al., WO98/31700; and Roberts & Szostak, Proc. Natl. Acad. Sci. USA (1997) vol. 94, p. 12297-12302. Briefly, the fragments were transcribed using an Ambion in vitro transcription kit, MEGAshortscript (Ambion, Austin, Tex.), and the resulting mRNA was gel-purified and ligated to a DNA-puromycin linker using DNA ligase. The mRNA-DNA-puromycin molecule was then translated using the Ambion rabbit reticulocyte lysate-based translation kit. The resulting mRNA-DNA-puromycin-protein PROfusion™ was purified using Oligo(dT) cellulose, and a complementary DNA strand was synthesized using reverse transcriptase and the RT primers described above (Unisplint-S or flagASA), following the manufacturer's instructions.
The PROfusion™ obtained for each fragment was next purified on the resin appropriate to its peptide purification tag, i.e., on Ni-NTA agarose for the His6-tag and M2 agarose for the FLAG-tag, following the procedure recommended by the manufacturer. The DNA component of the tag-binding PROfusions™ was amplified by PCR using Pharmacia Ready-to-Go PCR Beads, 10 pmol of 5′ and 3′ PCR primers, and the following PCR program (Pharmacia, Piscataway, N.J.): Step 1: 95° C. for 3 minutes; Step 2: 95° C. for 30 seconds, 58/62° C. for 30 seconds, 72° C. for 1 minute, 20/25/30 cycles, as required; Step 3: 72° C. for 5 minutes; Step 4: 4° C. until end.
The resulting DNA was cleaved by 5 U EarI (New England Biolabs) per1 ug DNA; the reaction took place in T4 DNA Ligase Buffer (New England Biolabs) at 37° C., for 1 hour, and was followed by an incubation at 70° C. for 15 minutes to inactivate Ear I. Equal amounts of the BC, DE, and FG fragments were combined and ligated to form a full-length 10Fn3 gene with randomized loops. The ligation required 10 U of fresh Earl (New England Biolabs) and 20 U of T4 DNA Ligase (Promega, Madison, Wis.), and took 1 hour at 37° C.
Three different libraries were made in the manner described above. Each contained the form of the FG loop with 10 randomized residues. The BC and the DE loops of the first library bore the wild type 10Fn3 sequence; a BC loop with 7 randomized residues and a wild type DE loop made up the second library; and a BC loop with 7 randomized residues and a DE loop with 4 randomized residues made up the third library. The complexity of the FG loop in each of these three libraries was 1013; the further two randomized loops provided the potential for a complexity too large to be sampled in a laboratory.
The three libraries constructed were combined into one master library in order to simplify the selection process; target binding itself was expected to select the most suitable library for a particular challenge. PROfusions™ were obtained from the master library following the general procedure described in Szostak et al., U.S. Ser. Nos. 09/007,005 and 09/247,190; Szostak et al., WO98/31700; and Roberts & Szostak, Proc. Natl. Acad. Sci. USA (1997) vol. 94, p. 12297-12302 (
The master library in the PROfusion™ form was subjected to selection for binding to TNF-α. Two protocols were employed: one in which the target was immobilized on an agarose column and one in which the target was immobilized on a BIACORE chip. First, an extensive optimization of conditions to minimize background binders to the agarose column yielded the favorable buffer conditions of 50 mM HEPES pH 7.4, 0.02% Triton, 100 μg/ml Sheared Salmon Sperm DNA. In this buffer, the non-specific binding of the 10Fn3 RNA fusion to TNF-α Sepharose was 0.3%. The non-specific binding background of the 10Fn3 RNA-DNA to TNF-α Sepharose was found to be 0.1%.
During each round of selection on TNF-α Sepharose, the Profusion™ library was first preincubated for an hour with underivatized Sepharose to remove any remaining non-specific binders; the flow-through from this pre-clearing was incubated for another hour with TNF-α Sepharose. The TNF-α Sepharose was washed for 3-30 minutes.
After each selection, the PROfusion™ DNA that had been eluted from the solid support with 0.3 M NaOH or 0.1M KOH was amplified by PCR; a DNA band of the expected size persisted through multiple rounds of selection (
In the first seven rounds, the binding of library PROfusions™ to the target remained low; in contrast, when free protein was translated from DNA pools at different stages of the selection, the proportion of the column binding species increased significantly between rounds (
Wild-type 10Fn3 contains an integrin-binding tripepetide motif, Arginine 78-Glycine 79-Aspartate 80 (the “RGD motif’H) at the tip of the FG loop. In order to avoid integrin binding and a potential inflammatory response based on this tripeptide in vivo, a mutant form of 10Fn3 was generated that contained an inert sequence, Serine 78-Glycine 79-Glutamate 80 (the “SGE mutant”), a sequence which is found in the closely related, wild-type 11Fn3 domain. This SGE mutant was expressed as an N-terminally His6-tagged, free protein in E. coli, and purified to homogeneity on a metal chelate column followed by a size exclusion column.
In particular, the DNA sequence encoding His6-10Fn3(SGE) was cloned into the pET9a expression vector and transformed into BL21 DE3 pLysS cells. The culture was then grown in LB broth containing 50 μg/mL kanamycin at 37° C., with shaking, to A560=1.0, and was then induced with 0.4 mM IPTG. The induced culture was further incubated, under the same conditions, overnight (14-18 hours); the bacteria were recovered by standard, low speed centrifugation. The cell pellet was resuspended in 1/50 of the original culture volume of lysis buffer (50 mM Tris 8.0, 0.5 M NaCl, 5% glycerol, 0.05% Triton X-100, and 1 mM PMSF), and the cells were lysed by passing the resulting paste through a Microfluidics Corporation Microfluidizer M 110-EH, three times. The lysate was clarified by centrifugation, and the supernatant was filtered through a 0.45 μm filter followed by filtration through a 0.2 μm filter. 100 mL of the clarified lysate was loaded onto a 5 mL Talon cobalt column (Clontech, Palo Alto, Calif.), washed by 70 mL of lysis buffer, and eluted with a linear gradient of 0-30 mM imidazole in lysis buffer. The flow rate through the column through all the steps was 1 mL/min. The eluted protein was concentrated 10-fold by dialysis (MW cutoff=3,500) against 15,000-20,000 PEG. The resulting sample was dialysed into buffer 1 (lysis buffer without the glycerol), then loaded, 5 mL at a time, onto a 16×60 mm Sephacryl 100 size exclusion column equilibrated in buffer 1. The column was run at 0.8 mL/min, in buffer 1; all fractions that contained a protein of the expected MW were pooled, concentrated 10× as described above, then dialyzed into PBS. Toxikon (MA) was engaged to perform endotoxin screens and animal studies on the resulting sample.
In these animal studies, the endotoxin levels in the samples examined to date have been below the detection level of the assay. In a preliminary toxicology study, this protein was injected into two mice at the estimated 100× therapeutic dose of 2.6 mg/mouse. The animals survived the two weeks of the study with no apparent ill effects. These results suggest that 10Fn3 may be incorporated safely into an IV drug.
Alternative Constructs for In Vivo Use
To extend the half life of the 8 kD 10Fn3 domain, a larger molecule has also been constructed that mimics natural antibodies. This 10Fn3-Fc molecule contains the —CH1 —CH2—CH3 (
In order to construct the murine form of the 10Fn3-CH1—CH2—CH3 clone, the —CH1—CH2—CH3 region was first amplified from a mouse liver spleen cDNA library (Clontech), then ligated into the pET25b vector. The primers used in the cloning were 5′Fc Nest and 3′ 5 Fc Nest, and the primers used to graft the appropriate restriction sites onto the ends of the recovered insert were 5′Fc HIII and 3′Fc Nhe:
Further PCR is used to remove the CH1 region from this clone and create the Fc part of the shorter, 10Fn3-CH2—CH3 clone. The sequence encoding 10Fn3 is spliced onto the 5′ end of each clone; either the wild type 10Fn3 cloned from the same mouse spleen cDNA library or a modified 10Fn3 obtained by mutagenesis or randomization of the molecules can be used. The oligonucleotides used in the cloning of murine wild-type 10Fn3 were:
The human equivalents of the clones are constructed using the same strategy with human oligonucleotide sequences.
10Fn3 Scaffolds in Protein Chip Applications
The suitability of the 10Fn3 scaffold for protein chip applications is the consequence of (1) its ability to support many binding functions which can be selected rapidly on the bench or in an automated setup, and (2) its superior biophysical properties.
The versatile binding properties of 10Fn3 are a function of the loops displayed by the Fn3 immunoglobulin-like, beta sandwich fold. As discussed above, these loops are similar to the complementarity determining regions of antibody variable domains and can cooperate in a way similar to those antibody loops in order to bind antigens. In our system, 10Fn3 loops BC (residues 21-30), DE (residues 51-56), and FG (residues 76-87) are randomized either in sequence, in length, or in both sequence and length in order to generate diverse libraries of mRNA-10Fn3 fusions. The binders in such libraries are then enriched based on their affinity for an immobilized or tagged target, until a small population of high affinity binders are generated. Also, error-prone PCR and recombination can be employed to facilitate affinity maturation of selected binders. Due to the rapid and efficient selection and affinity maturation protocols, binders to a large number of targets can be selected in a short time.
As a scaffold for binders to be immobilized on protein chips, the 10Fn3 domain has the advantage over antibody fragments and single-chain antibodies of being smaller and easier to handle. For example, unlike single-chain scaffolds or isolated variable domains of antibodies, which vary widely in their stability and solubility, and which require an oxidizing environment to preserve their structurally essential disulfide bonds, 10Fn3 is extremely stable, with a melting temperature of 110° C., and solubility at a concentration >16 mg/mL. The 10Fn3 scaffold also contains no disulfides or free cysteines; consequently, it is insensitive to the redox potential of its environment. A further advantage of 10Fn3 is that its antigen-binding loops and N-terminus are on the edge of the beta-sandwich opposite to the C-terminus; thus the attachment of a 10Fn3 scaffold to a chip by its C-terminus aligns the antigen-binding loops, allowing for their greatest accessibility to the solution being assayed. Since 10Fn3 is a single domain of only 94 amino acid residues, it is also possible to immobilize it onto a chip surface at a higher density than is used for single-chain antibodies, with their approximately 250 residues. In addition, the hydrophilicity of the 10Fn3 scaffold, which is reflected in the high solubility of this domain, leads to a lower than average background binding of 10Fn3 to a chip surface.
The stability of the 10Fn3 scaffold as well as its suitability for library formation and selection of binders are likely to be shared by the large, Fn3-like class of protein domains with an immunoglobulin-like fold, such as the domains of tenascin, N-cadherin, E-cadherin, ICAM, titin, GCSF—R, cytokine receptor, glycosidase inhibitor, and antibiotic chromoprotein. The key features shared by all such domains are a stable framework provided by two beta-sheets, which are packed against each other and which are connected by at least three solvent-accessible loops per edge of the sheet; such loops can be randomized to generate a library of potential binders without disrupting the structure of the framework (as described above).
Immobilization of Fibronectin Scaffold Binders (Fn-binders)
To immobilize Fn-binders to a chip surface, a number of exemplary techniques may be utilized. For example, Fn-binders may be immobilized as RNA-protein fusions by Watson-Crick hybridization of the RNA moiety of the fusion to a base complementary DNA immobilized on the chip surface (as described, for example, in Addressable Protein Arrays, U.S. Ser. No. 60/080,686; U.S. Ser. No. 09/282,734; and WO 99/51773). Alternatively, Fn-binders can be immobilized as free proteins directly on a chip surface. Manual as well as robotic devices may be used for deposition of the Fn-binders on the chip surface. Spotting robots can be used for deposition of Fn-binders with high density in an array format (for example, by the method of Lueking et al., Anal Biochem. 1999 May 15;270(1):103-11). Different methods may also be utilized for anchoring the Fn-binder on the chip surface. A number of standard immobilization procedures may be used including those described in Methods in Enzymology (K. Mosbach and B. Danielsson, eds.), vols. 135 and 136, Academic Press, Orlando, Fla., 1987; Nilsson et al., Protein Expr. Purif. 1997 October; 11(1): 1-16; and references therein. Oriented immobilization of Fn-binders can help to increase the binding capacity of chip-bound Fn-binders. Exemplary approaches for achieving oriented coupling are described in Lu et al., The Analyst (1996), vol. 121, p. 29R-32R; and Turkova, J Chromatogr B Biomed Sci App. 1999 Feb. 5; 722(1-2):11-31. In addition, any of the methods described herein for anchoring Fn-binders to chip surfaces can also be applied to the immobilization of Fn-binders on beads, or other supports.
Target Protein Capture and Detection
Selected populations of Fn-binders may be used for detection and/or quantitation of analyte targets, for example, in samples such as biological samples. To carry out this type of diagnostic assay, selected Fn-binders to targets of interest are immobilized on an appropriate support to form multi-featured protein chips. Next, a sample is applied to the chip, and the components of the sample that associate with the Fn-binders are identified based on the target-specificity of the immobilized binders. Using this technique, one or more components may be simultaneously identified or quantitated in a sample (for example, as a means to carry out sample profiling).
Methods for target detection allow measuring the levels of bound protein targets and include, without limitation, radiography, fluorescence scanning, mass spectroscopy (MS), and surface plasmon resonance (SPR). Autoradiography using a phosphorimager system (Molecular Dynamics, Sunnyvale, Calif.) can be used for detection and quantification of target protein which has been radioactively labeled, e.g., using 35S methionine. Fluorescence scanning using a laser scanner (see below) may be used for detection and quantification of fluorescently labeled targets. Alternatively, fluorescence scanning may be used for the detection of fluorescently labeled ligands which themselves bind to the target protein (e.g., fluorescently labeled target-specific antibodies or fluorescently labeled streptavidin binding to target-biotin, as described below).
Mass spectroscopy can be used to detect and identify bound targets based on their molecular mass. Desorption of bound target protein can be achieved with laser assistance directly from the chip surface as described below. Mass detection also allows determinations, based on molecular mass, of target modifications including post-translational modifications like phosophorylation or glycosylation. Surface plasmon resonance can be used for quantification of bound protein targets where the Fn-binder(s) are immobilized on a suitable gold-surface (for example, as obtained from Biacore, Sweden).
Described below are exemplary schemes for selecting Fn binders (in this case, Fn-binders specific for the protein, TNF-α) and the use of those selected populations for detection on chips. This example is provided for the purpose of illustrating the invention, and should not be construed as limiting.
Selection of TNF-α Binders Based on 10Fn3 Scaffold
In one exemplary use for fibronectin scaffold selection on chips, an 10Fn3-based selection was performed against TNF-α, using a library of human 10Fn3 variants with randomized loops BC, DE, and FG. The library was constructed from three DNA fragments, each of which contained-nucleotide sequences that encoded approximately one third of human 10Fn3, including one of the randomized loops. The DNA sequences that encoded the loop residues listed above were rebuilt by oligonucleotide synthesis, so that the codons for the residues of interest were replaced by (NNS)n, where N represents any of the four deoxyribonucleotides (A, C, G, or T), and S represents either C or G. The C-terminus of each fragment contained the sequence for the FLAG purification tag.
Once extended by Klenow, each DNA fragment was transcribed, ligated to a puromycin-containing DNA linker, and translated in vitro, as described by Szostak et al. (Roberts and Szostak, Proc. Natl. Acad. Sci USA. 94:12297, 1997; Szostak et al., U.S. Ser. No. 09/007,005 and U.S. Ser. No. 09/247,190; Szostak et al., WO98/31700), to generate an mRNA-peptide fusion, which was then reverse-transcribed into a DNA-mRNA-peptide fusion. The binding of the FLAG-tagged peptide to M2 agarose separated full-length fusion molecules from those containing frameshifts or superfluous stop codons; the DNA associated with the purified full-length fusion was amplified by PCR, then the three DNA fragments were cut by Ear I restriction endonuclease and ligated to form the full length template. The template was transcribed, ligated to puromycin-containing DNA linkers, and translated to generate a 10Fn3-PROfusion™ library, which was then reverse-transcribed to yield the DNA-mRNA-peptide fusion library which was subsequently used in the selection.
Selection for TNF-α binders took place in 50 mM HEPES, pH 7.4, 0.02% Triton-X, 0.1 mg/mL salmon sperm DNA. The PROfusion™ library was incubated with Sepharose-immobilized TNF-α; after washing, the DNA associated with the tightest binders was eluted with 0.1 M KOH, amplified by PCR, and transcribed, ligated, translated, and reverse-transcribed into the starting material for the next round of selection.
Ten rounds of such selection were performed (as shown in
Fn-binder Immobilization, Target Protein Capture, and MALDI-TOF Detection
As a first step toward immobilizing the Fn-binders to a chip surface, an oligonucleotide capture probe was prepared with an automated DNA synthesizer (PE BioSystems Expedite 8909) using the solid-support phosphoramidite approach. All reagents were obtained from Glen Research. Synthesis was initiated with a solid support containing a disulfide bond to eventually provide a 3′-terminal thiol functionality. The first four monomers to be added were hexaethylene oxide units, followed by 20 T monomers. The 5′-terminal DMT group was not removed. The capture probe was cleaved from the solid support and deprotected with ammonium hydroxide, concentrated to dryness in a vacuum centrifuge, and purified by reverse-phase HPLC using an acetonitrile gradient in triethylammonium acetate buffer. Appropriate fractions from the HPLC were collected, evaporated to dryness in a vacuum centrifuge, and the 5′-terminal DMT group was removed by treatment with 80% AcOH for 30 minutes. The acid was removed by evaporation, and the oligonucleotide was then treated with 100 mM DTT for 30 minutes to cleave the disulfide bond. DTT was removed by repeated extraction with EtOAc. The oligonucleotide was ethanol precipitated from the remaining aqueous layer and checked for purity by reverse-phase HPLC.
The 3′-thiol capture probe was adjusted to 250 μM in degassed 1×PBS buffer and applied as a single droplet (75 μL) to a 9×9 mm gold-coated chip (Biacore) in an argon-flushed chamber containing a small amount of water. After 18 hours at room temperature, the capture probe solution was removed, and the functionalized chip was washed with 50 mL 1×PBS buffer (2× for 15 minutes each) with gentle agitation, and then rinsed with 50 μL water (2× for 15 minutes each) in the same fashion. Remaining liquid was carefully removed and the functionalized chips were either used immediately or stored at 4° C. under argon.
About 1 pmol of 10Fn3 fusion pool from the Round 10 TNF-α selection (above) was treated with RNAse A for several hours, adjusted to 5×SSC in 70 μL, and applied to a functionalized gold chip from above as a single droplet. A 50 μL volume gasket device was used to seal the fusion mixture with the functionalized chip, and the apparatus was continuously rotated at 4° C. After 18 hours the apparatus was disassembled, and the gold chip was washed with 50 mL 5×SSC for 10 minutes with gentle agitation. Excess liquid was carefully removed from the chip surface, and the chip was passivated with a blocking solution (1×TBS+0.02% Tween-20+0.25% BSA) for 10 minutes at 4° C. Excess liquid was carefully removed, and a solution containing 500 μg/mL TNF-α in the same composition blocking solution was applied to the chip as a single droplet and incubated at 4° C. for two hours with occasional mixing of the droplet via Pipetman. After removal of the binding solution, the chip was washed for 5 minutes at 4° C. with gentle agitation (50 mL 1×TBS+0.02% Tween-20) and then dried at room temperature. A second chip was prepared exactly as described above, except fusion was not added to the hybridization mix.
Next, MALDI-TOF matrix (15 mg/mL 3,5-dimethoxy-4-hydroxycinnamic acid in 1:1 ethanol/I 0% formic acid in water) was uniformly applied to the gold chips with a high-precision 3-axis robot (MicroGrid, BioRobotics). A 16-pin tool was used to transfer the matrix from a 384-well microtiter plate to the chips, producing 200 micron diameter features with a 600 micron pitch. The MALDI-TOF mass spectrometer (Voyager DE, PerSeptive Biosystems) instrument settings were as follows: Accelerating Voltage=25 k, Grid Voltage=92%, Guide Wire Voltage=0.05%, Delay=200 on, Laser Power=2400, Low Mass Gate=1500, Negative Ions=off. The gold chips were individually placed on a MALDI sample stage modified to keep the level of the chip the same as the level of the stage, thus allowing proper flight distance. The instrument's video monitor and motion control system were used to direct the laser beam to individual matrix features.
Fn-binder Immobilization, Target Protein Capture, and Fluorescence Detection
Pre-cleaned 1×3 inch glass microscope slides (Goldseal, #3010) were treated with Nanostrip (Cyantek) for 15 minutes, 10% aqueous NaOH at 70° C. for 3 minutes, and 1% aqueous HCl for 1 minute, thoroughly rinsing with deionized water after each reagent. The slides were then dried in a vacuum desiccator over anhydrous calcium sulfate for several hours. A 1% solution of aminopropytrimethoxysilane in 95% acetone/5% water was prepared and allowed to hydrolyze for 20 minutes. The glass slides were immersed in the hydrolyzed silane solution for 5 minutes with gentle agitation. Excess silane was removed by subjecting the slides to ten 5-minute washes, using fresh portions of 95% acetone/5% water for each wash, with gentle agitation. The slides were then cured by heating at 110° C. for 20 minutes. The silane treated slides were immersed in a freshly prepared 0.2% solution of phenylene 1,4-diisothiocyanate in 90% DMF/10% pyridine for two hours, with gentle agitation. The slides were washed sequentially with 90% DMF/10% pyridine, methanol, and acetone. After air drying, the functionalized slides were stored at 0° C. in a vacuum desiccator over anhydrous calcium sulfate. Similar results were obtained with commercial amine-reactive slides (3-D Link, Surmodics).
Oligonucleotide capture probes were prepared with an automated DNA synthesizer (PE BioSystems Expedite 8909) using conventional phosphoramidite chemistry. All reagents were from Glen Research. Synthesis was initiated with a solid support bearing an orthogonally protected amino functionality, whereby the 3′-terminal amine is not unmasked until final deprotection step. The first four monomers to be added were hexaethylene oxide units, followed by the standard A, G, C and T monomers. All capture oligo sequences were cleaved from the solid support and deprotected with ammonium hydroxide, concentrated to dyrness, precipitated in ethanol, and purified by reverse-phase HPLC using an acetonitrile gradient in triethylammonium acetate buffer. Appropriate fractions from the HPLC were collected, evaporated to dryness in a vacuum centrifuge, and then coevaporated with a portion of water.
The purified, amine-labeled capture oligos were adjusted to a concentration of 250 μM in 50 mM sodium carbonate buffer (pH 9.0) containing 10% glycerol. The probes were spotted onto the amine-reactive glass surface at defined positions in a 5×5×6 array pattern with a 3-axis robot (MicroGrid, BioRobotics). A 16-pin tool was used to transfer the liquid from 384-well microtiter plates, producing 200 micron features with a 600 micron pitch. Each sub-grid of 24 features represents a single capture probe (i.e., 24 duplicate spots). The arrays were incubated at room temperature in a moisture-saturated environment for 12-18 hours. The attachment reaction was terminated by immersing the chips in 2% aqueous ammonium hydroxide for five minutes with gentle agitation, followed by rinsing with distilled water (3× for 5 minutes each). The array was finally soaked in 10× PBS solution for 30 minutes at room temperature, and then rinsed again for 5 minutes in distilled water.
Specific and thermodynamically isoenergetic sequences along the 10Fn3 mRNA were identified to serve as capture points to self-assemble and anchor the 10Fn3 protein. The software program HybSimulator v4.0 (Advanced Gene Computing Technology, Inc.) facilitated the identification and analysis of potential capture probes. Six unique capture probes were chosen and printed onto the chip, three of which are complementary to common regions of the 10Fn3 fusion pool's mRNA (CP3′, CP5′, and CPflag). The remaining three sequences (CPneg1, CPneg2, and CPneg3) are not complementary and function in part as negative controls. Each of the capture probes possesses a 3′-amino terminus and four hexaethylene oxide spacer units, as described above. The following is a list of the capture probe sequences that were employed (5′-3′):
About 1 pmol of 10Fn3 fusion pool from the Round 10 TNF-α selection was adjusted to 5× SSC containing 0.02% Tween-20 and 2 mM vanadyl ribonucleotide complex in a total volume of 350 μL. The entire volume was applied to the microarray under a 400 μL gasket device and the assembly was continuously rotated for 18 hours at room temperature. After hybridization the slide was washed sequentially with stirred 500 mL portions of 5×SSC, 2.5×SSC, and 1×SSC for 5 minutes each. Traces of liquid were removed by centrifugation and the slide was allowed to air-dry.
Recombinant human TNF-α (500 μg, lyophilized, from PreproTech) was taken up in 230 μL 1× PBS and dialyzed against 700 mL stirred 1× PBS at 4° C. for 18 hours in a Microdialyzer unit (3,500 MWCO, Pierce). The dialyzed TNF-α was treated with EZ-Link NHS-LC-LC biotinylation reagent (20 μg, Pierce) for 2 hours at 0° C., and again dialyzed against 700 mL stirred 1× PBS at 4° C. for 18 hours in a Microdialyzer unit (3,500 MWCO, Pierce). The resulting conjugate was analyzed by MALDI-TOF mass spectrometry and was found to be almost completely functionalized with a single biotin moiety.
Each of the following processes was conducted at 4° C. with continuous rotation or mixing. The protein microarray surface was passivated by treatment with 1× TBS containing 0.02% Tween-20 and 0.2% BSA (200 μL) for 60 minutes. Biotinylated TNF-α (100 nM concentration made up in the passivation buffer) was contacted with the microarray for 120 minutes. The microarray was washed with 1× TBS containing 0.02% Tween-20 (3×50 mL, 5 minutes each wash). Fluorescently labeled streptavidin (2.5 μg/mL Alexa 546-streptavidin conjugate from Molecular Probes, made up in the passivation buffer) was contacted with the microarray for 60 minutes. The microarray was washed with 1×TBS containing 0.02% Tween-20 (2×50 mL, 5 minutes each wash) followed by a 3 minute rinse with 1× TBS. Traces of liquid were removed by centrifugation, and the slide was allowed to air-dry at room temperature.
Fluorescence laser scanning was performed with a GSI Lumonics ScanArray 5000 system using 10 μM pixel resolution and preset excitation and emission wavelengths for Alexa 546 dye. Phosphorimage analysis was performed with a Molecular Dynamics Storm system. Exposure time was 48 hours with direct contact between the microarray and the phosphor storage screen. Phosphorimage scanning was performed at the 50 μM resolution setting, and data was extracted with ImageQuant v.4.3 software.
Other embodiments are within the claims.
All publications, patents, and patent applications mentioned herein are hereby incorporated by reference.