Title:
Universal fibronectin type III binding-domain libraries
Kind Code:
A1


Abstract:
Walk-through mutagenesis and natural-variant combinatorial fibronectin Type III (FN3) polypeptide libraries are described, along with their method of construction and use. Also disclosed are a number of high binding affinity polypeptides selected by screening the libraries against a variety of selected antigens.



Inventors:
Cappuccilli, Guido (San Mateo, CA, US)
Crea, Roberto (San Mateo, CA, US)
Shen, Randy (Sunnyvale, CA, US)
Hokanson, Craig A. (Pleasanton, CA, US)
Kirk, Peter B. (San Mateo, CA, US)
Liston, David R. (San Mateo, CA, US)
Application Number:
12/228404
Publication Date:
07/09/2009
Filing Date:
08/11/2008
Assignee:
ProtElix, Inc. (Hayward, CA, US)
Primary Class:
Other Classes:
506/14, 506/17, 506/18, 506/26, 530/324
International Classes:
C40B30/04; C07K14/00; C40B40/02; C40B40/08; C40B40/10; C40B50/06
View Patent Images:
Related US Applications:
20090227466SOL COMPOSITION FOR SOL-GEL BIOCHIP TO IMMOBILIZE PROBE ON SUBSTRATE WITHOUT SURFACE TREATMENT AND METHOD FOR SCREENING THEREOFSeptember, 2009Kim
20080227132Hydrolase Enzymes and Their Use in Kinetic ResolutionSeptember, 2008Short et al.
20100029497METHOD FOR ENGINEERING IMMUNOGLOBULINSFebruary, 2010Himmler et al.
20090104615PHENOTYPE PREDICTIONApril, 2009Godfrey et al.
20100021903Use of Genetic Determinants in Cardiovascular Risk AssessmentJanuary, 2010Johnson et al.
20090311697METHODS FOR SCREENING FOR GENETIC PREDISPOSITION TO TYPE I DIABETESDecember, 2009Hagopian et al.
20090178929SEPARATION DEVICE FOR ISOELECTRIC FOCUSINGJuly, 2009Broer et al.
20080176768HYDROGEL MICROARRAY WITH EMBEDDED METAL NANOPARTICLESJuly, 2008Zheng et al.
20090155784ASSESSMENT OF ASTHMA AND ALLERGEN-DEPENDENT GENE EXPRESSIONJune, 2009O'toole et al.
20080227656Biosurface Structure ArraySeptember, 2008Besenbacher et al.
20080311563Methods for selecting medicationsDecember, 2008Mrazek et al.



Primary Examiner:
BOESEN, CHRISTIAN C
Attorney, Agent or Firm:
King & Spalding LLP (P.O. Box 889, Belmont, CA, 94002-0889, US)
Claims:
It is claimed:

1. A method of forming a library of fibronectin Type 3 (FN3) domain polypeptides useful in screening for the presence of one or more polypeptides having a selected binding or enzymatic activity, comprising (i) aligning BC, DE, and FG amino acid loop sequences in a collection of native fibronectin Type 3 domain polypeptides, (ii) segregating the aligned loop sequences according to loop length, (iii) for a selected loop and loop length from step (ii), performing positional amino acid frequency analysis to determine the frequencies of amino acids at each loop position, (iv) for each loop and loop length analyzed in step (iii), identifying at each position a conserved or selected semi-conserved consensus amino acid and other natural-variant amino acids, (v) for at least one selected loop and loop length, forming: (1) a library of walk-through mutagenesis sequences expressed by a library of coding sequences that encode, at each loop position, the consensus amino acid, and if the consensus amino acid has a occurrence frequency equal to or less than a selected threshold frequency of at least 50%, a single common target amino acid and any co-produced amino acids, or (2) a library of natural-variant combinatorial sequences expressed by a library of coding sequences that encode at each loop position, a consensus amino acid and, if the consensus amino acid has a frequency of occurrence equal to or less than a selected threshold frequency of at least 50%, other natural variant amino acids, including semi-conserved amino acids and variable amino acids whose occurrence rate is above a selected minimum threshold occurrence at that position, or their chemical equivalents, (vi) incorporating the library of coding sequences into framework FN3 coding sequences to form an FN3 expression library, and (vi) expressing the FN3 polypeptides of the expression library.

2. The method of claim 1, wherein the given threshold frequency is 100%.

3. The method of claim 1, wherein the given threshold frequency is a selected frequency between 50-95%.

4. The method of claim 1, wherein step (ii) segregates the loops and loop lengths into the group consisting of BC/11, BC/14, BC/15, DE/6, FG/8, and FG/11.

5. The method of claim 5, wherein the library formed has a library of walk-through mutagenesis sequences formed at each of the loops and loop lengths selected from the group consisting of BC/11, BC/14, BC/15, DE/6, FG/8, and FG11, and from each of the common target amino selected from the group consisting of lysine, glutamine, aspartic acid, tyrosine, leucine, praline, serine, histidine, and glycine.

6. The method of claim 1, wherein the library formed has a library of natural-variant combinatorial sequences at a combination of loops and loop lengths selected from loops BC and DE, BC and FG, and DE and FG loops, where the BC loop is selected from one of BC/11, BC/14, and BC/15, the DE loop is DE/6, and the FG loop is selected from one of FG/8, and FG11.

7. The method of claim 6, wherein the given threshold is 100%, unless the loop amino acid position contains only one dominant and one variant amino, and the dominant and variant amino have side chains with similar physiochemical properties, in which case the given threshold is 90%.

8. The method of claim 6, wherein the two loops in the selected combination of loops and loop lengths has an average diversity of between 105 and 107, or where the average number of different amino acids at each position is equal to or less than 5.

9. The method of claim 1, wherein said polypeptides have the wildtype amino acid sequences in beta-strand regions A, AB, B, C, CD, D, E, EF, F, and G of the 14th fibronectin Type III module of human fibronecton.

10. A walk-through mutagenesis library of fibronectin Type 3 (FN3) domain polypeptides useful in screening for the presence of one or more polypeptides having a selected binding or enzymatic activity, said polypeptides comprising: (a) regions A, AB, B, C, CD, D, E, EF, F, and G having wildtype amino acid sequences of a selected native fibronectin Type 3 polypeptide, and (b) loop regions BC, DE, and FG having selected lengths, where at least one selected loop region of a selected length contains a library of walk through mutagenesis sequences expressed by a library of coding sequences that encode, at each loop position, a conserved or selected semi-conserved consensus amino acid and, if the consensus amino acid has an occurrence frequency equal to or less than a selected threshold frequency of at least 50%, a single common target amino acid and any co-produced amino acids.

11. The library of claim 10, wherein the given threshold frequency is 100%.

12. The library of claim 10, wherein the given threshold frequency is a selected frequency between 50-95%.

13. The library of claim 10, wherein loops and loop lengths are selected from the group consisting of BC/11, BC/14, BC/15, DE/6, FG/8, and FG11, and which has a library of walk-through mutagenesis sequences formed at each of the loops and loop lengths selected from the group consisting of BC/11, BC/14, BC/15, DE/6, FG/8, and FG11.

14. The library claim 13, which has a library of walk-through mutagenesis sequences formed from each of the common target amino selected from the group consisting of lysine, glutamine, aspartic acid, tyrosine, leucine, praline, serine, histidine, and glycine.

15. A natural-variant combinatorial library of fibronectin Type 3 (FN3) domain polypeptides useful in screening for the presence of one or more polypeptides having a selected binding or enzymatic activity, said polypeptides comprising: (a) regions A, AB, B, C, CD, D, E, EF, F, and G having wildtype amino acid sequences of a selected native fibronectin Type 3 polypeptide, and (b) loop regions BC, DE, and FG having selected lengths, where at least one selected loop region of a selected length contains a library of natural-variant combinatorial sequences expressed by a library of coding sequences that encode at each loop position, a conserved or selected semi-conserved consensus amino acid and, if the consensus amino acid has a frequency of occurrence equal to or less than a selected threshold frequency of at least 50%, other natural variant amino acids, including semi-conserved amino acids and variable amino acids whose occurrence rate is above a selected minimum threshold occurrence at that position, or their chemical equivalents.

16. The library of claim 15, which has a library of natural-variant combinatorial sequences at a combination of loops and loop lengths selected from loops BC and DE, BC and FG, and DE and FG loops, where the BC loop is selected from one of BC/11, BC/14, and BC/15, the DE loop is DE/6, and the FG loop is selected from one of FG/8, and FG11.

17. The library of claim 15, which has at two of the loop combinations BC and DE, BC and FG, and DE and FG, beneficial mutations identified by screening a universal combinatorial library containing amino acid variants in the two loop combination, and at the third loop, identified by FG, DE, and BC, respectively, a library of natural variant combinatorial sequences at a third loop and lop length identified by BC/11, BC/14, and BC/15, DE/6, or FG/8, and FG11.

18. The library method of claim 17, wherein each of the two selected loops have an average diversity of between 105 and 107.

19. The library of claim 15, which has a given threshold is 100%, unless the loop amino acid position contains only one dominant and one variant amino, and the dominant and variant amino have side chains with similar physiochemical properties, in which case the given threshold is 90%.

20. The library of claim 15, wherein said polypeptides have the wildtype amino acid sequences in beta-strand regions A, AB, B, C, CD, D, E, EF, F, and G of the 14th fibronectin Type III module of human fibronecton.

21. The library of claim 15, wherein said polypeptides have the wildtype amino acid sequences in regions A, AB, B, C, CD, D, E, EF, F, and G of the 10th fibronectin Type III module of human fibronecton.

22. The library of claim 15, wherein the BC loop length is 11, and has the amino acid sequence identified by SEQ ID NOS: 43 or 49.

23. The library of claim 15, wherein the BC loop length is 14, and has the amino acid sequence identified by SEQ ID. NOS: 44 or 50.

24. The library of claim 15, wherein the BC loop length is 15, and has the amino acid sequence identified by SEQ ID. NOS: 45 or 51.

25. The library of claim 15, wherein the DE loop length is 6, and has the amino acid sequence identified by SEQ ID. NOS: 46 or 52.

26. The library of claim 15, wherein the FG loop length is 8, and has the amino acid sequence identified by SEQ ID. NOS: 47, for the first N-terminal six amino acids, or SEQ ID NO:53.

27. The library of claim 15, wherein FG loop length is 11, and has the amino acid sequence identified by SEQ ID. NO: 48, for the first N-terminal nine amino acids, or SEQ ID NO:54.

28. The library of claim 15, wherein the polypeptides are encoded by an expression library selected from the group consisting of a ribosome display library, a polysome display library, a phage display library, a bacterial expression library, and a yeast display library.

29. An expression library of polynucleotides encoding the library of polypeptides of claim 15, and produced by synthesizing polynucleotides encoding one or more beta-strand framework regions and one or more loop regions wherein the polynucleotides are predetermined, wherein the polynucleotides encoding said regions further comprise sufficient overlapping sequence whereby the polynucleotide sequences, under polymerase chain reaction (PCR) conditions, are capable of assembly into polynucleotides encoding complete fibronectin binding domains.

30. A method of identifying a polypeptide having a desired binding affinity with respect to a selected antigen, comprising reacting the natural-variant combinatorial library of FN3 polypeptides of claim 15 with the selected antigen, and screening the FM3 polypeptides to select those having a desired binding affinity with respect to the selected antigen

31. The method of claim 30, wherein the method further comprises the step of identifying the polynucleotide that encodes the selected fibronectin binding domain.

32. A TNF-α binding protein having a Kd binding constant equal to or greaterthan 0.1 μM and having a sequence selected from SEQ ID NOS: 55-63.

33. A VEGF binding protein having a Kd binding constant equal to or greater than 0.1 μM and having a sequence selected from SEQ ID NOS: 64-67.

34. A HMGB1 binding protein having a Kd binding constant equal to or greater than 0.1 μM and having a sequence selected from SEQ ID NOS: 67-81.

Description:

This patent application claims priority to U.S. Provisional Patent Application No. 60/955,334 filed Aug. 10, 2007 and U.S. Provisional Patent Application No. 61/075,107 filed Jun. 24, 2008, both of which are incorporated herein by reference in their entirety.

BACKGROUND OF THE INVENTION

Scaffold based binding proteins are becoming legitimate alternatives to antibodies in their ability to bind specific ligand targets. These scaffold binding proteins share the quality of having a stable framework core that can tolerate multiple substitutions in the ligand binding regions. Some scaffold frameworks have immunoglobulin like protein domain architecture with loops extending from a beta sandwich core. A scaffold framework core can then be synthetically engineered from which a library of different sequence variants can be built upon. The sequence diversity is typically concentrated in the exterior surfaces of the proteins such as loop structures or other exterior surfaces that can serve as ligand binding regions.

Fibronectin Type III domain (FN3) was first identified as a one of the repeating domains in the fibronectin protein. The FN3 domain constitutes a small (˜94 amino acids), monomeric β-sandwich protein made up of seven β strands with three connecting loops. The three loops near the N-terminus of FN3, are functionally analogous to the complementarity-determining regions of immunoglobulin domains. FN3 loop libraries can then be engineered to bind to a variety of targets such as cytokines, growth factors and receptor molecules and other proteins.

One potential problem in creating these synthetic libraries is the high frequency of unproductive variants leading therefore, to inefficient candidate screens. For example, creating diversity in the variants often involves in vitro techniques such as random mutagenesis, saturation mutagenesis, error-prone PCR, and gene shuffling. These strategies are inherently stochastic and often require the construction of exceedingly large libraries to comprehensively explore sufficient sequence diversity. Additionally, there is no way to enumerate the number, what type and where in the protein the mutations have occurred. Furthermore, these random strategies create indiscriminate substitutions that cause protein architecture destabilization. It has been shown that improvement in one characteristic, such as affinity optimization, usually leads to decreased thermal stability when compared to the original protein scaffold framework.

Accordingly, a need exists for a fibronectin binding domain library that is systematic in construction. By bioinformatics led design, the loop candidates are flexible for insertion into multiple FN3 scaffolds. By specific targeted loop substitutions, overall scaffold stability is maximized while concurrently, non-immunogenic substitutions are minimized. Additionally, the library can be size tailored so that the overall diversity can be readily screened in different systems. Furthermore, the representative diversity of the designed loops are still capable of binding a number of pre-defined ligand targets. Moreover, the systematic design of loop still allows subsequent affinity maturation of recovered binding clones.

SUMMARY OF THE INVENTION

In one aspect, the invention includes a natural-variant combinatorial library of fibronectin Type 3 domain polypeptides useful in screening for the presence of one or more polypeptides having a selected binding or enzymatic activity. The library polypeptides include (a) regions A, AB, B, C, CD, D, E, EF, F, and G having wildtype amino acid sequences of a selected native fibronectin Type 3 polypeptide or polypeptides, and (b) loop regions BC, DE, and FG having selected lengths. At least one selected loop region of a selected length contains a library of natural-variant combinatorial sequences expressed by a library of coding sequences that encode at each loop position, a conserved or selected semi-conserved consensus amino acid and, if the consensus amino acid has a frequency of occurrence equal to or less than a selected threshold frequency of at least 50%, other natural variant amino acids, including semi-conserved amino acids and variable amino acids whose occurrence rate is above a selected minimum threshold occurrence at that position, or their chemical equivalents.

The library may have a given threshold is 100%, unless the loop amino acid position contains only one dominant and one variant amino, and the dominant and variant amino are chemically similar amino acids, in which case the given threshold may be 90%. In this embodiment, the library contains all natural variants or their chemical equivalents having at least some reasonable occurrence frequency, e.g., 10%, in the in the selected loop and loop position.

The natural-variant combinatorial sequences may be in a combination of loops and loop lengths selected from loops BC and DE, BC and FG, and DE and FG loops, where the BC loop is selected from one of BC/11, BC/14, and BC/15, the DE loop is DE/6, and the FG loop is selected from one of FG/8, and FG11.

The library may have at two of the loop combinations BC and DE, BC and FG, and DE and FG, beneficial mutations identified by screening a natural-variant combinatorial library containing amino acid variants in the two loop combination, and at the third loop, identified by FG, DE, and BC, respectively, a library of natural variant combinatorial sequences at a third loop and loop length identified by BC/11, BC/14, and BC/15, DE/6, or FG/8, and FG11.

In one embodiment, the library may have the wildtype amino acid sequences in regions A, AB, B, C, CD, D, E, EF, F, and G of the 14th fibronectin Type III module of human fibronecton. In another embodiment, the library may have the wildtype amino acid sequences in regions A, AB, B, C, CD, D, E, EF, F, and G of the 10th fibronectin Type III module of human fibronecton.

A natural-variant combinatorial library may have the following sequences for the indicated loops and loop lengths: (a) BC loop length of 11, and the amino acid sequence identified by SEQ ID NOS: 43 or 49; (b) BC loop length of 14, and the amino acid sequence identified by SEQ ID. NOS: 44 or 50; (c) BC loop length of 15, and the amino acid sequence identified by SEQ ID. NOS: 45 or 51; (d) DE loop length of 6, and the amino acid sequence identified by SEQ ID. NOS: 46 or 52; (e) FG loop length of 8, and the amino acid sequence identified by SEQ ID. NOS: 47, for the first N-terminal six amino acids, or SEQ ID NO:53, and (f) FG loop length of 11, and the amino acid sequence identified by SEQ ID. NO: 48, for the first N-terminal nine amino acids, or SEQ ID NO:54.

The library of polypeptides may be encoded by an expression library selected from the group consisting of a ribosome display library, a polysome display library, a phage display library, a bacterial expression library, and a yeast display library.

The libraries may be used in a method of identifying a polypeptide having a desired binding affinity, in which the natural-variant combinatorial library are screened to select for an fibronectin binding domain having a desired binding affinity. In particular, it has been found that the natural-variant combinatorial library provides high-binding polypeptides with high efficiency for a number of antigen targets, such as FNFα. VEGF, and HMGB1.

The screening may involve, for example, contacting the fibronectin binding domains with a target substrate, where the fibronectin binding domains being associated with the polynucleotide encoding the fibronectin binding domain. The method may further include identifying FN3 polynucleotides that encode the selected fibronectin binding domain.

Also disclosed is an expression library of polynucleotides encoding the above library of polypeptides, and produced by synthesizing polynucleotides encoding one or more framework regions and one or more loop regions wherein the polynucleotides are predetermined, wherein the polynucleotides encoding said regions further comprise sufficient overlapping sequence whereby the polynucleotide sequences, under polymerase chain reaction (PCR) conditions, are capable of assembly into polynucleotides encoding complete fibronectin binding domains.

In another aspect, the invention includes a walk-through mutagenesis library of fibronectin Type 3 domain polypeptides useful in screening for the presence of one or more polypeptides having a selected binding or enzymatic activity. The library polypeptides include (a) regions A, AB, B, C, CD, D, E, EF, F, and G having wildtype amino acid sequences of a selected native fibronectin Type 3 polypeptide or polypeptides, and (b) loop regions BC, DE, and FG having selected lengths. At least one selected loop region of a selected length contains a library of walk through mutagenesis sequences expressed by a library of coding sequences that encode, at each loop position position, a conserved or selected semi-conserved consensus amino acid and, if the consensus amino acid has a occurrence frequency equal to or less than a selected threshold frequency of at least 50%, a single common target amino acid and any co-produced amino acids.

The given threshold frequency may be 100%, or a selected frequency between 50-100%. The loops and loop lengths in the library may be selected from the group consisting of BC/11, BC/14, BC/15, DE/6, FG/8, and FG11.

The library may have a library of walk-through mutagenesis sequences formed at each of the loops and loop lengths selected from the group consisting of BC/11, BC/14, BC/15, DE/6, FG/8, and FG11, and for each common target amino selected from the group consisting of lysine, glutamine, aspartic acid, tyrosine, leucine, praline, serine, histidine, and glycine.

In one embodiment, the library may have the wildtype amino acid sequences in regions A, AB, B, C, CD, D, E, EF, F, and G of the 14th fibronectin Type III module of human fibronecton. In another embodiment, the library may have the wildtype amino acid sequences in regions A, AB, B, C, CD, D, E, EF, F, and G of the 10th fibronectin Type III module of human fibronecton.

In another aspect, the invention includes a method of forming a library of fibronectin Type 3 domain polypeptides useful in screening for the presence of one or more polypeptides having a selected binding or enzymatic activity. The method includes the steps of:

(i) aligning BC, DE, and FG amino acid loop sequences in a collection of native fibronectin Type 3 domain polypeptides,

(ii) segregating the aligned loop sequences according to loop length,

(iii) for a selected loop and loop length from step (ii), performing positional amino acid frequency analysis to determine the frequencies of amino acids at each loop position,

(iv) for each loop and loop length analyzed in step (iii), identifying at each position a conserved or selected semi-conserved consensus amino acid and other natural-variant amino acids,

(v) for at least one selected loop and loop length, forming:

(1) a library of walk-through mutagenesis sequences expressed by a library of coding sequences that encode, at each loop position, the consensus amino acid, and if the consensus amino acid has a occurrence frequency equal to or less than a selected threshold frequency of at least 50%, a single common target amino acid and any co-produced amino acids, or

(2) a library of natural-variant combinatorial sequences expressed by a library of coding sequences that encode at each loop position, a consensus amino acid and, if the consensus amino acid has a frequency of occurrence equal to or less than a selected threshold frequency of at least 50%, other natural variant amino acids, including semi-conserved amino acids and variable amino acids whose occurrence rate is above a selected minimum threshold occurrence at that position, or their chemical equivalents,

(vi) incorporating the library of coding sequences into framework FN3 coding sequences to form an FN3 expression library, and

(vi) expressing the FN3 polypeptides of the expression library.

The method may be employed in producing various type of walk-through mutagenesis and natural-variant combinatorial libraries, such as those described above.

Also disclosed is a TNF-α binding protein having a Kd binding constant equal to or greater than 0.1 μM and having a sequence selected from SEQ ID NOS: 55-63; a VEGF binding protein having a Kd binding constant equal to or greater than 0.1 μM and having a sequence selected from SEQ ID NOS: 64-67; and an HMGB1 binding protein having a Kd binding constant equal to or greater than 0.1 μM and having a sequence selected from SEQ ID NOS: 67-81.

These and other objects and features of the invention will become more fully apparent when the following detailed description is read in conjunction with the accompanying drawings.

BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a schematic diagram illustrating the method for constructing a fibronectin binding domain libraries using computer assisted genetic database biomining and delineation of beta-scaffold and loop structures.

FIG. 2A is a schematic representation of the FN3 binding domain illustrating the two antiparallel beta-sheets domain. One half is composed of beta strands (ABE) and the other half is composed of (CDFG). The 6 CDR like loops are also indicated: AB, BC, CD, DE, EF, and FG. Loops BC, DE and FG (dotted lines) are present at the N-terminus of the FN3 domain and are arranged to form ligand binding surfaces. The RGD sequence is located in the FG loop.

FIG. 2B shows is a ribbon diagram of the FN3 binding domain illustrating the BC, DE and FG loops (dotted lines) are present at the N-terminus of the FN3 domain and are arranged to form ligand binding surfaces

FIGS. 3A and 3B are (3A) a ribbon diagram of the structural overlay comparisons of the overall loop and beta-strand scaffolds between FN3 module 10 and module 14, and (3B) structural overlay comparisons of the FG loop and F and G beta-strand boundaries between FN3 module 10, 13, and 14 indicating that the position of the loop acceptor sites are well-conserved in the FN3 protein domain architecture even though their respective loops may be quite varied in topology.

FIG. 4 is a schematic representation of the FN3 loop and beta-strand amino acid numberings for the BC, DE and FG loop inserts (light shading).

FIG. 5B shows amino acid sequence alignment of the 1st-16th fibronectin Typw III modules of human fibronectin, the location of three loops: (BC, DE, and FG), and several highly conserved residues through out the fibronectin binding domain including W22, Y/F32, V50, A57, A74, and I/L88 are indicated above the alignment. The conserved amino acid are used as landmarks to aid in alignments the FN3 module and to introduce gaps where necessary.

FIG. 6 is a bar graph showing BC loop length diversity derived from bioinformatics analysis of all FN3 modules. BC loop lengths 11, 14 and 15 are the predominant sizes seen in expressed FN3 sequences.

FIG. 7 is a bar graph showing DE loop length diversity derived from bioinformatics analysis of all FN3 modules. DE loop length 6 is the single most predominant size seen in expressed FN3 sequences.

FIG. 8 is a bar graph showing FG loop length diversity derived from bioinformatics analysis of all FN3 modules. FG loop lengths 8 and 11 are the most predominant sizes seen in expressed FN3 sequences.

FIG. 9 shows sequence diversity of an exemplary loop region in the form of amino acid variability profile (frequency distribution) for BC loop length size 11.

FIG. 10 shows sequence diversity of an exemplary loop region in the form of amino acid variability profile (frequency distribution) for BC loop length size 14.

FIG. 11 shows sequence diversity of an exemplary loop region in the form of amino acid variability profile (frequency distribution) for BC loop length size 15.

FIG. 12 shows sequence diversity of an exemplary loop region in the form of amino acid variability profile (frequency distribution). DE loop length size 6.

FIG. 13 shows amino acid sequence diversity of FG loop region in the form of amino acid variability profile (frequency distribution) for FG loop length size 8.

FIG. 14 shows sequence diversity of FG loop region in the form of amino acid variability profile (frequency distribution) for FG loop length size 11.

FIGS. 15A-15I show the base fixed sequence and variable positions of a BC loop length size 11 and the amino acid matrix showing wild type, the WTM target positions and the extra potential diversity generated from the degenerate WTM codons for each of the selected amino acids K (15A), Q (15B), D (15C), Y (15D), L (15E), P (15F), S (15G), H (15H), and G (15I).

FIGS. 16A-16I show the base fixed sequence and variable positions of an BC loop length size 15 and the amino acid matrix showing wild type, the WTM target positions and the extra potential diversity generated from the degenerate WTM codons for each of the selected amino acids K (16A), Q (16B), D (16C), Y (16D), L (16E), P (16F), S (16G), H (16H), and G (16I).

FIGS. 17A-17I show the base fixed sequence and variable positions of an DE loop length size 6 and matrix showing wild type, the WTM target positions and the extra potential diversity generated from the degenerate WTM codons for each of the selected amino acids K (17A), Q (17B), D (17C), Y (17D), L (17E), P (17F), S (17G), H (17H), and G (17I).

FIGS. 18A-18B show the base fixed sequence and variable positions of an FG loop length size 11 amino acids and matrix showing wild type, the WTM target positions and the extra potential diversity generated from the degenerate WTM codons for each of the selected amino acids K (18A), Q (18B), D (18C), Y (18D), L (18E), P (18F), S (18G), H (18H), and G (18I).

FIG. 19 shows the degenerate and mixed base DNA oligonucleotide sequences for the fixed and variable positions of a BC loop length size 15. The amino acid matrix showing wild type, the WTM target positions and the extra potential diversity generated from the degenerate WTM codons.

FIG. 20 shows construction of the FN3 binding domain library using a combination of overlapping nondegenerate and degenerate oligonucleotides that can be converted to double-stranded nucleic acids using the single overlap extension polymerase chain reaction (SOE-PCR). Eight oligonucleotides are required for the entire gene. Each loop (BC, DE, and FG) is encoded by a separate series of degenerate oligonucleotides.

FIG. 21 shows loop diversity by WTM in the BC, DE, and FG variable positions and total fibronectin binding domain library size when combining the different FN3 loops.

FIG. 22 shows the modular construction using different FN III binding domains of module 14 and Tenascin using their respective set of overlapping non-degenerate and degenerate oligonucleotides. The same BC, DE and FG loop diversity library can be placed into their respective module 14 and Tenascin loop positions.

FIG. 23 shows the construction of the natural-variant amino acid library for loop BC or length 11.

FIGS. 24A-24C show ELISA with three selected anti-TNFα 14FN3 variants, A6, C10, and C5 for binding to TNFα (light bars) and VEGF (dark bars) (24A); binding specificity of anti-TNFα 14 FN3 variants with respect to TNFα (light bars), control (dark bars) and VEGF (white bars) (24B); and sequences of anti-TNFα 14FN3 variants (24C).

FIGS. 25A-25C show binding specificity of anti-VEGF 14 FN3 variants with respect to VEGF (light bars), TNFα (dark bars bars), control (white bars) (25A); sequences of three anti-VEGF 14FN3 variants (25B); and Octet analysis of variant R1D4 (25C).

FIGS. 26A-26C show sequences of anti-HMGB114FN3 sequences (26A); specific binding by anti-HMGB1 variants with respect to HMGB1 (light bars), TNF (dark bars), and control (white bars) (26B), and binding kinetics of HMGB1 variants (26C).

DETAILED DESCRIPTION OF THE INVENTION

I. Definitions

The terms below have the following meanings unless indicated otherwise in the specification:

“Fibronectin Type III (FN3) domain polypeptides” or “FN3 polypeptides” refer to polypeptides having the Fibronectin Type III domain or module discussed in Section II below, where one or more modules will make up a fibronectin-type protein (FN3 protein), such as the sixteen different FN3 modules making up human fibronectin (FN), and the 15 different FN3 modules making up tenascin. Individual FN3 domain polypeptides are referred to by module number and protein name, e.g., the 10th or 14th module of human fibronectin (10/FN or 14/FN) or the 1st module of tenascin (1/tenascin).

A “library” of FN3 polypeptides refers to a collection of FN3 polypeptides having a selected sequence variation or diversity in at least one of the BC, DE, and FG loops of a defined length (see Section II below). The term “library” is also used to refer to the collection of amino acid sequences within a selected BC, DE, or FG loop of a selkected length, and to the collection of coding sequences that encode loop or polypeptide amino acid libraries.

A “universal FN3 library” refers to a FN3 polypeptide library in which amino acid diversity in one or more of the BC, DE or FG loop regions is determined by or reflects the amino acid variants present in a collection of known FN3 sequences.

The term “conserved amino acid residue” or “fixed amino acid” refers to an amino acid residue determined to occur with a frequency that is high, typically at least 50% or more (e.g., at about 60%, 70%, 80%, 90%, 95%, or 100%), for a given residue position. When a given residue is determined to occur at such a high frequency, i.e., above a threshold of about 50%, it may be determined to be conserved and thus represented in the libraries of the invention as a “fixed” or “constant” residue, at least for that amino acid residue position in the loop region being analyzed.

The term “semi-conserved amino acid residue” refers to amino acid residues determined to occur with a frequency that is high, for 2 to 3 residues for a given residue position. When 2-3 residues, preferably 2 residues, that together, are represented at a frequency of about 40% of the time or higher (e.g., 50%, 60%, 70%, 80%, 90% or higher), the residues are determined to be semi-conserved and thus represented in the libraries of the invention as a “semi-fixed” at least for that amino acid residue position in the loop region being analyzed. Typically, an appropriate level of nucleic acid mutagenesis/variability is introduced for a semi-conserved amino acid (codon) position such that the 2 to 3 residues are properly represented. Thus, each of the 2 to 3 residues can be said to be “semi-fixed” for this position. A “selected semi-conserved amino acid residue” is a selected one of the 2 or more semi-conserved amino acid residues, typically, but not necessarily, the residue having the highest occurrence frequency at that position.

The term “variable amino acid residue” refers to amino acid residues determined to occur with a lower frequency (less than 20%) for a given residue position. When many residues appear at a given position, the residue position is determined to be variable and thus represented in the libraries of the invention as variable at least for that amino acid residue position in the loop region being analyzed. Typically, an appropriate level of nucleic acid mutagenesis/variability is introduced for a variable amino acid (codon) position such that an accurate spectrum of residues are properly represented. Of course, it is understood that, if desired, the consequences or variability of any amino acid residue position, i.e., conserved, semi-conserved, or variable, can be represented, explored or altered using, as appropriate, any of the mutagenesis methods disclosed herein, e.g., WTM and natural-variant combinatorial libraries. A lower threshold frequency of occurrence of variable amino acids may be, for example, 5-10% or lower. Below this threshold, variable amino acids may be omitted from the natural-variant amino acirds at that position.

A “consensus” amino acid in a BC, DE, or FG loop of an FN3 polypeptide is a conserved amino acid or a selected one of a semi-conserved amino acids.

“Natural-variant amino acids” include conserved, semi-conserved, and variable amino acid residues observed, in accordance with their occurrence frequencies, at a given position in a selected loop of a selected length. The natural-variant amino acids may be substituted by chemically equivalant amino acids, and may exclude variable amino acid residues below a selected occurrence frequency, e.g., 5-10%, or amino acid residues that are chemically equivalent to other natural-variant amino acids.

A “library of walk through mutagenesis sequences” refers to a library of sequences within a selected FN3 loop and loop length which is expressed by a library of coding sequences that encode, at each loop position, a conserved or selected semi-conserved consensus amino acid and, if the consensus amino acid has an occurrence frequency equal to or less than a selected threshold frequency of at least 50%, a single common target amino acid and any co-produced amino acids. Thus, for each of target amino acid, the library of walk-through mutagenesis sequences within a given loop will contain the target amino acid at all combinations of one to all positions within the loop at which the consensus amino acid has an occurrence frequence equal to or less than the given threshold frequency. If this threshold frequency is set at 100%, each position in the loop will be contain the target amino acid in at least one library member. The tem “library of walk-through mutagenesis sequences” also encompasses a mixture of walk-through mutagenesis libraries, one for each target amino acids, e.g., each of nine different target amino acids.

A “library of natural-variant combinatorial sequences” refers to a library of sequences within a selected FN3 loop and loop length which is expressed by a library of coding sequences that encode at each loop position, a conserved or selected semi-conserved consensus amino acid and, if the consensus amino acid has a frequency of occurrence equal to or less than a selected threshold frequency of at least 50%, other natural variant amino acids, including semi-conserved amino acids and variable amino acids whose occurrence rate is above a selected minimum threshold occurrance at that position, or their chemical equivalents. Thus, for each amino acid position in a selected loop and loop length, the library of natural variant combinatorial sequences will contain the consensus amino acid at that position plus other amino acid variants identified as having at least some minimum frequency at that position, e.g., at least 5-10% frequency, or chemically equivalent amino acids. In addition, natural variants may be substituted or dropped if the coding sequence for that amino acid produces a significant number of co-produced amino acids, via codon degeneracy. The average number of encoded amino acid variants in the loop region will typically between 3-5, e.g., 4, for loops having a loop length of 10 or more, e.g., BC/11, BC/14, BC/15, and may have an average number of substitututions of 6 or more for shorter loops, e.g., DE/6, FG/8 and FG/11, (where variations occurs only at six positions) such that the total diversity of a typical loop region can be maintained in the range preferably about 104-107 for an FN3 BC, DE, or FG loop, and the diversity for two of the three loops can be maintained in the range of about 1012 or less. It will be appreciated from Examples 9 and 10 below that the natural variants at any loop position can be limited to the topmost frequent 3-5 variants, where natural variants that are omitted are those for which the codon change for that amino acid would also a lead to a significant number of co-produced amino acids, where the variant is already represented in the sequence by a chemically equivalent amino acid, or where the frequency of that amino acid in the sequence profile is relativiely low, e.g., 10% or less.

The term “framework region” refers to the art recognized portions of a fibronectin beta-strand scaffold that exist between the more divergent loop regions. Such framework regions are typically referred to the beta strands A through G that collectively provide a scaffold for where the six defined loops can extend to form a ligand contact surface(s). In fibronectin, the seven beta-strands orient themselves as two beta-pleats to form a beta sandwich. The framework region may also include loops AB, CD, and EF between strands A and B, C and D, and E and F. Variable-sequence loops BC, DE, and FG may also be referred to a framework regions in which mutagnesis is introduced to create the desired amino acid diversity within the region.

The term “ligand” or “antigen” refers to compounds which are structurally/chemically similar in terms of their basic composition. Typical ligand classes are proteins (polypeptides), peptides, polysaccharides, polynucleotides, and small molecules. Ligand can be equivalent to “antigens” when recognized by specific antibodies.

The term “loop region” refers to a peptide sequence not assigned to the beta-strand pleats. In the fibronectin binding scaffold there are six loop regions, three of which are known to be involved in binding domains of the scaffold (BC, DE, and FG), and three of which are located on the opposite sided of the polypeptide (AB, EF, and CD). In the present invention, sequence diviersity is built into one or more of the BC, DE, and FG loops, whereas the AB, CD, and EF loops are generally assigned the wildtype amino sequences of the FN3 polypeptide from which other framework regions of the polypeptide are derived.

The term “variability profile” refers to the cataloguing of amino acids and their respective frequency rates of occurrence present at a particular loop position. The loop positions are derived from an aligned fibronectin dataset. At each loop position, ranked amino acid frequencies are added to that position's variability profile until the amino acids' combined frequencies reach a predetermined “high” threshold value.

The term “amino acid” or “amino acid residue” typically refers to an amino acid having its art recognized definition such as an amino acid selected from the group consisting of: alanine (Ala, A); arginine (Arg, R); asparagine (Asn, N); aspartic acid (Asp, D); cysteine (Cys, C); glutamine (Gln, Q); glutamic acid (Glu, E); glycine (Gly, G); histidine (His, H); isoleucine (Ile, I): leucine (Leu, L); lysine (Lys, K); methionine (Met, M); phenylalanine (Phe, F); proline (Pro, P); serine (Ser, S); threonine (Thr, T); tryptophan (Trp, W); tyrosine (Tyr, Y); and valine (Val, V) although modified, synthetic, or rare amino acids may be used as desired.

“Chemically equivalent amino acids” refer to amino acids that have similar steric, charge, and solubility properties. One common scheme groups amino acids in the following way: (1) glycine, having a hydrogen side chain; (2) alanine (Ala, A), valine (Val, V), leucine (Leu, L), and isoleucine (Iso, I), having hydrogen or an unsubstituted aliphatic side chain; (3) serine (Ser, S) and threonine (Thr, T) having an aliphatic side chain bearing a hydroxyl group; (4) aspartic (Asp, D) and glutamic acid (Glu, E), having a carboxyl containing side chain; (5) asparagine (Asn, N) and glutamine (Glu, Q), having an aliphatic side chain terminating in an amide group; (6) arginine (Arg, R) lysine (Lys, L) and histidine (His, H), having an aliphatic side chain terminating in a basic amino group; (7) cysteine (Cys, C) and methionine (Met, M), having a sulfur containing aliphatic side chain; (8) tyrosine (Tyr,Y) and phenylalanine (Phe, F), having an aromatic side chain; and (9) tryptophan (Trp, W), praline (Pro, P), and histidine (His, H), having a heterocyclic side chain.

The term “polynucleotide(s)” refers to nucleic acids such as DNA molecules and RNA molecules and analogs thereof (e.g., DNA or RNA generated using nucleotide analogs or using nucleic acid chemistry). As desired, the polynucleotides may be made synthetically, e.g., using art-recognized nucleic acid chemistry or enzymatically using, e.g., a polymerase, and, if desired, be modified. Typical modifications include methylation, biotinylation, and other art-known modifications. In addition, the nucleic acid molecule can be single-stranded or double-stranded and, where desired, linked to a detectable moiety. Polynucleotide basis and alternative base pairs are given their usual abbreviations herein: Adenosine (A), Guanosine (G), Cytidine (C), Thymidine (T), Uridine (U), puRine (R=A/G), pyrimidine (Y═C/T or C/U), aMino (M=A/C), Keto (K=G/T or G/U), Strong (S=G/C), Weak (W=A/T or A/U), V (A or C or G, but not T), N or X, (any base).

The term “mutagenesis” refers to, unless otherwise specified, any art recognized technique for altering a polynucleotide or polypeptide sequence. Preferred types of mutagenesis include walk-through mutagenesis (WTM), natural-variant combinatorial mutageneis, and beneficial natural-variant combinatorial mutagenesis, although other mutagenesis libraries may be employed, including look-through mutagenesis (LTM), improved look-through mutagenesis (LTM2), WTM using doped nucleotides for achieving codon bias, extended WTM for holding short regions of sequence as constant or fixed within a region of greater diversity, or combinations thereof.

The term “beneficial natural-variant combinational library” refers to a combination library of coding sequences that encode, in two of the three BC, DE, and FG loops of the polypeptide, beneficial mutations determined by screening natural-variant combinatorial libraries containing sequence diversity in those two loops, and natual-variant combinatorial amino acids in the third loop.

II. Overview of the Method and Libraries

Artificial antibody scaffolds that bind specific ligands are becoming legitimate alternatives to antibodies. Antibodies have been useful as both diagnostic and therapeutic tools. However, obtaining specific antibodies recognizing certain ligands have been difficult. Current antibody libraries are biased against certain antigen classes only after immunological exposure. Therefore it is frequently necessary to immunize a host animal with a particular antigen before recovery of specific antibodies can occur. Furthermore, these in vivo derived antibody libraries usually do not have candidates that recognize self antigens. These are usually lost in a expressed human library because self reactive antibodies are removed by the donor's immune system by negative selection. Furthermore, antibodies are difficult and expensive to produce requiring special cell fermentation reactors and purification procedures.

The limitations of antibodies has spurred the development of alternative binding proteins based on immunoglobulin like folds or other protein topologies. These non-antibody scaffold share the general quality of having a structurally stable framework core that is tolerant to multiple substitutions in other parts of the protein.

The present invention provides a universal fibronectin binding domain library that is more comprehensive and engineered to have artificial diversity in the ligand binding loops. By creating artificial diversity, the library size can be controlled so that they can be readily screened using, for example, high throughput methods to obtain new therapeutics. The universal fibronectin library can be screened using positive physical clone selection by FACS, phage panning or selective ligand retention. These in vitro screens bypass the standard and tedious methodology inherent in generating an antibody hybridoma library and supernatant screening.

Furthermore, the universal fibronectin library has the potential to recognize any antigen as the constituent amino acids in the binding loop are created by in vitro diversity techniques. This produces the significant advantages of the library controlling diversity size and the capacity to recognize self antigens. Still further, the fibronectin binding domain library can be propagated and re-screened to discover additional fibronectin binding modules against other desired targets.

IIA. Fibronectin Review: (FN)

Fibronectin Type III (FN3) proteins refer to a group of proteins composed of momomeric subunits having Fibronectin Type III (FN3) structure or motif made up of seven β-strands with three connecting loops. %-strands A, B, and E form one half β-sandwich and β-strands C,D, F, and G form the other half (see FIGS. 2a and 2b), and having molecular weights of about 94 amino acids and molecular weights of about 10 Kda. The overall fold of the FN3 domain is closely related to that of the immunoglobulin domains, and the three loops near the N-terminus of FN3, named BC, DE, and FG (as illustrated in FIG. 2b), can be considered structurally analogous to the antibody variable heavy (VH) domain complementarity-determining regions, CDR1, CDR2, and CDR3, respectively. Table 1 below shows several FN3 proteins, and the number of different FN3 modules or domains associated with each protein. Thus, fibronectin itself is composed of 16 different modules or domains whose amino acid sequences are shown in aligned form in FIG. 5. A given module of an FN3 protein is identified by module number and protein name, for example, the 14th FN3 module of human fibronectin (14/FN or 14/FN3), the 10th FN3 module of human fibronectin (10/FN or 10/FN3), the 1st FN3 module of tenascin (1/tenascin), and so forth.

TABLE 1
Representative FN3 protins and their modules
FN3 ProteinFN3 modules
Angiopoietin 1 receptor.3
Contactin protein4
Cytokine receptor common β chain2
Down syndrome cell adhesion protein6
Drosophila Sevenless protein7
Erythropoietin receptor1
Fibronectin16
Growth hormone receptor1
Insulin receptor2
Insulin-like growth factor I receptor3
Interferon-γ receptor β chain.2
Interleukin-12 β chain1
Interleukin-2 receptor β chain1
Leptin receptor (LEP-R)3
Leukemia inhibitory factor receptor (LIF-R)6
Leukocyte common antigen2
Neural cell adhesion protein L14
Prolactin receptor2
Tenascin protein15
Thrombopoietin receptor.2
Tyrosine-protein kinase receptor Tie-13

Fibronectin itself is involved in many cellular processes, including tissue repair, embryogenesis, blood clotting, by serving as a general cell adhesion molecule anchoring cells to integrin, collagen or other proteoglycan substrates. In addition, fibronectin also can serve to organize the extracellular matrix binding to different components, such as heparin, to membrane-bound receptors on cell surfaces. The amino acid sequence of fibronectin reveals three types of internally homologous repeats or modules separated by (usually) short connecting sequences. There are 12 type I, 2 type II and 16 type III modules, and referred to as FN I, FNII and FNIII respectively. Each FN module constitutes an independently folded unit, often referred to as a domain. As noted above, modules homologous to those in fibronectin are also found in other proteins, especially the FN3 motif which is one of the most ubiquitous of all modules, being found in extracellular receptor kinases, phosphatases, tenascin and others. Since its discovery, this FN3 domain has been found in many animal proteins and is estimated to occur in 2% of the proteins sequenced to date. Within fibronectin itself, there are sixteen FN3 domains and have remarkably similar tertiary structures. Interestingly, while FN3 conformation are highly conserved, the similarity between different modules of the same type within a given fibronectin protein is quite low typically less than 20%. In contrast, the amino acid sequence homology for the same FN-III modules across multiple species is notably higher, approximately 80%-90%.

Fibronectin modules fold independently and thus can exist in isolation from their neighbors. The three dimensional structures of several examples of each type of fibronectin module have been determined. As expected from the well-known relationship between amino acid sequence and 3D structure, modules of the same type have similar folds. All three types of module are composed almost exclusively of antiparallel 1 sheets and turns, with little or no alpha helix. In F3 modules, the top sheet contains four antiparallel beta strands and the bottom sheet is three-stranded. Disulphide bridges do not stabilize FN3 structure. Instead, this occurs solely through hydrophobic interactions in the module core.

IIB. Identifying and Selecting Fibronectin Scaffold and Loop Components Using Bioinformatics

The first step in building a universal fibronectin library of the invention is selecting sequences that meet certain predetermined criteria. PFAM, ProSite and similar databases were searched for sequences containing FN3 domains (FIG. 1). These electronic databases (box 30 in FIG. 1) contain catalogued expressed fibronectin and fibronectin-like protein sequences and can be queried for those FN3 module and similar sequences (using the BLAST search algorithm, box 32). The FN3 module sequences can then be grouped to predefined criteria such as module subclasses, sequence similarity or originating organism(s). The framework sequence selection can also be performed for scaffold proteins such as FN I, FN II or ankyrin and other proteins (box 34). Example 1 provides additional; details of the method.

Candidate FN3 β-strand scaffold framework sequences are then delineated whereupon the intervening loop regions and constituent amino acids are then identified (box 36). This then determines the length of the existing loops, the amino acid profiles for each loop length and, hence the physical size and amino acid diversity that can be accommodated within these frameworks. With reference to FIG. 1, once the loop are identified, sequences within each loop are aligned, at 38, 40, and 42, and the aligned sequences are then split into groups according to loop length, as at 44. The distribution of loop lengths for the BC, DE, and FG loops are shown at FIGS. 6-8, respectively. Using this information, the most common loop sizes are selected, at 46. In a general embodiment of the invention that will be iluustrated below, the selected loop lengths are BC/11, BC/14, BC15, DE/6, FG/8, and FG11, as shown at 48 for BC/11. For each β-strand, one can determine the preferred loop acceptor sites in the frameworks based on both comparative structural and sequence analysis (FIGS. 3 and 5). For example, FIG. 3a shows a structural overlay comparison of the overall loop and β strand scaffolds between the fibronectin 10/FN3 and 14/FN3. FIG. 3b shows the structural overlay comparison of the FG loop and F and G β strand boundaries between FN3 modules 10, 13 and 14. In identifying precise loop positions, the above step greatly minimizes necessary diversity loop mutations that would not result in functional ligand binding specificity. Additional details can be found in Example 2.

Once loop lengths are selected, a positional amino acid frequency analysis is performed at each loop position, to determine the frequency of occurrence, in a set of native FN3 modules, e.g., all human FN3 modules, at box 50. This method includes a frequency analysis and the generation of the corresponding variability profiles (VP) of existing loop sequences (See Example 2 and FIGS. 9-14). High frequency (e.g. >50%) positions are considered conserved or fixed. Moderately high frequency or “semi-conserved” amino acids or (when 2 or 3 are combined account for >40%) are chosen as “wildtype” at other positions. These wildtype amino acids are then systematically altered using, mutagenesis, e.g. walk-through mutagenesis (WTM), to generate the universal loop library (see Example 3). “Variable” positions are those where typically, no one amino acid accounts for more than 20% of the represented set.

The choice of candidate frameworks based on the criteria of the invention dictates both the loop sizes to be introduced and the initial amino acid sequence diversity.

A loop variability profile analysis of the FN3 databases allows identification of loop amino acid residue positions that fall within three categories, e.g., 1) positions that should be conserved or “fixed,” and 2) semi-conserved and/or 3) variable positions that are suitable for diversity generation. A variability profile analysis is performed and a threshold frequency is used to identify the most favorable sequences to be used in designating the overall loop diversity (box 52).

The conserved or a selected semi-conserved sequence (typically the most frequent amino acid in the semi-conserved residues) is considered the “wild type” or “cojnsensus” residue in the loop sequence. Surprisingly, this “consensus” or “frequency” approach identifies those particular amino acids under high selective pressure. Accordingly, these residue positions are typically fixed, with diversity being introduced into remaining amino acid positions (taking into account the identified preference for certain amino acids to be present at these positions). As will be seen below, the threshold for occurrence frequency at which amino acid variation will be introduced can vary between selected levels as low as 40%, preferably 50% to as high as 100%. At the 100% threshold frequency, WTM amino acids can be introduced at all positions of the loop, and the only constraints on natural-variant amino acids will be the total number of variants and whether chemical equivalents are available.

When designing the diversity for any of the above-mentioned loops, modified amino acid residues, for example, residues outside the traditional 20 amino acids used in most polypeptides, e.g., homocysteine, can be incorporated into the loops as desired. This is carried out using art recognized techniques which typically introduce stop codons into the polynucleotide where the modified amino acid residue is desired. The technique then provides a modified tRNA linked to the modified amino acid to be incorporated (a so-called suppressor tRNA of, e.g., the stop codon amber, opal, or ochre) into the polypeptide (see, e.g., Köhrer et al., Import of amber and ochre suppressors tRNAs into mammalian cells: A general approach to site-specific insertion of amino acid analogues into proteins, PNAS, 98, 14310-14315 (2001)).

The bioinformatic analysis focuses on FN3 modules genes for descriptive purposes, but it will be understood that genes for other FN modules and other scaffold protein are similarly evaluated.

IIC. Computer-Assisted Universal Fibronectin Library Construction

The universal fibronectin loop libraries of the invention and their construction is conducted with the benefit of sequence and structural information such that the potential for generating improved fibronectin binding domains is increased. Structural molecular replacement modeling information can also be used to guide the selection of amino acid diversity to be introduced into the defined loop regions. Still further, actual results obtained with the fibronectin binding domains of the invention can guide the selection (or exclusion), e.g., affinity maturation, of subsequent fibronectin binding domains to be made and screened in an iterative manner.

In the preferred embodiment, in silico modeling is used to eliminate the production of any fibronectin binding domains predicted to have poor or undesired structure and/or function. In this way, the number of fibronectin binding domains to be produced can be sharply reduced thereby increasing signal-to-noise in subsequent screening assays. In another particular embodiment, the in silico modeling is continually updated with additional modeling information, from any relevant source, e.g., from gene and protein sequence and three-dimensional databases and/or results from previously tested fibronectin binding domains, so that the in silico database becomes more precise in its predictive ability (FIG. 1).

In yet another embodiment, the in silico database is provided with the assay results, e.g., binding affinity/avidity of previously tested fibronectin binding domains and categorizes the fibronectin binding domains, based on the assay criterion or criteria, as responders or nonresponders, e.g., as fibronectin binding domain molecules that bind well or not so well. In this way, the affinity maturation of the invention can equate a range of functional responses with particular sequence and structural information and use such information to guide the production of future fibronectin binding domains to be tested. The method is especially suitable for screening fibronectin binding domains for a particular binding affinity to a target ligand using, e.g., a Biacore assay.

Accordingly, mutagenesis of noncontiguous residues within a loop region can be desirable if it is known, e.g., through in silico modeling, that certain residues in the region will not participate in the desired function. The coordinate structure and spatial interrelationship between the defined regions, e.g., the functional amino acid residues in the defined regions of the fibronectin binding domain, e.g., the diversity that has been introduced, can be considered and modeled. Such modeling criteria include, e.g., amino acid residue side group chemistry, atom distances, crystallography data, etc. Accordingly, the number fibronectin binding domains to be produced can be intelligently minimized.

In a preferred embodiment, one or more of the above steps are computer-assisted. In a particular embodiment, the computer assisted step comprises, e.g., mining the NCBI, Genbank, PFAM, and ProSite databases and, optionally, cross-referencing the results against PDB structural database, whereby certain criteria of the invention are determined and used to design the desired loop diversity (FIG. 1). The method is also amenable to being carried out, in part or in whole, by a device, e.g., a computer driven device. For example, database mining fibronectin module sequence selection, diversity design, oligonucleotide synthesis, PCR-mediated assembly of the foregoing, and expression and selection of candidate fibronectin binding domains that bind a given target, can be carried out in part or entirely, by interlaced devices. In addition, instructions for carrying out the method, in part or in whole, can be conferred to a medium suitable for use in an electronic device for carrying out the instructions. In sum, the methods of the invention are amendable to a high throughput approach comprising software (e.g., computer-readable instructions) and hardware (e.g., computers, robotics, and chips).

IID. Universal Walk-Through Mutagenesis Libraries

In one general aspect, the present invention includes a walk-through mutagenesis (WTM) library of fibronectin Type 3 domain polypeptides useful in screening for the presence of one or more polypeptides having a selected binding or enzymatic activity. The library polypeptides include (a) regions A, AB, B, C, CD, D, E, EF, F, and G having wildtype amino acid sequences of a selected native fibronectin Type 3 polypeptide or polypeptides, and (b) loop regions BC, DE, and FG having one or more selected lengths. At least one selected loop region of a selected length contains a library of WTM sequences encoded by a library of coding sequences that encode, at each loop position, a conserved or selected semi-conserved consensus amino acid and, if the consensus amino acid has a occurrence frequency equal to or less than a selected threshold frequency of at least 50%, a single common target amino acid and any co-produced amino acids (amino acids produced by the coding sequences at a given position as a result of codon degeneracy).

In constructing a WTM library within a given loop of a given loop length, the variability profile is used to define a sequence of fixed and “variable” positions, i.e., positions at which a target WTM amino acid can be introduced. The number of fixed positions (no substitutions made) will depend on the selected threshold frequency for the consensus amino acid at each position. Example 3 illustrates the design of WTM loop sequences for each of nine representative target amino acids for FG/11. Conserved or semi-conserved residues were observed for N79, G79, G/S 81 and S84 (two consensus amino acids were placed at position 81, reflecting the high frequency of both G (60%) and S (31%). These four positions (and the two terminal positions S and K) were fixed, and the WTM target amino acid was introduced at each of the other positions in the sequence. For the target amino acid lysine (K), the example shows the various WTM sequenes containing a K at from one up to all substitution positions in the loop. Table 5 is the example shows the substitution sequences for all six. BC/11, BC/14, BC/15, DE/6, FG/8, and FG11 loops selected.

In an alternative embodiment, the threshold for consensus sequences is set at 100%, so that all residue positions will be selected for introduction of a WTM target amino acid. This approach in illustrated in Example 5.

Once the WTM loop sequences are selected, a library of coding-sequence oligonucleotides encoding all of the identified WTM sequences is constructed, making codon substitutions as shown that are effective to preserve the existing consensus amino acid, but also encode the selected target amino acid, and any other co-product amino acids encoded by degenerate codons, as detailed in Examples 4 and 5 for the two different WTM libraries (with and without consensus threshold constraints). FIGS. 15A-15I show the base fixed sequence and variable positions of a BC loop length size 11 and the amino acid matrix showing wild type, the WTM target positions and the extra potential diversity generated from the degenerate WTM codons for each of the selected amino acids K (15A), Q (15B), D (15C), Y (15D), L (15E), P (15F), S (15G), H (15H), and G (15I). FIGS. 16A-16I, 17A-17I, 18A-18I provide similar tables for BC loop length 15 (BC/15), DE loop length 6 (DE/6), and FG loop length 11 (FG/11), respectiviely.

The library of coding sequences for the WTM loops is added to the framework sequences, as detailed in the section below and in Example 6, to construct the library of coding sequences for the WTM polypeptide libraries. In one preferred embodiment, the coding library includes coding sequences for each of the six different loop and loop lengths, and for each of the nine selected “representative” WTM target amino acids (see below).

The library of polypeptides may be encoded by an expression library format that includes a ribosome display library, a polysome display library, a phage display library, a bacterial expression library, or a yeast display library.

The libraries may be used in a method of identifying a polypeptide having a desired binding affinity, in which the natural-variant combinatorial library are screened to select for an fibronectin binding domain having a desired binding affinity. The efficiency of a 14/FN WTM library constructed in accordance with the present invention, for selecting FN3 polypeptides having a high binding affinity for a selected antigen, TNFα, can be appreciated from Example 8, with respect to FIGS. 24A-24C.

IIE. Universal Natural-Variant combinatorial Mutagenesis Libraries

In another general aspect, the invention includes a natural-variant combinatorial library of fibronectin Type 3 domain polypeptides useful in screening for the presence of one or more polypeptides having a selected binding or enzymatic activity. The library polypeptides include (a) regions A, AB, B, C, CD, D, E, EF, F, and G having wildtype amino acid sequences of a selected native fibronectin Type 3 polypeptide or polypeptides, and (b) loop regions BC, DE, and FG having selected lengths. At least one selected loop region of a selected length contains a library of natural-variant combinatorial sequences expressed by a library of coding sequences that encode at each loop position, a conserved or selected semi-conserved consensus amino acid and, if the consensus amino acid has a frequency of occurrence equal to or less than a selected threshold frequency of at least 50%, other natural variant amino acids, including semi-conserved amino acids and variable amino acids whose occurrence rate is above a selected minimum threshold occurrence at that position, or their chemical equivalents.

In constructing a natural-variant combinatorial library for a given loop and loop length, the variability profile is used to define a sequence of fixed and “variable” positions, i.e., positions at which amino acid variations can be introduced. As in the WTM libraries, the number of fixed positions (no substitutions made) will depend on the selected threshold frequency for the consensus amino acid at each position. If, for example, the selected frequency threshold was set at about 60%, the conserved or semi-conserved residues for FG/11 are N79, G79, G/S 81 and S84 (from Example 3) and natural-variant substitutions would not be made at these positions. Conversely, if the threshold frequency is set at 100%, all positions would be considered open to variation, recognizing that a single amino acid with a frequency of 100% at a loop position would not be substituted, and a position that had one very dominant amino acid, e.g., with a frequency of 90%, might be substituted only if the low-frequency variant(s) were chemically dissimilar to the dominant amino acid.

From the amino acid profile for a given loop and loop length, and knowing which of the positions will be held fixed and which will be admit variations, the amino acid substitutions at each variable position can be selected. In general, the number of variations that are selected (including co-produced amiono acids) will depend on the number of variable substitution positions in the loop and the average number of variations per substituted loop position. Ideally, the number of variations selected will be such as to maintain the diversity of the loop in the range 105-107 Isequences, allowing a library of two variable-sequence loops in the range of about 1012. As will be seen from Example 9 below, an FG/11 loop having amino acid variations at 10 of the 11 positions, will have an average of about 4 amino acid variations/position, whereas a shorter loop, such as DE/6 will admit up to 6 or more variations per position. Of course, if natural-variant substitutions are introduced into a single loop only, many more variations per position can be accommodated.

The particular natural variant amino acids that are selected for each position will generally include the amino acids having the highest frequencies, while limited the number of co-produced amino acids, and secondarily, preserving chemical diversity at each site. Thus, if the codon change for one variant amino acid would produce several co-produced amino acids, that variant would likely be omitted, and/or a chemically equivalent amino acid would be sought. Similarly, if one natural variant is chemically equivalent to another one, one of the two could be omitted. In summary, the natural-variant loop sequences are constructed to include the highest frequency natural variants, while minimizing co-produced amino acids and minimizing redundancy in amino acid side chain properties, and to limit the total diversity of the loop and loop length if necessary, i.e., where sequence variation is introduced into more than one loop. The application of these rules can be seen in the exemplary variant substitutions in a BC/11 loop (FIG. 25B) from the BC/11 loop profile given in FIG. 9. Similarly, the application of the rules can be seen in the exemplary variant substitutions in a DE/6 loop (FIG. 25B), from the DE loop profile in FIG. 12.

Once the natural-variant loop sequences are selected, a library of coding-sequence oligonucleotides encoding all of the identified natural-variant sequences is constructed, making codon substitutions that are effective to preserve the existing consensus amino acid, and encode the selected variant amino acids, including variants encoded encoded by degenerate codons.

The library of coding sequences for the natural-variants loops is added to the framework sequences, as detailed in the section below and in Example 6, to construct the library of coding sequences for the natural-variant polypeptide libraries. In one preferred embodiment, the coding library includes coding sequences for a pair of BC/DE, BC/FG or DE/FG loops, where each loop in the pair has one selected length, egg., BC/11 and DE/6. Such a two-loop library is described in Examples 9 and 10 below. After selecting high-affinity binding (or enzymatic) polypeptides from this library, a second “beneficial” library can be constructed that includes the beneficial mutations contained in one or both of original two-loop natural-variation library, and natural-variant amino acids in the third loop, i.e., the previously fixed-sequence loop.

Natural-variant combinatorial library may have the following sequences for the indicated loops and loop lengths: (a) BC loop length of 11, and the amino acid sequence identified by SEQ ID NOS: 43 or 49; (b) BC loop length of 14, and the amino acid sequence identified by SEQ ID. NOS: 44 or 50; (c) BC loop length of 15, and the amino acid sequence identified by SEQ ID. NOS: 45 or 51; (d) DE loop length of 6, and the amino acid sequence identified by SEQ ID. NOS: 46 or 52; (e) FG loop length of 8, and the amino acid sequence identified by SEQ ID. NOS: 47, for the first N-terminal six amino acids, or SEQ ID NO:53, and (f) FG loop length of 11, and the amino acid sequence identified by SEQ ID. NO: 48, for the first N-terminal nine amino acids, or SEQ ID NO:54.

The library of polypeptides may be encoded by an expression library that has the format of a ribosome display library, a polysome display library, a phage display library, a bacterial expression library, or a yeast display library.

The libraries may be used in a method of identifying a polypeptide having a desired binding affinity, in which the natural-variant combinatorial library are screened to select for an fibronectin binding domain having a desired binding affinity. The efficiency of a 14/FN natural-variant combinatorial library constructed in accordance with the present invention, for selecting FN3 polypeptides having a high binding affibity for each of two selected antigens, VEGF and HMGB1, can be appreciated from Example 9, with respect to FIGS. 25A-25C, and Example 10, with respect to FIGS. 26A-26C.

IIF. Synthesizing Universal Fibronectin Binding Domain Libraries

In one embodiment, the universal fibronectin binding domains of the invention are generated for screening by synthesizing individual oligonucleotides that encode the defined region of the polypeptide and have no more than one codon for the predetermined amino acid. This is accomplished by incorporating, at each codon position within the oligonucleotide either the codon required for synthesis of the wild-type polypeptide or a codon for the predetermined amino acid and is referred to as look-through mutagenesis (LTM) (see, e.g., U.S. Patent Publication No. 20050136428).

In another embodiment, when diversity at multiple amino acid positions is required, walk-through mutagenesis (WTM) can be used (see e.g., U.S. Pat. Nos. 6,649,340; 5,830,650; and 5,798,208; and U.S. Patent Publication No. 20050136428. WTM allows for multiple mutations to be made with a minimum number of oligonucleotides. The oligonucleotides can be produced individually, in batches, using, e.g., doping techniques, and then mixed or pooled as desired.

The mixture of oligonucleotides for generation of the library can be synthesized readily by known methods for DNA synthesis. The preferred method involves use of solid phase beta-cyanoethyl phosphoramidite chemistry (e.g., see U.S. Pat. No. 4,725,677). For convenience, an instrument for automated DNA synthesis can be used containing specified reagent vessels of nucleotides. The polynucleotides may also be synthesized to contain restriction sites or primer hybridization sites to facilitate the introduction or assembly of the polynucleotides representing, e.g., a defined region, into a larger gene context.

The synthesized polynucleotides can be inserted into a larger gene context, e.g., a single scaffold domain using standard genetic engineering techniques. For example, the polynucleotides can be made to contain flanking recognition sites for restriction enzymes (e.g., see U.S. Pat. No. 4,888,286). The recognition sites can be designed to correspond to recognition sites that either exist naturally or are introduced in the gene proximate to the DNA encoding the region. After conversion into double stranded form, the polynucleotides are ligated into the gene or gene vector by standard techniques. By means of an appropriate vector (including, e.g., phage vectors, plasmids) the genes can be introduced into a cell-free extract, phage, prokaryotic cell, or eukaryotic cell suitable for expression of the fibronectin binding domain molecules.

Alternatively, partially overlapping polynucleotides, typically about 20-60 nucleotides in length, are designed. The internal polynucleotides are then annealed to their complementary partner to give a double-stranded DNA molecule with single-stranded extensions useful for further annealing. The annealed pairs can then be mixed together, extended, and ligated to form full-length double-stranded molecules using SOE-PCR (see, e.g., Example 3). Convenient restriction sites can be designed near the ends of the synthetic gene for cloning into a suitable vector. The full-length molecules can then be ligated into a suitable vector.

When partially overlapping polynucleotides are used in the gene assembly, a set of degenerate nucleotides can also be directly incorporated in place of one of the polynucleotides. The appropriate complementary strand is synthesized during the extension reaction from a partially complementary polynucleotide from the other strand by enzymatic extension with a polymerase. Incorporation of the degenerate polynucleotides at the stage of synthesis also simplifies cloning where more than one domain or defined region of a gene is mutagenized or engineered to have diversity.

In another approach, the fibronectin binding domain is present on a single stranded plasmid. For example, the gene can be cloned into a phage vector or a vector with a filamentous phage origin of replication that allows propagation of single-stranded molecules with the use of a helper phage. The single-stranded template can be annealed with a set of degenerate polynucleotides representing the desired mutations and elongated and ligated, thus incorporating each analog strand into a population of molecules that can be introduced into an appropriate host (see, e.g., Sayers, J. R. et al., Nucleic Acids Res. 16: 791-802 (1988)). This approach can circumvent multiple cloning steps where multiple domains are selected for mutagenesis.

Polymerase chain reaction (PCR) methodology can also be used to incorporate polynucleotides into a gene, for example, loop diversity into β-strand framework regions. For example, the polynucleotides themselves can be used as primers for extension. In this approach, polynucleotides encoding the mutagenic cassettes corresponding to the defined region (or portion thereof) are complementary to each other, at least in part, and can be extended to form a large gene cassette (e.g., a fibronectin binding domain) using a polymerase, e.g., using PCR amplification.

The size of the library will vary depending upon the loop length and the amount of sequence diversity which needs to be represented using, e.g., WTM or LTM. Preferably, the library will be designed to contain less than 1015, 1014, 1013, 1012, 1011, 1010, 109, 108, 107, and more preferably, 106 fibronectin binding domain.

The description above has centered on representing fibronectin binding domain diversity by altering the polynucleotide that encodes the corresponding polypeptide. It is understood, however, that the scope of the invention also encompasses methods of representing the fibronectin binding domain diversity disclosed herein by direct synthesis of the desired polypeptide regions using protein chemistry. In carrying out this approach, the resultant polypeptides still incorporate the features of the invention except that the use of a polynucleotide intermediate can be eliminated.

For the libraries described above, whether in the form of polynucleotides and/or corresponding polypeptides, it is understood that the libraries may be also attached to a solid support, such as a microchip, and preferably arrayed, using art recognized techniques.

The method of this invention is especially useful for modifying candidate fibronectin binding domain molecules by way of affinity maturation. Alterations can be introduced into the loops and/or into the β-strand framework (constant) region of an fibronectin binding domain. Modification of the loop regions can produce fibronectin binding domains with better ligand binding properties, and, if desired, catalytic properties. Modification of the β-strand framework region can also lead to the improvement of chemo-physical properties, such as solubility or stability, which are especially useful, for example, in commercial production, bioavailabilty, and affinity for the ligand. Typically, the mutagenesis will target the loop region(s) of the fibronectin binding domain, i.e., the structure responsible for ligand-binding activity which can be made up of the three loop regions. In a preferred embodiment, an identified candidate binding molecule is subjected to affinity maturation to increase the affinity/avidity of the binding molecule to a target ligand.

IIG. Expression and Screening Systems

Libraries of polynucleotides generated by any of the above techniques or other suitable techniques can be expressed and screened to identify fibronectin binding domain molecules having desired structure and/or activity. Expression of the fibronectin binding domain molecules can be carried out using cell-free extracts (and e.g., ribosome display), phage display, prokaryotic cells, or eukaryotic cells (e.g., yeast display).

In one embodiment, the polynucleotides are engineered to serve as templates that can be expressed in a cell free extract. Vectors and extracts as described, for example in U.S. Pat. Nos. 5,324,637; 5,492,817; 5,665,563, can be used and many are commercially available. Ribosome display and other cell-free techniques for linking a polynucleotide (i.e., a genotype) to a polypeptide (i.e., a phenotype) can be used, e.g., Profusion™ (see, e.g., U.S. Pat. Nos. 6,348,315; 6,261,804; 6,258,558; and 6,214,553).

Alternatively, the polynucleotides of the invention can be expressed in a convenient E. coli expression system, such as that described by Pluckthun and Skerra. (Pluckthun, A. and Skerra, A., Meth. Enzymol. 178: 476-515 (1989); Skerra, A. et al., Biotechnology 9: 273-278 (1991)). The mutant proteins can be expressed for secretion in the medium and/or in the cytoplasm of the bacteria, as described by M. Better and A. Horwitz, Meth. Enzymol. 178: 476 (1989). In one embodiment, the fibronectin binding domain are attached to the 3′ end of a sequence encoding a signal sequence, such as the ompA, phoA or pelB signal sequence (Lei, S. P. et al., J. Bacteriol. 169: 4379 (1987)). These gene fusions are assembled in a dicistronic construct, so that they can be expressed from a single vector, and secreted into the periplasmic space of E. coli where they will refold and can be recovered in active form. (Skerra, A. et al., Biotechnology 9: 273-278 (1991)).

In another embodiment, the fibronectin binding domain sequences are expressed on the membrane surface of a prokaryote, e.g., E. coli, using a secretion signal and lipidation moiety as described, e.g., in US20040072740A1; US20030100023A1; and US20030036092A1.

In still another embodiment, the polynucleotides can be expressed in eukaryotic cells such as yeast using, for example, yeast display as described, e.g., in U.S. Pat. Nos. 6,423,538; 6,331,391; and 6,300,065. In this approach, the fibronectin binding domain molecules of the library are fused to a polypeptide that is expressed and displayed on the surface of the yeast.

Higher eukaryotic cells for expression of the fibronectin binding domain molecules of the invention can also be used, such as mammalian cells, for example myeloma cells (e.g., NS/0 cells), hybridoma cells, or Chinese hamster ovary (CHO) cells. Typically, the fibronectin binding domain molecules when expressed in mammalian cells are designed to be expressed into the culture medium, or expressed on the surface of such a cell. The fibronectin binding domain can be produced, for example, as single individual module or as multimeric chains comprising dimers, trimers, that can be composed of the same module or of different module types. (10FN3-10FN3: homodimer, 10FN3-5FN3: heterodimer)

The screening of the expressed fibronectin binding domain (or fibronectin binding domain produced by direct synthesis) can be done by any appropriate means. For example, binding activity can be evaluated by standard immunoassay and/or affinity chromatography. Screening of the fibronectin binding domain of the invention for catalytic function, e.g., proteolytic function can be accomplished using a standard hemoglobin plaque assay as described, for example, in U.S. Pat. No. 5,798,208. Determining the ability of candidate fibronectin binding domain to bind therapeutic targets can be assayed in vitro using, e.g., a Biacore instrument, which measures binding rates of a fibronectin binding domain to a given target or ligand. In vivo assays can be conducted using any of a number of animal models and then subsequently tested, as appropriate, in humans.

IIH. Analysis and Screening of FN3 WTM Libraries for Catalytic Function.

FN3 WTM libraries can also be used to screen for FN3 proteins that possess catalytic activity. The study of proteins has revealed that certain amino acids play a crucial role in their structure and function. For example, it appears that only a discrete number of amino acids participate in the catalytic event of an enzyme. Serine proteases are a family of enzymes present in virtually all organisms, which have evolved a structurally similar catalytic site characterized by the combined presence of serine, histidine and aspartic acid. These amino acids form a catalytic triad which, possibly along with other determinants, stabilizes the transition state of the substrate. The functional role of this catalytic triad has been confirmed by individual and by multiple substitutions of serine, histidine and aspartic acid by site-directed mutagenesis of serine proteases and the importance of the interplay between these amino acid residues in catalysis is now well established. These same three amino acids are involved in the enzymatic mechanism of certain lipases as well. FIG. 26 is a schematic depiction of a “walk-through” mutagenesis of a an active site that utilizes the serine, histidine and aspartic acid triad.

Similarly, a large number of other types of enzymes are characterized by the peculiar conformation of their catalytic site and the presence of certain kinds of amino acid residues in the site that are primarily responsible for the catalytic event. For an extensive review, see Enzyme Structure and Mechanism, 1985, by A. Fersht, Freeman Ed., New York.

Though it is clear that certain amino acids are critical to the mechanism of catalysis, it is difficult, if not impossible, to predict which position (or positions) an amino acid must occupy to produce a functional site such as a catalytic site. Unfortunately, the complex spatial configuration of amino acid side chains in proteins and the interrelationship of different side chains in the catalytic pocket of enzymes are insufficiently understood to allow for such predictions. Selective site-directed mutagenesis and saturation mutagenesis are of limited utility for the study of protein structure and function in view of the enormous number of possible variations in complex proteins.

Protein libraries generated by any of the above WTM/CBM/LTM techniques or other suitable techniques can be screened to identify variants of desired structure or activity.

By comparing the properties of a wild-type protein and the variants generated, it is possible to identify individual amino acids or domains of amino acids that confer binding and/or catalytic activity. Usually, the region studied will be a functional domain of the protein such as a binding domain. For example, the region can be the external BC, DE and FG loop binding regions of FN3 domain. The screening can be done by any appropriate means. For example, catalytic activity can be ascertained by suitable assays for substrate conversion and binding activity can be evaluated by standard immunoassay and/or affinity chromatography.

From the chemical properties of the side chains, it appears that only a selected number of natural amino acids preferentially participate in a catalytic event. These amino acids belong to the group of polar and neutral amino acids such as Ser, Thr, Asn, Gln, Tyr, and Cys, the group of charged amino acids, Asp and Glu, Lys and Arg, and especially the amino acid His. Typical polar and neutral side chains are those of Cys, Ser, Thr, Asn, Gln and Tyr. Gly is also considered to be a borderline member of this group. Ser and Thr play an important role in forming hydrogen-bonds. Thr has an additional asymmetry at the beta carbon, therefore only one of the stereoisomers is used. The acid amide Gln and Asn can also form hydrogen bonds, the amido groups functioning as hydrogen donors and the carbonyl groups functioning as acceptors. Gln has one more CH2 group than Asn which renders the polar group more flexible and reduces its interaction with the main chain. Tyr has a very polar hydroxyl group (phenolic OH) that can dissociate at high pH values. Tyr behaves somewhat like a charged side chain; its hydrogen bonds are rather strong.

Histidine (His) has a heterocyclic aromatic side chain with a pK value of 6.0. In the physiological pH range, its imidazole ring can be either uncharged or charged, after taking up a hydrogen ion from the solution. Since these two states are readily available, His is quite suitable for catalyzing chemical reactions. It is found in most of the active centers of enzymes.

Asp and Glu are negatively charged at physiological pH. Because of their short side chain, the carboxyl group of Asp is rather rigid with respect to the main chain. This may be the reason why the carboxyl group in many catalytic sites is provided by Asp and not by Glu. Charged acids are generally found at the surface of a protein.

Therefore, several different regions or loops of a FN3 protein domain can be mutagenized simultaneously. The same or a different amino acid can be “walked-through” each loop region. This enables the evaluation of amino acid substitutions in conformationally related regions such as the regions which, upon folding of the protein, are associated to make up a functional site such as the catalytic site of an enzyme or the binding site of an antibody. This method provides a way to create modified or completely new catalytic sites. As depicted in FIG. 26, the three loop regions of FN3, which can be engineered to confer target ligand binding, can be mutagenized simultaneously, or separately within the BC, DE and FG loops to assay for contributing catalytic functions at this binding site. Therefore, the introduction of additional “catalytically important” amino acids into a ligand binding region of a protein may result in de novo catalytic activity toward the same target ligand.

Hence, new structures can be built on the natural “scaffold” of an existing protein by mutating only relevant regions by the method of this invention. The method of this invention is suited to the design of de novo catalytic binding proteins as compared to the isolation of naturally occurring catalytic antibodies. Presently, catalytic antibodies can be prepared by an adaptation of standard somatic cell fusion techniques. In this process, an animal is immunized with an antigen that resembles the transition state of the desired substrate to induce production of an antibody that binds the transition state and catalyzes the reaction. Antibody-producing cells are harvested from the animal and fused with an immortalizing cell to produce hybrid cells. These cells are then screened for secretion of an antibody that catalyzes the reaction. This process is dependent upon the availability of analogues of the transition state of a substrate. The process may be limited because such analogues are likely to be difficult to identify or synthesize in most cases.

The method of this invention can be used to produce many different enzymes or catalytic antibodies, including oxidoreductases, transferases, hydrolases, lyases, isomerases and ligases. Among these classes, of particular importance will be the production of improved proteases, carbohydrases, lipases, dioxygenases and peroxidases. These and other enzymes that can be prepared by the method of this invention have important commercial applications for enzymatic conversions in health care, cosmetics, foods, brewing, detergents, environment (e.g., wastewater treatment), agriculture, tanning, textiles, and other chemical processes. These include, but are not limited to, diagnostic and therapeutic applications, conversions of fats, carbohydrates and protein, degradation of organic pollutants and synthesis of chemicals. For example, therapeutically effective proteases with fibrinolytic activity, or activity against viral structures necessary for infectivity, such as viral coat proteins, could be engineered. Such proteases could be useful anti-thrombotic agents or anti-viral agents against viruses such as AIDS, rhinoviruses, influenza, or hepatitis. In the case of oxygenases (e.g., dioxygenases), a class of enzymes requiring a co-factor for oxidation of aromatic rings and other double bonds, industrial applications in biopulping processes, conversion of biomass into fuels or other chemicals, conversion of waste water contaminants, bioprocessing of coal, and detoxification of hazardous organic compounds are possible applications of novel proteins.

Throughout the examples, the following materials and methods were used unless otherwise stated.

Materials and Methods

In general, the practice of the present invention employs, unless otherwise indicated, conventional techniques of chemistry, molecular biology, recombinant DNA technology, PCR technology, immunology (especially, e.g., antibody technology), expression systems (e.g., cell-free expression, phage display, ribosome display, and Profusion™), and any necessary cell culture that are within the skill of the art and are explained in the literature. See, e.g., Sambrook, Fritsch and Maniatis, Molecular Cloning: Cold Spring Harbor Laboratory Press (1989); DNA Cloning, Vols. 1 and 2, (D.N. Glover, Ed. 1985); Oligonucleotide Synthesis (M.J. Gait, Ed. 1984); PCR Handbook Current Protocols in Nucleic Acid Chemistry, Beaucage, Ed. John Wiley & Sons (1999) (Editor); Oxford Handbook of Nucleic Acid Structure, Neidle, Ed., Oxford Univ Press (1999); PCR Protocols: A Guide to Methods and Applications, Innis et al., Academic Press (1990); PCR Essential Techniques: Essential Techniques, Burke, Ed., John Wiley & Son Ltd (1996); The PCR Technique: RT-PCR, Siebert, Ed., Eaton Pub. Co. (1998); Current Protocols in Molecular Biology, eds. Ausubel et al., John Wiley & Sons (1992); Large-Scale Mammalian Cell Culture Technology, Lubiniecki, A., Ed., Marcel Dekker, Pub., (1990). Phage Display: A Laboratory Manual, C. Barbas (Ed.), CSHL Press, (2001); Antibody Phage Display, P O'Brien (Ed.), Humana Press (2001); Border et al., Yeast surface display for screening combinatorial polypeptide libraries, Nature Biotechnology, 15(6):553-7 (1997); Border et al., Yeast surface display for directed evolution of protein expression, affinity, and stability, Methods Enzymol., 328:430-44 (2000); ribosome display as described by Pluckthun et al. in U.S. Pat. No. 6,348,315, and Profusion™ as described by Szostak et al. in U.S. Pat. Nos. 6,258,558; 6,261,804; and 6,214,553, and bacterial periplasmic expression as described in US20040058403A1.

Further details regarding fibronectin and Fn3 sequence classification, identification, and analysis may be found, e.g., in SEQHUNT. A program to screen aligned nucleotide and amino acid sequences, Methods Mol. Biol. 1995; 51:1-15. and Wu et al. Clustering of highly homologous sequences to reduce the size of large protein databases. Bioinformatics. 2001 March; 17(3):282-3; Databases and search and analysis programs include the PFAM database at the Sanger Institute (pfam.sanger.ac.uk); the ExPASy PROSITE database (www.expasv.ch/prosite/); SBASE web (hydra.icgeb.trieste.it/sbase/); BLAST (www.ncbi.nlm.nih.gov/BLAST/); CD-HIT (bioinformatics.ljcrf.edu/cd-hi/); EMBOSS (www.hqmp.mrc.ac.uk/Software/EMBOSS/); PHYLIP (evolution.genetics.washington.edu/phylip.html); and FASTA(fasta.bioch.virginia.edu).

Briefly, a microbial expression and display system is used which has a demonstrated reliability for expressing fibronectin binding domain libraries. Typically, the fibronectin binding domain is joined together by a linker peptide to another surface molecule creating a fusion protein. A variety of methods are available for library expression and display including ribosome, phage, E. coli, and yeast surface display. All combine genotype-phenotype linkage to allow selection of novel target binding clones.

Yeast:

The fibronectin binding domain library (ie. FN3) is transfected into the recipient bacterial/yeast hosts using standard techniques. Yeast can readily accommodate library sizes up to 107, with 103-105 copies of each FNII fusion protein being displayed on each cell surface. Yeast cells are easily screened and separated using flow cytometry and fluorescence-activated cell sorting (FACS) or magnetic beads. The yeast eukaryotic secretion system and glycosylation pathways of yeast also allows FN3 type molecules to be displayed with N and O linked sugars on the cell surface.

The yeast display system utilizes the a-agglutinin yeast adhesion receptor to display proteins on the cell surface. The proteins of interest, in this case, FN3 WTM, LTM and CBM libraries, are expressed as fusion partners with the Aga2 protein.

These fusion proteins are secreted from the cell and become disulfide linked to the Aga1 protein, which is attached to the yeast cell wall (see Invitrogen, pYD1 Yeast Display product literature). The plasmid pYD1, prepared from an E. coli host by plasmid purification (Qiagen), is digested with the restriction enzymes, Bam HI and Not I, terminally dephosphorylated with calf intestinal alkaline phosphatase. Ligation of the pYD1 vector and the above SOE-PCR products WTM libraries (also digested by BamHI and NotI), E. coli (DH5a) transformation and selection on LB-ampicillin (50 mg/ml) plates were performed using standard molecular biology protocols to amplify the WTM libraries before electroporation into yeast cell hosts.

Methods for selecting expressed FN3 library variants having substantially higher affinities for target ligands (TNF, VEGF, VEGF-R etc), relative to the reference wild type FN3 domain, will now be described.

Candidate test ligands (TNF, VEGF, VEGF-R etc), are fluorescently labeled (either directly or indirectly via a biotin—streptavidin linkage as described above). Those library clones that efficiently bind the labeled antigens are then enriched for by using FACS. This population of yeast cells is then re-grown and subjected to subsequent rounds of selection using increased levels of stringency to isolate a smaller subset of clones that recognize the target with higher specificity and affinity. The libraries are readily amenable to high-throughput formats, using, e.g., FITC labeled anti-Myc-tag FN3 binding domain molecules and FACS analysis for quick identification and confirmation. In addition, there are carboxyl terminal tags included which can be utilized to monitor expression levels and/or normalize binding affinity measurements.

To check for the display of the Aga2-FN3 fusion protein, an aliquot of yeast cells (8×105 cells in 40 μl) from the culture medium is centrifuged for 5 minutes at 2300 rpm. The supernatant is aspirated and the cell pellet is washed with 200 μl of ice cold PBS/BSA buffer (PBS/BSA 0.5% w/v). The cells are re-pelleted and supernatant removed before re-suspending in 100 μl of buffer containing the biotinylated TNFα (200 nM). The cells were left to bind the TNFa at 20° C. for 45 minutes after which they were washed twice with PBS/BSA buffer before the addition and incubation with streptavidin-FITC (2 mg/L) for 30 minutes on ice. Another round of washing in buffer was performed before final re-suspension volume of 400 μl in PBS/BSA. The cells were then analyzed on FACSscan (Becton Dickinson) using CellQuest software as per manufacturers directions.

Kinetic selections of the yeast displayed TNF-α fibronectin binding domain libraries involve initial labeling of cells with biotinylated TNF-α ligand followed by time dependent chase in the presence of large excess of un-biotinylated TNF-α ligand. Clones with slower dissociation kinetics are identified by steptavidin-PE labeling after the chase period and sorted using a high speed FACS sorter. After Aga2-FN3 induction, the cells are incubated with biotinylated TNF-α at saturating concentrations (400 nM) for 3 hours at 25 C under shaking. After washing the cells, a 40 hour cold chase using unlabelled TNF-a (1 uM) at 25° C. was performed. The cells were then washed twice with PBS/BSA buffer, labeled with Streptavidin PE (2 mg/ml) anti-HIS-FITC (25 nM) for 30 minutes on ice, washed and re-suspended and then analyzed on FACS ARIA sorter.

Ribosome Display:

Ribosome display utilizes cell free in vitro coupled transcription/translation machinery to produce protein libraries. The FN3 library genes are inserted upstream to kappa light immunoglobulin gene that does not have a termination stop codon causing the ribosome to stall, but not release, when it reaches the end of the mRNA. Additionally, the kappa domain spacer serves to physically distance the FN3 protein from the ribosome complex so that FN3 binding domain has better accessibility to recognize its cognate ligand. The mRNA library is introduced into either S30 E. coli ribosome extract preparations (Roche) or rabbit reticulate lysate (Promega). In either case, the 5′ end of the nascent mRNA can bind to ribosomes and undergo translation. During translation, the ligand-binding protein remains non-covalently attached to the ribosome along with its mRNA progenitor in a macromolecular complex {He, 2005 #58; He, 1997 #59; He, 2007 #57}.

The functional FN3 proteins can then bind to a specific ligand that is either attached to magnetic beads or microtiter well surface. During the enrichment process, non-specific variants are washed away before the specific FN3 binders are eluted. The bound mRNA is detected by RT-PCR using primers specific to the 5′ FN3 and 3′ portion of the kappa gene respectively (FIG. 9). The amplified double stranded cDNA is then cloned into an expression vector for sequence analysis and protein production.

Prokaryotic translation reactions contained 0.2 M potassium glutamate, 6.9 mM magnesium acetate, 90 mg/ml protein disulfide isomerase (Fluka), 50 mM Tris acetate (pH 7.5), 0.35 mM each amino acid, 2 mM ATP, 0.5 mM GTP, 1 mM cAMP, 30 mM acetyl phosphate, 0.5 mg/ml E. coli tRNA, 20 mg/ml folinic acid, 1.5% PEG 8000, 40 ml S30 E. coli extract and 10 mg mRNA in a total volume of 110 ml. Translation was performed at 37 C for 7 min, after which ribosome complexes were stabilized by 5-fold dilution in ice-cold selection buffer [50 mM Tris acetate (pH 7.5), 150 mM NaCl, 50 mM magnesium acetate, 0.1% Tween 20, 2.5 mg/ml heparin].

For Eukaryotic Ribosome Display we Used the Flexi Rabbit Reticulocyte Lysate

System (Promega). Eukaryotic translation reactions contained 40 mM KCl, 100 mg/ml protein disulfide isomerase (Fluka), 0.02 mM each amino acid, 66 ml rabbit reticulocyte lysate and 10 mg mRNA in a total volume of 100 ml. Translation was performed at 30 C for 20 min, after which ribosome complexes were stabilized by 2-fold dilution in ice-cold PBS.

Affinity Selection for Target Ligands.

Stabilized ribosome complexes were incubated with biotinylated hapten [50 nM fluorescein-biotin (Sigma)] or antigen [100 nM IL-13 (Peprotech) biotinylated in-house] as appropriate at 4 C for 1-2 h, followed by capture on streptavidin-coated M280 magnetic beads (Dynal). Beads were then washed to remove non-specifically bound ribosome complexes. For prokaryotic selections, five washes in ice-cold selection buffer were performed. For eukaryotic selections, three washes in PBS containing 0.1% BSA and 5 mM magnesium acetate were performed, followed by a single wash in PBS alone. Eukaryotic complexes were then incubated with 10 U DNAse I in 40 mM Tris-HCl, 6 mM MgCl2, 10 mMNaCl, 10 mM CaCl2 for 25 min at 37 C, followed by three further washes with PBS, 5 mM magnesium acetate, 1% Tween 20.

Recovery of mRNA from Selected Ribosome Complexes

For analysis of mRNA recovery without a specific disruption step, ribosome complexes bound to magnetic beads were directly processed into the reverse transcription reaction. For recovery of mRNA from prokaryotic selections by ribosome complex disruption, selected complexes were incubated in EB20 [50 mM Tris acetate (pH 7.5), 150 mM NaCl, 20 mM EDTA, 10 mg/ml Saccharomyces cerevisae RNA] for 10 min at 4 C. To evaluate the efficiency of the 20 mM EDTA for recovery of mRNA from eukaryotic selections, ribosome complexes were incubated in PBS20 (PBS, 20 mM EDTA, 10 mg/ml S. cerevisae RNA) for 10 min at 4 C. mRNA was purified using a commercial kit (High Pure RNA Isolation Kit, Roche). For prokaryotic samples, the DNAse I digestion option of the kit was performed; however, this step was not required for eukaryotic samples, as DNAse I digestion was performed during post-selection washes. Reverse transcription was performed on either 4 ml of purified RNA or 4 ml of immobilized, selected ribosome complexes (i.e. a bead suspension).

For prokaryotic samples, reactions contained 50 mM Tris-HCl (pH 8.3), 75 mM KCl, 3 mMMgCl2, 10 mMDTT, 1.25 primer, 0.5 mM PCR nucleotide mix (Amersham Pharmacia), 1 URNAsin (Promega) and 5 U SuperScript II (Invitrogen) and were performed by incubation at 50 C for 30 min. For eukaryotic samples, reactions contained 50 mM Tris-HCl (pH 8.3), 50 mM KCl, 10 mM MgCl2, 0.5 mM spermine, 10 mM DTT, 1.25 mM RT primers, 0.5 mM PCR nucleotide mix, 1 U RNasin and 5 U AMV reverse transcriptase (Promega) and were performed by incubation at 48 C for 45 min.

PCR of Selection Outputs

End-point PCR was performed to visualize amplification of the full-length construct. A 5 ml sample of each reverse transcription reaction was amplified with 2.5 UTaq polymerase (Roche) in 20 mM Tris-HCl (pH 8.4), 50 mM KCl, 1 mM MgCl2, 5% DMSO, containing 0.25 mM PCR nucleotide mix, 0.25 mM forward primer (T7B or T7KOZ for prokaryotic and or eukaryotic experiments, respectively) and 0.25 mM RT primer. Thermal cycling comprised 94 C for 3 min, then 94 C for 30 s, 50 C for 30 s and 72 C for 1.5 min for 30 cycles, with a final step at 72 C for 5 min. PCR products were visualized by electrophoresis on an ethidium bromide stained agarose gels. The isolated PCR products can then be sub-cloned into a bacterial pBAD expression vector for soluble protein production (below).

Bacterial Expression and Production:

Competent E. coli host cells are prepared as per manufacturer's instructions (Invitrogen PBAD expression system). Briefly, 40 μl LMG 194 competent cells and 0.5 μl pBAD FN3 constructs (approximately 1 μg DNA) is incubated together on ice for 15 minutes after which, a one minute 42° C. heat shock was applied. The cells are then allowed to recover for 10 minutes at 37° C. in SOC media before plating onto LB-Amp plates and 37° C. growth overnight. Single colonies are picked the next day for small scale liquid cultures to initially determine optimal L-arabinose induction concentrations for FN3 production. Replicates of each clone after reaching an OD600=0.5 were test induced with serial (1:10) titrations of L-arabinose (0.2% to 0.00002% final concentration) after overnight growth at room temperature. Test cultures (1 ml) are collected, pelleted and 100 μl 1×BBS buffer (10 mM, 160 mM NaCl, 200 mM Boric acid, pH=8.0) added to resuspend the cells before the addition of 50 μl of lysozyme solution for 1 hour (37° C.). Cell supernatants from the lysozyme digestions are collected after centrifugation, and MgSO4 was added to final concentration 40 mM. This solution was applied to PBS pre-equilibrated Ni-NTA columns. His-tagged bound FN3 samples are twice washed with PBS buffer upon which elution was accomplished with the addition of 250 mM imidazole. Purity of the soluble FN3 expression is then examined by SDS-PAGE.

Larger scale E. coli cell culture, 100 ml, pellets are collected by centrifugation after overnight growth at 25° C. The pellets are then re-suspended in PBS buffer (0.1% tween) and subjected to 5 rounds of repeated sonication (Virtis Ultrasonic cell Disrupter) to lyse the bacterial cell membrane and release the cytoplasmic contents. The suspension is first clarified by high speed centrifugation to collect the supernatant for further processing. This supernatant is then applied to PBS pre-equilibrated Ni-NTA columns. His-tagged bound FN3 samples are twice washed with PBS buffer upon which elution is accomplished with the addition of 250 mM imidazole. The pH of the supernatant is then adjusted to 5.5 with 6M HCl and before loading onto a SP Sepharose HP cation exchange column (Pharmacia). The FN3 was eluted a salt (NaCl) gradient and fraction concentrations containing the FN3 were determined by optical density at 280 nm and verified by PAGE. Fractions containing FN3s are then pooled and dialyzed with PBS.

Octet Kinetic Analysis:

Binding affinities (KD=kd/ka=koff/kon) of the FN3 variants are calculated from the resultant association (ka=kon) and dissociation (kd=koff) rate constants as measured using a OCTET biolayer inferometry system (ForteBio, Inc). The ligand, e.g., TNF-α, is immobilized on the OCTET streptavidin capillary sensor tip surface and, in effect, allows monitoring of the monomeric FN3 kinetic binding. OCTET streptavidin tips are activated according to manufacturer's instructions using 50 uM TNF-a. A solution of PBS/BSA is also introduced as a blocking agent.

For association kinetic measurements, FN3 variants are introduced at a concentration of 10 ug/ml. Dissociation is observed in PBS buffer without the agents. The kinetic parameters of the binding reactions were determined using ForteBio software. FIG. 25 displays OCTET results from the reference D2E7 anti-TNF-a antibodies and 9-14 TNF FN3 clone.

Candidate clones are then isolated and plasmid preparations are performed to obtain fibronectin binding domain sequence information. The approach allows for a hypothesis-driven rational replacement of codons necessary to determine and optimize amino acid functionality in the loop of the fibronectin binding domain. Comparative sequence analysis and individual clone affinity/specificity profiles then determine which clones undergo affinity maturation.

Example 1

Methods for Bioinformatic-Guided Identification of Universal Fibronectin Binding Domain Library Sequences

In this example, universal BC, DE, and FG loop sequences for fibronectin binding domain library sequences are identified and selected using bioinformatics and the criteria of the invention. A generalized schematic of this process is presented in FIG. 1.

Briefly, the PFAM database was searched for a multiple sequence alignment containing only sequences belonging to the Fibronectin Type III family (FN3, PFAM ID: PF00041).

This search returned an initial dataset of 9321 protein sequences. It is noted, however, that this set of sequences can increase in number as additional sequences are cloned and entered into the database. The base dataset of ˜9321 FN3 superfamily sequences included FN3 sequences from multiple sources: human fibronectin, tenascin, drosophila sevenless, protein tyrosine phosphatase, EPH tyrosine kinases, and cytokine receptors and Z proteins (see Table 1). It is appreciated that gene sequences related to FN1, FN2 and other protein scaffold framework domains can also be searched in these databases and identified in similar manner.

TABLE 2
Distribution of analyzed proteins in the Fibronectin module III
superfamily.
Base dataset: All FN module III like proteins9321
First round filtered non redundant mammalian5955
FN module III
Non redundant human FN module III794

Within the starting FN3 sequences “base dataset”, the compiled loop sequences are likely to possess vastly different characteristics. The sequences will vary with respect to: originating species, originating modules, loop lengths, and other qualities. Due to originating sequence disparity among the “base dataset” members, trying to derive a coherent analysis may require further refinement from the starting “base dataset” such that subset datasets share one or more of elected properties of interest. Therefore, from the initial ˜9321 base dataset, redundant or duplicatively entered sequences were removed leaving 5955 filtered non-redundant FN3 sequences. From this second subset, another set of sequences were further parsed into a third subset made of 794 human FN3 sequences. Constituent subset members sharing those respective properties can produce a more “standardized” set of sequences for meaningful comparative and unskewed data analysis within the subgroup. This process can be iterated, resulting in the generation of smaller datasets with a higher degree of predefined relationships. Widening sequence variability may be achieved, for example, by including the datasets containing non-human modules.

The next step involved the designation of the β-scaffold framework and loop lengths and loop amino acid sequences, followed by a frequency analysis of these candidate loop sequences. Therefore, determination of variability profiles entails collection and selection of aligned amino acid sequences that shares one or more defined properties of interest to create a dataset. We developed a β-strand and loop positional classification system using the positional numbering architecture of the 10/FN3 domain as a reference point. β-strands were first identified based on the crystal structure of FN3 fibronectin-protein modules 1, 4, 7, 8, 9, 10, 12, 13, 14, and 15. We also examined their dihedral angles and hydrogen bonding patterns to assist with the assignment of β-strands. Amino acid sequence alignments were then performed to further aid in the determination of β-strands and loops. For example in Table 3, using 10FN3 as a reference sequence, the following amino acid positions for the of β-strands and loops were designated as follows:

TABLE 3
β-strands and loop amino acid positions of FN3
β-strandsLoops
A: 8-14AB: 15-16
B: 17-20BC: 21-31
C: 32-39CD: 40-44
D: 45-50DE: 51-56
E: 57-59EF: 60-67
F: 68-75FG: 76-86
G: 87-94

For example, the aligned analysis indicated that the tryptophan at position 22 (W22) is conserved between the different modules. Based on a crystallographic structural comparisons, W22 is located near the junction of the BC loop acceptor site and we chose position 21 as the loop starting point. There is a well conserved group of valine and isoleucine amino acids at position 20 between the FN3 domains (V20 in 10FN3) and appears to be part of the B 13-strand. Immediately following W22, we found that there are structural differences between the overlayed loops before the BC loops reenter the C β-strand at a well-conserved Y/F32 found between the members. By designating position 21 and 31 as the outer boundaries of the BC loop, we then segregated the “filtered dataset” into BC loop sizes and determined their amino acid frequency compositions. Similarly, this sequence/structure comparative analysis was applied to the DE and FG loops. We designated the DE loop as bordered by a conserved group of valine, leucine and isoleucine at position 50 (V50 in 10FN3) of the D β-strand and ending at position 56 (T56 in 10FN3) of the E β-strand. Likewise, we choose positions 76 (T76 in 10FN3) and 86 (K86 in 10FN3) as the F β-strand and G β-strand junctions for the FG loop.

Based on these aligned β strand boundary definitions, the loops were found not be of one size, but to occur in differing lengths. The frequency distribution of the BC, DE and FG loop sizes were analyzed and tabulated (FIGS. 6-8). The loop size classification and description follow the nomenclature of: FN3 LOOP/LENGTH. For example BC/15 refers to FN3 BC loop length of 15 amino acids. In the BC loop definition, the scaffold loop boundaries were originally set for an eleven amino acid length. However, the BC loop ranges from nine to eighteen amino acids in length, with length fourteen (BC/14), and fifteen (BC/15) also being quite common (FIG. 6). Alignments of the loop amino acids can be performed between the differing loop lengths. For example, the proline at position 25 (P25) in BC/15 was found quite conserved within constituent BC/15 members. P25 is also observed to be quite prevalent in BC/14 and BC/11 members. The loop amino acid positions were then aligned to best approximate these equivalent positions. Hence in BC/15 the extra comprising loop amino acids given the designation 26a, 26b, 26c, and 26d, as they are located between residues 26 and 27 of the BC/15 loop. In loop BC/14 the extra comprising loop amino acids are designated from 26a, 26b, and 26c, as they are between residues 26 and 27 of the BC/14 loop. As stated above, numbering is based upon 10/FN3 (the 10th FN3 module of human fibronectin) as a reference.

Example 2

Assessing Loop Variability Profiles Using Bioinformatics Through Filtering and Cluster Analysis of Gene Sequences

The universal fibronectin binding domain libraries were designed by determining the variability profiles for the loops expressed in vivo. The variability profiles represent the cataloging of the different amino acids, and their respective rates of occurrence, present at a particular position in a dataset of aligned sequences. Size related families of loop sequences using the parameters set forth above within this starting “base dataset” can be identified and delineated. Comparative analysis of these multiple aligned loops provide variability profile information as to the existing and “tolerated” diversity for introducing amino acid changes that can lead to potential ligand binding. The designation of loops and their comprising amino acids can also be described for other scaffold like proteins using similar definitions.

The frequency distribution of the six loop sizes, shown in FIGS. 6 to 8, were generated to determine if there was preferred BC, DE and FG loop sizes for FN3 sequences (Table 2). For the BC loop (FIG. 6), a fifteen amino acid loop size was the most common accounting for 30% of the BC loop population. BC loop sizes 14 and 11 were the next common sizes occurring at 18% and 16% of the BC loop population. BC loop sizes 11, 14 and 15 were then chosen for our variability profile analysis.

Frequency size analysis of the DE loop demonstrated that loop size 6 occurred in nearly 45% of the analyzed FN3 sequences (FIG. 7). The other DE loop sizes 4, 5 and 7 accounted for no more than 16% each. DE loop size 6 accounted for 55% of the members suggesting FN3 modules have more of a preference for DE loop size 6. In this case, only DE loop size 6 was chosen for further variability profile analysis. Frequency size analysis of the FG loop demonstrated that loop size 8 and 11 were the two most common loop sizes accounting for 58% and 18% of the analyzed dataset respectively. Both FG loop sizes 8 and 11 were therefore included for further variability profile analysis. In an earlier version of the analysis, the FG loop lengths FG/8 and FG/11 were identified as FG/6 and FG/9, respectively, the difference being the addition of two C-terminal amino acids S, K to each FG loop length, as seen in Table 5 below). In some examples, where an FG/8 loop is identified by six amino acids, it is understood that these amino acids represent the six N-terminal amino acids of FG/8. Similarly, where an FG/11 loop is identified by nine amino acids, it is understood that these amino acids represent the nine N-terminal amino acids of FG/11.

TABLE 4
Lengths of FN3 BC, DE, and FG loops analyzed along with
the number of sequences indicated
number of
LOOPlengthsequences
BC11885
141066
151732
DE62729
FG81057
113340

For each of the 6 selected loops (Table 4) a separate frequency analysis was executed to determine positional amino acid usage in the context of the selected loop. A simple frequency analysis using EMBOSS/prophecy (bioinfo.nhri.org.tw/cgi-bin/emboss/prophecy) was executed generating a matrix representing the positional amino acid usage. The output matrix was then parsed and filtered in order to have relative frequency data for each position. The parser provides a very simple filter based on two thresholds (low and high). For each position the parser processes only amino acids with relative frequency above the “low” threshold until the cumulative frequency reaches the “high” threshold. If the high threshold is not reached, then the parser also evaluates the amino acids with relative frequency below the low threshold. A good low-high threshold combination was 5-50 because it provides good sensitivity for position classification. The parser therefore limits the listing of only the most frequent, and not all, amino acids that occur at each position. The parser output is visualized as frequency charts and the results are shown in FIGS. 9-14.

Example 2A

Identifying Fixed and Non-Fixed Loop Positions with Thresholds

In one embodiment, a natural-variant combinatorial library with a conserved or selected semi-conserved consensus amino acid is designed as follows.

FN3 loop datasets are enumerated as above for amino acid variability and their relative frequencies at each aligned position (FIGS. 9-14). The above analysis identified positional preferences in all FN3 module loops and are termed “variability profiles.” For example, in BC loop size 11 (FIG. 9), a tryptophan (W) at position 22 is found at nearly 95% of all FN3 loop positions demonstrating high degree of selective pressure for its presence. Position 22 would be considered “conserved” for tryptophan (W22) (see below Example 3) as it occurred above a predetermined 40% threshold level and was more than twice as common as the next most frequent amino acid at that position. This “fixed” residue is seen as the dominant amino acid with respect to the other amino acids that occur at that loop position. A “fixed” position may not be subject to mutagenic diversification in first round library building. At other BC loop size 11 loop positions, it was evident that position 25 favored prolines (P) (approximately >45%) as it occurred twice as often compared to the next most frequent amino acids (<20%) in that same position. Hence, P25 position is considered “fixed” as the predominant amino acid was two fold more frequent than the next. In contrast, prolines were also the most favored amino acid at positions 24, 26 and 28, but the frequency of the other amino acids were also fairly abundant (8-19%). As positions 24, 26 and 28 have a wide range of populating amino acids with similar frequency of occurrences, and the predominant amino acid was not more than two fold more frequent, these positions would be considered “variable.”

In BC loop size 14 (FIG. 10), the positional prevalence of particular amino acids was pronounced. In BC/14, the W position 22 occurs at (>95%) while position 25 has proline occurring at 70%. This indicated that W22 and P25 are heavily favored in BC/14 loops so that both positions are considered “fixed.” Also fixed was isoleucine at position 29. Here in BC/14, isoleucine (I29) occurred at >52%, more than two-fold more than the next two most common amino acids, valine and leucine at 18% and 16% respectively.

For BC/15 loops (FIG. 11), W22, P25 and 129 were also found to be fixed positions similar to BC/11 and BC/14. Additionally, glycines (G) at positions 26e and 26f occur at frequencies >75% and can be also considered “fixed.” By first determining these “fixed” positions and non-fixed positions (indicated by “X”), B/15 would have a starting “fixed” sequence of: X21 W22 X23 X24 P25 X26 X26a D26b G26c G26d X27 X28 I29 X30 X31. Similarly, BC/4 and BC/1 would have starting “fixed” sequences of X21 W22 X23 X24 P25 X26 X26a X26b X26c X27 X28 I29 X30 X31 and X21 W22 X23 X24 P25 X26 X26 X27 X28 I29 X30 X31, respectively.

The same variability profile analysis was performed for all the other loops. Briefly, for DE loop size 6 (FIG. 12), there were no loop positions where a single amino acid accounted for more than 35% of the representative positional amino acids. There was a slight preference for glycine at position 52 (52G), threonine at position 55 (55T) and serine at position 56 (56S). In this case however, there were no predominant acids above a 40% threshold level to be considered “fixed” amino acids and all loop positions are considered “variable.” For library construction, the most common amino acid in each loop position serves as the starting amino acid for first round mutagenesis. The DE/loop therefore has an all “variable” starting loop sequence of: X51 X52 X53 X54 X55 X56.

For FG loop size 8 (FIG. 13), the serine (S) at position 84 was “fixed” as it occurred in 75% of the populating sequences. All other remaining FG loop size 8 positions were “variable” using the most frequently occurring amino acid yielding a FG/8 starting “fixed” sequence of X76 X77 X78 X79 X80 84S (the N-terminal six amino acids of FG/8). The FG/8 loop numbering jumps from 80 to 84 as the longer loop size nine is used as the reference size length based on 10/FN3. FG/11 contained four “fixed” amino acids with asparagine (56N), glycine (79G), serine (84S) and a “semi-fixed” glycine and serine at position 81. The three amino acids, 56N, 79G, 84S and semi-fixed 81 G/S occurred at greater than 40% frequency in the populating sequences. This analysis yielded a FG/11 starting “fixed” loop sequence of N76 X77 X78 79G X80 G/S81 X82 X83 S84 (the N-terminal nine amino acids of FG/11).

The variability profile for each loop dataset then identifies the desired characteristics of a given loop position for further introduction of diversity representation. These above results demonstrate that the diversity of loop amino acids introduced into a library can be “fine-tuned” depending on the threshold level of frequency of occurrence. These “fixed” loop positions attempt to replicate some of the natural diversity to promote possible structural stabilization effects.

This “fixing” of positions also has the effect of “narrowing” the diversity of variable positions in starting loop sequences. However, there can be the occasion to perform the reverse, that is, to obtain larger more diverse libraries. In this case, the “widening” effect is accomplished by raising the threshold of frequency of occurrence used to designate “fixed” amino acids. In this way, the variability profile will capture fewer of the most conserved loop positions and classify them as “fixed” positions. The remaining loop positions would be part of the broader “variable” amino acids that can be diversified.

Example 2B

Identifying Fixed and Non-Fixed Loop Positions without Thresholds

In another embodiment it is possible to design natural-variation combinatorial diversity in each of the six selected loops (Table 4) without defining a variability threshold, i.e., where the selected threshold is 100%. In this embodiment each mutated loop is designed to contain amino acids that mimics its variability profile in terms of both variability and chemistry characteristics. At each specific loop position, oligonucleotide synthesis is optimized to contain a degenerate codon that would match/mimic the chemistry and the variability at that position. Positions having two or more amino acids in their variability profiles will be mutated regardless of the degree of variability of that position.

In BC loop sizes 11 (FIG. 24) and 14 only the W at position 22 will be kept fixed resulting in the following mutagenesis pattern: BC/11=X21 W22 X23 X24 X25 X26 X27 X28 X29 X30 X31; BC/14=X21 W22 X23 X24 X25 X26 X26a X26b X26c X27 X28 X29 X30 X31. In BC loop size 15 there are 3 fixed positions (W22, P25 and G26c) resulting in the following mutagenesis pattern: BC/15=X21 W22 X23 X24 P25 X26 X26a X26b G26c X26d X27 X28 X29 X30 X31).

In DE loop size 6 all the positions are variable and therefore will have a starting sequence of DE6=X51 X52 X53 X54 X55 X56.

The FG loop size 8 will be encoded by the starting sequence of FG/8=X76 X77 X78 X79 X80 X84 S85 S86. Finally FG loop size 11 will be encoded as FG/11=X76×77×78 G79 X80 X81 X82 X83 X84 S85 S86.

Each of the positions denoted as X will encode for the amino acids present in the related variability profile or for a subset of chemically equivalent ones. In fact in some cases two or more amino acids present at a certain position are chemically very similar. In such situations it is possible to include in the mutagenesis design only a subset of the amino acids and still preserve the natural chemistry characteristics of that position. This will both reduce the total number of mutants and give more flexibility for the optimization of the oligonucleotide synthesis.

Example 3

Methods for Designing WTM Loop Diversity for FN3 Domain Libraries with Threshold Constraints

In this example, methods for optimizing the loop diversity of an FN3 binding domain library are presented. The choice of candidate frameworks, as previously noted, dictates both the loop sizes to be introduced and the initial amino acid sequence selection. The method is illustrated particularly for the FG/11 loop.

To design the FG/11 loop library for FN3 based scaffolds, the variability profile considerations are as follows: As stated above, a “fixed” amino acid residue is determined to occur with a threshold frequency that is typically at least 40% (typically at least 50%) and is twice fold more frequent than the next most frequent amino acid for a given loop position. Upon inspection of the FN3 FG/11 variability profile (FIG. 14), it can be seen that the Gly at position 79 (G79) was found to occur at nearly 85%. The next most frequently occurring amino acids, was at frequency of less than 5% and did not register on the minimum threshold value (cut off by parser). When a given residue, in this case G79, is determined to occur at such a high frequency, it is highly conserved and thus represented in the libraries of the invention as “fixed,” meaning that it will not be mutated in the first round library diversification. Similarly, N76, G81 and S84 are also “highly conserved” as they occur at a frequency rate of 66%, 47%, and 61% respectively and are “fixed” in the first diversity library. In some cases, it can be that there are two semi-conserved amino acid residue at a given loop position. For example, G81 and S81 are found to occur at nearly 66% and 30% respectively. Hence, in this case, both amino acids are considered “semi-fixed” at position 81 as it appears that there is strong selective pressure to maintain G81 and S81 and no other amino acids. The 66% G81 and 30% S81 ratios can be maintained by controlled “doping” of the incoming nucleotide mixture (see EXAMPLE 5). The number of WTM substitutions per loop can be readily achieved by oligonucleotide synthesis doping (see, e.g., US20040033569A1 for technical details).

Positions N76, G79, G/S81, and S84 are fixed candidate clones so that FG/11 has a starting “fixed” loop sequence of N76-77-78 79G-80 G/S81-82-83 S84 (underline indicating the fixed amino acid). The reason for not creating diversity at all sites is to restrict the initial diversity size of the library to facilitate efficient expression and display of all variants. These initially “fixed” positions indicate strong selective pressures for their preservation. However, they are still sites for look-through mutatgenesis (LTM) mutagenesis during affinity maturation. In other words, the initial “fixed” positions can be later mutated. The overall goal in “fixing” positions in the first round library diversification is two fold: 1) to maximize the number of functional clones while by incorporating most of the preferred loop residues and 2) to minimize the total library size. Subsequent LTM mapping of retrieved clones then precisely determines whether the “fixed” site is a “hot” or “cold” spot that can be mutated for further affinity maturation without losing overall binding function (see U.S. Patent Publication No. 20050136428).

The term “variable” amino acid residue refers to amino acid residues determined to occur with a lower frequency (less than the high threshold value of 20%) for a given residue position. Upon inspection of the FG/11 variability profile for example (FIG. 14), it can be seen that variable positions 77, 78, 80, 82 and 83 each have no single prevalent amino acid that occurs at higher than 40% frequency. For FG/11 positions 77, 78, 80, 82 and 83 each loop position had many different amino acids with occurring at a fairly low level frequency. Accordingly, in positions 77, 78, 80, 82 and 83 are sites for creation of initial loop sequence diversity by mutagenesis, e.g., WTM. In introducing subsequent amino acid diversity, the FG/11 sequence would then be N76 X77 X78 79G X80 G/S81 X82 X83 S84, where X denotes the variable position for initial mutagenesis (e.g., WTM) is conducted. The X position will be consist of a starting “wild type” amino acid, the WTM replacement or a “co-product” amino acid depending on the WTM designed codon changes.

For example, to construct a WTM library within the FG/11 loop using lysine (K) as the WTM target amino acid into the starting sequence N76 X77 X78 79G X80 G/S81 X82 X83 S84, lysine residues can be substituted in positions 77, 78, 80, 82, and 83. The FG/11 WTM(K): library variants will have resulting sequences that are comprised of single replacements:

N76 K77 X78 79G X80 G/S81 X82 X83 S84,

N76 X77 K78 79G X80 G/S81 X82 X83 S84,

N76 X77 X78 79G K80 G/S81 X82 X83 S84,

N76 X77 X78 79G X80 G/S81 K82 X83 S84, and

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

The FG/11: WTM(K): library variant sequences are additionally comprised of double replacements:

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

The FG/11: WTM(K): library variant sequences could be additionally comprised of triple replacements for example:

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

The FG/11: WTM(K): library variant sequences could be additionally comprised of quadruple replacements:

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

N76 X77 X78 79G X80 G/S81 X82 K83 S84

Finally, the FG/11: WTM(K): library variant sequences could be additionally comprised of quintuple replacements:

N76 X77 X78 79G X80 G/S81 X82 K83 S84,

For WTM™ diversity, each “X” position in the above example must have a starting “wild type” amino acid in order to perform the targeted codon changes. From the above variability profile analysis, FG/11 has a “wild type” consensus sequence of N76 A77 A78 79G V80 G/S81 P82 P83 S84. The residues A77, A78, P82, and P83 are chosen as starting “wildtype” amino acid as the variability profile (FIG. 14) indicated that these were the most frequent amino acids at their respective positions.

In this instance, the WTM (K) effects of introducing an amine side chain are explored in ligand binding. It is appreciated that the WTM (K) amino acids will occur only in those target “variable” positions and not in the “fixed” positions. The generated WTM diversity will include any “co-product” amino acids based on the degeneracy of the “wildtype” amino acid of the target codon. For example at position 77, the “wildtype” amino acid is alanine (Ala). To perform lysine WTM at position 77 so that both Ala and Lys are present, the degenerate codon used is RMG. The “coproduct” amino acids resulting from RMG will be threonine (T) and glutamate (E). The remaining WTM amino acids (G, S, H, L, P, Y, D, and Q) are then similarly replaced into the X variable positions of FN3 FG9 loop N76 X77 X78 79G X80 G/S81 X82 X83 S84 sequence. It is appreciated that the WTM amino acids will occur along with their “co-product” amino acids.

WTM for FG/11 using the nine pre-chosen WTM™ amino acids produces a library diversity of approximately 2×95 or 59049 (which includes WTM co-products). For comparative purposes, saturation mutagenesis of five FG/11 “variable” positions with all twenty amino acids would be 205 or 3.2×106. Accordingly, performing saturation mutagenesis at all of the FG/11 loop positions would create a library diversity of 209 or 5×1011 which is just at the limit of current library display and screening technology. However, it is important to note that is for an FN3 loop library with just one mutated loop, and if the other FN3 loops (e.g. BC/11 and DE/6) were diversified by saturation mutagenesis, the minimum FN3 library would then be 2026 or 6.7×1033. In contrast, the minimum and maximum FN3 library based on identifying the “variable” BC, DE and FG loop positions are 6.7×1033. This illustrates an advantage of the invention which, by contrast, allows for a smaller but more representative library to be constructed. Indeed, the methods of the invention provide for a manageable library for some loop positions in order to identify the first generation of binding molecules. Subsequent affinity maturation mutagenesis in the other CDR positions then optimizes those identified binding molecules.

In another approach, byproducts can be avoided by employing look-through mutagenesis (LTM) which typically requires the synthesis of an oligonucleotide for each desired change but eliminates any by-products by degenerate codons.

A summary of identified loop sequences for use in the FN3 binding domain library of the invention is set forth below in Table 5. The names of the loops, sizes, and residue positions are standardized as previously discussed. Single-letter positions are fixed positions; two-letters positions are semi-fixed positions where the synthesis is performed with a mix in order to have only 2 targeted amino acids (example: S-G); and two-letters at positions where the first is ‘X’ are variable positions for WTM diversity. The amino acid following the X is the most frequent amino acid based on that variability profile and considered to be “wild type” (example: X-V).

TABLE 5
In some example, where an FG/8 loop is identified by six amino
acids, it is understood that these amino acids represent the six N-terminal amino
acids of FG/8.
BC LOOP2122232425262728303031
BC_11X-SWX-TX-PPX-PX-GX-PX-VX-DX-G
21222324252626a26b26c2728293031
BC_14X-SWX-KX-PPX-DX-DX-PX-NX-GX-PIX-TX-G
21222324252626a26b26c26d2728293031
BC_15X-SWX-EX-PPX-EX-DDGGX-SX-PIX-TX-G
DE LOOP515253545556
DE_6X-PX-GX-TX-EX-TX-S
FG LOOP767778798084
FG_6XXXXXS
FG_87677787980848586
FG_8X-NX-GX-GX-GX-ESSK
767778798081828384
FG_9NXXGXG/SXXS
FG_117677787980818283848586
FG_11NX-AX-AGX-VG/SX-PX-PSSK

Example 4

Oligonucleotide Design for Introducing Loop Diversity Using WTM and Extended WTM

To design the BC/1 5 diversity (Table 5), a variability profile of the BC/15 loop positions from the filtered dataset was also performed using a threshold frequency of 50% (FIG. 11). The variability profile indicates that BC/15 would have a “wild type” starting sequence of: X-S21 W22 X-E23 X-P24 P25 X-E26 X-D26a D26b G26c G26d X-S27 X-P28 I29 X-T30 X-G31.

21222324252626a26b26c26d2728293031
BC/15X-SWX-EX-PPX-EX-DDGGX-SX-PIX-TX-G

To perform WTM using Lys (K), one would introduce selective degenerate codons to place both the starting “wild type” amino acid and the WTM (K) amino acid in the selected “variable” position. The information presented here is also illustrated in FIGS. 16a-16i.

WTM:K21222324252626a26b26c26d2728293031
BC/15K-SWK-EK-PPK-EK-DDGGK-SK-PIK-TK-G
WTM:KarmtggraammaccgraarawGatggcggcarmmmaattamarra

The WTM (K) oligonucleotide sequence for BC/15 would then be: 5′(arm)(tgg)(raa)(mma)(ccg)(raa)(raw)(gat)(ggc)(ggc)(arm)(mma)(att)(ama)(rra)-3′ (using degenerate base codes) (SEQ ID NO: 82), or 5′-(a a/g a/c)(tgg)(a/g a a)(a/c a/c a)(ccg)(a/g a a)(a/g a a/t)(gat)(ggc)(ggc)(a a/g a/c)(a/c a/c a)(att)(a a/c a)(a/g a/g a)-3′ using standard base codes.

The oligonucleotide to be used in the subsequent PCR reaction will contain both 5′ and 3′ flanking regions as shown here.

(SEQ ID NO:193)
ACCACCATCACAATTARMTGGRAAMMACCGRAARAWGATGGCGGCARMMM
AATTAMARRATTCCAAGTCGACGCA

The underlined portion encodes for the varied BC loop as described above. The 5′ flanking sequence encodes the C-terminal portion of the B β strand domain and the 3′ flanking sequence encodes the N-terminal portion of the C β strand domain.

For position 21, (arm) or (a a/g a/c): agc encodes S (wildtype) and aaa encodes K (WTM target). Additionally, aga encodes R and aac encodes N both of which are “co-products” from the WTM mixed base reaction.

For position 22, (tgg) encodes W.

For position 23, (raa) or (a/g a a) encodes E or K.

For position 24, (mma) or (a/c a/c a) encodes T, K, Q or P.

For position 25, (ccg) encodes P.

For position 26, (raa) or (a/g a a) encodes E or K.

For position 26a, (raw) or (a/g a a/t) encodes N, E, K or D.

For position 26b, (gat) encodes D.

For position 26c, (ggc) encodes G.

For position 26d, (ggc) encodes G.

For position 27, or arm or (a a/g a/c) encodes R, N, K or S.

For position 28, or mma (a/c a/c a) encodes T, K, Q or P.

For position 29, or (att) encodes I.

For position 30, or ama (a a/c a) encodes T or K.

For position 31, or rra (a/g a/g a) encodes R, E, G or K.

The WTM “co-products” contribute to extra library diversity. In following this exercise through for the remaining BC15 loop codons, we will generate a limited WTM (K) loop library of 8192 possible combination variants. We then continue applying WTM for the remaining WTM amino acids: G, S, H, L, P, Y, E, and Q (see Table 5). Overall, in completing the nine WTM amino acid series, a total BC15 loop library size of 105.

TABLE 6
WTM oligonucleotide design for BC/15
BC_15
21222324252626a26b26c26d2728293031
SEQ
X-SWX-EX-PPX-EX-DDGGX-SX-PIX-TX-GID NO:
WTM:Karmtggraammaccgraarawgatggcggcarmmmaattamarra82
WTM:GrgctgggrasscccggragrcGatggcggcrgcsscattrscggc83
WTM:Dkmttgggawsmtccggawgatgatggcggckmtsmtattrmcgrt84
WTM:Qmrstggsagcmgccgsagsakgatggcggcmrscmgattmmgsrg85
WTM:Sagctggrrmyctccgrrmrrcgatggcggcagcyctattascrgc86
WTM:Hymttggsawcmtccgsawsatgatggcggcymtcmtattmmcsrt87
WTM:Ytmttggtatymtccgtatkatgatggcggctmtymtattwmckrt88
gaagaa
WTM:Ltyatggswgcygccgswgswtgatggcggctyacygattmygskg89
WTM:Pycttggsmgccgccgsmgsmtgatggcggcyctccgattmcgssg90

SPLIT POOL Synthesis: WTM Y Oligonucleotides

For performing tyrosine (Y) WTM where the starting “wildtype” amino acid is glutamate (E), an unwanted side effect may be the introduction of a stop codon. Tyrosine is encoded by tat and tac, glutamate is encoded by gaa and gag, the amber stop codon is tag and the ochre stop codon is taa. Substituting a tyrosine for a glutamate, therefore, requires that the first position be at for the tyrosine or g for glutamate, the second position be an a, and the third position be at or c for tyrosine or an a or g for glutamate. Thus, a standard nucleotide mixture designed as described above may result in a taa or tag.

In this situation it can be preferable that alternate oligonucleotide synthesis procedure be utilized where pools of codons are synthesized separately then combined and split for the following round of synthesis (E A Peters, P J Schatz, S S Johnson, and W J Dower, J. Bacteriol. 1994 July; 176(14): 4296-4305.). This split pool synthesis allows for the introduction of these (Y) WTM residues without producing unwanted co-products such as the amber stop codon that occurs in regular single reaction oligonucleotide synthesis. In split pool synthesis process, two pools are utilized: the first pool utilizes the codon TAT, encoding Y, and the second pool utilizes the codon GAA, encoding E. In this split pool design, the amber stop codon is not made as each reaction pool does not code for the UAG or UAA stop codon. For BC15 the (Y) WTM oligonucleotide reaction is made until position 26 whereupon the reaction is split into two separate columns for further synthesis.

In order to produce the defined mixture of amino acids, two columns are utilized. In the first, the fixed 3′ portion of the oligonucleotides are synthesized as defined by the flanking regions and the fixed portion of the BC15 loop shown above. For position 26 in the example sequence above, the first column synthesizes the codon TAT (TAT in the 3′-5′ DNA synthesis). The second columns synthesize the codon GAA (AAG in the 3′-5′ DNA synthesis). After the three nucleotides are coupled, the two columns are opened, the synthesis support is removed by washing with acetonitrile, and the resins are pooled. After mixing, the resin is placed into one column. At this point, the next position, positions 25 and 24 are synthesized. The single column synthesizes the codon CGG and YMT as described above. The resin is then split for (Y) WTM position 23 and reapportioned as described for position 26. After which, the syntheses are pooled and continued until the 5′ fixed and flanking region are.

WTM oligonucleotide design for the BC/14 and BC/11 loops are the carried out in similar fashion.

In reference to FN3 BC/14 which the variability profile is described as:

BC_1421222324252626a26b26c2728293031
X-SWX-KX-PPX-DX-DX-PX-NX-GX-PIX-TX-G

TABLE 7
WTM oligonucleotide design for BC/14
BC_14
21222324252626a26b26c2728293031
SEQ
X-SWX-KX-PPX-DX-DX-PX-NX-GX-PIX-TX-GID NO
WTM:KarmtggaaammaccgrawRawmmaaawrrammaattamarra91
WTM:GrgctggrrasscccggrcGrcsscrrcggcsscattrscggc92
WTM:DkmttggrawsmtccggatGatsmtratgrtsmtattrmcgrt93
WTM:QmrstggmagcmgccgsakSakcmgmaksrgcmgattmmgsrg94
WTM:SagctggarmyctccgrrcRrcycyarcrgcycyattascrgc95
WTM:HymttggmawcmtccgsatSatcmtmatsrtcmtattmmcsrt96
WTM:YtmttggaaaymtccgkatKatymtwatkrtymtattwmckrt97
tat
WTM:LtyatggmwgcygccgswtSwtcygmwtskgcygattmygskg98
WTM:PycttggmmgccgccgsmtSmtccgmmtssgccgattmcgssg99

In reference to FN3 BC11 which the variability profile is described as:

BC_112122232425262728293031
X-SWX-TX-PPX-PX-CX-PX-VX-DX-G

TABLE 8
WTM oligonucleotide design for BC/11
BC_11
2122232425262728293031
SEQ ID
X-SWX-TX-PPX-PX-CX-PX-VX-DX-GNO
WTM:KarmtggamammaccgmmaRrammarwgrawrra100
WTM:GrgctggrscsscccgsscGgcsscgkcgrcggc101
WTM:DkmttggrmcsmtccgsmtGrtsmtgwtgatgrt102
WTM:QmrstggmmgcmgccgcmgSrgcmgswgsaksrg103
WTM:SagctggascyctccgyctRgcyctrkcrrcrgc104
WTM:HymttggmmccmtccgcmtSrtcmtswtsatsrt105
WTM:YtmttggwmcymtccgymtKrtymtkwtkatkrt106
WTM:LTyatggmygcygccgcygSkgcygstgswtskg107
WTM:PycttggmcgccgccgccgSsgccgsvgsmtssg108

The co-products formed in using the above oligonucleotide design for the BC11 loop are shown in FIGS. 15a-15i

In reference to 10FN3 DE6 which the variability profile is described as:

DE LOOP515253545556
X-PX-GX-TX-EX-TX-S

Walkthrough 10FN3 DE/66 is performed at the appropriate variable positions and listed Table 6. As above, in the sequence denoted with an X refers to the walkthrough amino acid, and the amino acid(s) following the (dash) - refer to the starting base wild type amino acid.

TABLE 9
WTM oligonucleotide design for DE/6
DE_6
515253545556
SEQ
X-PX-GX-TX-EX-TX-SID NO
WTM:KmmarraAmaraaamaarm109
WTM:GsscggcRscgrarscrgc110
WTM:DsmtgrtRmcgawrmckmt111
WTM:QcmgsrgMmgsagmmgmrs112
WTM:SyctrgcAscrrmascagc113
WTM:HcmtsrtMmcsawmmcymt114
WTM:YymtkrtWmckawwmctmt115
WTM:LcygskgMygswgmygtya116
WTM:PccgssgMcgsmgmcgyct117

The co-products formed in using the above oligonucleotide design for the DE/6 loop are shown in FIGS. 17a and 17b for walkthrough amino acids Q, K, H, G, S, L, P and Y in FIGS. 15a and 15b.

In reference to FN3 FG/11 which the variability profile is described as:

FG_9767778798081828384
X-NX-AX-AGX-VG/SX-PX-PS

TABLE 10
WTM oligonucleotide design for FG/11
FG_9
767778798081828384
SEQ ID
X-NX-AX-AGX-VG/SX-PX-PSNO
WTM:KaawrmgrmgggcRwgrgcmmammaAgc118
WTM:GrrcgscgscggcGkcrgcsscSscagc119
WTM:DratgmcgmcggcGwtrgcsmtSmtagc120
WTM:QmaksmgsmgggcSwgrgccmgCmgagc121
WTM:SarckctkctggcRkcrgcyctyctagc122
WTM:HmatsmcsmcggcSwtrgccmtcmtagc123
WTM:YwatkmckmcggcKwtrgcymtymtagc124
WTM:LmwtsygsygggcStgrgccygcygagc125
WTM:PmmtscgscgggcSygrgcccgccgagc126

The co-products formed in using the above oligonucleotide design for the FG9 loop are shown in FIGS. 18a-18c.

In reference to FN36 which the variability profile is described as:

767778798084
FG_6X-NX-GX-GX-GX-ES

TABLE 11
WTM oligonucleotide design for FG_6
767778798084SEQ ID
FG_6X-NX-GX-GX-GX-ESNO
WTM:KaawrrarrarraRaaarm127
WTM:GrrcggcggcggcGrargc128
WTM:DratgrtgrtgrtGawkmt129
WTM:QmaksrgsrgsrgSagmrs130
WTM:SarcrgcrgcrgcRrmagc131
WTM:HmatsrtsrtsrtSawymt132
WTM:Ywatkrtkrtkrttattmt133
gaa
WTM:LmwtskgskgskgSwgtya134
WTM:PmmtssgssgssgSmgyct135

All 20 amino acids and unnatural amino acids utilizing the amber codon can potentially be walked through at the appropriate shaded positions.

Oligo Construction using Table 11, was carried out using extended walkthrough and doping as follows.

Extended walkthrough (for loop) can also be performed at the appropriate loop positions where noted in the sequence denoted with an X. The X refers to the walkthrough amino acid, and the amino acid(s) following the (dash) - refer to the base wild-type amino acid and any required co-products denoted after a (slash)/. Extended walkthrough is another example of WTM codon design where the co-product can be predetermined to occur. For example, the variability profiles have illustrated that G and P are frequent amino acids in the loop positions. The frequency of some G and P amino acids do not characterize them as either “fixed” or “wild-type” amino acids in the starting WTM sequence. However, by extended walkthrough, the degenerate WTM codons can be biased to that either G or P can be a WTM co-product.

TABLE 12
SEQ ID
BC_1421222324252626a26b26c2728293031NO
RegularX-SWX-KX-PPX-DX-DX-PX-NX-GX-PIX-TX-G
WTM
ExtendedX-SWX-KX-PPX-DX-DX-P/GX-N/GX-GX-PIX-TX-G
WTM
WTM:KarmtggaaamaaccgraaRawgatGgcarmmmaattamarra136
ExtendedarmtggaaamaaccgraaRawmcaRrmarmmmaattamarra137
WTM:K

Here, a second BC/14 loop variation is given to exemplify the use of extended walkthrough mutagenesis with required co-products. The regular WTM and extended WTM design in Table 11 for BC14 is shown above. In this case, regular WTM has P and N as the starting “wildtype” amino acid at positions 26b and 26c respectively. For WTM using K as the target amino acid, the required co-products are G at both positions.

For position 26b, we use the degenerate codon v(acg)ca. The codon aca codes for the WTM K amino acid, the codon cca codes P, the wildtype amino acid, and gca codes for G the required coproduct amino acid. Therefore a, c, and g are in the first position, c is in the second position, and a is in the third codon position.

Similarly, the degenerate codon rrm is used at position 26c. The codon aaa codes for K, aac codes for N, and gga codes for G. Therefore a, and g are in the first position, a and g are in the second position, and c and a are in the third codon position.

These extended WTM (K) results are summarized below and the 5′ and 3′ flanking codons are the same as the regular WTM (K) oligonucleotides. This principle would be applied to the remaining WTM amino acids (G, S, H, L, P, Y, D, and Q) so that degenerate codons are designed to incorporate the WTM, wildtype and any required co-products.

BC_1421222324252626a26b26c2728293031
RegularX-SWX-KX-PPX-DX-DX-PX-NX-GX-PIX-TX-G
WTM
ExtendedX-SWX-KX-PPX-DX-DX-P/GX-N/GX-GX-PIX-TX-G
WTM
WTM:KarmtggAaamaaccgraaRaagatarmmmaattamarra

TABLE 13
Doping of base mixture incorporation of extended BC_14 WTM (K)
oligonucleotide (SEQ ID NO: 138).
252626a26b26c27
5′-ccgaaaaaaacaaacaaa-3′
ggcggagc
g
705050504040405050
305050256060605050
25

Doping of the incoming base mixture can be performed to achieve predetermined amino acid ratios.

In position 26 of this example, it is desired that the WTM K (lysine) be incorporated into the loop 70% of the time as compared to the 30% of starting D (aspartic acid) wildtype amino acid. The usage of lysine is defined by the percentage of “a” utilized in the first codon position. Therefore, to achieve 70% lysine incorporation, the percentage of “a” in the mixture was introduced at 70% while “g” was introduced at 30% to achieve 30% aspartic acid.

In position 26a of this example, it is desired that the WTM K (lysine) be incorporated into the loop in equal 50% with 50% of starting D (aspartic acid) wildtype amino acid. Therefore, to achieve 50% lysine incorporation and 50% aspartic acid, the percentage of “a” and “g” were both adjusted to be introduced at 50% for oligonucleotide incorporation.

Similarly, in positions 96-99, the level of glycine incorporation was tuned to achieve an approximately 25% level of glycine incorporation while decreasing the level of co-product incorporation.

In the preferred usage, flanking regions are added to the 5′ and 3′ oligonucleotide regions to facilitate incorporation of the BC, DE and FG loops into the β-scaffold FN3 domain. The flanking regions have complementary base pairing with the oligonucleotides encoding other β-strands so that SOE-PCR can be performed (see FIG. 20). The following flanking scaffold 5′ and 3′ sequences are used for the BC, DE and FG loops:

B′ β-strand1617181920BC loop3132333435C′ β-strand
ACCTCTTTGCTTATATCGGATCACGTAC (SEQ ID NO: 194)
D′ β-strand4647484950DE loop5657585960E′ β-strand
F′ β-strand7172737475FG loop8586878889G′ β-strand

Example 5

WTM Oligonucleotide Design (No Threshold)

In another embodiment it is possible to execute WTM mutagenesis for the entire length of each loop, all the positions in the loops were considered as variable (threshold=100%). Therefore this library is a superset of the one in the previous example. In addition, because of how WTM works, the most abundant amino acid at each position is encoded at each and every position in the loops with a frequency between 25% and 50%. Thus, the libraries will cover a much larger sequence space, yet still be biased towards the amino acid pattern found in the database.

Positions at the strand-loop boundary require special handling. In fact residues at this positions might very important for the stability of the framework. Therefore if such positions are highly conserved in the related variability profile, they will be kept fixed (see S21, W22 and G31 in loop BC11; S21 and W22 in loops BC14 and BC15; N76 in loops FG8 and FG11).

TABLE 14
Summary of loop sequences for the universal Fibronectin III binding
domains library
BC LOOP
2122232425262728293031
BC_11SWX-TX-PX-PX-PX-GX-PX-VX-DG
21222324252626a26b26c2728293031
BC_14SWX-KX-PX-PX-DX-DX-PX-NX-GX-PX-IX-TX-G
21222324252626a26b26c26d2728293031
BC_15SWX-EX-PX-PX-EX-DX-DX-GX-GX-SX-PX-IX-TX-G
DE LOOP
515253545556
DE_6X-PX-GX-TX-EX-TX-S
FG LOOP
7677787980848586
FG_8NX-GX-GX-GX-EX-SX-SX-K
7677787980818283848586
FG_11NX-AX-AX-GX-VX-GX-PX-PX-SX-SX-K

By following Table 14, the design of loop BC15 will be as described here below:

21222324252626a26b26c26d2728293031
BC_15SWX-EX-PX-PX-EX-DX-DX-GX-GX-SX-PX-IX-TX-G

To perform WTM using Lys (K), one would introduce selective degenerate codons to place both the starting “wild type” amino acid and the WTM (K) amino acid in the selected “variable” position.

21222324252626a26b26c26d2728293031
BC_15SWX-EX-PX-PX-EX-DX-DX-GX-GX-SX-PX-IX-TX-G
WTM:KagctggraammammaraarawRawrrarraarmmmaawaamarra

The WTM (K) oligonucleotide sequence for BC15 would then be: 5′-(agc)(tgg)(raa)(mma)(mma)(raa)(raw)(raw)(rra)(rra)(arm)(mma)(awa)(ama)(rra)-3′ (SEQ ID NO:82) (using degenerate base codes)

or
5′-(agc)(tgg)(a/g a a)(a/c a/c a)(a/c a/c a)(a/g a a)(a/g a a/t)(a/g a a/t)(a/g a/g a)(a/g a/g a)(a a/g a/c)(a/c a/c a)(a a/t a)(a a/c a)(a/g a/g a)-3′ using standard base codes.

The oligonucleotide to be used in the subsequent PCR reaction will contain both 5′ and 3′ flanking regions as shown here.

(SEQ ID NO:195)
ACCACCATCACAATTAGCTGGRAAMMAMMARAARAWRAWRRARRAARMMM
AAWAAMARRATTCCAAGTCGACGCA

The underlined portion encodes for the varied BC loop as described above. The 5′ flanking sequence encodes the C-terminal portion of the B β strand domain and the 3′ flanking sequence encodes the N-terminal portion of the C β strand domain.

For position 21, (agc) encodes S.

For position 22, (tgg) encodes W.

For position 23, (raa) or (a/g a a) encodes E or K.

For position 24, (mma) or (a/c a/c a) encodes P, K, Q or T.

For position 25, (mma) or (a/c a/c a) encodes P, K, Q or T.

For position 26, (raa) or (a/g a a) encodes E or K.

For position 26a, (raw) or (a/g a a/t) encodes D, E, K or N.

For position 26b, (raw) or (a/g a a/t) encodes D, E, K or N.

For position 26c, (rra) or (a/g a/g a) encodes G, R, E or K.

For position 26c, (rra) or (alg a/g a) encodes G, R, E or K.

For position 27, or arm or (a a/g a/c) encodes S, N, K or R.

For position 28, or mma (a/c a/c a) encodes P, K, Q or T.

For position 29, or (awa) or (a a/t a) encodes I or K.

For position 30, or ama (a a/c a) encodes T or K.

For position 31, or rra (a/g a/g a) encodes G, R, E or K.

This approach would be continued for the remaining WTM amino acids: G, S, H, L, P, Y, E, and Q (see Table 5).

TABLE 15
WTM oligonucleotide design for BC/15
21222324252626a26b26c26d2728293031SEQ ID
BC_15SWX-EX-PX-PX-EX-DX-DX-GX-GX-SX-PX-IX-TX-GNO
WTM:KagctggraammammaraaRawRawrrarraarmmmaawaamarra139
WTM:GagctgggrasscsscgraGrcGrcggcggcrgcsscrkcrscggc140
WTM:DagctgggawsmtsmtgawGatGatgrtgrtkmtsmtrwtrmcgrt141
WTM:QagctggsagcmgcmgsagSakSaksrgsrgmrscmgmwammgsrg142
WTM:SagctggrrmyctyctrrmRrcRrcrgcrgcagcyctakcascrgc143
WTM:HagctggsawcmtcmtsawSatSatsrtsrtymtcmtmwtmmcsrt144
WTM:YagctggtatymtymttatKatKatkrtkrttmtymtwwtwmckrt145
gaagaa
WTM:LagctggswgcygcygswgSwtSwtskgskgtyacygwtamygskg146
WTM:PagctggsmgccgccgsmgsmtSmtssgssgyctccgmytmcgssg147

WTM oligonucleotide design for the BC14 and BC11 loops are the carried out in similar fashion.

In reference to FN3 BC14 which the variability profile is described as:

21222324252626a26b26c2728293031
BC_14SWX-KX-PX-PX-DX-DX-PX-NX-GX-PX-IX-TX-G

TABLE 16
WTM oligonucleotide design for BC/14
21222324252626a26b26c2728293031SEQ ID
BC_14SWX-KX-PX-PX-DX-DX-PX-NX-GX-PX-IX-TX-GNO
WTM:Kagctggaaammammarawrawmmaaasrrammaawaamarra148
WTM:Gagctggrrasscsscgrcgrcsscrrtggcsscrkcrscggc149
WTM:Dagctggrawsmtsmtgatgatsmtratgrtsmtrwtrmcgrt150
WTM:Qagctggmagcmgcmgsaksakcmgmcwsrgcmgmwammgsrg151
WTM:Sagctggarmyctyctrrcrrcyctarcrgcyctakcascrgc152
WTM:Hagctggmawcmtcmtsatsatcmtmatsrtcmtmwtmmcsrt153
WTM:Yagctggaaaymtymtkatkatymtwatkrtymtwwtwmckrt154
tat
WTM:Lagctggmwgcygcygswtswtcygmwcskgcygwtamygskg155
WTM:Pagctggmmgccgccgsmtsmtccgmmtssgccgmytmcgssg156

In reference to FN3 BC11 which the variability profile is described as:

2122232425262728293031
BC_11SWX-TX-PX-PX-PX-GX-PX-VX-DG

TABLE 17
WTM oligonucleotide design for BC/11
2122232425262728293031SEQ ID
BC_11SWX-TX-PX-PX-PX-GX-PX-VX-DGNO
WTM:Kagctggamammammammarrammarwgrawggc157
WTM:Gagctggrscsscsscsscggcsscgkcgrcggc158
WTM:Dagctggrmcsmtsmtsmtgrtsmtgwtgatggc159
WTM:Qagctggmmgcmgcmgcmgsrgcmgswgsakggc160
WTM:Sagctggascyctyctyctrgcyctrkcrrcggc161
WTM:Hagctggmmccmtcmtcmtsrtcmtswtsatggc162
WTM:Yagctggwmcymtymtymtkrtymtkwtkatggc163
WTM:Lagctggmygcygcygcygskgcygstgswtggc164
WTM:Pagctggmcgccgccgccgssgccgsygsmtggc165

The co-products formed in using the above oligonucleotide design for the BC11 loop are shown in FIGS. 15a and 15b.

In reference to DE6 the variability profile is described as:**

515253545556
DE/6X-PX-GX-TX-EX-TX-S

Walkthrough DE/6 is performed at the appropriate variable positions and listed Table 6. As above, in the sequence denoted with an X refers to the walkthrough amino acid, and the amino acid(s) following the (dash) - refer to the starting base wild type amino acid.

TABLE 18
WTM oligonucleotide design for DE/6
515253545556SEQ ID
DE/6X-PX-GX-TX-EX-TX-SNO
WTM:KmmaRraAmaraaamaarm166
WTM:GsscGgcRscgrarscrgc167
WTM:DsmtGrtRmcgawrmckmt168
WTM:QcmgSrgMmgsagmmgmrs169
WTM:SyctRgcAscrrmascagc170
WTM:HcmtSrtMmcsawmmcymt171
WTM:YymtKrtWmckawwmctmt172
WTM:LcygSkgMygswgmygtya173
WTM:PccgSsgMcgsmgmcgyct174

The co-products formed in using the above oligonucleotide design for the DE/6 loop are shown in FIGS. 17a and 17b.

In reference to the FG/11 loop the variability profile is described as:

7677787980818283848586
FG/11NX-AX-AX-GX-VX-GX-PX-PX-SX-SX-K

TABLE 19
WTM oligonucleotide design for FG/11
7677787980818283848586SEQ ID
FG/11NX-AX-AX-GX-VX-GX-PX-PX-SX-SX-KNO
WTM:KaatrmgrmgRraRwgRrammammaarmarmaaa175
WTM:GaatgscgscGgcGkcGgcsscsscrgcrgcrra176
WTM:DaatgmcgmcGrtGwtGrtsmtsmtkmtkmtraw177
WTM:QaatsmgsmgSrgSwgSrgcmgcmgmrsmrsmag178
WTM:SaatkctkctRgcRkcRgcyctyctagcagcarm179
WTM:HaatsmcsmcSrtSwtSrtcmtcmtymtymtmaw180
WTM:YaatkmckmcKrtKwtkrtymtymttmttmtaaa181
tat
WTM:LaatsygsygSkgStgSkgcygcygtyatyamwg182
WTM:PaatscgscgSsgSygSsgccgccgyctyctmmg183

The co-products formed in using the above oligonucleotide design for the FG/9 loop are shown in FIGS. 18a-18c.

In reference to the FG/8 loop the variability profile is described as:

7677787980848586
FG/8NX-GX-GX-GX-EX-SX-SX-K

TABLE 20
WTM oligonucleotide design for FG/8
7677787980848586SEQ ID
FG/8NX-GX-GX-GX-EX-SX-SX-KNO
WTM:Kaatrrarrarraraaarmarmaaa184
WTM:Gaatggcggcggcgrargcrgcrra185
WTM:Daatgrtgrtgrtgawkmtkmtraw186
WTM:Qaatsrgsrgsrgsagmrsmrsmag187
WTM:Saatrgcrgcrgcrrmagcagcarm188
WTM:Haatsrtsrtsrtsawymtymtmaw189
WTM:Yaatkrtkrtkrttattmttmtaaa190
gaatat
WTM:Laatskgskgskgswgtyatyamwg191
WTM:Paatssgssgssgsmgyctyctmmg192

Example 6

Methods for Genetically Engineering a Fibronectin Binding Domain Library

In this example, the steps for making and assembling a universal fibronectin binding domain library using genetic engineering techniques are described. The approach in making the WTM FN3 libraries is summarized below and in FIG. 20.

Briefly, the fibronectin modules are cloned using standard molecular biology techniques. The oligonucleotides encoding the beta-strand scaffold framework and diversity loops of the variable regions are assembled by the single overlap extension polymerase chain reaction (SOE-PCR) as illustrated in FIG. 20. The full-length molecules are then amplified using flanking 5′ and 3′ primers containing restriction sites that facilitate cloning into the expression-display vector(s). The total diversity of the libraries generated depends on the number of framework sequences and number of positions in the loops chosen for mutagenesis, e.g., using WTM.

For example, the diversity of the FN3 BC/11, FN3 DE/6 and FN3 FG/11 library, using 9 amino acids to conduct WTM™, is 3.5×106 (22 for FN3 BC/11×26 for FN3 DE/6×28 for the 13 amino acids FN3 FG/9) (see FIG. 21). The diversity of the is an upper limit and the diversity library from 1010 to 1011 which is within the range of the transformation efficiencies of bacterial systems.

Accordingly, 5 oligonucleotides are synthesized to encompass the β-strand frameworks (strands A, B, C, D, E, F, and G) for the fibronectin module libraries. To build a library based on the module 10FN3 scaffold, the β-strand oligonucleotides are listed below:

A′ beta-strand (sense strand encoding the
N-terminus of A beta strand with flanking
region containing a BamHI restriction site):
(SEQ ID NO:196)
5′-CATATTGCTGGTACTCAGGGATCCGTTAGTGACGTCCCA-3′
B′- beta-strand (antisense strand that spans
the AB loop):
(SEQ ID NO:197)
TATAAGCAAAGAGGTAGGAGTCGCTGCAACCACT
TCCAGATCCCGTGGGACGTCACTAAC-3′
C′- beta-strand (antisense strand that spans
the CD loop):
(SEQ ID NO:198)
5′-AACTGTGAATTCTTGAACTGGTGAGTTGCCACCTGTCTCTCCGTACG
TGATCCGATA-3′
D′- beta-strand (antisense strand that spans
the EF loop):
(SEQ ID NO:199)
5′-CACTGCATAGACAGTAATGGTATAGTCGACACCGGG
CTTAAGTCCAGAGATAGTTGC-3′
F′ beta-strand (antisense strand that encodes the
C-terminus of the F beta-strand, with flanking
region containing an XhoI restriction site):
(SEQ ID NO:200)
5′-ATGAACTGGGTATACTCTCTCGAGCGTCCGATAGTTGATGGA-3′

The B′, C′, D′, and F′ oligonucleotides described above are depicted in FIG. 20.

In addition, there are two flanking oligonucleotides that code for the N-terminal linker region and a C-terminal linker that have His and Myc immunotags respectively. In addition, a subset of 30-60 degenerate oligonucleotides in FN3 BC/11, FN3 DE/6 and FN3 FG/9 loops are synthesized (total 90-180). These oligonucleotides are assembled by the SOE-PCR method to generate the libraries that include the necessary FN3 BC/11, FN3 DE/6 and FN3 FG/9 combinations.

In the SOE-PCR reactions, the oligonucleotides for β-strands and the designed combinations of the BC, DE and FG loop libraries are added. PCR reactions used: 5 μl of 10 μM oligonucleotide mix, 0.5 μl Pfx DNA polymerase (2.5 U/μl), 5 μl Pfx buffer (Invitrogen), 1 μl 10 mM dNTP, 1 μl 50 mM MgSO4 and 37.5 μl dH20 at 94 C for 2 min, followed by 24 cycles of 30 sec at 94 C, 30 sec at 50 C, and 1 min at 68 C and then incubated for a 68 C for 5 min. The reactions were performed using a programmable thermocycler (MJ Research). An aliquot of the above SOE-PCR reaction is then taken to which the 5′ and 3′ primer pairs are added to a new PCR reaction.

Random clones from each library are then chosen for sequence verification and assessment of library quality with respect to desired mutational diversity, unintended point mutations, deletions, and insertions. This efficiency contrasts with random/stochastic mutagenesis strategies where uncontrolled introduction of various bases produces higher levels of undesired base change effects leading to low expression or fibronectin binding domains functionality due to unfavorable amino acid usage and inadvertent stop codons.

Individual FN3 modules can then be subsequently linked with together with natural FN linking sequences or use of synthetic poly-Gly-Ser linker (typically GGGGSGGGGSGGGGS) (SEQ ID NO:201) to generate multimeric binding domains.

Example 7

Methods for Performing High-Throughput Affinity Maturation of Candidates from a Universal Fibronectin Binding Domain Library

In this example, the steps for identifying and improving a candidate fibronectin binding domain using CBM (combinatorial benefical mutations) /WTM/LTM affinity maturation is described.

Briefly, a known 9-14 TNFα fibronectin binding domain can be designated as a test clone and mutagenized (using, e.g., CBMWTMLTMtechnology), expressed, displayed, and improved according to the methods of the invention. Regarding loop diversity, LTM is used to explore small perturbations within the loops (e.g., one change per loop). For further improvement, combinatorial beneficial mutagenesis (CBM), is subsequently used to exhaustively incorporate the individual LTM changes into the loop landscape(s).

The test lysozyme fibronectin binding domain was expressed and displayed using phage display, although any of the above-mentioned yeast/bacterial display systems can also be used.

Example 8

Screening and Analysis of WTM 14/FN3 Libraries for Anti-TNFα Binding Activity

WTM Library Construction

The WTM fibronectin libraries were assembled by polymerase cycling assembly (PCA) and cloned into a phagemid vector. The oligonucleotides share fifteen base pair homology with its reverse orientated neighboring oligonulceotide. The complete fibronectin gene was assembled using eight oligonucleotides. Oligo 3 contains the eleven amino acid BC loop, Oligo 5 encodes the six amino acid DE loop and Oligo 7 contains the 9 amino acid FG loop.

The 8 framework oligos for wildtype 14FN3 are:

Oligo 1: 14FN3_FMK_1_S
(SEQ ID NO:202)
CATATTGCTGGTACTCAGGGATCCAACGTTAGTCCGCCT
Oligo 2: 14FN3_FMK_2_S
(SEQ ID NO:203)
AATTGTGATGGTGGTCTCGGTGGCGTCTGTGACGCGGGCACGCCGAGGCG
GACTAACGTT
Oligo 3: 14FN3_FMK_3_S
(SEQ ID NO:204)
ACCACCATCACAATTAGCTGGCGCACTAAAACGGAAACAATCACCGGTTT
CCAAGTCGACGCA
Oligo 4: 14FN3_FMK_4_A
(SEQ ID NO:205)
GATTGTCCGCTGAATGGGAGTTTGGCCGTTGGCTGGGACTGCGTCGACTT
GGAA
Oligo 5: 14FN3_FMK_5_S
(SEQ ID NO:206)
ATTCAGCGGACAATCAAGCCTGACGTTCGTTCTTATACCATTACCGGG
Oligo 6: 14FN3_FMK_6_A
(SEQ ID NO:207)
CAAAGTATACAGGTAGATTTTATAATCAGTTCCTGGCTGTAACCCGGTAA
TGGTATA
Oligo 7: 14FN3_FMK_7_S
(SEQ ID NO:208)
TACCTGTATACTTTGAATGACAACGCGCGTAGCAGTCCGGTGGTTATAGA
TGCCAGC
Oligo 8: 14FN3_FMK_8_A
(SEQ ID NO:209)
TGAACTGGGTATACTCTCTCGAGCGTGCTGGCATCTATAAC

The UF3L-WTM-TYR construction use Oligo 1, Oligo 2, Oligo 4, Oligo 6, and Oligo 8 of the wildtype 14FN3 framework sequence. Oligo 3, Oligo 5A/Oligo 5B, and Oligo 7A/Oligo 7B are replace with the following:

Oligo 3: 14FN3UL_w3_S_WTM-TYR
(SEQ ID NO:210)
ACCACCATCACAATTAGCTGGWMCYMTYMTYMTKRTYMTKWTKATGGCTT
CCAAGTCGACGCA
Oligo 5A: 14FN3UL_w5_S_WTM_TYR1
(SEQ ID NO:211)
ATTCAGCGGACAATCYMTKRTWMCTATWMCTMTTATACCATTACCGGG
Oligo 5B: 14FN3UL_w5_S_WTM_TYR2
(SEQ ID NO:212)
ATTCAGCGGACAATCYMTKRTWMCGAAWMCTMTTATACCATTACCGGG
Oligo 7A: 14FN3UL_w7_S_WTM_TYR1
(SEQ ID NO:213)
TACCTGTATACTTTGAATKMCKMCKRTKWTKRTYMTYMTTMTTMTAAACC
GGTGGTTATAGATGCCAGC
Oligo 7B: 14FN3UL_w7_S_WTM_TYR2
(SEQ ID NO:214)
TACCTGTATACTTTGAATKMCKMCKRTKWTKRTYMTYMTTMTTMTTATCC
GGTGGTTATAGATGCCAGC

The 14 amino acid extended BC loop library, UF3L-WTM-TYR-BC_extended, construction use Oligo 1, Oligo 2, Oligo 4, Oligo 6, and Oligo 8 of the wildtype 14FN3 framework sequence. Oligo 3A, Oligo 3B, Oligo 3C, Oligo 3D, Oligo 5A/Oligo 5B, and Oligo 7A/Oligo 7B are the same as in the UF3L-WTM-TYR library.

Oligo 3A BC ext: 14FN3UL_w3_15_S_WTM_TYR1
(SEQ ID NO:215)
ACCACCATCACAATTAGCTGGTATYMTYMTTATKATKATKRTKRTTMTYM
TWWTWMCKRTTTCCAAGTCGACGCA
Oligo 3B BC ext: 14FN3UL_w3_15_S_WTM_TYR2
(SEQ ID NO:216)
ACCACCATCACAATTAGCTGGTATYMTYMTGAAKATKATKRTKRTTMTYM
TWWTWMCKRTTTCCAAGTCGACGCA
Oligo 3C BC ext: 14FN3UL_w3_15_S_WTM_TYR3
(SEQ ID NO:217)
ACCACCATCACAATTAGCTGGGAAYMTYMTTATKATKATKRTKRTTMTYM
TWWTWMCKRTTTCCAAGTCGACGCA
Oligo 3D BC ext: 14FN3UL_w3_15_S_WTM_TYR4
(SEQ ID NO:218)
ACCACCATCACAATTAGCTGGGAAYMTYMTGAGKATKATKRTKRTTMTYM
TWWTWMCKRTTTCCAAGTCGACGCA

All eight oligos are mixed together at 10 μM to assemble the fibronectin gene by PCA. The 50 μl Assembly PCR 1 mixture contains 5 μl pooled oligos (10 μM), 10 μl 5× Buffer (Finnzymes), 1 μl 10 mM dNTPs, 0.5 μl Phusion Polymerase (Finnzymes), 33.5 μl diH2O. The PCR protocol is one cycle of 98° C. for 30 sec, followed by 30 cycles of 98° C. for 7 sec, 50° C. for 20 sec, and 72° C. for 15 sec. The final cycle is 72° C. for 1 min.

The 100 μl Rescue PCR mixture contains 2.5 μl of the PCR 1 reaction, 2.5 μl of Oligo 1: 14FN3_FMK1_S, 2.5 μl of Oligo 8: 14FN3_FMK8_A, 2 μl 10 mM dNTPs, 20 μl 5× Buffer (Finnzymes), 1 μl Phusion Polymerase (Finnzymes), 69.5 μl diH2O. The PCR protocol is one cycle of 98° C. for 30 sec, followed by 30 cycles of 98° C. for 7 sec, 50° C. for 20 sec, and 72° C. for 15 sec. The final cycle is 72° C. for 1 min.

The PCR amplicons were purified using the Qiagen Qiaquick PCR Clean-up Kit by manufacturer's protocol and eluted off the spin column with 80 μl diH2O. The PCR fragment was digested at 37° C. with the restriction enzymes BamHI (NEB) and XhoI (NEB), gel purified on a 1.2% agarose gel and the DNA band was purified using the Qiagen Gel Extraction Kit.

The backbone DNA for the phage display libraries is a pBluescript-based phagemid vector. This vector features: (1) a N-terminal signal sequence from the E. Coli DsbA gene for periplasmic export, (2) a multiple cloning site to insert the FN3-based scaffolds, (3) a c-myc epitope tag for monitoring protein expression, (4) a hexahistidine tag for protein purification, (5) the C-terminal domain (aa250-406) of the M13 p3 protein for display of the FN3 on the phage surface, and (6) an amber stop codon (TAG) inserted between the FN3 and the M13 p3 CTD which allows switching between expression of membrane anchored FN3-p3 fusion proteins and soluble FN3 proteins simply by using different E. Coli strains. Expression of fusion proteins is under the control of an inducible lac promoter.

The 20 μg of phagemid vector was digested at 37° C. with the restriction enzymes BamHI (NEB), XhoI (NEB), and Calf Intestinal Alkaline Phosphatase (CIP from NEB) until completion. The backbone DNA was gel purified on a 1.2% agarose gel and the DNA band was purified using the Qiagen Gel Extraction Kit. Overnight ligations were set-up with the phagemid backbone DNA and the gene synthesized PCR fragment. The ligations were purified by phenol/chloroform extraction and ethanol precipitated and electroporated into E. Coli XL-1 Blue cells (Stratagene). Library transformation efficiency was determined by serial dilution tittering.

The effective library sizes for UF3L-WTM-TYR-BC_extended was 8.0e7 variants and for UF3L-WTM-TYR was 1.2e8 variants.

Phage Display

Phage particle library stocks were produced by starting with an initial inoculum of library TG1 cells at ten times the size of the library (media=2YT,1% glucose, 50 ug/ml ampicillin). When cells reached an OD600=0.5, an MOI=20 of M13K07 helper phage (Invitrogen) is added and the cells are incubated at 37° C. for 30 minutes. Next the cells shake for an additional 30 minutes at 37° C. The cells are spun down and resuspended in induction media (2YT, 50 ug/ml ampicillin, 25 ug/ml kanamycin, and 0.1 mM IPTG) and shaked overnight at 30° C. Phage is purified from the supernatant by standard 20% PEG/2.5M NaCl precipitation protocol and tittered by serial dilution in TG1 cells.

The following conditions were used for the panning of the library against the target TNF-α by phage display. The first three rounds of panning were preformed on Reacti-Bind™ NeutrAvidin™ Coated High Binding Capacity (HBC) Clear 8-well Strips with Superblock® Blocking Buffer (Pierce) and round 4 was done on a MaxiSorp plate (Nunc). This will eliminate any nuetravidin binding phage and phage that only recognize the biotintylated TNF-α target protein. TNF-Negative controls were carried out with no biotintylated-TNFα (R&D Systems) added. Different blocking reagents will be used including BSA, ovalbumin, and instant milk for each round of panning.

For Round 1, 0.25 ug of biotintylated TNF-α in 0.5% BSA blocking solution (100 ul) bound to the plate for two hours and then washed with PBS-0.05% Tween-20. Next the overnight binding buffer was added which contains 1e1013 library phage particles, 1 uM biotin, and 0.2% BSA, 0.1% Tween-20 in PBS. The next day the wells were washed 5 times with 0.2% BSA, 0.1% Tween-20 in PBS and five more times with 0.05% Tween-20 in PBS. The phage was eluted with 100 ul 100 mM HCL and neutralized in 1 ml of 1.0 M Tris-HCl, pH 8.0. The phage was tittered on TG1 cells. For round 1, the output phage was 1.7e7 phage/ml and the negative control was at 2.4e6 phage/ml. This is a 7.1 fold increase in enrichment.

For Round 2, 0.25 ug of biotintylated TNF-α in 0.5% ovalbumin blocking solution (100 ul) bound to the plate for two hours and then washed with PBS-0.05% Tween-20. Next the overnight binding buffer was added which contains 1e1013 Round 1 phage particles, 1 uM biotin, 0.1 mg/ml streptavidin, and 0.2% ovalbumin, 0.1% Tween-20 in PBS. The next day the wells were washed 5 times with 0.2% ovalbumin, 0.1% Tween-20 in PBS and five more times with 0.05% Tween-20 in PBS. The phage was eluted with 100 ul 100 mM HCl and neutralized in 1 ml of 1.0 M Tris-HCl, pH 8.0. The phage was tittered on TG1 cells. For round 2, the output phage was 3.0e6 phage/ml and the negative control was at 1.36e6 phage/ml. This is a 2.2 fold increase in enrichment.

For Round 3, 0.25 ug of biotintylated TNF-α in 0.5% instant milk blocking solution (100 ul) bound to the plate for two hours and then washed with PBS-0.05% Tween-20. Next the overnight binding buffer was added which contains 1e1013 Round 2 phage particles, 1 uM biotin, 0.1 mg/ml streptavidin, and 0.2% instant milk, 0.1% Tween-20 in PBS. The next day the wells were washed 5 times with 0.2% instant milk, 0.1% Tween-20 in PBS and five more times with 0.05% Tween-20 in PBS. The phage was eluted with 100 ul 100 mM HCl and neutralized in 1 ml of 1.0 M Tris-HCl, pH 8.0. The phage was tittered on TG1 cells. For round 3, the output phage was 5.66e7 phage/ml and the negative control was at 4.16e7 phage/ml. This is a 1.4 fold increase in enrichment.

For Round 4, 0.25 ug of biotintylated TNF-α in 0.5% instant milk blocking solution (100 ul) was bound to a MaxiSorp plate for two hours and then washed with PBS-0.05% Tween-20. Next the overnight binding buffer was added which contains 1e1013 Round 3 phage particles and 0.2% instant milk, 0.1% Tween-20 in PBS. The next day the wells were washed 5 times with 0.2% instant milk, 0.1% Tween-20 in PBS and five more times with 0.05% Tween-20 in PBS. The phage was eluted with 100 ul 100 mM HCl and neutralized in 1 ml of 1.0 M Tris-HCl, pH 8.0. The phage was tittered on TG1 cells. For round 4, the output phage was 1.0e6 phage/ml and the negative control was at 3.56e5 phage/ml. This is a 2.8 fold increase in enrichment.

ELISA Screening

After the third and fourth rounds of selection, a total of 373 clones were analyzed for TNF-α binding using phage ELISA. Briefly, phage from individual clones was produced by co-infection with M13K07 helper phage in 1 ml cultures in 96-deep well blocks. Phage-containing culture supernatants were then added to microtiter plates coated with the target protein. Following washes with PBS-0.1% Tween-20, bound phage was detected with a mouse anti-M13 antibody horseradish peroxidase conjugate (GE Biosciences) and TMB substrate (Pierce). Positive variants were then screened for binding specificity against VEG-F and nuetravidin plate only. From these assays, 48 positive clones were identified. As shown in FIG. 24b, these clones bound TNF-α but not VEGF or a blank well, indicating specific binding to TNF-α.

Three of these clones were screened for further analysis. The variants were expressed as soluble protein. Transforming the phagemid clone into the non-suppressing TOP10 E. Coli strain produced soluble protein. Protein expression was induced with 1 mM IPTG for 4 hours at 30° C. and the 6×His-tagged protein was purified by Ni-NTA spin columns (Qiagen). ELISA determined the binding specificity of these variants. As seen in FIG. 24a, all three variants bound TNFα, but did not cross-react with VEGF, indicating specific binding to TNFα. The results of this screen demonstrated that the 14FN3 scaffold is suitable to generate specific binding proteins.

Sequencing

After sequencing the 48 positive clones, 9 unique variants were identified. Sequencing of the anti-TNFα variants revealed significant convergence of the sequences, particularly in the BC and FG loops, as several of the clones contained the same FG loop. Interestingly, all the variants recovered were from the tyrosine WTM libraries, with a clear preference for a BC loop length of 14 amino acids. In addition, all of the variants contained cysteine residues in the BC and FG loops, with a definite selection for cysteines at positions 26C in the extended BC loop and 82 in the FG loop. The presence of these cysteines suggests that a disulfide bond may form between the BC and FG loop in these variants. Formation of a disulfide bridge between the BC and FG loops could serve to stabilize the 14FN3 structure and bring the loops closer together, resulting in a larger contact surface with TNFα and a higher binding affinity.

The sequences of 8 high-affinity binders are shown in FIG. 24C, and the nine high-affinity binders are identified as SEQ ID NOS: 55-63.

Cloning 14FN3 TNFα Binder as Fc Construct

To examine the properties of a 14FN3-Fc fusion, the best anti-TNFα variant, R41A6, was expressed as an Fc fusion protein. The template DNA for the Fc was from clone MGC:39273 IMAGE:5440834 9 (Ref ID: BC024289). The following oligonucleotides were used to PCR amplify the R41A6 gene with NcoI and XhoI restriction sites and the Fc region of human IgG1 with XhoI ends.

14FN3 NcoI For:
(SEQ ID NO:219)
ATCGCCATGGAAAACGTTAGTCCGCCTCGGCGT
14FN3 FMK 8 A:
(SEQ ID NO:220)
ATGAACTGGGTATACTCTCTCGAGCGTGCTGGCATCTATAAC
Fc XhoI For:
(SEQ ID NO:221)
ATCGCTCGAGCCCAAATCTTGTGACAAAAC
Fc XhoI Rev:
(SEQ ID NO:222)
CGATCTCGAGTTTACCCGGAGACAGGGAGA

The PCR reaction for A6 amplification contained 2 ul phagemid R41A6 mini-prep DNA (100 ng), 1 ul of 14FN3 NcoI For primer (20 uM), 1 ul of 14FN3 FMK 8 A primer (20 uM), 1 ul 10 mM dNTPs (NEB), 5 ul 10× Standard Taq Reaction Buffer (NEB), 1 ul Taq DNA Polymerase (NEB), and 39 ul diH2O. The PCR reaction for Fc amplification contained 2 ul Origene clone BC024289.1 mini-prep DNA (100 ng), 1 ul of Fc XhoI For primer (20 uM), 1 ul of Fc XhoI Rev primer (20 uM), 1 ul 10 mM dNTPs, 5 ul 10× Standard Taq Reaction Buffer (NEB), 1 ul Taq DNA Polymerase (NEB), and 39 ul diH2O. The PCR protocol is one cycle of 94° C. for 2 min., followed by 25 cycles of 94° C. for 30 sec, 60° C. for 30 sec, and 75° C. for 4 min. The final cycle is 75° C. for 3 min.

First the TNF-α binder R41A6 gene was cloned into the pET-20b expression vector (Novagen). This vector contains the pelB signal sequence for periplasmic targeting of the protein and also a C-terminal hexahistidine tag for purification. The 20 μg of pET-20b vector was digested at 37° C. with the restriction enzymes NcoI (NEB), XhoI (NEB), and Calf Intestinal Alkaline Phosphatase (CIP from NEB) until completion. The backbone DNA was gel purified on a 1.2% agarose gel, and the DNA band was purified using the Qiagen Gel Extraction Kit. The PCR amplicons were purified using the Qiagen Qiaquick PCR Clean-up Kit by manufacturer's protocol and eluted off the spin column with 80 μl diH2O. The R41A6 PCR fragment was digested at 37° C. with the restriction enzymes NcoI (NEB) and XhoI (NEB), gel purified on a 1.2% agarose gel and the DNA band was purified using the Qiagen Gel Extraction Kit. Overnight ligations were set-up with the digested pET-20b backbone DNA and R41A6 PCR fragment. The ligation DNA (1 ul) was electroporated into TOP10 cells (Invitrogen) and clones selected on LB-AMP plates. Clone pET-20b+TNF A6 #27 was recovered after screening mini-prep DNA by restriction digest and sequencing.

Next the Fc fragment is cloned into the XhoI site of the vector pET-20b+TNF A6 #27. The 20 μg of pET-20b+TNF A6 vector was digested at 37° C. with the restriction enzymes XhoI (NEB), and Calf Intestinal Alkaline Phosphatase (CIP from NEB) until completion. The backbone DNA was gel purified on a 1.2% agarose gel, and the DNA band was purified using the Qiagen Gel Extraction Kit. The PCR amplicons were purified using the Qiagen Qiaquick PCR Clean-up Kit by manufacturer's protocol and eluted off the spin column with 80 μl diH2O. The Fc PCR fragment was digested at 37° C. with the restriction enzymes XhoI (NEB), gel purified on a 1.2% agarose gel and the DNA band was purified using the Qiagen Gel Extraction Kit. Overnight ligations were set-up with the digested pET-20b+TNF A6 #27 (A6) backbone DNA and Fc PCR fragment. The ligation DNA (1 ul) was electroporated into TOP10 cells (Invitrogen) and clones selected on LB-AMP plates. Clone pET-20b+TNF A6+Fc #39 (A6+Fc) was recovered after screening mini-prep DNA for the correct orientation by restriction digest and sequencing.

The A6 and A6+Fc clones were expressed in BL-21(DE3) cells (Invitrogen) and purified using Talon Magnetic beads (Invitrogen). The A6 and A6+Fc clones in BL-21 were grown to a density of OD600=0.6 in LB media with 100 ug/ml ampicillin. At this density, IPTG was added to a final concentration of 1 mM, and the cells were put in the shaker at 30° C. for 4 hours. Cell pellets were frozen overnight at −80° C.

Thawed cell pellets were resuspended in lysis buffer (50 mM Sodium phosphate, pH 8, 500 mM NaCl, 40 mM Imidazole, 5% glycerol, 1× Bugbuster detergent (EMD), 1× protease inhibitors (Sigma), and 1 ul Lysonase (EMD)) and mix for 20 minutes at 4° C. Lysate is spun down at 14,000 rpm for 20 minutes at 4° C. and mixed with Talon magnetic beads for 1 hour at 4° C. Beads were pelleted on a magnet and washed five times with wash buffer (50 mM Sodium phosphate, pH 8, 500 mM NaCl, 40 mM Imidazole, 5% glycerol, and 0.01% Tween-20). Protein was eluted two times with 350 ul elution buffer (50 mM Sodium phosphate, pH 8, 500 mM NaCl, 150 mM Imidazole, 5% glycerol, and 0.01% Tween-20) and buffer exchanged into PBS using a 2 ml Zeba desalting column (Pierce). Protein concentration was quantified using the CBQCA Protein Quantification Kit (Invitrogen).

The Ka, Kdis, and KD values for soluble protein A6 and A6+Fc were determined using an Octet instrument (ForteBio), which measures refracted surface light interference. The target protein was biotintylated TNF-α soluble protein (R&D Systems) and is at a concentration of 0.25 ug/ml for loading onto the streptavidin sensor tips. The A6 protein was loaded at 290 nM and the A6+Fc protein was loaded a 370 nM. The Octet was run using the manufacturers protocol for Basic Kinetic Assay. The KD for A6 is between 20-100 nM and the KD for A6+Fc is between 12-25 nM. The Koff for A6+Fc was improved 2 fold as compared to the monomeric A6 protein. These results demonstrate that adding the Fc region to a 14FN3 binder can improve its binding properties.

Kinetic Binding Summary

Kon [1/Ms]Koff [1/s]KD
A6 + Fc9.12E+032.26E−0424.8 nM
A6 + Fc1.49E+042.20E−0414.9 nM
A6 + Fc1.95E+042.29E−0411.7 nM
A62.37E+044.38E−0418.4 nM
A62.74E+032.67E−0497.4 nM

Example 9

Screening and Analysis of Fn3 Libraries for Anti-VEGF Binding

In order to test the functionality of our combinatorial library design, a 14FN3 library with 2 natual-variant combinatorial loops (BC/11 andf DE/6) for binders designed. The library was constructed to contain the natural-variant amino acids or chemical equivalents shown at the top in FIG. 25B for the BC loop. A comparison of the natural amino acid variants given in FIG. 9 and those incorporated into the BC loop (FIG. 25A) shows that in general, the highest occurring 1-6 amino acid variants or their chemical equivalents were selected for at each position, where the number selected depended on the frequency distribution of the variants, minimizing where necessary codon changes leading to co-produced amino acids, such that the average number of variants per residue was somewhat greater than 4, yielding a total diversity of about 106. Similarly, the DE/6 loop was constructed to contain the natural-variant amino acids or chemical equivalents shown at the top in FIG. 25B for the BC loop. A comparison of the natural amino acid variants given in FIG. 12 and those incorporated into the DE loop (FIG. 25A) shows that in general, the highest occurring 6 amino acid variants or their chemical equivalents were used at each position, exept for position 1 where all nine variants were used. The greater average number of variants at each position was guided, in part, on the fewer number of residue positions, allowing more residues at each position while still retaining a total deversity of less than about 105.

The library was constructed in the phagemid vector as described above. To screen the library, phage were incubated with biotinylated VEGF(121) that had been pre-bound to streptavidin-coated magnetic beads for 1 hr at room temperature in PBS-2.5% nonfat dry milk. Following washes with PBS-0.1% Tween-20, bound phage were eluted with 0.1N HCl and propagated in TG1 cells for the next round of selection. In round 1 the VEGF concentration used was 1 μM and this was reduced 5-fold per round such that the final round of selection contained 20 nM VEGF. To measure relative enrichment, phage prepared from a non-displaying control phagemid (pBC), which confers chloramphenicol resistance, were also included in the selection reactions. In each round a 2 to 5-fold enrichment of library phage as compared to the non-displaying control phage was observed (Table 20).

TABLE 20
VEGF 14FN3 Combinatorial Library Selection
Round[VEGF]Phage TiterEnrichment
1  1 μM1.6 × 105ND
20.2 μM1.9 × 1055
3 40 nM4.0 × 1052
4 20 nM4.2 × 1053

After the selections, individual clones were screened for binding to VEGF by phage ELISA. A total of 380 clones were screened from rounds 3 and 4, identifying 37 positive clones (FIG. 25a). These positive clones were rescreened by phage ELISA for binding specificity against VEGF, TNFα or a blank plate

Of these 37 clones, 35 specifically bound VEGF. Sequencing of these clones identified 3 unique variants (FIG. 25b). One variant, R1D4, was found in 29/35 clones, while the other two were seen only once. This indicates significant convergence on a single sequence from the combinatorial library.

The most abundant clone, R1D4, was selected for further analysis. R1D4 was expressed as a soluble protein and purified using a Ni-chelating column. Using the Octet, the dissociation rate for binding to VEGF was measured as 1.7×10−4 (FIG. 25c). This is comparable to an anti-VEGF antibody which had a Koff value of 1.1×10−4. These findings validate the combination of combinatorial library design and 14FN3 scaffold as able to generate binders with significant affinity to their target.

Example 10

Screening and Analysis of FN3 Libraries for Anti-HMGB1 Binding Activity

HMGB1 Screen of 14FN3 Combinatorial Universal Library.

The 14FN3 natural-variant combinatorial library from Example 9 was also against another therapeutic target, HMGB1. (FIG. 26A shows the sequence variation at each position of the BC/11 and DE/6 loops). HMGB1 was initially described as a ubiquitous DNA-binding protein critical for chromatin stabilization and transcriptional regulation. More recently, extracellular HMGB1, when released by either necrotic cells or activated immune cells, has been shown to be an important pro-inflammatory cytokine. To identify 14FN3-based antibody mimics against HMGB1, the 14FN3 2 loop (BC/DE) combinatorial library for HMGB1 binders. Three rounds of phage display selection were performed as described above for VEGF. In round 1, either 0.5 or 0.1 μM HMGB1 was used. This was reduced 2.5-fold each round such that the final round selections contained either 80 or 16 nM HMGB1. In the final round of selection a 900 to 1300-fold enrichment of library phage over a non-displaying control phage was observed(Table 11).

TABLE 11
HMGB1 14FN3 Combinatorial Library Selection
[HMGB1]Phage TiterEnrichment
Round 10.5 μM6.3 × 104ND
0.1 μM1.2 × 105
Round 20.2 μM5.4 × 10516
 40 nM5.6 × 1057
Round 3 80 nM2.6 × 1071300
 16 nM4.4 × 107898

Following the selections, 439 individual clones from rounds 2 and 3 were screened by phage ELISA for binding to HMGB1. This screen resulted in 143 positive clones. Sequencing of 50 of these clones identified 15 unique 14FN3 variants (FIG. 26B, having sequences identified by SEQ ID NOS: 67-81. The most common variant was found in 22/50 clones, while 4 other sequences were also isolated multiple times. Interestingly, we observed significant similarity between the sequences, indicating convergence on a common set of traits. There was a clear selection for basic amino acids in the DE loop, in particular a KR(S/T)-H motif. The BC loop also contains several conserved residues: such as the valine or isoleucine at positions 24 and 29 and the arginine at positions 21 and 31. The similarities between the variants suggest that they may recognize a common epitope in HMGB1.

To analyze these variants effectively, a small-scale expression system to produce soluble protein was developed. Each clone was transformed into the non-suppressing E. coli strain NEB Express Iq. Next, a 2 ml culture was induced with 1 mM IPTG for 6 hrs at 30° C. The 6×His-tagged 14FN3 variants were then purified using Talon magnetic beads (Invitrogen). This procedure yielded from 5 to 10 μg of each soluble protein.

To determine the relative affinities of the variants, the dissociation rates of the individual proteins were measured using the Octet. Since the dissociation rate is the largest contributor to binding affinity and is not dependent on protein concentration, this analysis allowed us to quickly screen the variants and select the best proteins for further characterization. The clones displayed a range of dissociation rates from 1.4×10−2 to 1.0×10−6, with the majority between 1.5 to 2.0×10−4. Dissociation rates in this range suggest that these binders have a high affinity for HMGB1. We then selected six candidates for further investigation. We measured association and disassociation rates using the Octet (FIG. 26C). Table 22 shows that the obtained Kd values are in the low-nanomolar range.

Variantkon/104 (M−1 s−1)koff/10−4 (s−1)Kd (nM)
R2E68.35.06.0
R2G23.72.66.9
R2F89.92.22.2
R2B58.22.53.1
P2C61.86.235
Kinetic constants were determined from measurements using an Octet instrument with biotinylated HMGB1 immobilized on streptavidin biosensors.
Measurements were performed at 30° C.
Protein concentrations were determined using a CBQCA protein quantitation assay (Invitrogen) with a BSA standard curve.
The equilibrium dissociation constant, Kd, is calculated from the ratio of the rate constants, koff/kon.

Example 11

Screening and Analysis of FN3 Libraries for Catalytic Function

Protease Activity Plate Assays

In the case where the activity to be assayed is a proteolytic activity, substrate-containing nutrient plates can be used for screening for colonies which secrete a protease. Protease substrates such as denatured hemoglobin can be incorporated into nutrient plates (Schumacher, G. F. B. and Schill, W. B., Anal. Biochem., 48: 9-26 (1972); Benyon and Bond, Proteolytic Enzymes, 1989 (IRL Press, Oxford) p. 50). FN3 variant libraries can be displayed on the yeast cell surfaces or bacterial cell surfaces {Rutherford, 2006 #60} {Varadarajan, 2005 #61} as fusion proteins.

Alternatively, when bacterial colonies capable of secreting a protease are grown on these plates, the colonies are surrounded by a clear zone, indicative of digestion of the protein substrate present in the medium. A protease must meet several criteria to be detected by this assay. First, the protease must be secreted into the medium where it can interact with the substrate. Second, the protease must cleave several peptide bonds in the substrate so that the resulting products are soluble, and a zone of clearing results. Third, the cells must secrete enough protease activity to be detectable above the threshold of the assay. As the specific activity of the protease decreases, the threshold amount required for detection in the assay will increase. Proteases that are produced as phage p3 fusion proteins are also capable of enzymatic function (REF).

One or more protease substrates may be used. For example, hemoglobin (0.05-0.1%), casein (0.2%), or dry milk powder (3%) can be incorporated into appropriate nutrient plates. Colonies can be transferred from a master plate using and inoculating manifold, by replica-plating or other suitable method, onto one or more assay plates containing a protease substrate. Following growth at 37° C. (or the appropriate temperature), zones of clearing are observed around the colonies secreting a protease capable of digesting the substrate. Four proteases of different specificities and reaction mechanisms were tested to determine the range of activities detectable in the plate assay. The enzymes included elastase, subtilisin, trypsin, and chymotrypsin. Specific activities (elastase, 81 U/mg powder; subtilisin, 7.8 U/mg powder; trypsin, 8600 U/mg powder; chymotrypsin, 53 U/mg powder) were determined by the manufacturer. A dilution of each enzyme, elastase, subtilisin, trypsin, and chymotrypsin, was prepared and 5 μl aliquots were pipetted into separate wells on each of three different assay plates.

Plates containing casein, dry milk powder, or hemoglobin in a 1% Difco bacto agar matrix (10 ml per plate) in 50 mM Tris, pH 7.5, 10 mM CaCl2 buffer were prepared. On casein plates (0.2%), at the lowest quantity tested (0.75 ng of protein), all four enzymes gave detectable clearing zones under the conditions used. On plates containing powdered milk (3%), elastase and trypsin were detectable down to 3 ng of protein, chymotrypsin was detectable to 1.5 ng, and subtilisin was detectable at a level of 25 ng of protein spotted. On hemoglobin plates, at concentrations of hemoglobin ranging from 0.05 and 0.1 percent, 1.5. ng of elastase, trypsin and chymotrypsin gave detectable clearing zones. On hemoglobin plates, under the conditions used, subtilisin did not yield a visible clearing zone below 6 ng of protein.