Solid-Phase Synthesis and Combinatorial Technologies

a:incubation with biotinylated paclitaxel bound onto streptavidin plates; b: washing of the unbound clones;

c:elution of the bound phages; d: amplification of the selected population in E. coli;

e:steps a-d, two additional cycles; f: DNA sequencing, identification of peptide sequences.

CONSENSUS MOTIFS:

Figure 10.15 Screening of the phage library L6 to find macromolecular binding partners for paclitaxel: identification of Bcl-2 as a binding partner.

gene fragments). Both vector libraries have been processed, then ligated to assemble a VH–VL genetic library, which was eventually translated and displayed on a phage surface. The careful construction of the vectors and their ligation (C1–C4, linker structure, Fig. 10.17) ensured that only full mAbs-displaying phage particles maintained their infectivity, thus allowing a degree of self-control through the discarding of nonfunctional products of the biological machinery. Phagemids containing foreign DNA sequences fused onto the pIII gene and expressing one copy of mAbs on each phage capsid have been commonly used as genetic vectors.

Steinberger et al. (100) assembled a light-chain gene library and ligated it onto the pIII coat gene, then processed the construct and ligated a library of VH chains genes onto it (Fig. 10.18) to produce a focused library L8 of ≈5 × 107 individual antibodyclones. The lightand heavy-chain libraries were assembled into L8, ensuring the presence of a cleavable site between the recombinant antibody and the phage coat protein (Fig. 10.18). The encoding nucleotide sequences were derived from and biased toward IgE (immunoglobulin E) obtained from a grass pollen allergic patient. L8 was incubated onto coated plates with several purified recombinant grass pollen allergens for 2 h, binding phages were selected, and the amplification/selection cycle was repeated four times (steps a–e, Fig. 10.18). The binding clones were solubilized by cleavage of the junction to the pIII protein (step f) and 20 of them tested for their specificity for group 5 grass pollen allergens, and four specific sequences were identified and structurally characterized after DNA sequencing (steps g and h, Fig. 10.18). The VH fragment for all the isolated clones was identical, showing a selection consensus during the iterative cycles, and, although the VL fragments were extremely similar, several different patterns were observed, especially in CDR1 and CDR3

a: incubation with grass pollen allergens-coated plates, rt, 2 hrs; b: washing of the unbound clones; c: elution of the bound phages, pH 2.2; d: amplification of the selected population in E. coli;

e:steps a-d, four additional cycles; f: cleavage of the library; g: check of mAb specificity;

h:DNA sequencing, identification of peptide structures.

Figure 10.18 Screening of the IgE-biased phage library L8 of antibodies: the selection/amplification process.

(complementarity-determining regions). The four clones were tested on a panel of clinically relevant group 5 allergens from different grass species, their group 5 specificity and their recognition of allergens from different grass species were observed, and their potential use as therapeutic tools in allergology was forecasted. Many other reports of combinatorial antibody libraries, originating either from the optimization of known antibody structures (101–110) or from random assembly of heavyand light-chain fragments (111–120), have been reported together with several comprehensive reviews (121–125).

A specific class of antibodies selected to perform chemical transformations, the catalytic antibodies (126), has also been the target of phage display libraries. Fujii et al. (127) reported the affinity maturation of the esterolytic antibody Mab 17E11, generated by immunization with a transition-state analogue (TSA) 10.9 to allow the isolation of regioselective esterolytic activities at C4 (128). Mab 17E11 regiochemically deacylated the 4-position of the sugar 10.10 to give compound 10.11 (Fig. 10.19). The same catalytic antibody showed a greatly reduced activity when the C6-hindered analogue 10.12 (Fig. 10.19) was used as a substrate. An improvement of catalytic properties was sought via phage display of catalytic antibody libraries because a previous site-directed mutagenesis approach based on the antibody modeling in the presence of the substrate did not produce better catalytic antibodies for the hydrolysis of 10.12. The library L9 (107 clones) was generated by randomizing six amino acid positions in the heavy-chain CDR3 loop, known by molecular modeling to interact with the C6 region of the TSA. The library was screened on plates coated with bovine serum albumin–coupled TSA 10.13 (Fig. 10.20). The selected population was iterated for four rounds; then 124 randomly picked clones were rescreened and checked for their specific binding to 10.13 (steps a–h, Fig. 10.20). The 24 resulting clones were sequenced, and six of them showed different amino acid sequences compared with Mab 17E11 (three to six-residues changed). Two of these clones were found to be better catalysts than mAb 17E11 for the regiochemical deprotection of 10.12, the best

526 BIOSYNTHETIC COMBINATORIAL LIBRARIES

AcNH 10.11

AcNH 10.10

AcNH

NHAc

AcNH 10.12

poor substrate for mAb 17E11

Figure 10.19 Catalytic antibodies: structures of the transition state analogue/selection substrate (10.9), the catalyzed reaction substrate (10.10) and product (10.11), and a modified poor substrate for the esterolytic antibody Mab 17E11.

being roughly 12 times more efficient than the parent antibody. Four other reports describing libraries of catalytic antibodies have been published recently (86, 129– 131), including an intriguing report on the synthesis of selection reagents for the discovery of catalytic metalloantibodies.

Selection of a catalytic antibody via its binding to a TSA assumes that this binding will be predictive of the catalytic efficiency of the selected antibody. An intriguing application in which the selection of an antibody via a process based directly on catalytic efficiency was reported by Gao et al. (132). The acylated tripeptide amides 10.14a,b and esters 10.15a,b were used as substrates for hydrolysis to acids 10.16a,b (Fig. 10.21), and a catalytic antibody was looked for through the generation a display

6-mer random sequence in the CDR3 domain

a: incubation with conjugated albumin-10.13-coated plates, rt, 2 hrs; b: washing of the unbound clones; c: elution of the bound phages, pH 2.2; d: amplification of the selected population in E. coli; e: steps a-d, three additional cycles; f: affinity measurement with free 10.13; g: DNA sequencing, identification of peptide structures; h: selection of the most active mutated clones.

Figure 10.20 Screening of the mAb 17E11–biased antibody phage library L9 to find optimized catalytic antibodies for the transition state analogue/selection substrate 10.13: the selection/amplification process.

library L10 (2 × 108 members) starting with cDNA sequences from mouse spleen cells immunized with the boronic acid transition state analogue 10.17 (Fig. 10.21). Four rounds of selection/amplification produced several positive clones (steps a–f, Fig. 10.21) from which one was isolated, sequenced, and purified having full catalytic specificity for the amide 10.14a (4 × 104 enhancement of the reaction rate), with no effect on the hydrolysis of the stereoisomer 10.14b or the ester analogue 10.15a. The same authors reported two other examples of joint catalysis and infectivity in the selection of catalytic antibodies by phage display (133, 134); other researchers used similar approaches where the selected phage clones and the catalytic reagents and products were connected (135, 136).

Phage libraries of proteins have also been produced by insertion of DNA sequences from various species into the phage genome with expression of the foreign proteins as single copies per phage clone fused onto the pIII coat protein. The proteins displayed on phage capsids have been used to study potentially new protein–protein interactions to clarify relevant biological processes or to identify unknown homologue proteins in different organisms. Yamabhai et al. (137) screened a phage-expressed library of frog cDNA L11 for its binding to a biotinylated peptide 10.18 (Fig. 10.22) known to bind to Src homology 3 (SH3) domains of proteins. Identification of positive clones should have led to novel frog proteins involved in signal transduction (see the original paper for more details). This approach, named cloning of ligand targets (COLT) had been

528 BIOSYNTHETIC COMBINATORIAL LIBRARIES

L10

recombinant antibody

a: incubation with conjugated albumin-10.17-coated plates, 37°C, 2 hrs; b: washing of the unbound clones; c: elution of the bound phages, pH 2.2; d: amplification of the selected population in E. coli; e: steps a-d, three additional cycles; f: DNA sequencing, identification of peptide structures.

10.17

Figure 10.21 Direct selection of catalytic antibodies from phage libraries: the selection/amplification process to an improved stereoselective hydrolytic antibody from the libraryL10 using the transition state analogue/selection substrate 10.17.

previously used to identify other SH3-binding proteins (138). Screening of L11 (steps a–d, Fig. 10.22) produced a single clone that eventually led to the sequence of a novel protein named intersectin by the authors. Structural analysis highlighted the presence of five SH3 domains as well as two Eps15 homology (EH) domains in the structure of intersectin. These latter domains are known to be involved in protein–protein interaction.

The binding specificity of these EH modules was checked with two phage libraries L12 (linear 9-mers on pVIII coat protein) and L13 (11-mers with a terminal cysteine on pIII coat protein) that were incubated with plates coated with constructs containing the intersectin EH domains EHa and EHb (Fig. 10.23). After selection and amplification of L12 and L13, many positives containing the known NPF (asparagine–proline–

a: incubation with 10.18; b: washing of the unbound library individuals; c: hybridization; d: DNA sequencing, identification of the peptide structures.

10.18

Figure 10.22 Screening of the frog cDNA phage-expressed libraryL11 to find macromolecular binding partners for SH3 protein domains: the selection/amplification process using the biotinylated selection substrate 10.18.

L12

L13

a: incubation with EHa and EHb intersectin domains-coated plates; b: washing of the unbound

clones; c: elution of the bound phages; d: amplification of the selected population in E. coli; e: DNA sequencing, identification of the peptide structures.

10.21

CLONE FROM L13

BINDING WITH INTERSECTIN: CONFIRMED

Figure 10.23 Screening of the phage display libraries L12 and L13 to find macromolecular binding partners for intersection: the selection/amplification process and the structure of selected binding sequences 10.19–10.21 for the EH domains of intersectin.

530 BIOSYNTHETIC COMBINATORIAL LIBRARIES

phenylalanine) binding motif for EH domains were obtained. Further refinements of binding specificities and affinities showed the constrained peptide 10.21 from L13 to be a strong binder, while two linear examples from L8 (10.19 and 10.20) failed to show significant binding to intersectin (Fig. 10.23). These findings confirmed both the importance of constrained sequences (e.g., 10.21 completely lost its affinity for EH upon substitution of the cysteines with serines) and the amplification of weak affinities by pVIII display (2700 copies of pVIII-displayed, weakly actives 10.19 and 10.20 versus five copies of 10.21 by pIII display). The bound intersectin EH domains were also screened, albeit unsuccessfully, using the phage-expressed frog cDNA library L11 in the search for intersectin protein binders. However, a different cDNA library L14 from mouse embryo produced four positive clones. These were in turn connected to three protein sequences out of which two novel proteins, named intersectin-binding proteins (Ibp), were characterized. The characterization of SH3 and EH domains of intersectin was then completed using other phage libraries. Homology searches found similar proteins from other organisms and other candidate proteins potentially able to interact with the SH3 domains of frog intersectin. Two recent reviews (139, 140) covered extensively the use and applications of phage displayed cDNA libraries.

10.2 BIOSYNTHETIC OLIGONUCLEOTIDE LIBRARIES

10.2.1 General Considerations

Biosynthetic polypeptide libraries are the translation of a genetic, oligonucleotidebased diversity information into a biologically relevant pool of potential binders, catalysts, or receptors. The previous section illustrated how the availability of complex DNA sequences and of the appropriate techniques for their synthesis and manipulation allows the biological production of diverse polypeptide libraries. However, the intrinsic properties of nucleotide sequences have themselves stimulated researchers into building a variety of oligonucleotide (ON) libraries per se for a variety of applications.

ON libraries have been employed in two main fields. First, the ability of nucleotides to bind strongly to a variety of partners either as small oligomers or as larger DNA or RNA sequences is well known. Thus, various examples of ON libraries targeted toward the discovery of DNA, or more often RNA, ligands (aptamers) (141, 142) for other nucleotide sequences, small molecules, drugs, or proteins are described below. Second, the discovery of naturally occurring catalytic RNA sequences (ribozymes) (143, 144) has led both to the optimization of natural ribozyme sequences and to the quest for novel artificial ribozymes and deoxyribozymes via combinatorial ON libraries capable of catalyzing reactions such as oligonucleotide cleavage and ligation, peptide bond formation, ester hydrolysis, and C–C bond formation.

The first part of this section is devoted to the description of the in vitro selection process, which enables the iterative selection/amplification cycles to reliably and accurately furnish ON sequences.

10.2.2 In Vitro Selection of ON Sequences

The state of the art in the SPS of oligonucleotides (see Section 2.2) has reached excellent levels of reliability and performance, which can be summarized as follows:

•Sequences of >100 oligomers can be routinely synthesized with extremely high purity and yield.

•Libraries of up to 1016 individuals can be made and handled using various techniques.

•Four commercially available building blocks with similar reactivity and with orthogonal protection allow the construction of any target ON sequence.

•The reduced number of building blocks allows the complete randomization of long ONs (426 ≈ 1016 individuals).

•Efficient, fully automated SP ON synthesizers are commercially available.

Comparing the SP synthesis of ONs versus peptides, the latter does not exceed the complete randomization of octamers (208 ≈ 2.6 × 1010 individuals) and cannot attain the same yield and purity in each coupling step, thus preventing even the synthesis of single >100-mer target sequences. Therefore, the synthesis of large, randomized peptides is better carried out using biological tools, whereas the SPS of ONs produces large, high-quality libraries that are not biased by any of the potential problems of using the biosynthetic machinery of a cell, such as viability, infectivity, and depletion of amino acid pools. Only when the randomized sequence is extremely long, as is the case for ribozymes, are the DNA strands ligated using biochemical tools or prepared by genetic mutations of preexisting sequences (vide infra).

The main steps involved in the synthesis, selection, and amplification of ON libraries are reported in Fig. 10.24. Automated synthesis of libraries of DNA fragments of a given length usually uses the phosphoramidite SP strategy (Section 2.2.1) with equimolar quantities of each nucleotide at each elongation step to give equally represented ON libraries (step a). The random ssDNA fragments, which usually vary from 30 to 300 nucleotides, are flanked on both sides by constant regions that allow their amplification (step c) and transcription into the corresponding RNA sequences using RNA polymerase (step d) after conversion to double-strand DNA (dsDNA, step b). The RNA sequences are submitted to a target-assisted screening, which selects the active library components through their immobilization with the support-bound target (step e) and discards the unbound library individuals by washing (step f), as already seen for biosynthetic peptide libraries. The main principles encountered in the selection of phage libraries such as stringency and yields are also important here, but the increased quality of the library due to the use of equimolar quantities of library individuals and the lack of constraints related to the anchorage onto the phage capsid make the process extremely reliable and also allow further modification of the stringency conditions by fine tuning the selection parameters in vitro. After recovery of the selected RNAs by disruption of the target–RNA complex (step g), they are transcribed into ssDNA using reverse transcriptase (step h) and then converted to dsDNA (step b). This material is then amplified to allow iterative selection cycles to

532 BIOSYNTHETIC COMBINATORIAL LIBRARIES

flanking region

PA-ONs = phosphoramidite oligonucleotides

RNA sequence

cDNA sequence

a:chemical synthesis of ssDNA pools; b: dsDNA conversion; c: amplification;

d:transcription into RNA strands; e: incubation with target B; f: elimination of unbound RNAs;

g:disruption of target-RNA complex; h: reverse transcription into ssDNA;

i:PCR amplification; j: transcription into RNA; k: iterative cycle e-j, n times.

Figure 10.24 Synthesis, selection, and amplification of biosynthetic ON libraries: the whole process.

arrive at the best library individuals. At first, amplification was carried out using in vivo biological tools (145), as already described for the phage display techniques, but later more efficient in vitro systems were developed and among them the polymerase chain reaction (PCR, step i) (146) is by far the most used because of its fidelity and efficiency. The flanking regions of the DNA strand are kept constant and are used to drive the amplification of the random portion of the DNA. The amplified dsDNA sequences are then transformed into the corresponding RNA strands (step j), and a second iterative cycle is carried out (step k, Fig. 10.24). Typically, up to 15 cycles with increasing stringency are used to select up to 1000 best sequences.

The length of the random ON sequence of a library must be carefully considered in order to extract meaningful results from the screening as longer sequences may be the optimal binding partners for a target but their relative abundance in an ON library may be too low. In a hypothetical example of a 1014-member library of 20-mers, only one library individual contains a given 20-mer motif (420 = 1014), but around