Development of Bioactive Foldamers Using Ribosomally Synthesized Nonstandard Peptide Libraries

Abstract

Genetic code manipulation enables the ribosomal synthesis of peptide libraries bearing diverse nonproteinogenic amino acids, which can be applied to the discovery of bioactive peptides in combination with screening methodologies, such as mRNA display. Despite a tremendous number of successes in incorporation of l-α-amino acids with non-proteinogenic sidechains and N-methyl-l-α-amino acids into nascent peptide chains, d-, β-, and γ-amino acids have suffered from low translation efficiency. This obstacle has been hindering their integration into such peptide libraries. However, the use of engineered tRNAs, which can effectively recruit EF-Tu or/and EF-P, has recently made possible significant improvement of their incorporation efficiency into nascent peptides. This article comprehensively summarizes advances in such methodology and applications to the discovery of peptide ligands against target proteins of interest.

1. Introduction

Recently, peptides have attracted attention as a platform for developing novel drug leads. In terms of their molecular weights (typically 500–5,000 Da), peptides can be classified into mid-sized molecules compared with conventional small organic molecules (<500 Da) and large molecules like proteins, which include antibodies, and nucleic acids (>5,000 Da). The advantages of peptide drugs over small molecules are their higher target specificities and their abilities to regulate protein–protein interactions. This is due to the larger interaction surface areas of peptides against target proteins than is the case with small molecules. Moreover, various screening methodologies, such as mRNA display, ribosome display, and phage display have been developed to date and can be applied to facilitate the quick discovery of active peptide species.^1–5 Compared with large molecules, peptides have a potential to permeate cell membranes depending on their structures, which enables them to approach intracellular targets. In addition, their synthetic costs are generally much lower than those of large molecules.

In nature, peptides are generally created by ribosomal synthesis. A ribosome is able to utilize only the 20 kinds of canonical l-α-amino acids known as proteinogenic amino acids (PAA). However, the bioactive peptides developed to date consist of not only PAA but also diverse non-PAA (nPAA) such as N-alkyl-, d-, β-, and γ-amino acids as well as macrocyclic scaffolds.^6–12 They contribute to the unique folding propensities, improved serum stability, membrane permeability, and binding affinity of peptides. Therefore, there are increasing demands for methodologies to ribosomally introduce diverse nPAA into peptide libraries, which can be applied to the screening methodologies, such as mRNA display.

The relationship between mRNA codons and the corresponding amino acids is defined by the conserved genetic code in which the 20 PAA are exclusively assigned. Therefore, in order to introduce nPAA beyond the 20 PAA, the relationship must be artificially changed via manipulation of the genetic code. To date, various genetic code manipulation technologies have been developed to enable the incorporation of diverse nPAA into peptides. However, some kinds of inefficient nPAA (like d-, β-, and γ-amino acids) cannot be easily introduced.^13–18 In particular, the consecutive incorporation of such nPAA is a formidable challenge, indicating that the preparation of peptide libraries containing those amino acids is almost impossible. Thus, further engineering of the ribosomal translation system to broaden its substrate tolerance is essential.

This article summarizes the authors’ recent achievements in improving the ribosomal translation system to enable the consecutive incorporation of d- and β-amino acids and the non-consecutive multiple incorporation of γ-amino acids. A ribosomally synthesized macrocyclic peptide library containing three types of cyclic β-amino acids was applied to an mRNA display, by which we obtained strong binder peptides against human FXIIa and IFNGR1 with remarkably high binding affinity, inhibitory activity, serum stability, and unique folding propensity.¹⁹

2. Genetic Code Manipulation Methodologies for Incorporation of nPAA

2.1 Preparation of nPAA-tRNA

In mRNA, a codon comprises a triplet of the four types of nucleotides (A, U, G, and C) and designates one amino acid. In the canonical genetic code, 61 sense codons designate the 20 PAA, and the remaining three codons function as stop codons. As aminoacyl-tRNA is the key molecule for decoding, substituting a canonical PAA-tRNA with a wrongly charged nPAA-tRNA enables incorporation of the nPAA at the cognate codon designated by the anticodon of tRNA.

For preparation of such nPAA-tRNA, Suga et al. developed a series of versatile aminoacylation ribozymes, which are referred to as flexizymes: e.g. enhanced flexizyme (eFx) and dinitro-flexizyme (dFx).^20,21 The eFx can acylate amino acids activated as cyanomethyl ester or p-chlorobenzyl thioester onto arbitrary tRNA, whereas dFx utilizes amino acids activated as 3,5-dinitrobenzyl ester. Owing to the broad substrate scope of the flexizymes, l-α-amino acids with side chain modification, N-alkyl-amino acids, d-α-amino acids, β-amino acids, γ-amino acids, and even α-hydroxyl acids have been successfully charged onto tRNA.^{19,20,22–24} Moreover, since the flexizymes recognize only the conserved 5′-NCCA-3′ sequence at the 3′-terminus of tRNA, virtually any tRNA species can be applied to the flexizyme reaction. Specifically, a combination of aminoacyl-tRNA prepared using flexizyme technology with a reconstituted in vitro translation system is referred to as the flexible in vitro translation (FIT) system and has been applied to the incorporation of diverse nPAA into peptides.⁷

2.2 Genetic Code Reprogramming

When manipulating the genetic code, the artificial nPAA-tRNA should not conflict with the endogenous PAA-tRNA, as this would cause the nPAA and PAA to be competitively introduced into the same codon. In order to circumvent this, several methodologies have been developed. In the non-sense codon suppression method, nPAA is introduced at one or two of the three stop codons.^25,26 Therefore, there are no competing PAA-tRNA. However, a possible drawback of this method is the competition of the nPAA-tRNA against the release factors, by which premature translation termination is induced. Moreover, only up to two kinds of nPAA can be introduced to the translation system, because at least one of the three stop codons should remain for use in translation termination.

Conversely, the genetic code reprogramming method assigns nPAA to the 61 sense codons. In order to prevent the conflict between nPAA-tRNA and canonical PAA-tRNA, the corresponding PAA are removed from the translation system. For instance, Forster et al. reprogrammed the genetic code in an E. coli cell-free reconstituted translation system in which all amino acids, aminoacyl-tRNA synthetases, and tRNA were omitted and instead, pre-charged nPAA-tRNA were introduced.²⁷ By using this strategy, they introduced 2-amino-4-pentenoic acid, O-methylserine, and 2-amino-4-pentenoic acid to Asn AAC, Thr ACC, and Val GUU codons, respectively. However, the main drawback of this method is that the number of available PAA is reduced from 20 due to their removal from the translation system.

To overcome this issue, this current research recently developed a new method to artificially divide a codon box into two to which two amino acids, the original PAA and an additional nPAA, are assigned (Figure 1).²⁸ An E. coli cell-free translation system was reconstituted using 32 in vitro-transcribed tRNA bearing SNN anticodons (S = G or C) to decode the 20 PAA. In this system, the tRNAs with GNN and CNN anticodons decode the NNY (Y = U or C) and NNG codons, respectively, whereas no tRNAs that decode NNA codons are present, and consequently the NNA codons remain unassigned. Some NNY codons were reprogrammed to nPAA by pre-charged nPAA-tRNA_GNN, whereas NNG codons in the same codon box were not reprogrammed to retain the original PAA in the codon table. For instance, Val GUY, Arg CGY, and Gly GGY codons were reprogrammed to citrulline (Cit), 4-iodophenylalanine (^IodoF), and N-ε-acetyllysine (^AcK), respectively, whereas Val GUG, Arg CGG, and Gly GGG codons remained to assign the original PAA (Figure 1a). This resulted in the successful translation of a 32-mer peptide, rP1 consisting of 23 different amino acids (3 nPAA and 20 PAA) under the reprogrammed genetic code without sacrificing any of the 20 PAA (Figures 1b, c).

3. Improvement of Inefficient nPAA Incorporation

3.1 Inefficient Amino Acid Substrates for Ribosomal Incorporation

Despite the development of above-mentioned genetic code manipulation methodologies, not all nPAA can be easily introduced into peptides and proteins. For instance, although the single incorporation of d- and β-amino acids has been previously demonstrated, two or more consecutive incorporations of those amino acids are extremely inefficient compared with single incorporation.^16,17 Moreover, even the single incorporation of γ-amino acids was not accomplished until very recently.^18,29d-α-amino acids can be classified into three groups depending on their single incorporation efficiencies using the FIT system: Group I with a yield of over 40% relative to the incorporation of l-amino acid counterparts (Ala, Ser, Cys, Met, Thr, His, Phe, and Tyr), Group II with a yield of 10%–40% (Asn, Gln, Val, and Leu), and Group III with a yield of less than 10% (Arg, Lys, Asp, Glu, Ile, Trp, and Pro).¹⁶ The difficulties in incorporating these amino acids are attributed to the following two major reasons: (1) the slow accommodation of the nPAA-tRNA onto the ribosomal A-site mediated by EF-Tu and (2) the slow peptide bond formation between the P-site peptidyl-nPAA-tRNA and the A-site nPAA-tRNA catalyzed by the peptidyl transferase center of the ribosome. The slow accommodation and peptide bond formation induce ribosomal stalling and eventually the mistranslocation of peptidyl-tRNA from the P site to the E site, resulting in peptidyl-tRNA drop-off. Incorporation of nPAA at the initiator AUG codon is generally easier than that at elongator codons, because the amino group of initiator substrate is not involved in ribosomal peptide bond formation, and therefore its reactivity is not necessarily high. This chapter mainly focuses on the elongation event rather than the initiation, and summarizes recent advances in engineering ribosomal translation systems to enhance nPAA incorporation efficiency by improving the accommodation and peptide bond formation of nPAA-tRNA.

3.2 Accelerating Accommodation of nPAA-tRNA onto the Ribosome

During the accommodation process, EF-Tu recognizes the T-stem region and the amino acid moiety of aminoacyl-tRNA.^30–32 However, in the case of nPAA-tRNA, the binding affinity of the amino acid moiety toward EF-Tu is often too weak to conduct the efficient accommodation of nPAA-tRNA onto the ribosomal A-site. Therefore, to accelerate the accommodation of nPAA-tRNA, the binding affinity of nPAA-tRNA toward EF-Tu must be increased. There are two major approaches to achieving this: 1) introduction of point mutations into the amino acid binding site of EF-Tu and 2) optimization of T-stem sequence of the tRNA required for EF-Tu binding.

For instance, incorporation of some phosphorylated amino acids, such as l-phosphoserine (Sep) and l-phosphotyrosine (pTyr) are inefficient due to the low affinity of aminoacyl-tRNA toward EF-Tu (Figure 2). Thus, Park et al. developed a mutant EF-Tu, referred to as EF-Sep, to improve the binding affinity to Sep-tRNA.³³ They substituted the negatively charged Glu and Asp at the binding pocket with uncharged Asn and Gly residues, respectively. In total, EF-Sep has six mutations around the binding pocket (H67R, E216N, D217G, F219Y, T229S, and N274W). Consequently, EF-Sep showed higher binding affinity toward Sep-tRNA and improved the Sep incorporation efficiency compared with the wild-type EF-Tu.

The binding affinity of aminoacyl-tRNA toward EF-Tu is also regulated by the sequence of the T-stem region of tRNA. Dale et al. reported that the binding affinities (ΔG⁰ values) of l-Val-tRNA^Asn, l-Val-tRNA^Gly, and l-Val-tRNA^Glu to EF-Tu were estimated to be −8.8, −10.7, and −11.7 kcal/mol, respectively, indicating that the binding of l-Val-tRNA^Glu was 110-fold tighter than that of l-Val-tRNA^Asn.³⁴ The difference in affinity is attributed to their T-stem structures. On this basis and to compensate the weak binding affinity of the amino acid moiety of nPAA-tRNA, tRNA whose T-stems have stronger affinity toward EF-Tu, like tRNA^Glu, can be used. This study devised an engineered tRNA, named tRNA^GluE2, which was based on the structure of E. coli tRNA^Glu (Figure 3b).³⁵ The expression level of rP2 (a model peptide containing two consecutive d-Ala or β-homomethionine [βhMet]) was significantly improved using tRNA^GluE2 (by 3- and 5-fold, respectively) compared with the use of a weaker E. coli tRNA^Asn-based tRNA, named tRNA^AsnE2 (Figures 2, 3a, 4a, d, e).^36,37

The hydrolytic stability of aminoacyl-tRNA is also related to its accommodation efficiency. In the case of γ-aminoacyl-tRNA, intramolecular nucleophilic attack of the γ-amino group to the esterified carboxyl group induces the self-deacylation of γ-aminoacyl-tRNA. Therefore, γ-aminoacyl-tRNA cannot be efficiently accommodated onto ribosomes, which makes the ribosomal incorporation of γ-amino acids extremely difficult. To overcome this issue, Ohshiro et al. utilized dipeptidyl-tRNA comprising γ- and α-amino acids as a translation initiator as γ,α-dipeptidyl-tRNA was not easily self-cyclized.¹⁸ The γ,α-dipeptide was introduced at the N-terminus followed by a thioester of diketopiperadine at the C-terminal region, which was installed by the self-rearrangement of the Pro-Cys-^HOGly (α-glycolic acid) sequence. Then, the intramolecular ligation occurred between the N-terminal γ-amino group and the thioester to give a backbone cyclic structure, by which the γ-amino acid was relocated in the middle of the peptide.

3.3 Accelerating Peptide Bond Formation between nPAA

Peptide bond formation between some inefficient nPAA (like d-, β-, and γ-amino acids) is intrinsically very slow. Among PAA, Pro is the only secondary amino acid and thus has lower reactivity in peptide bond formation than the other 19 PAA. To overcome the low reactivity, nature utilizes a Pro-specific translation factor (called EF-P) that accelerates peptide bond formation.^38,39 EF-P binds to the P-site peptidyl-Pro-tRNA^Pro by recognizing the tRNA^Pro-specific D-arm motif and accelerates the transfer of the nascent peptide onto the A-site Pro-tRNA^Pro. The authors reported that the conserved D-arm motif found in the tRNA^Pro isoacceptors consists of a 9-nt D-loop closed by a stable 4-bp D-stem containing two G/C base pairs at positions 12/23 and 13/22.⁴⁰ On this basis, it was hypothesized that the slow peptide bond formation between the inefficient nPAA could also be accelerated by EF-P if those nPAA were pre-charged on the tRNA^Pro isoacceptors bearing the D-arm motif. To verify this concept, the authors conducted an incorporation test of d-Ala and β-hMet into the model peptide, rP2, using tRNA^Pro1 (one of the three tRNA^Pro isoacceptors) in combination with EF-P (Figure 3c).^37,41 Consequently, EF-P was found to enhance the expression level of those model peptides by 3.2- and 3.6-fold (Figure 4d, e). Additionally, the introduction of a point mutation, C13G, in the D-arm motif significantly decreased the EF-P-dependent enhancement of peptide expression, indicating that EF-P requires the D-arm motif of the nPAA-tRNA^Pro1.

Given that EF-P and EF-Tu recognize different parts of a tRNA, D-arm and T-stem, respectively, both of the D-arm and T-stem motifs can be combined into one tRNA to simultaneously accelerate both peptide bond formation and accommodation. tRNA^Pro1E2 is a designer tRNA that has both the D-arm and T-stem motifs derived from tRNA^Pro1 and tRNA^GluE2, respectively (Figure 3d).^41,42 To demonstrate the ability of tRNA^Pro1E2, d-Ala and β-hMet were pre-charged on this tRNA and were used for two consecutive incorporations into the same model peptide, rP2. In the presence of EF-P, d-Ala- and β-hMet-tRNA^Pro1E2 exhibited an 18- and 28-fold enhancement in the peptide expression level, respectively, compared with those charged on tRNA^AsnE2 (Figures 4d, e). The enhancement effects of tRNA^Pro1E2 were much higher than those of tRNA^Pro1 and tRNA^GluE2. The study also analyzed how many β-hMet could be consecutively introduced into model peptides using tRNA^Pro1E2 (Figure 4a, rP3). In the presence of EF-P, up to seven consecutive β-hMet residues could be incorporated into full-length peptides (Figure 4f). Conversely, in the absence of EF-P, no more than three consecutive incorporations could be achieved. These results clearly demonstrate the powerful properties of the engineered tRNA^Pro1E2 for incorporation in exotic nPAA.

3.4 Optimizing EF-G Concentrations for Suppression of Mistranslocation

As mentioned above, slow accommodation and peptide bond formation eventually induce ribosomal stalling followed by the mistranslocation of peptidyl-tRNA from the P site to the E site. In the canonical translocation, the elongation factor G (EF-G) is responsible for moving the P-site deacylated tRNA and A-site peptidyl-tRNA into the E site and P site, respectively; this should occur only after the peptidyl transfer is complete. However, if mistranslocation occurs before the peptidyl transfer reaction completes, the peptidyl-tRNA at the P site moves to the E site and eventually detaches from the ribosome. It was assumed that this mistranslocation event is also triggered by EF-G. This hypothesis provided us with the motivation to optimize the concentration of EF-G in the FIT system to prevent the mistranslocation event and enhance the synthesis of full-length peptides without truncation caused by peptidyl-tRNA drop-off.³⁶ Indeed, excessive EF-G significantly decreased the expression level of a peptide containing two consecutive d-Ala when the concentration of EF-G was higher than 0.1 µM, suggesting that the EF-G concentration must be carefully optimized for translation of peptides containing inefficient nPAA such as d-amino acids.

4. Ribosomal Synthesis of Foldamer Peptides Containing Cyclic β- and γ-Amino Acids

When developing novel peptide drugs, researchers are often encouraged to rigidify the peptide scaffolds by creating some folding propensities; tightly folded and rigid peptides generally have higher binding affinity to the target molecules due to the entropic effect. Such peptides also tend to exhibit higher proteolytic resistance as they are protected from proteases in the cell by binding to the target molecules. Folding into a compact scaffold should also contribute to the membrane permeability of the peptide, especially when hydrogen bonding donors and acceptors are consumed by the intramolecular interactions through peptide folding. Typically, inductions of helix, turn, and macrocyclic structures are often attempted. β-amino acids are known to induce unique helical structures, such as 10-, 12-, and 14-helices whose stabilities are generally higher than those of α-helices formed by α-peptides.⁴³ β-amino acids can also induce specific turn structures, such as β-turns and γ-turns.^44,45 Owing to such folding propensities, peptides containing β-amino acids are often referred to as foldamers.

Among the various β-amino acids, we focused on some cyclic β^2,3-amino acids (cβAA), including 2-aminocyclopentanecarboxylic acid (2-ACPC) and 2-aminocyclohexanecarboxylic acid (2-ACHC); they are stronger helix/turn inducers due to their constrained cyclic structures (Figure 2).^46–49 Four stereoisomers of 2-ACPC [(1R,2R)-2-ACPC, (1R,2S)-2-ACPC, (1S,2R)-2-ACPC, and (1S,2S)-2-ACPC] and four 2-ACHC [(1R,2R)-2-ACHC, (1R,2S)-2-ACHC, (1S,2R)-2-ACHC, and (1S,2S)-2-ACHC] were tested for incorporation into model peptides using tRNA^Pro1E2 in the presence of EF-P.¹⁹ Surprisingly, EF-P did not always improve peptide expression levels and in some cases gave a negative effect depending on the conformation of 2-ACPC and 2-ACHC. The incorporation of (1R,2S)-2-ACPC, (1R,2R)-2-ACHC, and (1S,2S)-2-ACHC was improved by EF-P, whereas (1R,2R)-2-ACPC, (1S,2R)-2-ACPC, and (1S,2S)-2-ACPC were inhibited. The incorporation of (1R,2S)-2-ACHC and (1S,2R)-2-ACHC was not affected by EF-P. Among those substrates, (1S,2S)-2-ACPC exhibited a remarkable incorporation efficiency even without EF-P resulting in a successful expression of peptides containing up to ten consecutive (1S,2S)-2-ACPC—both in the presence and absence of EF-P (Figure 4g). Peptides consisting of four or more consecutive (1S,2S)-2-ACPC folded into 12-helices were reported in earlier studies;^50,51 the peptides expressed in this experiment with 10 consecutive (1S,2S)-2-ACPC should also fold into a 12-helix structure. We tested not only saturated cβAA but also aromatic cβAA such as aminobenzoic acid (Abz), in which the amino nitrogen and carbonyl carbon are both directly attached to an aromatic ring. In spite of the low nucleophilicity of their amino groups due to the resonance effect, various Abz derivatives such as 3-aminopyridine-4-carboxylic acid (Apy), 3-aminothiophene-2-carboxylic acid (Atp), and 5-aminothiazole-4-carboxylic acid (Atz) could be successfully introduced into nascent peptide chains taking advantage of tRNA^Pro1E2 in combination with EF-P (Figure 2).⁵²

Likewise, γ-amino acids are also attractive building blocks for developing novel peptide foldamers because of their folding abilities (like the induction of stable helices and turns).^53–63 In particular, cyclic γ-amino acids (cγAA) constrain structures like cβAA; therefore, strong helix/turn-inducing abilities are expected (Figure 2). As discussed above, γ-aminoacyl-tRNA suffer from self-deacylation caused by the intramolecular nucleophilic attack of the γ-amino group to the esterified carboxyl group. However, in the case of cγAA, the constrained cyclic structures make it possible to suppress the deacylation of cγAA-tRNA because the intramolecular nucleophilic attack of the γ-amino group should be inefficient. We examined the single incorporation of two 3-aminocyclobutane carboxylic acid (cis-3-ACBC and trans-3-ACBC); four 3-aminocyclopentane carboxylic acid [(1R,3R)-3-ACPC, (1R,3S)-3-ACPC, (1S,3R)-3-ACPC, and (1S,3S)-3-ACPC]; and four 3-aminocyclohexane carboxylic acid [(1R,3R)-3-ACHC, (1R,3S)-3-ACHC, (1S,3R)-3-ACHC, and (1S,3S)-3-ACHC].²⁹ Among them, eight substrates, cis-3-ACBC, trans-3-ACBC, (1R,3R)-3-ACPC, (1R,3S)-3-ACPC, (1S,3R)-3-ACPC, (1S,3S)-3-ACPC, (1R,3R)-3-ACHC, and (1R,3S)-3-ACHC were successfully incorporated into the model peptide, whereas (1S,3R)-3-ACHC and (1S,3S)-3-ACHC were not. Because of the intrinsically low incorporation efficiency, none of these substrates were successfully consecutively incorporated; however, the non-consecutive multiple incorporation of two trans-3-ACBC with two PAA inserted in between was successfully demonstrated.

Those cβAA and cγAA could also be introduced into macrocyclic scaffolds. The translation of model peptide rP4 (Figures 4a, b) was initiated with N-chloroacetyl-d-phenylalanine (^ClAcd-Phe) followed by four consecutive (1S,2S)-2-ACPC and d-Cys. Then, the sulfhydryl group of the d-Cys side chain spontaneously reacted with the N-terminal chloroacetyl group to form a thioether bond and produced a macrocyclic scaffold. The peptide rP5 was also closed by the N-terminal ^ClAcd-Phe and the downstream d-Cys. In addition, N-methyl-l-α-Ser (^MeSer), (1S,2S)-2-ACPC, and cis-3-ACBC were embedded in the macrocyclic scaffold (Figure 4a, c). As EF-P suppresses (1S,2S)-2-ACPC incorporation, (1S,2S)-2-ACPC was pre-charged on tRNA^GluE2 so that the (1S,2S)-2-ACPC-tRNA^GluE2 was unrecognized by EF-P, whereas the other nPAA (^MeSer, cis-3-ACBC, and d-Cys) were pre-charged on tRNA^Pro1E2, except for the initiator ^ClAcd-Phe, which was charged on tRNAⁱⁿⁱ. These are the first examples of the ribosomal synthesis of macrocyclic peptides containing cβAA and cγAA.

5. In Vitro Selection of Foldamer Peptides

Ribosomally synthesized peptide libraries are compatible with mRNA display-based screening methodologies, such as the random nonstandard peptides integrated discovery (RaPID) system. A macrocyclic peptide library containing three types of cβAA, (1S,2S)-2-ACHC, (1R,2R)-2-ACPC, and (1S,2S)-2-ACPC, was applied to the RaPID screening of peptide ligands binding to two targets: human FXIIa and IFNGR1 (Figure 5).¹⁹ The library was cyclized via a thioether bond between the N-terminal ^ClAcd-Tyr and the downstream d-Cys and had a repeat of 6–15 random residues encoded by NNU codons in which the three types of cβAA appeared. The C-terminal (Gly-Thr)₃ linker was conjugated to the template mRNA via a puromycin linker. Release factor-1 (RF1) was removed from the translation system so that ribosome stalled at the UAG stop codon, where puromycin was introduced instead of RF1 and conjugated with the C-terminus of peptide. A significant increase in peptide recovery rate was observed during seven rounds of affinity selection. Deep sequencing of the cDNA revealed that the library was enriched with several families of peptide sequences containing one or more cβAA after the fourth round. The investigation focused on several FXIIa and IFNGR1 binders and analyzed their binding affinity, inhibitory activity, and serum stability. Among the FXIIa binders, peptide F3 exhibited both a remarkably strong binding affinity (0.98 nM K_D) and an inhibitory activity (1.02 nM K_i) against the serine protease activity of FXIIa (Table 1). Similarly, I1-5 demonstrated a very strong binding affinity to IFNGR1 (1.87 nM K_D). The half-lives of F3 and I1-5 in human serum at 37 °C were 59 and 66 h, respectively. Such extremely high serum stabilities can be attributed to the existence of cβAA; substitution of the peptides’ cβAA with PAA significantly shortened their half-lives down to 13 and 1.4 h, respectively (Table 1, F3A and I1-1A).

Encouraged by the excellent affinity, activity, and stability of those peptides, the study then analyzed the co-crystal structure of the peptide F3 bound to FXIIa (Figure 6). Remarkably, F3 is folded into an anti-parallel β-sheet structure, which is induced by two γ-turns, a pseudo γ-turn and an inverse γ-turn, formed by a (1S,2S)-2-ACHC located at the edge of the β-sheet (ACHC8). The cyclic side chain of ACHC8 binds to a hydrophobic pocket of FXIIa formed by Tyr515 and His507. Another (1S,2S)-2-ACHC at position 13 (ACHC13) is also involved in a β-turn formation. Furthermore, the side chain of ACHC13 constructs intramolecular van der Waals interactions with Ala3 and Tyr16 to stabilize the peptide folding. Taken together, both (1S,2S)-2-ACHC play important roles in peptide folding as well as binding to the target FXIIa.

As mentioned above, one of the advantages of peptide drugs over small molecule-based ones is the relatively large interaction surface area between a peptide and the target molecules, which is comparable with that of protein binders with much larger molecular sizes. Therefore, the inhibition or promotion of protein–protein interaction is possible by means of peptide drugs. Indeed, the interaction surface area of F3 to FXIIa is 703 Ų, which is comparable with those of the protein-based trypsin inhibitors (700–900 Ų) like STI, BPTI, and Ovomucoid.

6. Summary and Perspective

In summary, we have recently developed a novel designer tRNA called tRNA^Pro1E2, which has specific D-arm and T-stem motifs for improved EF-P and EF-Tu binding. Using tRNA^Pro1E2, the successful acceleration of both accommodation and peptide bond formation of nPAA was confirmed. Then, the investigation demonstrated the successful incorporation of consecutive d- and β-amino acids into macrocyclic scaffolds as well as non-consecutive multiple incorporations of γ-amino acids; this was previously unprecedented until our technologies were established.

One of the advantages of the ribosomal synthesis of peptides is that random peptide libraries containing those nPAA can be applied to screening methods, such as mRNA display. We conducted an mRNA display-based screening of a peptide library containing three types of cβAA against two target proteins, human FXIIa and IFNGR1, and successfully obtained extremely strong binders against those targets. X-ray crystallography of one of the peptides (F3) bound to FXIIa revealed that F3 was folded into an anti-parallel β-sheet structure, which was induced by two γ-turns around a (1S,2S)-2-ACHC. This result is encouraging for the introduction of not only (1S,2S)-2-ACHC but also other types of turn/helix inducers into peptide libraries for the screening of more diverse foldamer peptides—like F3. So far, only three types of saturated cβAA have been tested as turn/helix inducers: (1S,2S)-2-ACHC, (1R,2R)-2-ACPC, and (1S,2S)-2-ACPC. However, aromatic cβAA and cγAA would also be suitable building blocks for such foldamer peptide libraries. Further improvement of the ribosomal translation systems to enhance the substrate tolerance would be required for the incorporation of more diverse turn/helix inducers other than cβAA and cγAA.

Acknowledgment

This work was supported by the Japan Science and Technology Agency (JST) PRESTO of Molecular Technology and Creation of New Functions (JPMJPR14K3); the Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research (B) (18H02080) to T.K.; JST CREST of Molecular Technologies (JPMJCR12L2) to H.S.

References

Takayuki Katoh

Takayuki Katoh was born in Mie, Japan in 1975. He received his MS and PhD in Engineering from the University of Tokyo under the guidance of Prof. K. Watanabe and Prof. T. Suzuki. He studied the mechanism of RNA interference and biogenesis of mammalian microRNAs. After postdoctoral research at Japan Biological Informatics Consortium, he was appointed as an assistant professor at the University of Tokyo in 2009, and promoted to associate professor in 2018. His research now focuses on the development of artificial ribosomal translation systems compatible with diverse nonproteinogenic amino acids.

Hiroaki Suga

Hiroaki Suga is a Professor of the Department of Chemistry, Graduate School of Science at the University of Tokyo. He received his PhD at MIT (1994) followed by a post-doctoral fellowship at Massachusetts General Hospital (1997). He was Assistant and tenured Associate Professor at the SUNY at Buffalo (1997–2003) and Professor at the Research Center for Advanced Science and Technology at the University of Tokyo (2003–2010). Since 2010, he has held his present position. He is the recipient of the Akabori Memorial Award 2014, Max-Bergmann Medal 2016, Nagoya Medal Silver 2017, Bohlmann Lecture, Vincent du Vigneaud Award 2019, and T.Y. Shen Lecture 2020.