Fine-tuning the tRNA anticodon arm for multiple/consecutive incorporations of β-amino acids and analogs

Abstract Ribosomal incorporation of β-amino acids into nascent peptides is much less efficient than that of the canonical α-amino acids. To overcome this, we have engineered a tRNA chimera bearing T-stem of tRNAGlu and D-arm of tRNAPro1, referred to as tRNAPro1E2, which efficiently recruits EF-Tu and EF-P. Using tRNAPro1E2 indeed improved β-amino acid incorporation. However, multiple/consecutive incorporations of β-amino acids are still detrimentally poor. Here, we attempted fine-tuning of the anticodon arm of tRNAPro1E2 aiming at further enhancement of β-amino acid incorporation. By screening various mutations introduced into tRNAPro1E2, C31G39/C28G42 mutation showed an approximately 3-fold enhancement of two consecutive incorporation of β-homophenylglycine (βPhg) at CCG codons. The use of this tRNA made it possible for the first time to elongate up to ten consecutive βPhg's. Since the enhancement effect of anticodon arm mutations differs depending on the codon used for β-amino acid incorporation, we optimized anticodon arm sequences for five codons (CCG, CAU, CAG, ACU and UGG). Combination of the five optimal tRNAs for these codons made it possible to introduce five different kinds of β-amino acids and analogs simultaneously into model peptides, including a macrocyclic scaffold. This strategy would enable ribosomal synthesis of libraries of macrocyclic peptides containing multiple β-amino acids.


Introduction
Ribosomal incorporation of various nonproteinogenic amino acids into peptides and proteins has been enabled by development of genetic code manipulation methodologies, such as nonsense codon suppression ( 1 ), sense codon reassignment ( 2 ) and quadruplet codon suppression ( 3 ).lα-Amino acids with sidechain modifications can be relatively efficiently introduced into nascent peptide chains, whereas incorporation of backbone modified amino acids, such as β-amino acids, is generally much less efficient (4)(5)(6)(7)(8)(9).Moreover, multiple and / or consecutive incorporation of diverse β-amino acids at once is by far more difficult compared to incorporation of a single, selected β-amino acid ( 8 ).In fact, multiple / consecutive incorporation of β-amino acids had been considered impossible for a long time.On the other hand, it has been known that β-amino acids possess stronger turn / helix inducing abilities than α-amino acids, and thereby peptides comprised of β-amino acids can fold into unique and stable structures, referred to as foldamers (10)(11)(12).Such peptides could exhibit enhanced binding affinity and specificity to target molecules (13)(14)(15), membrane permeability ( 16 ,17 ), and proteolytic stability (18)(19)(20)(21).Thus, β-amino acids are an attractive set of building blocks to develop novel bioactive peptides.
Selection approach by a combination of genetic code reprogramming powered by the flexizyme technology ( 22 ) and mRNA display ( 23 ,24 ), referred to as RaPID ( Ra ndom nonstandard P eptides I ntegrated D iscovery) system ( 25 ), has led to the discovery of de novo macrocycles containing nonproteinogenic amino acids (25)(26)(27)(28)(29)(30).During the course of these studies, however, we have also observed some nonproteinogenic amino acids, such as β-amino acids and dαamino acids, are poorly incorporated using well-established suppressor tRNAs (e.g.tRNA AsnE2 ), particularly when their consecutive incorporation occurs ( 8 ,31 ).With this reason, we have been devoting our efforts to improve their incorporation efficiency, thereby allowing us to construct libraries containing such nonproteinogenic amino acids and to discover de novo bioactive peptides using the RaPID system.
The inefficiency of β-amino acid incorporation could be mainly attributed to the following two reasons: (i) the inefficient accommodation of β-aminoacyl-tRNA onto the ribosomal A site and (ii) the slow peptidyl transfer reaction of β-aminoacyl-tRNAs ( 32 ).To overcome these issues, we have previously developed a designer tRNA, referred to as tRNA Pro1E2 , bearing specific T-stem and D-arm motifs that can be recognized by EF-Tu and EF-P, respectively (Figure 1 A) ( 33 ,34 ).β-Aminoacyl-tRNA Pro1E2 can be efficiently accommodated onto ribosome by EF-Tu and the peptidyl transfer reaction of them can be accelerated by EF-P.By using the tRNA Pro1E2 body sequence, we have demonstrated consecutive incorporation of β-amino acids into peptides ( 26 , 29 , 32 ).Moreover, tRNA Pro1E2 has allowed us to incorporate β-amino acid analogs, such as α-aminoxy acids and α-hydrazino acids, into nascent peptide chain, yielding unique backbone alternative peptides in ribosomal expression ever ( 35 ).However, even if tRNA Pro1E2 is used, simultaneous introduction of multiple kinds of β-amino acids into nascent peptide is still poorly achieved, and thus it remains a major challenge.
Although the sequences of T-stem and D-arm of tRNA Pro1E2 have been optimized for binding to EF-Tu and EF-P, respectively, other regions of this tRNA have not yet been finetuned for β-amino acid incorporation.It has been previously suggested that the sequence of anticodon arm could affect the efficiency of translation; for instance, Yarus et al. investigated amber suppressor tRNA variants bearing anticodon stem / loop mutations and observed significant differences of amber codon suppression level among the variants (36)(37)(38)(39).Recently, we also reported that anticodon stem mutations of initiator tRNA also affect the efficiency of introducing noncanonical initiator building blocks, such as Nacetylproline and N -acetylβ-homophenylglycine, at the peptide N-terminus ( 40 ).Therefore, here we hypothesize that the efficiency of β-amino acid elongation can also be finetuned by introducing mutations into the anticodon arm.Such an optimization of tRNA Pro1E2 would enable more efficient β-amino acid incorporation, leading to simultaneous introduction of multiple kinds of β-amino acids.

Preparation of tRNAs and flexizymes
tRNAs (tRNA Pro1E2 variants and tRNA iniP ) and flexizymes (dFx and eFx) used in this study were synthesized by in vitro transcription.The DNA templates for transcription were prepared by extension and PCR using the forward and reverse primers shown in Supplementary Table S1 .In vitro transcription was carried out at 37 • C for overnight in the following solution: 40 mM Tris-HCl (pH 8.0), 3.75 mM nucleo-side triphosphates (NTPs), 5 mM guanosine monophosphate (GMP), 22.5 mM MgCl 2 , 0.01% Triton X-100, 1 mM spermidine, 1 mM dithiothreitol, 0.04 U / μl RNasin RNase inhibitor (Promega) and 0.12 μM T7 RNA polymerase.For transcription of flexizymes, the NTP concentration was increased to 5 mM and instead GMP was omitted.Then, the reaction mixture was treated with RQ1 DNase (Promega) for 30 min at 37 • C. The resulting tRNA and flexizymes were purified by 8% or 12% denaturing polyacrylamide gel electrophoresis.

T ricine SDS-P AGE of translated peptides and their quantification by autoradiography
Translation was conducted in the presence of 0.05 mM [ 14 C]-Asp in place of cold Asp, quenched by adding the same volume of stop solution [0.9 M Tris-HCl (pH 8.45), 8% SDS, 30% glycerol, and 0.001% xylene cyanol], and incubated at 95 • C for 3 min.4 μl of the sample was subjected to 15% tricine SDS-PAGE and analyzed by autoradiography using a Typhoon FLA 7000 (Cytiva).The relative translation efficiencies of peptides were estimated by the autoradiographic intensity of the peptide band in comparison with that of the control experiment, whose relative level was defined as 1.

Identification of translated peptides by MALDI-TOF mass spectrometry
5 μl of translation reaction mix was desalted using SPE C-tip (Nikkyo Technos) and co-crystallized with α-cyano-4-hydroxycinnamic acid on a sample plate.The sample was analyzed by UltrafleXtreme (Bruker Daltonics) in a reflector / positive mode.Peptide calibration standard II (Bruker Daltonics) was used for external mass calibration.

Evaluation of anticodon stem mutations for consecutive incorporation of β-amino acids at CCG codon
For fine-tuning tRNA Pro1E2 , we first focused on mutations at the anticodon stem.Two consecutive β-homophenylglycine ( βPhg) residues were incorporated into a model peptide P2 at CCG codons of a template mRNA, mR2 (Figures 1 B   and 2A , The P2 peptide containing βPhg is referred to as P2-βPhg).The original tRNA Pro1E2 bearing CGG anticodon (Figure 1 A, tRNA Pro1E2 CGG ) and 11 anticodon stem mutants were evaluated for βPhg incorporation.The original tRNA Pro1E2 CGG has 5 bp in the anticodon stem (Figure 1 A, Supplementary Figure S1 , C27G43, U28A42, U29A41, C30G40 and G31C39) and the compensatory mutations were introduced at one of the five base pairs, where the combinations of nucleotides were arbitrarily chosen (Figure 2 B, Supplementary Figure S1 , A31U39, C31G39, U31A39, A30U40, U30A40, A29U41, C29G41, A28U42, C28G42, G28C42 and U27A43).βPhg was precharged onto the tRNA Pro1E2 CGG variants by means of flexizyme ( 22 ) and then subjected to translation.For translation of peptides, we used a custom-made FIT system that contains a minimal set of amino acids required for translation of P2, i.e.Asp, Gly, Lys, Met and Tyr.The peptides were expressed in the presence of [ 14 C]-Asp, subjected to 15% tricine SDS-PAGE, and quantified by autoradiography ( Supplementary Figure S2 A).The identities of peptides expressed in the presence of cold Asp instead of [ 14 C]-Asp were also confirmed by MALDI-TOF mass spectrometry (MS) ( Supplementary Figure S3 A).The relative translation efficiencies of P2-βPhg using the anticodon stem mutants were estimated, where the efficiency of the original tRNA Pro1E2 CGG was defined as 1 (Figure 2 B).Among the 11 single base pair mutants, A31U39, C31G39, U31A39, A30U40, C29G41 and C28G42 showed higher expression levels of P2-βPhg compared to the use of the original tRNA Pro1E2 CGG (Figure 2 B, 1.4-, 1.7-, 1.7-, 1.2-, 1.2-and 1.3fold, respectively).All mutations at the N31N39 base pair resulted in higher level of P2-βPhg, indicating that the original G31C39 pair is not optimal for βPhg incorporation at CCG codon.C31G39 showed the highest translation efficiency of P2-βPhg, a 1.7-fold enhancement compared to the original tRNA Pro1E2 CGG .
To further improve the translation efficiency, we next tried combinations of mutations that resulted in high relative translation efficiencies (Figure 2    Here we evaluated five double base pair mutants (C31G39 / A30U40, C31G39 / C29G41, C31G39 / A28U42, C31G39 / C28G42 and C31G39 / U27A43) and two triple base pair mutants (C31G39 / A30U40 / C29G41 and C31G39 / C29G41 / C28G42).As a result, C31G39 / C28G42 and C31G39 / A30U40 / C29G41 showed the highest and second highest P2-βPhg levels (Figure 2 B, Supplementary Figure S2 A, S3 A, 2.9-and 2.6-fold enhancement compared to the original tRNA Pro1E2 CGG , respectively).The relative P2-βPhg levels of the 19 tRNA Pro1E2 CGG variants tested in this study ranged from 0.05 to 2.9, and their average and variance values were 1.5 and 0.5, respectively, showing the impact of anticodon stem mutation on the efficiency of βPhg incorporation.
To evaluate the applicability of these anticodon stem mutants to introduce other β-amino acids and β-amino acid analogs, two β 3 -amino acids ( βMet and βGln), two cyclic β 2,3 -amino acids [(1 S ,2 S )-2-ACPC and (1 R ,2 R )-2-ACPC] and one α-aminoxy acid ( NO Ala) were precharged on C31G39 / C28G42 and C31G39 / A30U40 / C29G41 as well as the original tRNA Pro1E2 CGG as a control and introduced into P2 or P3 at CCG codons (Figures 1 B  and 2C , Supplementary Figure S2 B).Consequently, both C31G39 / C28G42 and C31G39 / A30U40 / C29G41 showed significantly higher translation efficiencies of peptides compared to the use of the original tRNA Pro1E2 CGG for all substrates (Figure 2 C, 1.5 −20-fold enhancement), indicating that these mutant tRNAs can generally enhance incorporation of β-amino acids / analogs when introduced at CCG codon.The identities of these peptides were also confirmed by MALDI-TOF MS, showing that all these β-amino acids / analogs could be correctly introduced without misincorporation ( Supplementary Figure S3 B).
By using C31G39 / C28G42, we next tried three or more consecutive incorporations of βPhg into model peptides P3 −P10 at CCG codon (Figure 3 A).Although their translation efficiencies gradually decreased for more consecutive βPhg incorporations, up to ten consecutive incorporations of βPhg into P10 was confirmed (Figure 3 B, C, Supplementary Figure S2 C, the translation level of P3 was defined as 1, and the relative levels of P4 − P10 were calculated).The identities of the peptides were also confirmed by MALDI-TOF MS (Figure 3 C, Supplementary Figure S3 C).This is the first demonstration of ten consecutive incorporation of a β 3 -amino acid to the best of our knowledge.

Anticodon loop fine-tuning for βPhg incorporation
Although we focused on anticodon stem sequences in the series of experiments described above, it is known that the anticodon loop sequence is also involved in regulating translation efficiency (36)(37)(38)(39).Therefore, we next introduced anticodon loop mutations to the tRNA Pro1E2 variants bearing the optimal anticodon stem sequences to further enhance translation efficiency.The anticodon loop of the original tRNA Pro1E2 consists of 5 -UU-XXX-GA-3 , which is corresponding to the nucleotide positions of 32 to 38.In this mutation study, the nucleotides of positions 32, 33, 37 and 38 are changed to UU-AU, UU-AA, CU-AA and CU-AC.These sequences were chosen because the anticodon loop sequences of E. coli native tRNA Pro1 CCG , tRNA His QUG , tRNA Gln2 CUG , tRNA Thr3 GGU and tRNA Trp CCA are UU-GA, UU-AU, UU-AU, UU-AA and CU-AA, respectively (Figure 5 A).CU-AC was tested because this sequence is also found in other E. coli native tRNAs.These anticodon loop mutations were introduced into the anticodon stem variants of C31G39 / C28G42, C29G41 / C28G42, C29G41, C31G39 and A30U40 / C29G41 / C28G42 for two consecutive incorporations of βPhg into P2 at CCG, CAU,  contrast, the desired peptide was not observed when the original tRNA Pro1E2 set was used for incorporation of β-amino acids / analogs, showing the advantage of the anticodon arm-tuning strategy in incorporation of multiple β-amino acids / analogs.

Discussion
In summary, we succeeded in fine-tuning anticodon arm structure of tRNA Pro1E2 , thereby enhancing β-amino acid incorporation.The impact of anticodon arm mutation differed de-pending on the codons used for β-amino acid incorporation.Therefore, the anticodon arm sequence must be optimized individually for each codon.
In 1980s, Yarus et al. reported that substitution of anticodon loop nucleotides located at positions 32, 33, 37 and 38 affects in vivo translation efficiency and fidelity in E. coli (36)(37)(38)(39).Later studies showed that these residues regulate the stability of codon-anticodon interaction in vivo and in vitro (43)(44)(45)(46)(47)(48).Particularly, the 32-38 pair is a critical determinant of the binding affinity of aminoacyl-tRNA to the ribosomal A site; the affinity decreases by introducing a base pair at the 32-38.For instance, in the case of tRNA Ala2 GGC , base pair formation at this position, such as A-U and C-G, is preferred for efficient decoding with high fidelity ( 46 ), whereas tRNA Ala1based amber suppressor tRNAs bearing CUA anticodon exhibit higher suppression efficiencies with non-base paired 32-38 in the order of C-A > C-C > U-C > U-U ( 45 ).This is likely because the 32-38 pair contributes to fine-tuning the affinity to a uniform range.tRNA Ala2 GGC forms three G-C pairs between the codon and anticodon, resulting in high affinity, which is downregulated by introducing a base pair at the 32-38.On the other hand, the amber suppressor tRNA Ala1 bearing CUA anticodon forms only one G-C pair and thus the low affinity should be compensated for by the non-base-paired 32-38 for efficient translation.Oleiniczak et al. proposed that native tRNAs have evolved to have optimal 32-38 pairs for each anticodon so that the tRNAs have uniform affinities to the codon of mRNA ( 44 ,45 ).
However, these preceding studies focused on incorporation of canonical α-amino acids and thus the role of anticodon loop in β-amino acid incorporation had remained to be elucidated.As β-amino acids are extremely inefficient substrates for translation compared to the canonical α-amino acids due to their longer backbone, the anticodon loop sequences evolved for α-amino acid incorporation would not be necessarily optimal for β-amino acid incorporation.In fact, we revealed in this study that the native anticodon loop sequences were not optimal for β-amino acid incorporation (Figure 5 ).U-U, C-A, C-C, C-A and U-A were preferred for CGG, GUG, CUG, GGU and CCA anticodons, respectively, over the native loop sequences, U-A, U-U, U-U, U-A and C-A, respectively.
Although the role of anticodon stem in regulation of translation efficiency has not been fully elucidated to date, it is known that the anticodon stem is involved in modulating structural flexibility of tRNA.As tRNAs undergo conformational changes during translation on the ribosome, tRNA flexibility must be related to translation efficiency ( 49 ,50 ).We have speculated that, by modulating the flexibility of tRNA with anticodon stem mutations, β-amino acid could be placed at a preferable and reactive position at the peptidyl transferase center (PTC) of ribosome.The optimal location of β-amino acid for the reaction at the PTC may be different from that of the canonical α-amino acid due to the longer backbone of βamino acid by one methylene group.Thus, we confirmed that the sequences of the anticodon stem / loop mutants obtained in this study were not shared by any E. coli native tRNAs that are used for α-amino acid incorporation.Another possible explanation would be stabilization of codon-anticodon interaction by the anticodon stem mutations, similar to the function of anticodon loop nucleotides as mentioned above.Further structural / biochemical studies are required to elucidate the mechanism how anticodon stem mutations regulate the β-amino acid incorporation efficiency.
By using the optimized anticodon arm variants, we have demonstrated ribosomal incorporation of five different types of β-amino acids / analogs at once as well as ten consecutive incorporation of βPhg for the first time, to the best of our knowledge.The advantage of ribosomal incorporation of β-amino acids is that we can easily prepare random peptide libraries bearing various β-amino acids using randomized mRNA templates.Such peptide libraries can be readily combined with mRNA display-based screening methodologies, such as the RaPID system, to efficiently screen bioactive peptides that bind to specific target molecules.By introducing β-amino acids into the libraries, we can expect enhanced binding affinity and specificity to target molecules, membrane permeability and proteolytic stability of the screened peptides.

Figure 2 .
Figure 2. Ribosomal incorporation of β-amino acids into model peptides using tRNA Pro1E2 CGG and its anticodon stem variants.( A ) mRNAs, mR2 and mR3, and the corresponding peptide sequences, P2 and P3, used for β-amino acid ( βaa) incorporation.The amino acid sequence of 'flag' is A sp-Tyr-Lys-A sp-A sp-A sp-A sp-Lys.( B ) Relative translation efficiency of peptide P2-βPhg.βPhg was assigned at the CCG codons of mR2 using pre-charged βPhg-tRNA, where the original tRNA Pro1E2 CGG (Figure 1 A) and its anticodon stem mutants were used.Mutations at the anticodon stem are indicated at the bottom.The efficiency of the original tRNA Pro1E2 CGG was defined as 1.The variants that showed the highest and the second highest P2-βPhg le v els are indicated b y orange and blue, respectiv ely.Error bars, s.d. ( n = 3).See also Supplementary Figure S2 A f or the ra w data of tricine SDS-PAGE analyses of peptides.( C ) Relative translation efficiency of peptide P2 and P3 bearing various β-amino acids.β-amino acids were assigned at the CCG codons of mR2 or mR3 using pre-charged β-aminoacyl-tRNAs.The original tRNA Pro1E2 CGG (black), C31G39 / C28G42 (orange), and C31G39 / A30U40 / C29G41 (blue) were tested for their incorporations.Values for the original tRNA Pro1E2 CGG were defined as 1.Error bars, s.d. ( n = 3).See also Supplementary Figure S2 B for the raw data of tricine SDS-PAGE analyses of peptides.

Figure 4 .
Figure 4. Ribosomal incorporation of βPhg into P2 using tRNA Pro1E2 and its anticodon stem variants.Introduction of βPhg at CAU ( A ), CAG ( B ), ACU ( C ) and UGG ( D ) codons of mR2 using pre-charged βPhg-tRNA Pro1E2 bearing GUG, CUG, GGU and CCA anticodons, respectively.The original tRNA Pro1E2 and its anticodon stem mutants were used.Mutations at the anticodon stem are indicated at the bottom.Peptides were analyzed by tricine SDS-PAGE and their translation efficiencies were quantified by autoradiography.Error bars, s.d. ( n = 3).The variants that showed the highest P2-βPhg levels are indicated by red.

Figure 5 .
Figure 5. Ribosomal incorporation of βPhg into P2 using anticodon loop variants.( A ) Combinations of tRNA anticodons and anticodon loop sequences tested in this study.The anticodon loop of the original tRNA Pro1E2 is UU-GA, which is indicated by 'original'.The anticodon loop sequences of native E. coli tRNA Pro1 CGG , tRNA His GUG , tRNA Gln2 CUG , tRNA Thr3 GGU and tRNA Trp CCA are indicated by 'native'.The sequences that showed the highest translation efficiencies are indicated by 'optimal'.( B ) Relative translation efficiency of P2-βPhg using the anticodon-loop variants.The efficiencies of the original UU-G A w ere defined as 1.Error bars, s.d. ( n = 3).T he anticodon-loop sequences of the nativ e E. coli tRNAs are indicated b y asterisks.T he mutations that sho w ed the highest P2-βPhg le v els are indicated b y red.