New Designs for Aminoacyl-Donors

Supplementary Note

New designs for aminoacyl-donors

Our previous investigations1 with Fx have revealed that it is able to accept three distinct types of leaving groups: cyanomethyl esters (CME), thioesters, and adenosine monophosphate. This suggests that the leaving group is not involved in recognition of the substrate for Fx. Moreover, Fx accepts aminoacyl-donors with not only amino groups (including ammonium as well as N-acylated forms) but also hydroxy groups (D. Kourouklis, H.S., unpublished data), and therefore the amino group is also not a critical element for recognition. These observations have suggested that the essential recognition elements of Fx toward the acyl-donor substrate might reside on the methylene–aromatic moiety of the sidechain and the carbonyl group of the ester.

With this hypothesis in mind, we redesigned the substrate into one that can be widely recognized by Fx, or its mutants if necessary, without dependence upon the methylene–aromatic group in the sidechain. One possible solution was to alter the recognition element from the sidechain to a generic group in the substrate, such as the leaving group, and thereby Fx is able to react with substrates bearing various types of sidechains. We thus considered a substrate that has a benzyl ester as a leaving group (Fig. 1c). When such a substrate was compared structurally with the original substrate (Phe-CME), the sidechain and amino group in the new substrate was positioned similar to the CME leaving group in Phe-CME. In addition, based on the knowledge that the CME group could be altered to other leaving groups, the substitution of CME with a benzyl ester would not eliminate the reactivity toward Fx. More importantly, its reactive center could be superimposed on that of Phe-CME, and Fx might then be able to retain reactivity toward this substrate.

In vitro selection of Fx mutants

In order to rapidly and successfully isolate the desired Flexizymes, we set two seeds for the selection design. The first one was set in the RNA pool design. Only the bases involved in amino acid recognition were randomized without or with a single random base insertion (Supplementary Figure 2 online); there were a total of 10 or 11 nucleotides with a complexity of approximately one to four million (= 410 to411)]. Moreover, a microhelix RNA, an analog of tRNA that is also a good substrate for Fx, was chosen as the acyl-acceptor domain. To this end, each RNA pool, of which the random domain was connected to the microhelix via a 10-mer poly-adenosine linker, was constructed and the resulting two pools were mixed as a single set of pool. The second seed was set in the substrate, where -(N-biotinyl)amino--(S)-hydroxybutanoic acid DBE (Hbi-DBE, Supplementary Figure 2 online) was used as a substrate. The -hydroxy acid ester, instead of a protonated -amino acid ester, was chosen to aim at increasing the potential for selecting Flexizymes with broader specificity for acid esters with -substituted heteroatoms. In parallel to the above selection, we also carried out in vitro selection using a similar RNA pool containing Phe-CME, with the aim of enhancing the original activity of Fx (Supplementary Figure3 online).

After cloning of the enriched pool (Supplementary Figures 2 and 3 online) and testing each clone for trans-activity toward tRNAAsnCUA, we selected the single best catalyst from each selection, named dFx (d: dinitro) and eFx (e: enhanced) as shown in Fig. 1a and 1b, respectively. Both Flexizymes contained several base mutations from the original Fx; dFx had two mutations (G20U and G21C) and one mutation-insertion (U34CA), while eFx contained four mutations (A19G, G21C, U32G and G36A). The dFx was able to charge Hbi using Hbi-DBE onto tRNAAsnCUA with a 10% yield whereas Fx was nearly inactive (Supplementary Figure 1 online, lanes 9 and 7). The most striking observation for dFx was that when Aly-DBE was used Aly-tRNAAsnCUA was obtained in 17% yield, whereas Fx produced less than a 3% yield (Supplementary Figure 1 online, lanes 8 and 10). Destruction of the 3'-terminal diol of tRNA by NaIO4-oxidation completely eliminated the activity, indicating that aminoacylation occurred at the 3'-terminal hydroxyl group (Supplementary Figure 1 online, lanes 12 and 14).

Reaction toward a wide collection of DBE and CBT substrates

Based on our past experience, an approximate 20% yield of aminoacyl-tRNAAsnCUA was the threshold for obtaining over 20% suppression efficiency using suppressor tRNAAsnCUA under our translation conditions. The efficiencies observed in most reactions for natural amino acids met this requirement but unfortunately some amino acids did not behave as anticipated. Such amino acids fell into two groups based on the characteristics of their sidechains. The members of the first group are Cys, Asn, and Asp, in which sidechains bear nucleophiles (Fig. 1d, lanes 10, 11 and 13; see also SupplementaryTable 1 online). We suspected that the ester bond of these aminoacyl-tRNAs was hydrolyzed during the selective biotinylation of the -amino group performed for the conventional streptavidin-dependent gel-shift assays. In addition, the secondary amino group on Pro (lane 7) might be inefficient for the biotinylation. To investigate whether these are the reasons for the modest yields, we performed the aminoacylation of the microhelix RNA, the same RNA that was used for the selection, and analyzed the products directly on acid PAGE (aPAGE) without the biotinylation process (Fig. 1e). Because of its short length, this aminoacyl-acceptor with homogenous 3'-ends could be obtained following PAGE purification. This allowed us to assess more accurately the aminoacylation abilities of dFx for 3'-end aminoacylation compared to in vitro transcription of tRNAs, bearing heterogeneous 3'-ends due to difficulty in their purification. Indeed, this microhelix–aPAGE analytical method generally resulted in aminoacylation yields which were approximately 2-fold higher than when tRNA was used (Fig. 1e, lanes 26 and 32 for Leu and Lys, and SupplementaryTable 1 online, compare their yields with aPAGE). Upon analyzing the dFx-catalyzed aminoacylation of microhelix RNA, we found that all amino acids in the first group (Cys, Asn, Asp, and Pro) produced yields of more than 20% (Fig. 1e, lanes 27–31, and SupplementaryTable 1 online labeled with aPAGE), and thus these amino acids also met the threshold requirement.

We next extended our test to two sets of nonstandard amino acids (Supplementary Figure 5 online). The first set included those having altered sidechains (Cit, Aly, Bly, and Iph) and D-configurations (Dle). The second set included those that alter the ordinary -peptide bond to N-methyl-peptide (Mle), -peptide (Bal), or -ester (Hle and Hph) bonds, upon incorporation into proteins. As expected, all nonnatural amino acids in the first set could be efficiently charged by dFx or eFx onto tRNA (Fig. 1d, lanes 18–23 and Supplementary Table 1 online). Due to the difficulty of selective biotinylation on the secondaryamino group in Mle (see Fig. 1d, lane 21) and the -hydroxy group in Hle and Hph, the microhelix–aPAGE method was utilized to observe directly the acylated-RNA. Again these acyl-donors were charged efficiently (Fig. 1e, lanes 33, 34 and 37). One exception was Bal for which the yield was only 18% when the generic dFx system was used, probably because the change in position of the protonated amino (i.e. ammonium) group from  to  resulted in less activation of the carbonyl group. However, when the corresponding CBT was used the yield increased to 40% (Fig. 1f, lane 47).

Amber and programmed frame-shift suppression for site-specific protein mutagenesis

To verify whether the acyl-tRNAs prepared by dFx and eFx function in translation, we demonstrated site-specific mutagenesis of a model protein, green fluorescent protein (GFP), using amber suppression in an E. coli high-performance cell-free translation system (RTS-100). The same expression system9 has been previously utilized for evaluating acyl-tRNAAsnCUA (where a few mutations are introduced into the wildtype tRNAAsnCUA in order to increase the orthogonality of endogenous ARSs; see Supplementary Figure 6a, lane 3) prepared by the original Fx. Therefore, this system is very reliable for testing the function of acyl-tRNAAsnCUA prepared by the new Flexizymes.

Initially, the tRNAs charged with the 20 natural amino acids were individually tested for suppression of the amber codon at position 151 on the GFP mRNA. As expected, most amino acids were incorporated with over 20% suppression efficiencies (a dozen of the amino acids showed near or above 50% efficiencies, Supplementary Figure 6a, lanes 5–23). One exception was Gly, which was not incorporated into GFP at all (lanes 4 and 24). Because it was confirmed that Gly was attached to the 3'-end of tRNA like the other amino acids, the complete lack of incorporation was unanticipated. We thus suspected that unknown mechanisms were operating in RTS-100 to prevent the use of Gly-tRNAAsnCUA for elongation, e.g. its selective hydrolysis. To verify this, we used a reconstituted cell-free translation apparatus, the PURE system (see also ref. 13 in main text). Because the PURE system contains minimal numbers of recombinant protein enzymes and factors that are essential for translation, the contaminants responsible for the lack of incorporation of Gly may not be present. Indeed, Gly was incorporated into GFP in the PURE system (Supplementary Figure 6b), suggesting that there could be unknown factors present in RTS-100 which prevented amber suppression by Gly.

Next, we tested tRNAAsnCUA charged with nonnatural amino acids that were prepared by dFx or eFx (Supplementary Figure 6c). These were divided into two groups; (1) derivatives of -amino acids, consisting of Cit, Aly, Bly, and Iph (lanes 25–27 and 31, respectively), and (2) more exotic nonnatural amino acids, consisting of Mle, Dle, and Bal (lanes 28–30). In the first group, the suppression efficiency of Cit, Aly, and Iph exceeded 50% as expected. Bly was also incorporated but its efficiency was lower than the others, probably because the biotinyl group hindered the interaction with a component(s) of the translation machinery, such as EF-Tu or/and the ribosome peptidyl transfer center. On the other hand, the nonnatural amino acids in the second group were not incorporated into GFP using the RTS-100 (lanes 28–30) or PURE systems (data not shown). Our results suggested that these nonnatural amino acids are poor substrates for the E. coli cell-free translation systems tested in this study.

The Fx system is also able to charge -hydroxy acids onto tRNA as with amino acids. Therefore, three hydroxy acids, Hph, Hle, and Hbi, were chosen for the incorporation test into GFP (Supplementary Figure 6d, lanes 36–38). To also demonstrate the versatility of the Fx system toward tRNA bearing different anticodons, we performed programmed frame-shift suppression at position 178 using tRNAAsnACCC charged with the hydroxy acids. Both Hph and Hle were incorporated into GFP efficiently. In contrast, Hbi was incorporated poorly, most likely for the same reasons as the inefficient incorporation of Bly.

References

1.Saito, H., Kourouklis, D. & Suga, H. EMBO J.20, 1797-806 (2001).

1