BioFiles Volume 4, Number 8 — Metabolomics

Download Biofiles v4 n8 (13.1 Mb PDF)


Table of Contents


An Expanded Set of Amino Acid Analogs for the Ribosomal Translation of Unnatural Peptides

An excerpt taken from PLoS ONE 2(10): e972. doi:10.1371/journal.pone.0000972. The complete reference and accompanying citations are available online at

Matthew, C. T. Hartman, Kristopher Josephson, Chi-Wang Lin, Jack W. Szostak

Howard Hughes Medical Institute, Department of Molecular Biology, Center for Computational and Integrative Biology, Simches Research Center, Massachusetts General Hospital, Boston, Massachusetts, United States of America

Copyright: © 2009 Matthew C. T. et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.



The application of in vitro translation to the synthesis of unnatural peptides may allow the production of extremely large libraries of highly modified peptides, which are a potential source of lead compounds in the search for new pharmaceutical agents. The specificity of the translation apparatus, however, limits the diversity of unnatural amino acids that can be incorporated into peptides by ribosomal translation. We have previously shown that over 90 unnatural amino acids can be enzymatically loaded onto tRNA.

Methodology/Principal Findings
We have now used a competition assay to assess the efficiency of tRNA-aminoacylation of these analogs. We have also used a series of peptide translation assays to measure the efficiency with which these analogs are incorporated into peptides. The translation apparatus tolerates most side chain derivatives, a few α,α-disubstituted, N-methyl- and α-hydroxy-derivatives, but no β-amino acids. We show that over 50 unnatural amino acids can be incorporated into peptides by ribosomal translation. Using a set of analogs that are efficiently charged and translated we were able to prepare individual peptides containing up to 13 different unnatural amino acids.

Our results demonstrate that a diverse array of unnatural building blocks can be translationally incorporated into peptides. These building blocks provide new opportunities for in vitro selections with highly modified druglike peptides.

The recent development of translation systems reconstituted entirely from purified components [1-4] has enabled the ribosomal synthesis of peptides composed primarily of unnatural (i.e. nonproteinogenic) amino acids. The major roadblock to the ribosomal synthesis of highly modified drug-like peptides is the limited number of unnatural building blocks known to be compatible with the translation apparatus. This dearth of unnatural building blocks results in part from difficulties in loading unnatural amino acids onto tRNA, the key first step in translation. Several techniques for charging tRNAs with unnatural amino acids have been developed such as chemoenzymatic tRNA acylation [5-7], ribozyme acylation [8], chemical acylation using PNA thioesters [9], and the use of engineered aminoacyl-tRNA synthetases (AARSs) [10]. Unfortunately all of these approaches require specialized reagents and/or multistep syntheses. In contrast, an all-enzymatic system using the wild-type AARSs would be as convenient as a standard translation experiment using the twenty proteinogenic amino acids. To this end we have recently developed a MALDI-TOF MS screening assay [11] that has uncovered over 90 unnatural building blocks that are AARS substrates. The incorporation of these amino acids into peptides, however, depends on the efficiency of aminoacylation, and on whether or not they are compatible with the translational steps subsequent to aminoacylation. The specificity of many of these steps with respect to unnatural amino acids is unknown. It is clear; however, that the ribosome does not tolerate certain very sterically demanding aminoacyl-tRNAs [12,13], and recent studies suggest that a threshold EF-Tu affinity is required for aminoacyl-tRNAs to enter the A-site of the ribosome [14]. Particular types of backbone analogs also seem to be excluded during these translational steps [15], although exactly how they are rejected is not well-understood. Our newly defined pool of AARS substrates is ideal for addressing these specificity questions because it contains a diverse collection of side chain, N-methyl-, α,α-disubstituted and β-aminoacyl-tRNAs. Here we describe a straight forward means of determining whether an enzymatically charged unnatural amino acid can be effectively incorporated into a peptide. The results of these experiments have allowed us to identify over 50 unnatural amino acids that can be incorporated at a single defined position in a peptide with high efficiency using an all-enzymatic translation system. Our work identifies building blocks that can be combined to prepare highly modified peptides that contain multiple unnatural amino acids. We also identify many cases where the engineering of synthetases, EF-Tu or the ribosome itself may be required to enable efficient incorporation of interesting analogs.

back to top

Aminoacylation Competition Assay
The MALDI-TOF MS assay that we used to identify unnatural amino acids that are AARS substrates [11] does not provide a quantitative measure of tRNA charging efficiency. The rate of synthesis of aminoacylated tRNA, and the degree to which an analog competes effectively against traces of contamination with the natural amino acid, can greatly affect the yield and purity of analog-containing peptides synthesized by in vitro translation. However, detailed kinetic analysis of large numbers of analogs is not feasible. We therefore devised an AARS competitive inhibition assay in which the charging of a natural amino acid in the presence and absence of an analog could be quantitatively compared by MS through the use of an isotopically labeled derivatizing reagent [11]. In order to compare analog inhibition among different AARSs, we set the natural amino acid at roughly 2x its Km and then added the unnatural amino acid analog at 1000-fold higher concentration. As an example, the assays for the inhibition measurement of valine analog 3-fluorovaline (N15) are shown in Figure 1. For each assay an identical amount of nonisotopically derivatized Val-AMP was added as an internal standard, and experiments in the presence or absence of the analog were performed using the deuterium (d15) labeled derivatizing reagent [11]. The presence of competing N15 clearly led to a strong decrease in the intensity of the d15-Val-AMP peak (Figure 1B) when compared to the assay with valine alone (Figure 1A). Comparison of the ratio of Val-AMP to d15-Val-AMP in the presence and absence of the analog allowed determination of the % inhibition value. The inhibition data for the analogs is presented in Table 1. As expected, different analogs vary widely in their ability to competitively inhibit the aminoacylation of the natural amino acids. In general, we expect good competitive inhibitors to be good substrates, although this correlation need not always hold. Nevertheless, as discussed below, analogs that are good competitive inhibitors tend to be translated more efficiently than analogs that are poor competitors, supporting the idea that the synthesis of high levels of aminoacylated tRNA during a translation reaction facilitates the synthesis of analog-bearing peptides.

back to top

Translation Efficiency of Analogs
We have previously defined a set of 90 amino acid analogs that can be enzymatically charged onto tRNA [11]. To determine which of these could be incorporated into peptides, we used an assay based on the translation of short peptides containing a single unnatural amino acid. We prepared 5 separate mRNA templates containing a C-terminal FLAG or His6 tag for purification; this set of templates is similar to but more complete than that previously described [1].

Each template was designed to test analogs of 4 of the natural amino acids (Figure 2, a-e). The translation reactions were then carried out in the PURE translation system [1,4] with 35S-methionine as the N-terminal radiolabel, except for Met analogs which were tested using 3H-His and template e. In each test, one of the natural amino acids was omitted and a corresponding analog added in its place; thus synthesis of radiolabeled tagged peptide required readthrough of the codon assigned to the unnatural amino acid [1]. The peptide yield for each unnatural amino acid was determined relative to the peptide yield for the corresponding natural amino acid, by comparison of the peptide radioactivity for the all-natural peptide with that containing the unnatural analog. To verify that the translated peptide contained the unnatural amino acid, the mass of the peptide was determined by MALDI-MS. For unnatural amino acids that differed in mass by +/–1 amu from their natural counterparts, we verified incorporation using templates requiring multiple (4-7) incorporations. For instance, incorporation of the tryptophan analog A18 was tested using a template with 4 tryptophan codons (see supporting information Figure S1) and was unambiguously identified due to its +4 mass difference from the tryptophan containing peptide. We grouped the amino acid analogs that we studied into three classes based on translation efficiency and specificity. Class I monomers resulted in a peptide yield that was >75% of that obtained with the corresponding natural amino acid, and the translated peptide was homogeneous by MALDI-MS analysis. Class II monomers were translated less efficiently, with a peptide yield ranging from 25% to 75% of that obtained with the natural amino acid, but the translated peptide was still homogeneous. Class III monomers were either poorly translated (<25% peptide yield) and/or led to a heterogeneous mixture of peptides. In many cases inefficient translation correlated with poor aminoacylation, suggesting that the synthesis of aminoacylated tRNA may have limited peptide yield. Peptide heterogeneity was most often due to the misincorporation of natural amino acids in place of the desired analog. We discuss these analogs in groups based on their side chain and backbone compositions. We used the same amino acid abbreviations used in our previous paper [11] (C = Charged, P= Polar, A =Aromatic, N= Nonpolar, α = α,α disubstituted, β = beta amino acid, M= N-methyl). Since our first report we have found two additional unnatural amino acids that are efficiently aminoacylated and translated; these compounds ('Photo-Leu', N25; 'Photo-Met', N26) are also discussed below. Finally, we have described the translation results for twelve of these analogs in a preliminary report [1]; because these analogs are included in the other experiments described herein, we have included them here.

back to top

Correlation Between Aminoacylation and Translation Assay
A scatter plot comparing inhibition of the AARS reaction containing the cognate, natural amino acid and translation yield (single incorporation in templates a-e) is shown in Figure 10. Analogs that are efficient AARS competitors are located in the top half of the plot, and analogs that are translated efficiently appear on the right. From the plot it is clear that there is a correlation between competition with the natural AA for the AARS and translation efficiency—there are no highly efficient AARS competitors that translate poorly and thus would appear in the upper left quadrant of the plot. There are, on the other hand, a few analogs that translate well, but are inefficient AARS competitors and appear in lower right quadrant. These include A19, A21-A25, C9, C11, N11, N20, and P1. Several of these (A24, A25, C11, 11, N20, and P1) were among the more inefficient analogs with the 20 AA template experiments. Thus it seems reasonable to conclude that for these analogs the inefficiency is occurring during the AARS charging step. The synthetases (LeuRS and TrpRS) activating the remaining analogs in the lower right quadrant of the graph (A19, A21-23 and C9) have especially low Kms (1.5 µM [59] and 12 µM [60] respectively) for their cognate AAs. This might allow them to compensate for their relative inefficiencies of charging since the analogs were tested in these particular translation experiments at 400 µM.

Specificity of the Translation Apparatus
As a step towards the synthesis of large libraries of NRP-like molecules using an in vitro translation system, we have studied the tolerance of the translation apparatus to a wide variety of amino acid analogs. The translation experiments are summarized by analog class in Table 2. Most side chain analogs are efficiently incorporated into peptides, but the backbone analogs are for the most part poorly translated. Of the 71 analogs tested, 41 side chain analogs are efficiently translated with high fidelity along with the other 19 AAs, and another 8 side chain and 2 backbone analogs should be efficiently translated if contamination of the translation system with natural, cognate AAs can be removed. The remaining analogs (mostly backbone), although AARS substrates, are poorly incorporated into peptides during translation. Considering the side chain analogs as a class, poor charging onto tRNA appears to be the most common explanation for inefficient translation. This is supported by the correlation between competitive aminoacylation inhibition (a surrogate for aminoacylation activity) and translation efficiency (peptide yield and homogeneity) (Figure 10). For certain side chain analogs this explanation is certainly correct because they have been incorporated into peptides with good efficiency using stop codon suppression systems. For example, p-nitrophenylalanine (A14) has been efficiently incorporated into proteins in E. coli using an engineered AARS [61]. Similarly β-cyclopentyl alanine (N19) [51] and L-C-propargyl glycine (N9) [3] have been efficiently added using E. coli in vitro translation systems. The correlation between charging efficiency and translatability suggests that enzyme engineering of the AARSs to better recognize certain analogs will be an important step to improve translation yield. Alternatively, enzymatic pre-charging of the tRNAs at high enzyme and amino acid concentrations prior to addition to the translation mixture may improve yields while minimizing product heterogeneity due to competition from efficiently acylated contaminating natural amino acids.

For the backbone analogs, the relative importance of aminoacylation efficiency vs. later steps in translation varies with the type of backbone modification. In the case of the α-hydroxy acids, we cannot detect tRNA acylation directly. However, the ribosomes from E. coli are known to be able to generate ester linked products for the α-hydroxy analogs of Phe, Leu, Ile, Ala, [15,51,57,62] and now Met. Since only α-hydroxy methionine was incorporated into peptides in our assays of 13 α-hydroxy acids, either the synthetases or EF-Tu must discriminate against the other α-hydroxy acids. Given that many α-hydroxy acids are common metabolites (e.g. glycolate, lactate, glycerate, malate) while others may be generated at low levels in cells by hydrolysis of acyl-CoA intermediates or reduction of a-keto acids, it would not be surprising if many synthetases and perhaps EF-Tu have evolved the ability to discriminate against α-hydroxy acids. If this is the case, it will be necessary to engineer modified synthetases and perhaps EF-Tu to enable the efficient enzymatic incorporation of α-hydroxy acids into peptides. Similar synthetase engineering may be needed for the N-methyl amino acids, which, except for N-methyl His and Asp were poorly incorporated into peptides in our experiments and were poor AARS competitors as well. Previous experiments show that chemically generated N-methyl aminoacyl-tRNAs can be used to generate methylated peptides with reasonable efficiency [15,63], suggesting that the main block to incorporation is not at the level of EF-Tu or the ribosome. By default, poor efficiency of tRNA charging may be the limiting factor. In contrast, the β-amino acids are likely to be discriminated against by the active site of the peptidyl transferase center, possibly due to a higher pKa of the amino group [53].

Click on image for larger view.

Figure 10. Comparison of % AARS competition and % translation yield for all of the analogs shown in Table 1. The symbols are as follows: blue circles-side chain analogs, green squares-α,α disubstituted amino acids, black open triangles-N-methyl amino acids, red diamonds-β-amino acids. Translation yields for analogs that gave multiple peaks in the translation experiment were set to zero for this graph.

back to BioFiles v4 n8
back to top