Introduction

Proteins present in existing organisms are composed of 20 different amino acids, encoded by the universal genetic code. This code is considered to have evolved from simpler forms (Crick 1968; Wong 1975; Fournier and Gogarten 2007), such as one that encoded 19 amino acids. Similarly, genes encoding proteins composed of 20 unique amino acids are considered to have evolved from those encoding proteins composed of 7–13 unique amino acids (Wong 1976; Osawa et al. 1992; Baumann and Oro 1993) available in the primitive earth conditions (Miller 1953). Along with the establishment of novel cellular biosynthetic pathways, new amino acids were gradually introduced into the repertoire for synthesizing diverse proteins (Wong 1976). This theory is supported by the phylogenies of the aminoacyl-tRNA synthetases (aaRSs) (Woese et al. 2000).

Tryptophan (Trp) is thought to be the last amino acid included within the universal genetic code. Biosynthetic pathway of Trp pointed to the possibility that this amino acid was introduced during the late stages of the evolution of genetic code (Wong 1975). Indeed, Trp is hardly generated during chemical evolution experiments imitating primordial conditions (Miller 1953). Further, sequence analysis of aaRSs showed that tryptophanyl-tRNA synthetase (TrpRS), as well as tyrosyl-tRNA synthetase (TyrRS), are the youngest among 20 aaRSs (Ribas de Pouplana et al. 1996). At present, Trp is believed to be the last amino acid introduced during evolution (Trifonov 2004).

Proteins composed of 19 amino acids would have been able to evolve by accumulating beneficial mutations without using the 20th canonical amino acid. A previous report has described the experimental evolution of two naturally occurring proteins, GroEL and GroES, composed of 19 unique amino acids without Trp (Wang et al. 2002). Directed evolution improved the ability of E. coli-derived GroEL and GroES to assist folding of GFP. In spite of random mutagenesis to the DNAs, mutants with improved properties did not harbor Trp. Directed evolution has been successfully used for improving the properties of 19-amino-acid proteins without Cys. A single-chain-dimer streptavidin had been evolved to improve binding activity to biotin, without introducing Cys (Aslan et al. 2005).

To further understand the evolution of proteins composed of 19 distinct amino acids, we created active 19-amino-acid green fluorescent proteins (GFPs) without Trp using directed evolution with repeated rounds of random mutagenesis and selection. Here, we describe a less active mutant composed of 19 amino acids without Trp that generated by replacing Trp57. This mutant gained activity by accumulating beneficial mutations through two rounds of experimental evolution. We found that the combined effect of the two beneficial mutations on the initial mutant was additive, like in the case of GroEL, where the stability of the protein improved because of the combined effect of mutations (Wang et al. 1999). Recording the improvement in the activity of the 19-amino-acid protein without Trp using experimental evolution, where it is easy to observe the accumulation of beneficial mutations, would provide additional insights into the late stages of protein evolution.

Materials and Methods

DNA Construction

The sequences of the primers used in DNA construction are shown in Table S1. Site-directed mutagenesis was performed using the Gene Tailor site-directed mutagenesis system (Life Technologies).

For construction of the plasmid DNA encoding the GFP mutants, a coding region of GFPS1 on pGFPS1 (Seki et al. 2008) was amplified using primers named Placq-Bgl_II-RBS-F and Placq-Spe_I-R (Table S1). The amplified fragment was cloned into Placq-GFP (Ayukawa et al. 2010) between the Bgl II and Spe I sites to replace the GFP coding sequence with GFPS1. We called this construct Placq-GFPS1. Site-directed mutagenesis on Placq-GFPS1 was performed to create Placq-S1-W57F and Placq-S1-W57Y by using forward primers named W57F-F and W57Y-F, respectively, and a common reverse primer named W57FY-R. Saturation mutagenesis using degenerate oligonucleotides (NNK codon with N = G/A/T/C and K = G/T) was performed on S1-W57F to introduce mutations to amino acids adjacent to Phe57. For the mutation at Val29, we used a pair of primers named V29NNK-F and V29NNK-R. Similarly, for other single-point mutations at Phe46, Leu53, Leu60, and Leu64, we used pairs of primers named F46NNK-F and F46NNK-R, L53NNK-F and L53NNK-R, L60NNK-F and L60NNK-R, L64NNK-F and L64NNK-R, respectively. For the mutation at the three consecutive positions, Asp216, His217, and Met218, we used a pair of primers named D216H217M218NNK-F and D216H217M218NNK-R.

For the construction of the first and second round plasmid libraries, error-prone PCR was performed as previously described (Cadwell and Joyce 1994) on the S1-W57F coding region. Typically, a 50-μL reaction mixture contained 10 mM Tris–HCl (pH 8. 3), 50 mM KCl, 3 mM MgCl2, 0.5 mM MnCl2, 0.2 mM of each dATP and dGTP, 1.0 mM of each dTTP and dCTP, 0.2 μM of each primer set, 1 ng of template DNA, and 1.25 U of rTaq DNA polymerase (TOYOBO). The primers and procedure used for cloning were the same as those for Placq-GFPS1 construction. We found that mutation frequency was 6.7 mutations per 1,000 base pairs. The mixture of the random mutant plasmids was introduced into DH5α by electroporation. Mutant libraries were screened by FACS method. Typical mutant libraries for each FACS screening consisted of approximately 106 independent clones.

For the purpose of measuring the fluorescence intensity of the purified mutant proteins, we introduced a C-terminal His tag sequence YRYEFQLSAASKLAAALEHHHHHH into the sequences. In order to introduce a His tag sequence into the mutants derived from Placq-GFPS1, we applied the following three steps.

In step 1, we prepared the coding sequence of GFPS1-Cla_I-His that featured GFPS1 sequence followed by a Cla I recognition sequence and a His tag sequence. For this purpose, pGFPS1-Cla_I-His was constructed by introducing the Cla I site and His tag sequence simultaneously to pGFPS1 by site-directed mutagenesis with primers named pGFP-Cla_I-His-F and pGFP-Cla_I-His-R. The forward primer contained the Cla I site located between the GFPS1 coding region and the His tag to enable cloning as desired in the final step. By insertion of an additional nucleotide between the GFPS1 coding region and an out-of-frame His tag sequence, the primers generate the His tag sequence in frame.

In step 2, the GFPS1 coding sequence on Placq-GFPS1 was replaced with GFPS1-Cla_I-His. The GFPS1-Cla_I-His PCR fragment was amplified from pGFPS1-Cla_I-His by using primers named pGFPS1-F and GFP-His Spe_I-R. The fragment contained the NdeI site at the start codon of the GFPS1-Cla_I-His and Spe I site next to the stop codon of GFPS1-Cla_I-His. The PCR fragment was cloned into Placq-GFPS1 between the NdeI and SpeI sites to replace the GFPS1 coding sequence on Placq-GFPS1 with the GFPS1-Cla_I-His. We called this construct Placq-GFPS1-Cla_I-His.

In the third and final step, the coding sequences of the EGFP and GFPS1 mutants obtained from the selection in this work were amplified and cloned into Placq-GFPS1-Cla_I-His between the NdeI and ClaI sites to replace the GFPS1 coding region with the mutant sequences. For this purpose, we used Placq-Bgl_II-RBS-F, the same forward primer for the construction of Placq-GFPS1, and GFP-cloning-Cla_I-R.

Flow Cytometric Analysis and Cell Sorting

All flow cytometric analyses and cell sorting were carried out using a Becton-Dickinson FACSCalibur instrument with a 488-nm laser and a 515- to 545-nm emission filter. The throughput rate of cells was adjusted to 3,000 events per second. For the library constructed by saturation mutagenesis, 106 cells were analyzed. For the library constructed by random mutagenesis, 2 × 107 cells were sorted in exclusion mode. A gate in the fluorescence channel was set to recover highly fluorescent cells. Additional gates were set in the forward- and side-scatter channels to exclude events arising from large particles. The collected bacterial cells were centrifuged and rescued in LB medium. The cells were plated on LB dishes supplemented with 30 μg/mL kanamycin and incubated at 37 °C overnight. Colonies were picked, cultured, and analyzed using CellQuest (Becton Dickinson) and WinMDI 2.9 (http://en.bio-soft.net/other/WinMDI.html).

Protein Expression and Purification

Plasmids containing the GFP mutants bearing the His-tag-coding sequence were transformed into DH5α. Single colonies were picked and cultured at 37 °C overnight in 3 mL LB medium supplemented with 30 μg/mL kanamycin. Overnight culture (1 mL) was transferred into 100 mL of fresh LB medium containing 30 μg/mL kanamycin, and incubation was continued at 37 °C. When the optical density of cultures reached 0.6 at 590 nm, the temperature was lowered to 26 °C, and protein expression was continued for an additional 4 h. Following this, cells were collected by centrifugation for 5 min at 7,000 × g and 4 °C. The harvested cells were suspended in 4 mL of buffer A (40 mM Tris–HCl [pH 7.6], 300 mM NaCl, 10 mM imidazole, 1 mM DTT) and disrupted by sonication. The disrupted cells were centrifuged for 10 min at 12,000 × g and 4 °C. Aliquots (4 mL) of the supernatant were mixed with 6 mL buffer A, and filtered through Millex-GV 0.45-μm filters (Millipore). The filtrate was incubated with 500 μL bed volume of TALON resin (Clontech) at 4 °C for 1 h. The resin was then retrieved and washed with 30 mL buffer A. Bound proteins were eluted with elution buffer (buffer A containing 300 mM imidazole).

Fluorescence Intensity Measurement on Purified Mutant Proteins

For fluorescence measurements, 15.0 μg of purified protein was diluted in 2,000 μL PBS and analyzed at room temperature with an FP-6500 spectrofluorometer (JASCO). The excitation wavelength was set at 488 nm, and the emission wavelength between 500 and 730 nm (bandwidth =1 nm for both excitation and emission) was recorded.

Results

The loss of fluorescence due to mutation of a unique Trp in the parent GFP was not restored by a mutation of the Trp position itself or adjacent amino acids

As a starting point for mutagenesis in this work, we used a simplified GFP mutant named S1-W57F, in which Phe replaced the unique Trp57 (Fig. 1) of a brightness- and maturation speed-improved mutant GFPS1 (Seki et al. 2008). Enhanced GFP has been reported to not to allow Trp57 replacement by any of the other 19 amino acids (Steiner et al. 2008) or by non-canonical Trp-like amino acids (Budisa et al. 2004) without abolishing fluorescence. Accordingly, we confirmed that the replacement of Trp57 by the structurally similar residue Phe or Tyr abolished the fluorescence of GFPS1 (Fig. S1). We then attempted saturation mutagenesis at each position spatially close to Phe57 of S1-W57F to rescue the protein from structural perturbation introduced by the W57F mutation. Therefore, we prepared mutants in which codons for one of the amino acids Val29, Phe46, Leu53, Leu60, or Leu64 (Fig. S2) was replaced by NNK. Fluorescence intensities of E. coli libraries, consisting of 105 colonies expressing each mutant were then measured by FACS method after 488-nm excitation. We found that none of the mutations made for correcting Trp57 mutation improved the fluorescence of S1-W57F. Further, we introduced saturation mutagenesis simultaneously at three sequential positions, Asp216, His217, and Met218, which are also close to Phe57 and attempted to select active mutants from 106 cells for 105 clones, an adequate library size for the number of residues randomized. The strategy, however, did not improve the fluorescence of S1-W57F.

Fig. 1
figure 1

Ribbon structure of EGFP (PDB entry 2y0g). The Trp residue and the central chromophore are highlighted as stick models. The Trp residue is shown in orange and the chromophore is shown in yellow. Molecular graphics were illustrated with CueMol 2 (http://www.cuemol.org/)

Full size image

Repeated Rounds of Directed Evolution on 19-Amino-Acid GFP Without Trp

Because saturation mutagenesis at positions spatially close to Phe57 failed to yield any mutations that improved the fluorescence of S1-W57F, we used an error-prone PCR (epPCR) on the whole S1-W57F coding sequence. The epPCR products of the S1-W57F coding sequence were then cloned into Placq-GFP (Ayukawa et al. 2010). From an E. coli mutant library consisting of 106 mutants carrying plasmids that harbor the epPCR product, we selected clones that showed higher fluorescence than S1-W57F mutant cells. Plasmids were isolated from 10 colonies, and their sequences were analyzed (Fig. S3 and Table 1). We found that these sequences contained 13 amino acid changes, distributed throughout the S1-W57F gene. Because some clones had the same amino-acid sequence, in all, 6 sequences were found to be varied in the analyzed clones. All of the 6 mutant S1-W57F sequences lacked Trp in their whole coding region including at position 57. The selected clones lacked deletions, insertions, or additional stop codons. The most frequent mutation was S205T, which is adjacent to the chromophore, neither adjacent nor close to Phe57. This mutation was reported as one of the mutations that enhanced folding (Cubitt et al. 1999). A comparison of the fluorescence intensity of the mutant clones corresponding to the 6 distinct sequences revealed that mutants containing S205T mutation produced a protein with higher fluorescence intensity than the rest of the mutants (Table 1).

Table 1 Amino acid mutations, fluorescence intensities, and the number of clones for the mutants from the first round of the selection
Full size table

Additional active 19-amino-acid mutants without Trp were generated from a second round library prepared from the R1-1 mutant harboring a single S205T mutation. After mutagenesis and selection as described for the first round, we sequenced 3 mutants and found that each contained 1 additional amino acid mutation (Table 2). One of the substitutions, N105I, was found at the same residue as the N105T on the superfolder mutant (Pedelacq et al. 2006) and N105Y on the superfast mutant (Fisher and DeLisa 2008). Further, these mutations were not close to the chromophore, Phe57, or Thr205. All three mutants had higher fluorescence intensity than that of the predecessor R1-1.

Table 2 Amino acid mutations, fluorescence intensities, and the number of clones for the mutants from the second round of the selection
Full size table

Fluorescence Intensities of Purified 19-Amino-Acid Mutants Without Trp

To purify mutant proteins and evaluate the fluorescence intensity, we introduced His-tag to 7 representative clones: previously reported GFPs (EGFP and GFPS1), the initial simplified mutant (S1-W57F), and selected mutants (R1-1, R2-1, R2-2, and R2-3). Upon excitation at 488 nm, all of the simplified mutants showed similar fluorescence spectra to that of EGFP and GFPS1 (Fig. S5). The order of the fluorescence intensity among the selected mutants was the same regardless of whether quantification was done using the E. coli cells expressing the GFP mutants or purified proteins. We also found that the fluorescence intensity of R1-1 was higher than that of S1-W57F and that of R2-3 was 4.3 times higher than that of R1-1 (Fig. S5). The most active 19-amino-acid mutant without Trp showed 32 and 50 % fluorescence intensity of GFPS1 and EGFP, respectively, which are proteins that consist of the 20 canonical amino acids.

To further evaluate the effect of the beneficial mutations on 19-amino-acid mutants without Trp, we constructed additional mutants N105I and/or S205T of the 19-amino-acid protein S1-W57F and the 20-amino-acid parental protein GFPS1. As expected, each of N105I and S205T improved fluorescence of S1-W57F (Fig. 2). Further, we found that the combined effect of N105I and S205T on S1-W57F was nearly additive. However, the beneficial effects of these mutations, individually and in combination, were not as apparent on GFPS1 as they were on S1-W57F.

Fig. 2
figure 2

Evaluation of the effect of mutations that improved the activity of 19-amino-acid mutants without Trp. a) Relative fluorescent intensity of the GFP mutants. N105I and S205T, the two mutations appeared in R2-3, were introduced both individually and in combination on the 19-amino-acid mutant S1-W57F and the 20-amino-acids parental mutant GFPS1. The purified proteins were excited at 488 nm, and emission intensity at 510 nm was recorded. Fluorescence intensity of GFPS1 was set at 1. The protein concentration for all measurements was set to 7.5 μg/ml in PBS. Results are presented as means ± SD (n = 3). b) Schematic diagram illustrating that beneficial mutation is detected more easily on a poorly active mutant than on a highly active mutant. The corners of the cube represent the genotypes of GFP mutants. The sizes of the spheres represent fluorescence activities. The bottom face indicates 19 amino-acid mutants. The top face indicates 20 amino-acid mutants

Full size image

Discussion

In this report, we provide an additional experimental example of the accumulation of beneficial mutations on a 19-amino-acid protein without Trp, which is thought to be the last amino acid to be included within the universal genetic code. First, W57F mutation in GFPS1 abolished fluorescence (Fig. S1). Following the W57F replacement, using directed evolution with two rounds of random mutagenesis, we showed that S1-W57F, which is the initial 19-amino-acid mutant without Trp, could evolve and gain activity. Trp57, a unique Trp residue in Aequorea-derived GFPs, is one of the residues responsible for chromophore formation, which forms the origin of green fluorescence. The process of chromophore formation requires a hydrophobic environment to allow the autocatalytic cyclization of the residues at positions 65–67. Trp57 is a part of the proline-rich “PVPWP” pentapeptide motif of GFP and contributes to maintaining the hydrophobic environment required to protect the chromophore from collisional quenching by oxygen or other diffusible ligands (Steiner et al. 2008). Fluorescence loss caused by the W57F mutation on GFPS1 may be attributed to a perturbation of the hydrophobic environment, provided that the mutations introduced for constructing GFPS1 (Seki et al. 2008) did not change the hydrophobic environment drastically. The reduction of chromophore formation observed when the W57F mutation was introduced in wild-type GFP (Bell et al. 2003) can also be attributed to the same reason.

Moreover, we observed an additive effect of the beneficial mutations on the 19-amino-acid protein without Trp using experimental evolution involving random mutagenesis to the whole gene sequence. Both the single mutations, N105I and S205T, improved the activity of S1-W57F (Fig. 2). When a double mutation (N105I and S205T) was introduced to S1-W57F, the fluorescence intensity was enhanced in a nearly additive manner. A previous report describes S205T as a mutation that enhances folding (Cubitt et al. 1999). N105I was found to occur at the same site as N105T and N105Y mutations, both of which showed increased tolerance against denaturants during a refolding experiment on the superfolder (Pedelacq et al. 2006) and superfast mutants (Fisher and DeLisa 2008), respectively. Therefore, S205T, and probably N105I also, might contribute to improve the folding of S1-W57F that had been disrupted by the W57F mutation. Although the combination of two individual beneficial mutations does not always show additive effects (Weinreich et al. 2005), we observed that S1-W57F accumulated two individual mutations and evolved to R2-3 with a simple uphill walk, which is one of the key features of directed evolution (Tracewell and Arnold 2009). Even on 19-amino-acid proteins, additivity of beneficial mutations in directed evolution was also suggested by the accumulation of beneficial amino acid replacements although this study selected the replacements by comparison of protein family sequences (Wang et al. 1999). Further, our present study in fact demonstrated the additivity of beneficial mutations that had accumulated by directed evolution involving repeated rounds of random mutagenesis to the whole gene sequence.

The active 19-amino-acid GFP mutants without Trp created in this study could be used as scaffolds that allow the introduction of Trp at any position. Such experiments would provide insights into the folding process of GFP. Because of its unique fluorescence properties, Trp has been utilized as a folding reporter to trace refolding processes. The fluorescence peak of Trp shifts to a longer wavelength when the protein of interest is denatured, because of the exposure of a Trp residue from the protein’s interior to solvent water. The refolding process of GFP variants has been investigated using Trp57 and a chromophore (Enoki et al. 2004). Studies that involve the construction of various single Trp mutants by introducing a unique Trp at various positions throughout the structure into a Trp-free background might provide useful information about the protein folding process, as was shown on another protein (Maki et al. 2007). Re-introduction of Trp on a 19-amino-acid GFP without Trp would help understand the process of protein folding in detail. A 19-amino-acid GFP mutant without Cys has been created by combining two mutations, C48S and C70M, which were found by site-directed mutagenesis and saturation mutagenesis on Cys residues, respectively (Suzuki et al. 2012). This mutant improved the diffusional property of GFP in an oxidative environment and enhanced the applicability of the GFP tag in the analysis of the secretory pathway of proteins.

Because beneficial mutations on poorly active proteins can be readily detected when compared to that on highly active proteins containing 20 unique amino acids (Kotsuka et al. 1996; Zhou et al. 1996), we first created a 19 amino-acids mutant protein by site-directed mutagenesis. We then selected active mutants from a library prepared by random mutagenesis of the initial mutant. Indeed, N105I and S205T mutations found in R2-3 improved fluorescence of S1-W57F even when introduced individually (Fig. 2). In contrast, these mutations showed no clear improvement on the parental GFPS1. This asymmetric effect between different genetic backgrounds, known as epistasis (Lehner 2011; de Visser et al. 2011), is the basis of the “suppressor mutation method” in protein engineering (Kotsuka et al. 1996).

Trp was introduced into the genetic code likely because a bulky hydrophobic amino acid was needed for proteins to efficiently form a hydrophobic core. Although the universal genetic code has utilized Trp as one of the key building blocks, proteins can artificially evolve without Trp. For example, previous reports showed direct evolution of 19-amino-acid proteins, by introducing random mutagenesis to whole gene sequences (Wang et al. 2002; Aslan et al. 2005). Similarly, another report suggested that even an 18-amino-acid protein could evolve, while only part of the gene sequence was mutagenized in the report (Moroz et al. 2013). Proteins composed of 19 amino acids without Trp have also emerged through the natural evolution. For example, the percentage of E. coli proteins composed of 19 amino acids other than Trp is 11 %, which is the second largest number among E. coli protein composed of 19 amino acids, following 15 % for the percentage of proteins composed of 19 amino acids other than Cys (Han and Lee 2006). Furthermore, Trp is not essential for the catalytic moiety of E. coli enzymes (Pezo et al. 2013). If Trp does not provide catalytic activity, the question arises as to what is the significance of Trp addition to the genetic code. Introduction of Trp, which is the most bulky amino acids among the canonical ones, can strengthen hydrophobic interaction, as seen in another GroEL study (Muller et al. 2013). Fournier and Gogarten also reported that one of the most significant trends identified during the expansion of genetic code is an increase in the number and kind of aromatic amino acids, which can contribute to the packing of the protein hydrophobic core and helped in folding (Fournier and Gogarten 2007). If genes evolved to be longer to provide a new function, there will be a greater need for more amino acids that form the core of the protein than that form the surface to ensure proper folding (Supplementary Text). Therefore, the bulky hydrophobic amino acids would have contributed to occupy large volume of protein core (Alvarez-Carreno et al. 2013). With the length of the protein held constant, the occupation of the core by bulky hydrophobic amino acids allows allocation of more hydrophilic amino acid residues to the protein surface that is responsible for protein-protein interactions (Fig. S6). In this regard, addition of Trp, which has bulkiest side chain among the 20 canonical amino acids, was more significant than the addition of the other canonical amino acids.

Reappearance of excluded amino-acids often complicates the process of selecting an active simplified protein from the library that is constructed by random mutagenesis (Yamamoto et al. 2003). In our study, the reappearance of the UGG codon through random mutagenesis was relatively rare, since only one codon specifies Trp in the genetic code. However, if one attempts to exclude various types of amino acids through directed evolution, the reappearance of the codons for the specific amino acids through random mutagenesis would be inevitable. Reappearance of codons for the specific amino acids decreases the effective library size, thereby crippling the process of accumulating beneficial mutations. Thus, previous studies did not use directed evolution for the construction of simplified proteins. Reappearance of the codons for the specific amino acids during the directed evolution involving repeated rounds of random mutagenesis can be minimized using a simplified genetic code which assigns fewer than 20 amino acids to the sense codons (Kawahara-Kobayashi et al. 2012). Because the simplified genetic code completely excludes the specific amino acid from the genetic code, it is not incorporated into proteins translated from any mRNA. Although this code is currently an in vitro technique only, combination of this code with in vitro directed evolution method involving large library size, such as ribosome display, would lead efficient creation of simplified proteins. Therefore, the simplified code allows for the application of a conventional directed evolution strategy for the creation of simplified proteins to assess the functions of primordial proteins as well as to improve the utility of pharmaceuticals.

Creation of proteins composed of 19 unique amino acids could contribute to future creation of an organism that utilizes only 19 kinds of amino acids in its genetic code. To address the question whether the number of unique amino acids used in the proteome of an organism can be lowered, Pezo tried to eliminate Trp from E. coli proteome using a missense suppressor tRNAHis bearing the anticodon for Trp (Pezo et al. 2013). If their trial would have succeeded and if one additionally eliminates TrpRS and/or tRNATrp from the organism, the UGG codon becomes completely reassigned. Strategies involving global elimination of a specific amino acid in vivo could have lethal effects due to the severe damage to the host proteome. To create an organism that utilizes only 19 kinds of amino acids in its genetic code in future, we should create functional protein sequences without the specific amino acid and combine them into one genome by using genome engineering technology (Isaacs et al. 2011). In studies that use a combination of the simplified genetic code and directed evolution to facilitate selection, an active 19-amino-acid GFPs could be used as a folding reporter to distinguish folded or aggregated mutants of a target protein, when expressed as a C-terminal fusion construct of the target protein (Waldo et al. 1999). This would facilitate selection, especially if the initial 19-amino-acid inactive mutant or the mutants in the randomized library are aggregated. The creation of an organism that utilizes only 19 kinds of amino acids in its genetic code would give further implications about the significance of the appearance of the amino acids in the genetic code.

In conclusion, we improved a 19-amino-acid GFP without Trp through two rounds of random mutagenesis and selection. The combined effect of beneficial mutations in the poorly fluorescent 19-amino-acid GFP without Trp were additive. In contrast, these mutations showed no clear improvement individually or in combination on the parental GFP composed of 20 unique amino acids. The experimental evolution of proteins composed of fewer than 20 amino acids, in this and previous studies (Wang et al. 2002), suggested that proteins translated by genetic codes just before the completion of the universal genetic code would have evolved in the same manner as proteins composed of 20 amino acids. On the other hand, a study that severely limited the kinds of amino acids at non-conserved regions of a protein showed that some amino-acid sets were less effective than others in generating functional proteins (Tanaka et al. 2011). What amino-acid set is enough to retain evolvability equivalent to that of the 20-amino-acid set is an interesting question that warrants future research.