Combinatorial mutagenesis to restrict amino acid usage in an enzyme to a reduced set

Abstract

We developed an effective strategy to restrict the amino acid usage in a relatively large protein to a reduced set with conservation of its in vivo function. The 213-residue Escherichia coli orotate phosphoribosyltransferase was subjected to 22 cycles of segment-wise combinatorial mutagenesis followed by 6 cycles of site-directed random mutagenesis, both coupled with a growth-related phenotype selection. The enzyme eventually tolerated 73 amino acid substitutions: In the final variant, 9 amino acid types (A, D, G, L, P, R, T, V, and Y) occupied 188 positions (88%), and none of 7 amino acid types (C, H, I, M, N, Q, and W) appeared. Therefore, the catalytic function associated with a relatively large protein may be achieved with a subset of the 20 amino acid. The converged sequence also implies simpler constituents for proteins in the early stage of evolution.

Proteins are composed of 20 or nearly 20 kinds of naturally occurring amino acids. Does a protein need the full set of 20 amino acids to encode a specific protein function? Only a few studies have addressed this question experimentally. De novo protein-designing experiments have demonstrated that entire helical bundle architectures can be constructed from a reduced amino acid alphabet (1–3). In the cases of small proteins, the amino acid usage can be restricted with retention of their biological functions. First, for the 57-residue, predominantly β-sheeted Src SH3 protein, 70% of an active variant selected from a combinatorial library by the phage display approach was composed of just five amino acid residues: Ala, Gly, Glu, Lys, and Ile (4). In the cases of the 53-residue Arc repressor (5) and the 58-residue bovine pancreatic trypsin inhibitor (6), multiple alanine substitutions were made, and the proteins retained their specific functions. In contrast, no study has demonstrated that the function of a large protein can also be fabricated with a limited set of amino acids. It is still unclear, therefore, whether the principles deduced from the studies with the small proteins can be applied to much larger and topologically more complex proteins. Thus, we have developed an effective strategy to simplify the entire sequence of a large protein with conservation of its in vivo function.

To fabricate a simplified protein with a catalytic function, the 213-residue Escherichia coli orotate phosphoribosyltransferase (OPRTase, EC 2.4.2.10) was used as a template for restricting the building units to a limited set of amino acids while retaining the metabolic function. This enzyme is the product of the pyrE gene and catalyzes the Mg²⁺-dependent formation of orotidine 5′-monophosphate from orotate and α-d-5-phosphoribosyl-1-pyrophosphate, which is involved in the de novo pyrimidine biosynthesis as well as in the biosyntheses of histidine and tryptophan (7). An E. coli strain carrying a lethal mutation within the chromosomal pyrE gene is not able to grow in a minimum medium without uracil (8). The uracil auxotrophic phenotype is fully complemented by a plasmid carrying the wild-type pyrE gene. This complementation by the exogenous genes allows simple, rapid, and sensitive phenotype selection for OPRTase variants with catalytic activity. We combined this phenotype selection with a combinatorial mutagenesis to fabricate simplified amino acid sequences that are composed largely of a reduced set of amino acids and are able to manifest the OPRTase activity in vivo.

Experimental Procedures

Choice of Reduced Amino Acid Alphabet.

Leu and Ala were added initially to the amino acid subset because these two residues occur most frequently in E. coli OPRTase as well as in naturally occurring proteins (9). Because it seemed unlikely that the two hydrophobic residue types are sufficient to achieve suitable core packing of this relatively large enzyme, Val was also chosen for another hydrophobic amino acid. Two Thr residues are involved in a cluster of invariant residues, Asp¹²⁴-Asp-Val-Ile-Thr-Ala-Gly-Thr-Ala¹³², among the OPRTase family (10, 11). In addition, an unusual amino acid stretch of three contiguous Thr residues is present at positions 79–81. These facts suggest that the Thr residues specifically contribute to the structure and function of the enzyme. For this reason, Thr was adopted as a reduced-set residue. Asp, Arg, and Tyr were also added to the reduced amino acid set as representatives of acidic, basic, and aromatic amino acids, respectively. In addition to these seven representative residues, Gly and Pro were added to the subset because the flexibility of Gly and the rigidity of Pro make important contributions to the conformation of proteins. Thus, the present simplification experiment used the nine amino acid residues, Ala, Leu, Val, Thr, Asp, Arg, Tyr, Gly, and Pro (defined as the reduced set of amino acids; Fig. 1).

Figure 1.

Amino acid substitutions to reduce the amino acid types contained in E. coli OPRTase.

Plasmid Construction.

The pyrE gene was initially amplified from genomic DNA of E. coli JM109 by the PCR method using a pair of synthetic oligonucleotides containing HindIII and EcoRI restriction sites, respectively. The amplified pyrE gene was digested with HindIII and EcoRI and cloned into pUC119. The resulting plasmid was used to construct the combinatorial libraries.

Estimation of the Sensitivity of Phenotype Selection.

To estimate what level of activity is detectable by using the in vivo complementation of the E. coli pyrE-deficient strain RK1032 (F⁻, thr,leu,argE3,proA2,his,pyrE60,thi), two mutant enzymes, K73A+K103A and K73A+K100A+K103A, were prepared by oligonucleotide-directed mutagenesis (12). The importance of the three Lys residues for the catalytic activity was reported previously in the study of the Salmonella typhimurium OPRTase (13). It was confirmed that the uracil-auxotrophic growth of RK1032 could be complemented by the mutated pyrE gene encoding K73A+K103A but not by that encoding K73A+K100A+K103A. The activity of the mutant enzymes then was evaluated in vitro. Expression of the mutated genes and purification of the products were carried out as described (14). The purity of the enzymes used was >95% as judged by SDS-gel electrophoresis. The specific activity of each purified enzyme was determined at 37°C as described (13). The purified K73A+K103A and K73A+K100A+K103A proteins had specific activities that were 10,000- and 100,000-fold lower than that of the wild-type enzyme, respectively. Therefore, the in vivo complementation assay can detect activity that is ≈1/10,000 or greater relative to that of the wild-type enzyme.

Design of Oligonucleotide Primers for Randomization.

To generate the amino acid substitutions shown in Fig. 1, synthetic oligonucleotides were designed and synthesized as follows: the codons for Ala and Gly in the wild-type sequence were replaced by GSA, GSC, GSG, or GST (S = C + G); the Thr codons were replaced by RCA, RCC, or RCG (R = A + G); the Ser codons were replaced by DCC, DCG, or DCT (D = A + G + T); the Cys codons were replaced by KSC (K = G + T); the Leu and Val codons were replaced by STG; the Ile codons were replaced by VTC or VTT (V = A + C + G); the Met codons were replaced by VTG; the Lys codons were replaced by ARA or ARG; the Glu codons were replaced by GAW (W = A + T) or GAS; the Asn codons were replaced by RAC or RAT; the Gln codons were replaced by SAW or SAS; the His codons were replaced by YRC or YRT (Y = C + T); and the Phe codons were replaced by TWA or TWT. At the mixed positions, the two or three bases were so mixed that their encoding letters should appear with equal probability. The sequences with the base mixtures were flanked by unmutated 9- to 14-base stretches so as to ensure that the modified segments properly hybridized to the template single-stranded DNAs. These synthetic primers were used for the construction of semirandomly mutated plasmid DNA libraries as shown in Fig. 2.

Figure 2.

Construction of a combinatorial library for the pyrE gene and selection in E. coli strain RK1032. The entire sequence of the E. coli OPRTase gene was simplified stepwise in 22 rounds of mutagenesis and phenotype selection. In steps 1–3, a genetic library was constructed by a strategy similar to that of Huang et al. (15). (Step 1) A target nucleotide sequence on pUC119 was replaced with a linker containing a SalI restriction site to eliminate wild-type DNA contamination in the following library constructions. E. coli CJ236 was transformed with the resulting plasmid, and then uracil-containing single-stranded DNA was isolated. A synthetic oligonucleotide designed to replace the linker sequence and generate several types of amino acid substitutions within the target segment was annealed to the single-stranded DNA. (Step 2) Second strands were synthesized, and the resulting double-stranded plasmid mixture was amplified in E. coli JM109. (Step 3) In rounds 1–15, the amplified plasmid DNA variants were digested with HindIII and EcoRI to excise the pyrE gene variants with the combinatorial sequence, and the HindIII–EcoRI fragments were inserted into low copy-number plasmid pSTV29. The resulting plasmids were amplified again in E. coli JM109. The phenotype selection was through steps 4–5. (Step 4) The amplified plasmid mixture was used to transform the ΔpyrE E. coli strain RK1032. (Step 4′) In rounds 16–22, the pyrE gene variants were not cloned into the low copy-number plasmid. The mutated pyrE genes inserted into pUC119 were used to transform E. coli RK1032. The RK1032 transformants were spread on M9 medium agar plates without uracil and incubated for 2 days at 37°C. (Step 5) On average, 12 of the largest colonies then were selected from the plates. The plasmid DNA variants were prepared from the colonies and used for the subsequent PCR amplification of the pyrE genes. The amplified PCR products were sequenced directly to determine the amino acid sequence within the mutagenized region. Based on the sequence analyses, the positions where the reduced-set residues are tolerated were determined. Thus a simplified amino acid sequence was generated within the targeted segment on pUC119. (Step 6) The resulting sequence variant on the plasmid was subjected to the next round of mutagenesis.

Selection and Sequence Analysis of Phenotypically Active Mutants.

E. coli RK1032 was transformed with the respective plasmid DNA libraries. The transformed RK1032 then was spread on M9 medium agar plates supplemented with 0.2 mg⋅ml⁻¹ glucose, 10 μg⋅ml⁻¹ thiamine, 20 μg⋅ml⁻¹ each of threonine, leucine, arginine, proline, and histidine, and either 100 μg⋅ml⁻¹ ampicillin or 20 μg⋅ml⁻¹ chloramphenicol. On average 12 of the relatively large colonies were selected from the plates after 2 days of incubation at 37°C and cultured in 3 ml of LB medium supplemented with either 100 μg⋅ml⁻¹ ampicillin or 20 μg⋅ml⁻¹ chloramphenicol. The plasmid DNA variants were prepared from the cultures and used for the subsequent PCR amplification of the pyrE genes. The amplified PCR products were sequenced directly to determine the amino acid sequence within the region randomized. DNA sequencing was carried out with an ABI Prism BigDye Terminator cycle-sequencing ready-reaction kit and an ABI Prism 310 genetic analyzer.

Further Simplification of Simp-1.

To simplify the Simp-1 sequence further, the tolerance of changes of M197, N41, Q117, Q141, Q168, and Q5 with the nine reduced-set amino acids was tested one by one in that order. An inverse PCR method was used to randomize each position according to ExSite mutagenesis protocol (Stratagene) with slight modifications. To amplify the pyrE gene with base substitutions within the codon for Met-197, four mutagenic PCR primers, each of which contained a two-base mixture within the codon, were designed, synthesized, and mixed such that eight amino acid residues (Ala, Asp, Leu, Met, Arg, Thr, Val, and Tyr) could appear at this position with equal probability. The second primer was designed such that the mutagenic and the second primers annealed to the next sequences on opposite strands of the template plasmid. The following program was used: step 1, 95°C, 30 sec; step 2, 95°C, 30 sec; step 3, 55°C, 1 min; step 4, 68°C, 8 min; steps 2–4 were repeated 16 times. After amplification, the PCR products were treated with the DpnI restriction enzyme to digest the methylated, nonmutagenized parent DNA template and then were purified by agarose-gel electrophoresis. The purified DNAs were self-ligated, and the resulting ligation products were amplified in E. coli JM109. The amplified DNAs were used to transform E. coli RK1032, and the transformants were screened for the residual OPRTase activity in vivo on a plate of selective medium without uracil. After 2 days of incubation, an average of 12 relatively large colonies were selected from the plate. The plasmid DNA variants were isolated from the colonies, and the inserts were sequenced to identify amino acid substitutions. Allowable amino acid substitutions at positions 41, 117, 141, 168, and 5 also were determined in a similar manner.

Results and Discussion

Combinatorial Mutagenesis to Restrict the Amino Acid Usage.

Some segments of a protein tolerate extensive amino acid substitutions without compromising their structure and function (16–21), whereas a few amino acid substitutions can exert large effects on protein structure (22) and function (23, 24). To select allowable substitutions at permissive positions, we used the growth complementation of the pyrE-deficient E. coli strain RK1032 by exogenous genes for functional variants. This phenotype selection is sensitive enough, because a plasmid gene encoding a specific activity of 1/10,000 of the wild-type OPRTase can allow the host cells to undergo colony growth on a minimum medium without uracil (see Experimental Procedures). To make reasonably small libraries, the substitutions were restricted among similar amino acids (Fig. 1). In addition, the entire amino acid sequence of the enzyme was divided into 22 segments, each consisting of 8–13 residues. In the first trial, we simplified the 22 segments independently by selecting permissive substitutions in a mutated segment of the genes in which the other segments were left unchanged. Then, the resulting simplified segments were assembled into a single molecule. The advantage of this “independent simplification” strategy is that all the segments can be simplified at one time. Indeed, a simplified SH3 variant was generated by this strategy (4). However, in the present case, concatenation of the independently simplified fragments failed to produce a phenotypically active enzyme. Therefore, we used an alternative strategy in which the 22 segments were simplified in a stepwise manner, and the simplified sequences were fixed in the following rounds for the remaining segments. The entire sequence of E. coli OPRTase thus was simplified in 22 rounds of mutagenesis, selection, and introduction of the reduced alphabet residues into the relevant positions (Fig. 2). Fig. 3 shows the stepwise increase in the number of substitutions along the OPRTase sequence. The amino acid sequence of the final product, designated as Simp-1, contains a large number of the reduced set of amino acids; 182 (85%) of the overall 213 positions are occupied by the nine amino acid types.

Figure 3.

Amino acid sequences of E. coli OPRTase and its simplified variants. The nonreduced set and the substituting amino acids are shown in cyan and red, respectively. Orange shading indicates the region semiselectively randomized in each round.

Advantage of Stepwise Simplification.

The accumulative simplification is superior to the independent simplification, probably because the interference between mutations in different segments could be avoided in the course of the accumulative processes. A typical example is found in positions 9 and 182. The wild-type OPRTase Ile-9 is buried completely within an α-helix and interacts with Ile-182 by means of a hydrophobic interaction (Fig. 4a). When the Ile residue is located at position 182, Val seems to be allowed at position 9. Reasonably, the relevant independently simplified segment (residues 2–11) had Val at this position. However, in the case where Ile was replaced by Val at 182, the further introduction of Val at 9 was not tolerated, presumably because of the generation of a large cavity caused by the double Ile → Val substitutions. In our successful example, Val was introduced at position 182 in round 3, and the later combinatorial mutagenesis and phenotype selection in round 21 chose Leu for position 9; the terminal methyl group of the introduced Leu may fill a part of the cavity produced by the Ile-182 → Val substitution (Fig. 4b). Thus, our success in the simplification relied on the stepwise strategy as reported for minimizing a protein (25).

Figure 4.

Avoidance of the interference between mutations at different segments by stepwise mutagenesis. (a) The wild-type enzyme structure around positions 9 and 182. (b) Schematic illustration of amino acid substitutions at positions 9 and 182.

Further Simplification.

Despite the significant simplification of the amino acid sequence, Simp-1 still has a number of amino acid residues to be eliminated in the simplified sequence. For example, Simp-1 has a Met residue at position 197, as hydrophobic amino acids, Leu and Val, were not tolerated at this position. The Met-197 side chain, however, is somewhat exposed to the solvent in the wild-type enzyme structure, suggesting that some hydrophilic residues may be allowed at this position. Similarly, although Asn-41, Gln-5, Gln-117, Gln-141, and Gln-168 were resistant to replacement by Asp, it remained to be examined whether the neutral or basic polar reduced-set residues are allowed at these positions. Additional single-point random mutagenesis therefore was applied to positions 197, 41, 141, 168, 117, and 5. Randomization was done to generate the amino acid mixture indicated in Table 1. Sequence data of the functional variants also are summarized in Table 1. Based on the results, we introduced Tyr in place of Met at position 197, Arg at 41 and 141, and Thr at 117, 168, and 5 in place of their respective neutral hydrophilic residues. A further simplified variant thus was obtained and designated Simp-2.

Table 1.

Sequence data of functional variants obtained by additional mutagenesis experiments

Wild-type residue	Candidate amino acids in additional mutagenesis	Amino acids observed in functional variants*	Simp-2
M197	A, D, L, M, R, T, V, Y†	M(9), Y(3)	Y
N41	D, N, R, T, Y‡	N(9), R(2), T(1)	R
Q117	D, Q, R, T, Y§	D(2), Q(4), R(2), T(4)	T
Q168	D, Q, R, T, Y§	D(2), Q(6), R(2), T(3)	T
Q141	D, Q, R, T, Y§	D(1), R(5), Q(3), T(2), Y(1)	R
Q5	D, Q, R, T, Y§	Q(8), T(4)	T

^*An average of 12 functional variants were analyzed. These variants were obtained by in vivo complementation of the pyrE-defective E. coli following the site-directed random mutagenesis at the indicated positions. The frequency of an amino acid is given in parentheses next to the amino acid. 

^† The codon was randomized by using four mutagenic primers that contained RCG, CKG, KAC, and RTG, respectively (R = A + G; K = G + T). 

^‡ The codon was randomized by using two mutagenic primers that contained DAT and ASA, respectively (D = A + G + T; S = C + G). 

^§ Each of these codons was randomized by using three mutagenic primers that contained KAT, CRA, and ACT, respectively (K = G + T; R = A + G). 

In Fig. 5, the amino acid sequence of Simp-2 is compared with those of the wild-type enzyme and Simp-1. Fig. 6 indicates the positions of the simplified residues in Simp-2 superimposed on the structure of the wild-type OPRTase (27). In total, 65 of 87 positions at which simplification was attempted are occupied by the reduced set of nine residues. Consequently, 88% of Simp-2 is composed of the nine amino acids. More importantly, seven residue types (Cys, His, Ile, Met, Asn, Gln, and Trp) are absent from the simplified enzyme (the N-terminal Met is not taken into account). Hence, the entire sequence of Simp-2 is composed of just 13 amino acid types. Despite this dramatic change in its amino acid sequence, Simp-2 exhibits sufficient activity to complement the uracil auxotrophic growth of the pyrE-deficient E. coli strain even though the mutant's activity is not as high as that of the wild type. Thus, we have shown that the reduced amino acid set protein (RASP) contains the minimal elements required for achievement of the OPRTase activity. Our result exemplifies that the amino acid usage of a relatively large enzyme can also be restricted by this experimental method with retention of its catalytic function.

Figure 5.

Sequence alignment of the wild-type enzyme and simplified variants. The nonreduced-set amino acid residues are highlighted with black boxes. The N-terminal and catalytically important residues that were not subjected to mutagenesis are indicated with open letters. The secondary structures of the wild-type OPRTase are shown below their corresponding sequences. H, α-helix; S, β-strand.

Figure 6.

MOLSCRIPT (26) representation of a model of the monomer structure of the simplified OPRTase variant Simp-2. Eighty-eight percent of Simp-2 is composed of the reduced-set of nine amino acids, which are shown as a gray backbone ribbon. The residues not simplified in Simp-2 are rendered in black.

Simp-2 is the most simplified variant obtained thus far, but it may not represent the simplest possible OPRTase sequence. It is expected that the application of the successive segment-wise combinatorial mutagenesis reported here to the Simp-2 gene again would further restrict the amino acid usage in the enzyme molecule. Future attempts to generate more simplified enzymes should be done in this direction.

Primordial Proteins.

The results presented here provide some implications for understanding early protein evolution. The ancestor of the phosphoribosyltransferase family is thought to have arisen early in evolution, because its members are distributed in a wide variety of organisms from bacteria to mammals. Several authors (28–30) have proposed that early protein synthesis was much simpler and involved only 7–13 amino acids, and that the existing system for protein synthesis has evolved progressively from the primordial one by gradually obtaining new amino acids for the repertoire in protein synthesis. Such primitive proteins, which might have comprised a reduced set of amino acids, are not thought to have been optimized fully in terms of their stability and activity, because such optimizations may be the results of a long history of evolution. However, the primitive proteins must have had a sufficiently adequate structure for functional interactions and catalysis. Here we show that the full amino acid alphabet set is not necessarily essential, and only 13 amino acid types are sufficient to achieve protein functions in vivo. This number of amino acid types is quite consistent with the above-mentioned propositions. Accordingly, our result supports the hypothesis that proteins in the early stage of evolution were made from a limited set of amino acids.

Abbreviation

OPRTase

orotate phosphoribosyltransferase