A conserved histidine residue drives extein dependence in an enhanced atypically split intein

Abstract

Split intein-mediated protein trans splicing (PTS) is widely applied in chemical biology and biotechnology to carry out traceless and specific protein ligation. However, the external residues immediately flanking the intein (exteins), can reduce the splicing rate, thereby limiting certain applications of PTS. Splicing by a recently developed intein with atypical split architecture (‘Cat’) exhibits a stark dependence on the sequence of its N-terminal extein residues. Here, we further developed Cat using error prone PCR and a cell-based selection assay to produce Cat* having greatly enhanced PTS activity in the presence of unfavorable N-extein residues. We then applied solution nuclear magnetic resonance spectroscopy and molecular dynamics simulations to explore how dynamics of a conserved B-block histidine residue (His₇₈) contribute to this extein dependence. The enhanced extein tolerance of Cat* reported here should expand the applicability of atypically split inteins, and the mechanism highlights common principles that contribute to extein dependence.

Introduction

Protein splicing is a biochemical reaction wherein an intervening protein (intein) undergoes a posttranslational auto-processing event in which it self-excises while simultaneously ligating its flanking external residues (exteins) (Fig. S1) ^1,2. Although most identified inteins are contiguous domains and undergo cis-splicing, some are endogenously found as fractured genes. The resulting naturally split inteins are expressed as two separate domains, the N-intein (Int^N) and C-intein (Int^C) which condense prior to undergoing protein trans-splicing (PTS) ^3,4. Some naturally split inteins splice with exceptionally fast rates (t_½ <1 min), and these ultrafast split inteins serve as powerful tools for protein engineering and synthetic biology ^5–8. For example, ultrafast split inteins have been applied to synthetically incorporate chemical modifications such as fluorescent probes or posttranslational modifications within recombinantly expressed proteins, both in vitro and in vivo^9–11. Moreover, split inteins have enabled the generation of cyclic peptide libraries for screening inhibitors of protein-protein interactions, and gene therapy strategies to deliver large genetic cargo^12–14.

Naturally split inteins canonically fracture to form an Int^N and Int^C of approximately 100 and 35 amino acids respectively. However, inteins with atypical split sites (small Int^N and large Int^C) have also been discovered ^15–17. Since the shorter intein fragment is readily accessible through chemical synthesis, atypically split inteins are especially suited for the N-terminal modification of recombinant proteins ¹⁸. Previously, we applied consensus design of inteins from the T4-bacteriophage-type DNA-packaging terminase large subunit (TerL) to engineer a consensus atypical intein (Cat) ¹⁷. Cat exhibits enhanced splicing activity and stability in ambient conditions compared to other members of the TerL family. However, despite these advantages, substitutions at the N-extein −1 position from the endogenous glutamate residue profoundly decrease splicing activity. Such extein dependence is common among inteins evolved within the context of specific flanking residues, and can significantly impair applications of PTS.

Prior efforts have both elucidated the molecular mechanism by which extein residues impact splicing and engineered inteins with minimal extein dependence ^19–23. Inherently, a Cys, Thr, or Ser residue is mandatory at the first position of the C-extein (+1) as the nucleophile for formation of the branched intermediate. Additional residues resembling the native proximal N- and C- extein sequence are often favored as well. For example, naturally split inteins from the α subunit of DNA Polymerase III (DnaE) strictly depend on the +2 C-extein residue^19,21,22. This +2-position residue interacts with a catalytic F-block histidine (His₁₂₅), and mutations that augment His₁₂₅ dynamics can partially restore splicing activity in unfavorable extein contexts ^20–22. Furthermore, the atypically split VidaL intein favors glycine at the −1 position, which may facilitate scissile amide bond distortion through hydrogen bonding interactions with the side chains of a B-block Histidine and F-block Tyrosine^23,24. However, efforts to relieve N-extein dependence in VidaL have been unsuccessful, and a naturally split intein with an atypical split site that exhibits minimal extein dependence has yet to be identified or engineered.

Here, through employing an unbiased approach using error prone polymerase chain reaction (PCR) and a cell-based selection, we identify mutations in the Cat intein that accelerate splicing in nonnative extein contexts. The enhanced sequence tolerance is validated in vitro, and the mechanism is explored through NMR spectroscopy and molecular dynamics (MD) simulations. A key catalytic histidine residue (His₇₈) and a single substitution at an adjacent non-catalytic residue (S79P) are found to dramatically impact the activity, structure, and dynamics of the intein active site.

Results

Identification of a Cat mutation that imparts enhanced N-extein tolerance

Although Cat exhibits ultrafast splicing with its native Glu₋₁ residue (1.17± 0.10 ×10⁻² s⁻¹), we previously showed that N-extein substitutions such as Gly₋₁ (7.92 ± 0.33 ×10⁻⁵ s⁻¹) or Ala₋₁ (2.14 ± 0.08 ×10⁻⁴ s⁻¹) drastically reduce reaction rates (Fig. 1a, Table S1)¹⁶. This dependence is mainly localized to the −1 Ext^N residue, as −2, +2 and +3 C-extein mutations are tolerated. To engineer enhanced activity under nonnative N-extein residues, we adapted a previously reported cell selection system in which kanamycin resistance is coupled to PTS activity through reconstitution of a split aminoglycoside-3’-phosphotransferase (Kan^R) from two individually transcribed fragments (Fig. 1b)²⁰. We carried out error prone PCR on the Cat^N and Cat^C inteins, transformed the plasmid library into DH5α E. coli cells, and selected for increased survival in the presence of an unfavorable Ala₋₁ residue. A colony containing the mutations S70T and S79P (Cat*) was identified and shown to infer increased survival against kanamycin (Fig. 1c).

Figure 1.

Extein-dependent protein trans-splicing of the Cat TerL split intein. (a) Left: Schematic depicting Cat ultrafast trans-splicing under a favorable extein context. Cat is embedded within a target protein of interest (POI) and flanked by its native extein residues (glutamate at the N-extein and cysteine at the C-extein). The −1 position of the N-extein is highlighted in red. Right: Schematic depicting Cat splicing when embedded within a POI in the context of an unfavorable extein context. Glycine is present at the −1 N-extein position and is highlighted in red. (b) Depiction of the PTS-dependent E. coli selection system. Top: The kanamycin resistance protein, Kan^R, is split and fused to N- and C-intein fragments (Cat^N and Cat^C). The region subject to error prone PCR is highlighted in blue. Bottom: Depiction of the selection system, wherein transformed E. coli cells are subject to increasing concentrations of kanamycin (Kan) on an LB-Agar plate. (c) LB-Agar selection of E. coli transformed with split Kan^R plasmids containing Cat with −1 glutamate (top), Cat Glu₋₁, Cat Ala₋₁ (middle), and Cat* Ala₋₁ (bottom).

We next validated the enhanced activity of the Cat* intein using an in vitro splicing assay (Fig. 2a). We individually expressed and purified intein fragments fused to the model N- and C-extein domains, Maltose Binding Protein (MBP) and Green Fluorescent protein (GFP) respectively. For both Cat and Cat*, three residues from the endogenous N- and C- exteins were preserved as linkers, with mutations present at the −1 N-extein position. The cognate intein fragments were mixed to a final concentration of 2 μM, incubated at 30 ˚C, and reaction progress was monitored by gel electrophoresis. Enhanced activity for Cat* is observed in the presence of Ala₋₁ (10-fold increase in splicing rate, t_½ = 4 m), Gly₋₁ (16-fold increase, t_½ = 9.2 m), Ser₋₁ (2.5- fold increase, t_½ = 3 m), and Pro₋₁ (1.5-fold increase, t_½ = 98 m) N-exteins compared to Cat (Fig. 2b, 2c, supplementary Table S1). Although activity decreases by 2-fold for Glu₋₁, Cat* splicing still proceeds at an exceptionally fast rate (t_½ = 2 m, Fig. 2b, 2c).

Figure 2.

Characterization of the enhanced splicing activity of Cat*. (a) Schematic of the in vitro trans-splicing assay. An N-extein maltose binding protein (MBP) is fused to Cat^N, and a C-extein green fluorescent protein (GFP) is fused to Cat^C. The N-extein −1 residue is shown as a red X. (b) Reaction progress curves for splicing of Cat and Cat* at 30 °C with the indicated −1 N-extein residue. (c) In vitro splicing half-lives of Cat (red) and Cat* (blue) with indicated −1 N-extein residues (mean ± SD, n = 3). (d) In vitro splicing half-lives of Cat Gly₋₁ and the indicated mutations derived from Cat* (mean ± SD, n = 3). (e) Depiction of the Cat active site (PDB 6DSL) highlighting the proximity of the S79P mutation (cyan) to the active site B-block loop. The catalytic residues Thr₇₅ and His₇₈ are rendered as sticks. The values for Cat are reproduced from previously reported data¹⁶ in all panels.

To delineate the contribution of each mutation in Cat*, we measured splicing of Cat S79P and Cat S70T in the presence of Gly₋₁ (Fig. 2d). Introduction of S79P into Cat largely accounts for the enhanced activity of Cat* (t_½ = 11.8 m), while no increase in splicing rate is observed for Cat S70T (t_½ = 152 m). Mapping these mutations onto the previously reported structure of Cat (PDB 6DSL), Ser₇₀ is located on the periphery of the intein, while Ser₇₉ is present within the intein functional site (Fig. 2e). Importantly, Ser₇₉ sits directly adjacent to the B-block intein loop, which contains both catalytic Ser₇₅ and His₇₈ residues responsible for activation of the N-splice junction.

N-extein substitutions in Cat impact B-block catalytic residues

To understand how the S79P mutation restores activity in Cat*, we began by investigating the structural consequences of −1 N-extein substitutions. Cat Glu₋₁, Cat Gly₋₁, and Cat S79P Gly₋₁ constructs containing inactivating C1A and N134A mutations were expressed in ¹⁵N, ¹³C isotopically enriched media, complexed, purified to homogeneity by size-exclusion chromatography, and analyzed by NMR spectroscopy. Backbone resonances for the three constructs were assigned, and chemical shift perturbations (CSP) between each substitution were calculated (Table S2, Fig. S2). The ¹H-¹⁵N HSQCs of all complexes display very similar fingerprints, which demonstrates that large structural changes such as misfolding do not account for the observed differences in activity.

The CSP values for the Cat E-1G substitution were next mapped onto the previously reported Cat structure (PDB ID 6DSL) (Fig. 3a). Residues near the site of the N-extein substitution, particularly the catalytic residues C1A and His_78, show significant values of CSP. In addition, a well resolved H^δ2/C^δ2 crosspeak assigned for His₇₈ in Cat Glu₋₁ could not be determined for Cat Gly₋₁ in the ¹H-¹³C aromatic HSQCs (Fig 3b). Correspondingly, one less crosspeak is also observed for histidine H^ε1/C^ε1 resonances in Cat Gly₋₁ than Cat Glu₋₁, indicating line broadening of His₇₈ imidazole resonances. Increased local dynamics, such as loss of a steric restriction resulting from the E-1G mutation, can account for these line broadening effects. The B-block histidine is highly conserved and contributes to both the linear intermediate formation and the branched intermediate resolution splicing steps²⁵. Thus, mutations that alter His₇₈ dynamics may in turn augment splicing kinetics.

Figure 3.

Structural and dynamic effects of Cat N-extein substitutions. (a) Differences in backbone chemical shifts between Cat Glu₋₁ and Cat Gly₋₁. The weighted average chemical shift perturbation (CSP) was calculated for each residue and is mapped onto the solution structure of Cat (PDB ID: 6DSL). The substituted −1 residue is depicted in green, and active site residues that exhibit significant chemical shifts are rendered as sticks. (b) ¹H −¹³C-HSQC spectra of Cat in the context of Glu ₋₁(left), Gly ₋₁ (middle), or S79P Gly ₋₁ (right). (c) Differences in backbone chemical shifts between Cat Gly₋₁ and Cat S79P Gly₋₁. The weighted average chemical shift perturbation (CSP) was calculated for each residue and is mapped onto the solution structure of Cat (PDB ID: 6DSL). The S79P mutation is depicted in green, and reactive site residues that exhibit significant chemical shifts are rendered as sticks

The CSP values between Cat Gly₋₁ and Cat S79P Gly₋₁ were then mapped on the Cat structure with significant chemical shift perturbations predominantly localized in the active site (Fig. 3c). S79P strongly impacts the backbone chemical shifts of Asp₁₁₅, Tyr₁₂₅, and the adjacent block B loop (Asp₇₆, Asn₇₇, Phe₈₀, Gly₈₁). While His₇₈ H^δ2/C^δ2 cross peaks are not observed in the Cat Gly₋₁ ¹H-¹³C aromatic HSQC spectrum due to line broadening, the S79P mutation reverses this effect (Fig. 3b). Moreover, a peak in the Cat S79P Gly₋₁ histidine H^ε1/C^ε region is also restored. This observation suggests that the dynamic state of the His₇₈ side chain in Cat S79P Gly₋₁ more resembles that of Cat Glu₋₁ than Cat Gly₋₁.

Sidechain conformational dynamics of His₇₈ modulate reaction rate in Cat

Molecular dynamics simulations were carried out to explore the molecular basis for −1 N-extein dependence. In the NMR structure ensemble of Cat (PDB 6DSL), which contains Glu₋₁, the side chain of His₇₈ occupies two main conformations. The imidazole group N^δ1 is oriented proximal to the −1 N-extein scissile peptide bond in the first conformation but sits distal in the second (Fig S3). Prior studies suggest that the proximal orientation of N^δ1 is optimal for catalysis through interaction with the Ext^N −1 backbone carbonyl oxygen. Since the −1 Ext^N side chain interacts with His₇₈, we hypothesized that N-extein substitution could modify His₇₈ conformational dynamics. MD simulation trajectories were thus calculated from structures originating from both configurations for Cat Glu₋₁, Cat Gly_−1, and Cat S79P Gly₋₁. When the His₇₈ sidechain starts in the proximal configuration, limited χ² torsion angle change is observed irrespective of the −1 residue or the S79P mutation (Fig. S4). However, when the simulation originates from the distal configuration for Cat Glu₋₁, the imidazole ring undergoes a flip and reorients itself to achieve the proximal conformation thereby reducing the distance between His₇₈ N^δ1 and Glu₋₁ carbonyl oxygen. (Fig 4a, 4b, S5). The ring flip is preceded by a restructuring of the loop spanning residues 78 to 90 wherein the loop extends to open the active site and enable the His₇₈ sidechain to reorient itself (Fig. 4a, Movie S1). Distances calculated from the trajectories between Cα atoms of Gly₈₁ and Ser₁₃₃ increase by about 4 Å prior to the ring flip, indicating an extension of the loop away from the active site. (Fig 4c, S5). The retraction of the loop residues also corresponds with an increase in the void volume around the His₇₈ side chain, possibly to accommodate its reorientation (Fig. 4d, Table S3). In the Cat Gly₋₁ simulation, however, the His₇₈ side chain remains in the distal configuration with no observed β-hairpin reconfiguration. This unfavorable distal configuration may thus contribute to the decreased splicing activity observed for the Gly₋₁ extein (Movie S2). For Cat S79P Gly₋₁, the His₇₈ sidechain undergoes a reorientation like that in the Cat Glu₋₁ simulation. Flipping of the imidazole ring is once again preceded by restructuring of the residue 78–90 loop region and an increase in void volume around the His₇₈ sidechain and results in a reduced distance between His₇₈ N^δ1 and the Gly₋₁ carbonyl oxygen. (Fig 4c, 4d, S5, Movie S3). Interestingly in proximal configuration simulations, this loop restructuring is observed for Cat S79P in proximal configuration simulations as well, but not in any of other cases (Fig S4). Thus, the in silico observation that S79P augments residue 78–90 loop dynamics and facilitates His₇₈ reorientation to the proximal configuration provides a possible mechanism for the enhanced Cat* extein tolerance.

Figure 4.

Structural and dynamic effects of Cat S79P mutation. (a) Structural overlay of residues 74–89 from the MD simulation for Cat Glu₋₁ (black), Cat Gly₋₁ (gray), and Cat S79P Gly₋₁ (orange) at t = 0, 150, and 600 ns. His₇₈ is represented as a stick and an arrow is depicted adjacent to the Gly₈₁-Asp₈₃ loop. (b) His₇₈ χ² angles calculated from individual MD simulation trajectories for Cat Glu₋₁ (left), Cat Gly₋₁ (center), and Cat S79P Gly₋₁ (right). (c) Distances between Gly₈₁ (Cα) and Ser₁₃₃ (Cα) computed from MD simulation trajectories for Cat Glu₋₁ (left), Cat Gly₋₁ (center), and Cat S79P Gly₋₁ (right) (d) CASTp void volume calculations surrounding His₇₈ for Cat Glu₋₁ (left), Cat Gly₋₁ (center), and Cat S79P Gly₋₁ (right). An arrow is shown depicting the point of His₇₈ rotation.

Discussion and Conclusions

Extein dependence can hinder applications of split intein-mediated protein trans-splicing and necessitate insertion of nonnative residues into an exogenous protein of interest. This limitation motivates efforts to both understand and resolve extein dependence within split inteins. Here, we applied a cell-based selection assay to identify a variant of the atypically split Cat TerL intein (Cat*) with enhanced extein tolerance at the −1 N-extein position. An S79P mutation accounts for this enhanced activity and is adjacent to His₇₈. We then applied NMR spectroscopy to demonstrate that S79P acts locally to augment active site residues, including His₇₈ side chain dynamics. Lastly, MD simulations suggest these His₇₈ dynamics may directly alter splicing activity by impacting its distance to the −1-residue backbone carbonyl oxygen.

The enhanced promiscuity of Cat* expands the intein toolkit available to protein engineers by enabling splicing under previously unfavorable extein environments and should facilitate the N-terminal modification of proteins due to its synthetically accessible N-intein. Specifically, applications of naturally split inteins that could benefit from this enhanced promiscuity include efforts overcome cargo limitations of adeno-associated virus gene therapies, generate cyclic peptide libraries for drug screening, and synthetically modify proteins in cells^23,26–28. Each of these applications of protein trans-splicing could benefit from a greater ability to conserve native residues adjacent to the intein insertion site.

Examining this work in the context of prior efforts to engineer splicing promiscuity potentially reveals common principles to design split inteins with minimal extein dependence. The observation that Glu₋₁ interacts with catalytic residues within the active site and steers His₇₈ is reminiscent of prior investigations into the C-extein dependence of DnaE inteins ²¹. In Npu DnaE intein, a phenylalanine at the +2 C-extein position interacts with and guides the position of a catalytic F-block histidine (His₁₂₅) to facilitate succinimide formation. For both Cat and Npu, loss of a steric interaction from introduction of an extein residue such as glycine or alanine leads to unfavorable orientation of active site catalytic residues. Interestingly, much like the restoration of activity from the S79P mutation directly adjacent to His₇₈ in Cat reported here, introduction of a “GEP” substitution in the loop adjacent to His₁₂₅ (residues 122–124) in DnaE inteins enhances activity in unfavorable extein contexts²². In both cases, NMR and MD simulations suggest the substitutions compensate for losing the steric extein interaction by augmenting dynamics and conformation of the nearby catalytic histidine. However, some efforts to overcome extein dependence have been more challenging, such as for the atypically split VidaL intein, which requires a −1 Gly²³. In VidaL, Gly₋₁ is thought to facilitate distortion of the scissile amide bond. It is thus possible that inteins in which extein dependence results from loss of a side chain interaction, such as Npu and Cat, may be more suitable for engineering than those whose extein dependence is caused by introduction of unfavorable steric bulk, as is the case for VidaL.

Intriguingly, the identified mutations in Cat* (S70T, S79P) correspond to the native residues in the previously reported AceL TerL intein¹⁶. In addition, both Thr₇₀ and Pro₇₉ are the second most prevalent residues at their respective positions within the TerL alignment from which the Cat sequence was derived (Fig. S6). Cat, which contains Ser₇₀ and Ser₇₉, may therefore be less promiscuous than AceL (and other uncharacterized TerL inteins) despite its increased activity and stability. Consensus protein engineering assumes that the most common residues within a multiple sequence alignment (MSA) are selected to improve fitness, but because inteins evolve within the context of a specific insertion site (e.g. TerL), a consensus intein may optimize activity for the endogenous extein and therefore not favor extein substitutions. This finding agrees with the general observation that enzyme specialization often accompanies enhanced catalytic efficiency²⁹. The counterbalance between splicing efficiency and promiscuity is further noted in the decreased WT extein splicing rates for both the Cat* (~2 fold with WT Glu₋₁ N-extein) and the “GEP” DnaE mutations (~2 fold with Phe₊₂ C-extein). Future studies that identify and engineer highly efficient and robust inteins through consensus design should therefore benefit from additional characterization and expansion of extein promiscuity.

There have been few inquiries into why certain split inteins naturally exhibit partial N- or C-extein promiscuity, although significant effort has been invested to understand the cause of extein dependence and to increase splicing promiscuity. For example, while the splicing rate of Cat depends on the identity of the −1 N-extein position, it is largely unperturbed by C- extein substitutions¹⁷. Conversely, splicing in Npu depends on the identity of the +2 position, but tolerates substitutions at the −1 position³⁰. For Cat, which lacks an F-block histidine, branch resolution may be accelerated by a penultimate G-block histidine that does not interact with the C-extein. For Npu, however, it is unclear why N-extein mutations do not result in a significant loss of activity. A thorough understanding of natural intein promiscuity could enable grafting of endogenous motifs to facilitate promiscuity or identifying novel promiscuous inteins from sequence databases.

Materials and Methods

NMR Spectroscopy

NMR experiments were performed on a Cat^N /Cat^C complex (500 μM) enriched with ¹³C and ¹⁵N isotopes. The intein fragments were rendered inactive through Cat^N(C1A) and Cat^C(N134A) mutations. Polypeptides were individually expressed as SUMO fusion constructs in isotopically enriched M9 minimal media, purified and the fusion tag was removed by Ulp1 protease. Intein fragments were combined in 1.5:1 ratio of Cat^N: Cat^C and the complex formed was purified by size exclusion chromatography using a GE HiPrep 16/60 Sephacryl S-200 column. NMR experiments were performed on a Bruker DRX600 NMR spectrometer equipped with a TCI cryoprobe.

Molecular Dynamics Simulation

Starting structures for Molecular Dynamics (MD) simulation were obtained from the previously determined NMR structure of the Cat intein complex (PDB ID: 6DSL). The first and the fifth structure from the PDB file were used as starting structures which have the major and minor His₇₈ configurations. Cys₁ and Asn₁₃₄ residues were modeled to replace the inactivating alanine mutations in the PDB structure file using Chimera³¹. Additionally, the different mutants investigated in this work were modeled using Chimera to obtain respective starting structures. Standard all-atom molecular dynamics simulations in explicit water were performed using the GPU enhanced Amber 18 software code. The ff14SB force field was used for the protein molecules. TIP-3P explicit water model was used with a box having a 10 Å buffer limit to the solute molecule. Molecules were neutralized using additional Na⁺ ions as required. The system was prepared by performing energy minimization, heating, and density equilibration. Positional restraints were applied on the backbone of the protein using the SHAKE algorithm. The system was further equilibrated for 25 ps before the production was set up for a total simulated time of 500–700 ns. Void volume calculations were performed using the CastP web server ³².

Splicing Assay

Splicing reactions of Cat^N carrying maltose binding protein as the N-extein (MBP-Cat^N) and Cat^C with GFP as the C-extein (Cat^C-GFP) were performed as follows. Equal volumes of MBP-Cat^N and Cat^C-GFP (4 μM stocks) were incubated at 30 °C and mixed to start the PTS reaction. 10 μL aliquots were removed at the required time points and quenched by mixing with 10 μL of 2x Laemmli sample buffer (BioRad). Samples were analyzed by SDS-PAGE. Gel photographs were captured using the BioRad GelDoc imaging system and ImageJ software was used to quantify the bands in each lane.

Error Prone PCR Mutagenesis and Kanamycin Resistance Screening

Fast acting mutants were selected from a kanamycin resistance screen coupled to the intein splicing activity. Kan^R was encoded as two split intein precursor genes Kan^RN-Cat^N and Kan^RC-Cat^C using a pBlusecript plasmid under the T3 promoter (referred to hence as pKan^R). The pKan^R plasmid was transformed into 40 μL E. coli DH^5α ultracompetent cells (NEB) and outgrown in 800 μL of SOC medium at 37 °C. 160 μL of the outgrown cells were plated on LB agar plates containing 100 mg/L of ampicillin and the required concentration of kanamycin (0, 5, 7.5, 10, 20 and 30 μg/L). The plates were incubated overnight at 37 °C.

Error prone PCR was performed on a pKan^R plasmid carrying an FA extein using the Diversify PCR random mutagenesis kit (Takara, Mountain View, CA). The primer design was done such that the Cat^N and Cat^C sequences alone were subject to the mutagenesis and amplification by PCR while leaving the Kan^R gene and the ribosome binding sites of the plasmid unperturbed. The fragments amplified by error prone PCR were cloned into plasmids to obtain a pKan^R mutant library. The library was transformed into DH^5α cells and plated in increasing kanamycin concentrations as described above. Colonies viable in higher kanamycin concentration were selected for subsequent gene sequencing and characterization.