Optimization of Expression and Thermostability of Terminal Deoxynucleotidyl Transferase through Iterative Mutagenesis and Computational Design

doi:10.21203/rs.3.rs-7436431/v1

Download PDF

Research Article

Optimization of Expression and Thermostability of Terminal Deoxynucleotidyl Transferase through Iterative Mutagenesis and Computational Design

https://doi.org/10.21203/rs.3.rs-7436431/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Terminal deoxynucleotidyl transferase (TdT) is a template-independent polymerase that catalyzes the addition of deoxynucleoside triphosphates to the 3'-terminus of a DNA strand. While TdT is a key enzyme for developing enzymatic DNA synthesis technologies, its inherent low thermal stability presents a significant limitation.

This study aims to improve the thermostability of TdT from Mus musculusthrough a combination of site-saturation mutagenesis and rational design. Residues for saturation mutagenesis were identified using B-factor analysis and the B-FITTER program, while promising substitutions for rational design were selected using Foldit, ProteinMPNN and CARBonAra.

Through several iterations of mutagenesis, we obtained two highly promising variants. The first one, dubbed A4, obtained solely through saturation mutagenesis, showed a 26-fold higher expression level than the WT protein. The second one, dubbed mutant 275, demonstrated exceptional stability, showing no significant loss of activity after 180 minutes of incubation at 45°C— conditions under which the wild-type enzyme's half-life was less than 2 minutes. This corresponds to a >120-fold increase in stability. Additionally, its melting temperature (Tm) was increased by 6.5°C. Moreover, mutant 275 demonstrated a 4- to 6-fold increasein catalytic activity at 37 °C.

This significant enhancement in thermostability was achieved after four rounds of iterative mutagenesis and is attributed to the formation of a stabilizing salt bridge network on the protein surface, distant from the active site. The obtained thermostable TdT variants serve as robust scaffolds for further engineering to improve activity towards the 3'-reversibly blocked nucleotides required for next-generation enzymatic DNA synthesis.

Terminal deoxynucleotidyl transferase

Protein engineering

Thermostability

Enzymatic DNA synthesis

Site-saturation mutagenesis

Computational protein design

Terminal deoxynucleotidyl transferase (TdT) was first described in the 1960s, initially named "Thymus Polymerase," when its unique ability to catalyze template-independent nucleotide addition was demonstrated [1]. The physiological role of TdT was later understood in the context of V(D)J recombination, where it generates junctional diversity by adding non-templated (N) nucleotides to gene segments [2].

Subsequently, advancements were made in its production, including expression in the baculovirus-insect cell system [3] and the development of methods for high-yield recombinant production in Escherichia coli [4, 5]. During this period, different protein isoforms of TdT were also identified [6].

Further structural and mechanistic insights were gained once the crystal structure of the murine TdT catalytic core was solved. This, combined with site-directed mutagenesis, helped to elucidate the enzyme's mechanism of action [7, 8].

From the moment of its discovery, TdT found wide application in molecular biology. The enzyme was used for elongating the 3'-ends of DNA, introducing the ³²P radioactive label, and creating probes by attaching various labeled nucleotides as substrates [9–11].

By the 2000s, the growing needs of synthetic biology surpassed the capabilities of the standard phosphoramidite method of DNA synthesis. Since further increasing the length of oligonucleotides was technologically challenging, researchers focused on increasing their quantity. This was achieved by increasing the synthesis density through the use of DNA microchips and microfluidic platforms [12]. In parallel, methods for error correction during gene assembly from short oligonucleotides were developed [13].

The need to synthesize long genetic constructs and entire genomes created a demand for a new technology that would meet several key criteria:

a) the ability to synthesize oligonucleotides significantly longer than those obtained by chemical methods;

b) high synthesis fidelity, as correcting accumulated errors becomes critically complex with increasing construct length;

c) the use of accessible reagents while adhering to principles of environmental safety.

The technology of enzymatic oligonucleotide synthesis using TdT potentially meets all these requirements. Within this approach, a scheme for de novo enzymatic DNA synthesis was proposed using nucleotides containing a reversible acetyl protecting group at the 3'-hydroxyl [14].

To ensure controlled, stepwise addition of nucleotides during enzymatic synthesis, it is necessary to use their modified analogs containing easily removable terminating groups at the 3'-hydroxyl. A number of such groups were adapted from next-generation sequencing (NGS) technologies.

These include nucleotides with aminoalkoxyl [15], photocleavable [16], thermolabile [17], azidomethyl [18], as well as phosphate and other protecting groups [19, 20].

The aminoalkoxyl group gained the most widespread use due to its relatively simple synthesis, small steric hindrance, and mild conditions for deprotection. Its key advantage is that TdT efficiently incorporates nucleotides with this modification into the growing DNA strand, after which reversible termination of synthesis occurs.

Based on this approach, [21] developed a platform for enzymatic DNA synthesis using TdT, nucleotides with aminoalkoxyl protection, and inkjet printing technology. The researchers achieved reagent stability ("inks") and demonstrated the controlled synthesis of 8-mer oligonucleotides. The time for a single coupling cycle was 10 minutes at 20°C.

To synthesize longer oligonucleotides, it is necessary to significantly reduce the cycle time and increase its efficiency. The main obstacles are the thermolability and relatively low native activity of TdT, which limits the possibility of increasing the reaction temperature.

The problem of low TdT thermostability has been the subject of research by several scientific groups. For instance, [22] applied error-prone PCR combined with FRET-based screening to find thermostable variants of the enzyme. Analysis of 10,000 mutants allowed them to identify variants with thermostability increased by 10°C.

Another approach was used by [23], who applied the FireProt algorithm to predict substitutions that increase protein thermostability. Combining 15 point mutations led to the creation of a TdT variant whose stability was 8.8°C higher. An important result of this work was the demonstration that the formation of a 3'-terminal hairpin as short as four nucleotides almost completely inhibits the enzyme's activity. Increasing the reaction temperature by 10°C and optimizing the concentration of divalent cations allowed them to increase the reaction rate 10-fold even for primers prone to forming eight-nucleotide hairpins.

An alternative method that solves the problem of reversible termination was proposed by [24]. It is based on the use of covalent TdT-dNTP conjugates. In this case, termination of synthesis is achieved because the enzyme becomes chemically "stitched" to the substrate after its incorporation into the growing strand. Further elongation is impossible until the enzyme is cleaved off. This approach provides high efficiency and reduces the time of a single coupling cycle to 20 seconds. Although this leaves modified phosphodiester bonds ("scars") in the DNA strand, it has been shown that they do not hinder subsequent enzymatic manipulations, such as PCR.

Thus, this method effectively solves the problem of controlled termination and significantly reduces synthesis time, but it is associated with additional costs for producing modified nucleotides and does not eliminate the problem of primer secondary structures.

Thus, increasing the thermostability of TdT is a key task for the further development of enzymatic DNA synthesis technology. The formation of secondary structures (hairpins) is already a significant obstacle, reducing the efficiency of synthesizing oligonucleotides around 100 nucleotides long. When moving to the synthesis of longer sequences (> 1000 nt) or regions with high GC content, this problem becomes critically important.

Recently, impressive advances in TdT engineering have been reported. These include resource-intensive directed evolution campaigns that yielded enzymes with a 20°C increase in thermostability and dramatically enhanced activity towards modified nucleotides [25]. Concurrently, large-scale computational design has produced variants with a melting temperature increase of up to 24.3°C and a 77-fold longer half-life [26]. While these studies demonstrate powerful routes to highly stable and active enzymes, the challenge of creating a technologically viable TdT that combines these improved properties with high recombinant expression yield using accessible engineering strategies remains.

To address this challenge, we adopted a "stability-first" approach, based on the rationale that enhancing a protein's foundational thermostability should precede active site modifications, which are often destabilizing [22]. In this study, we pursued a synergistic strategy combining iterative saturation mutagenesis (ISM) of structurally flexible regions, identified by B-factor analysis, with rational design guided by modern neural network algorithms. Our goal was to achieve a comprehensive improvement of TdT's properties, with a dual focus on enhancing thermostability and, critically, maximizing the protein expression level—a key parameter for cost-effective production. Here, we report the engineering of two distinct variants with superior properties: mutant A4, with a remarkable 26-fold increase in expression yield, and mutant 275, with a significant 6.5°C increase in thermostability. These variants serve as robust and highly manufacturable scaffolds for applications in enzymatic de novo DNA synthesis.

Plasmids construction

The gene encoding wild-type (WT) Terminal deoxynucleotidyl transferase (TdT) from Mus musculus was codon-optimized for expression in Escherichia coli and designed with an N-terminal SUMO tag. The complete DNA sequence was synthesized using a PCR-based two-step DNA synthesis (PTDS) strategy [27].

The resulting gene was cloned into the pET-28a(+) vector using a ligation-independent method based on the Hot Fusion protocol [28] with in-house prepared Phusion DNA polymerase and T5 exonuclease. A map of the final plasmid construct is provided in Fig. S1. The sequence of the entire construct was confirmed by Sanger sequencing.

Protein Expression

Escherichia coli BL21(DE3) cells were transformed using the pET-28a(+) plasmid containing the gene for SUMO-tagged TdT. Starter cultures were grown overnight at 37°C in Luria-Bertani (LB) medium supplemented with 50 µg/mL kanamycin, with shaking at 200 rpm. The following day, fresh LB medium was inoculated with the overnight culture at a 1:100 (v/v) ratio.

The expression cultures were grown at 37°C with shaking until they reached an optical density at 600 nm (OD₆₀₀) of 0.6. The cultures were then chilled on ice for 20 minutes before protein expression was induced by the addition of isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.2 mM. The cells were incubated overnight at 16°C with shaking. Finally, cells were harvested by centrifugation at 5,000 x g for 10 minutes, and the resulting cell pellets were stored at -80°C until purification.

Protein Purification

Cell pellets were thawed and resuspended in lysis buffer (20 mM Tris-HCl pH 8.0, 1 M NaCl, 5% glycerol, 2 mM PMSF). The cells were then lysed by ultrasonication using a Scientz-IIID sonicator. The lysate was clarified by centrifugation at 20,000 x g for 30 minutes at 4°C.

The resulting supernatant was loaded onto a HiSep Ni SepFast IMAC Column (BioToolomics, UK) for affinity chromatography. After washing the column to remove unbound proteins, the target TdT protein was eluted using a buffer containing 20 mM Tris-HCl (pH 8.0), 0.2 M NaCl, 5% glycerol, and 700 mM imidazole.

The purified protein fractions were buffer-exchanged into a 2X storage buffer (100 mM KPO₄ pH 7.3, 200 mM NaCl, 2.86 mM β-mercaptoethanol, 0.2% Triton® X-100) and concentrated to approximately 2 mg/mL using Vivaspin 15 centrifugal concentrators (3 kDa MWCO, Sartorius, Germany). An equal volume of sterile glycerol was then added to the concentrated protein (for a final concentration of 50% glycerol in 1X storage buffer), and the final aliquots were stored at -80°C.

Site-Directed and Saturation Mutagenesis

To introduce specific point mutations and to generate saturation mutagenesis libraries, a one-step PCR-based mutagenesis protocol was employed [29].

For saturation libraries, primers were designed with NNK codon degeneracy to target the desired residue. A complete list of primers used for saturation mutagenesis is available in Table S1. Primers for rational design followed the same structural design but incorporated the specific codon for the desired amino acid substitution instead of the degenerate NNK codon.

Following PCR amplification, the reaction products were verified by agarose gel electrophoresis. The amplified products were then purified using a DNA purification kit (Biolabmix, Russia). To remove the parental methylated template plasmid, 300 ng of the purified product was digested with 10 units of DpnI restriction enzyme (ABclonal) in a 20 µL reaction volume for 3 hours at 37°C. Subsequently, 10 µL of the DpnI-treated reaction was transformed into chemically competent Escherichia coli TOP10 cells.

For site-directed mutagenesis, five individual colonies were selected from each transformation. Plasmids were isolated from overnight cultures using a plasmid purification kit (Biolabmix, Russia), and the entire gene sequence was verified by Sanger sequencing to confirm the desired mutation and the absence of any off-target mutations.

For saturation mutagenesis, all resulting colonies from a single transformation (a minimum of 96) were pooled together to ensure library diversity. The mixed plasmid library was then isolated from the pooled cell culture using a plasmid purification kit. The quality of the library and the successful incorporation of degenerate codons at the target position were confirmed by Sanger sequencing of the plasmid pool.

Prediction of single mutations ProteinMPNN and CARBonAra

We used PDB ID 4I2C (Ternary complex of mouse TdT with ssDNA and AMPcPP) as the model [30]. Missing residues 420–422 were added using the Swiss-Model Expasy Server [31, 32].

We used ProteinMPNN [33] and CARBonAra v1.0.0 [34] to predict possible mutations. In each prediction, all positions except the analyzed one were fixed. For ProteinMPNN, we used the soluble v_48_020 model with the temperature set at 0.3. We generated 50 sequences, obtained probability distributions using the --save_probs parameter and averaged the respective values. For CARBonAra, we set the imprint ratio at 0.5, generated 100 sequences per residue and calculated the respective probabilities for each residue using the alignment_to_matrix function of logomaker v. 0.8 [35].

Analysis of Site-Directed Mutants

E. coli BL21(DE3) cells were transformed with sequence-verified plasmids for individual mutants. Protein expression was carried out in 600 µL LB cultures following the protocol described previously. After incubation, cells were harvested and resuspended in 100 µL of reaction buffer (50 mM Potassium Acetate, 20 mM Tris-acetate, 10 mM Magnesium Acetate, pH 7.9) supplemented with 2 mM PMSF and 0.1 mg/mL lysozyme. Cell lysis was performed in an ultrasonic bath (Sapfir 9.5L, Russia) for 15 minutes at 4°C and maximum power. The lysate was then clarified by centrifugation at 20,000 x g for 30 minutes.

To assess thermostability, the clarified supernatant was first subjected to a thermal challenge by incubating it for 10 minutes at a specific temperature (ranging from 37°C to 60°C). Following the heat treatment, residual enzymatic activity was measured by adding dNTPs (to 0.5 mM) and 5 ng of DNA primer (5'-FAM-AGT CAT TGC GTT TGA G-3') to 9 µL of the treated supernatant. The reaction mixture was incubated at 37°C for 30 minutes, and the reaction was subsequently terminated by heating at 72°C for 5 minutes.

The reaction products were analyzed by electrophoresis on an 8% denaturing polyacrylamide gel (PAGE) containing 8 M urea. The residual activity of each mutant was determined by the length of the primer extension products visible on the gel (Fig. S2, Fig. S3).

Screening of Saturation Mutagenesis Libraries

E. coli BL21(DE3) cells were transformed with the plasmid library, and individual clones were inoculated into the wells of 96-deep-well plates, each containing 600 µL of LB medium. For libraries targeting a single residue, 96 clones were screened to achieve > 95% library coverage. For libraries targeting two residues simultaneously, 192 clones were screened. The subsequent small-scale expression, cell lysis, and thermostability screening for each clone were performed as described above for the site-directed mutants.

Thermostability and Activity Assays with Purified Protein

Thermostability of TdT variants was initially assessed by measuring residual activity after heat treatment. The assay utilized a molecular beacon with the sequence: 5'-FAM-TCCCGCTGCT-(dT)₁₈-GCAGCGGGT-BHQ1-3'. To perform the analysis, aliquots of the purified enzyme in reaction buffer (50 mM Potassium Acetate, 20 mM Tris-acetate, 10 mM Magnesium Acetate, pH 7.9) were subjected to a thermal challenge. Samples were incubated for 10 minutes at various temperatures ranging from 37°C to 53.5°C. Afterwards, to measure residual activity, a reaction mixture containing the molecular beacon (20 nM), dATP (0.25 mM), and T7p primer (0.5 µM) was added to the samples. The TdT-catalyzed elongation of the 3'-end of the T7p primer produces a poly(A) tail. This newly synthesized strand then hybridizes with the complementary loop region of the molecular beacon, "opening" the beacon and leading to an increase in fluorescence intensity, which was monitored in real-time. The reaction was conducted at 37°C for 30 minutes, recording the fluorescence signal (λₑₓ/λₑₘ = 485/525 nm) using a Tecan Infinite 200 PRO plate reader (Austria).

The activity of TdT variants was measured using the same molecular beacon-based protocol. The key difference was the absence of a prior heat treatment step. The reaction was performed at a constant temperature of 37°C directly in the plate reader, initiated by the addition of dATP, and the increase in fluorescence was monitored in real-time.

The half-life of the enzyme at elevated temperature was determined using the same fluorescent method. The key difference was that aliquots of the purified enzymes were incubated at a constant temperature of 45°C for various time intervals (from 1 to 180 minutes). After each time interval, the residual enzymatic activity was measured as described previously. The half-life was calculated as the time required to reduce the initial enzyme activity by 50%.

Protein Thermal Unfolding

Protein thermal stability was analyzed using the Tycho NT.6 (NanoTemper, USA). Measurements were performed in the instrument’s disposable glass capillaries using 10 µL of sample. Protein samples were heated from 35°C to 95°C at a rate of 30°C/min. Intrinsic fluorescence of protein tryptophan/tyrosine residues was recorded at wavelengths of 330 nm and 350 nm. The ratio of fluorescence intensity at 350 nm to intensity at 330 nm was calculated to monitor protein unfolding. Δ Ratio was determined as the difference between the initial ratio (at 35°C) and the final ratio (at 95°C). Inflection Temperature (Ti) was automatically determined by the instrument as the inflection point on the F350/F330 ratio versus temperature curve. Each sample was measured at least 5 times. Protein samples were tested in dilution series using reaction buffer (50 mM Potassium Acetate; 20 mM Tris-acetate; 10 mM Magnesium Acetate, pH 7.9) at concentrations of 0.1 mg/mL, 0.05 mg/mL, and 0.01 mg/mL. Unfolding curves and parameters (Ti, Δ Ratio) were generated automatically by the instrument. Protein similarity was analyzed by comparing the obtained parameters.

Differential Scanning Calorimetry (DSC) was performed on a NanoDSC (TA Instruments, USA) to verify the thermostability of WT TdT and its mutants at a 1°C/min heating rate and excess pressure of 3 bar in the temperature range (20 to 65)^oC. The WT and mutants samples were stored in 50 мМ K-Ac, 20 mM Tris-Ac, 10 mM Mg-Ac, 1 mM DTT, pH 7.9 and dialyzed against reaction buffer (50 мМ K-Ac, 20 mM Tris-Ac, 10 mM Mg-Ac, pH 7.9) just before measurements. Protein concentrations in TdT samples were (0.5–1.4) mg/mL. DSC measurements were performed as described in [36].

Analysis of Kinetic Parameters

The kinetic parameters of the enzymatic reaction were determined by high-performance liquid chromatography (HPLC) with a fluorescence detector (FLD). The analysis monitored the addition of ddCTP to a 5'-end fluorescently labeled primer (5'-FAM-AGT CAT TGC GTT TGA G-3'). The reaction rate for each TdT variant was calculated based on the ratio of the peak areas corresponding to the initial primer and the reaction product on the resulting chromatograms.

Selection of residues for Saturation Mutagenesis

Since a naturally thermostable TdT is unlikely to exist — the enzyme is found only in vertebrates, and related Polymerase X family enzymes from thermophiles lack terminal transferase activity — protein engineering is required to enhance its stability [37, 38]. Therefore, we employed a multi-faceted strategy to identify candidate residues for saturation mutagenesis.

Our primary approach involved identifying residues with high conformational flexibility based on crystallographic B-factors. We used the B-FITTER program, which ranks amino acids by their cumulative B-factor values, with the TdT structure PDB: 4i27 serving as the primary template [39]. This initial analysis highlighted residues located within two large loops, one of which is known to be critical for the enzyme's terminal transferase activity [40].

To expand the pool of candidates beyond these regions, we extended our analysis to include other TdT structures (PDB: 1JMS, 4QZ9) and structures of the related Polymerase Mu (PDB: 6VEZ, 6IPJ). Additionally, we used the FireProt web server for in silico stability predictions [41]. The results from these computational analyses were then cross-referenced with previously reported thermostabilizing mutations [22, 23]. All data were compiled and this comprehensive analysis guided our final selection of residues for saturation mutagenesis (Fig. 1).

For the first round of iterative saturation mutagenesis (ISM), residues E177, S223, E230/D231, Q287, A353, and T354 were selected. The standard ISM protocol involves sequential saturation of selected positions, where the mutant with the best performance from one round becomes the template for the next [42].

Initially, we adhered to this approach, but its effectiveness proved limited. Saturation at positions E177, A353, and T354 did not lead to a significant increase in thermostability. The most significant result was achieved by simultaneously saturating two adjacent residues—E230 and D231. Despite screening less than 6% of the theoretical library, a number of thermostable variants were identified.

The best of these, containing the E230A and D231T substitutions, was used as a basis for subsequent saturation at position Q287, which led to a further, albeit small, increase in thermostability. However, the next step—saturation at position S223—again yielded only a minimal effect.

In light of this, we revised the strategy for selecting residues for subsequent rounds. A new set of positions was formed based on the following criteria: a more stringent selection based on the B-factor of the TdT structure (PDB: 4i27), inclusion of residues from the N- and C-termini of the protein (K149/I150, A510), and targeting the most flexible regions—Loop 1 (V395) and Loop 2 (R414, E418/K419), as well as positions R182/E183 and Q371/Q372.

As a result of mutating residues located at the N- and C-termini of the protein, as well as in the most flexible loops, no noticeable increase in thermostability was achieved. The only exception was the double mutation R182S/E183R, which led to a moderate increase in both the stability and activity of the enzyme. The summary of the mutagenesis campaign is provided in Fig. 2.

Rational Design and its Results

Simultaneously with ISM, rational design of mutants was carried out using the Foldit program [43]. The selection of amino acid substitutions was based on several criteria: formation of new intramolecular ionic bonds (salt bridges), replacement of hydrophobic residues with hydrophilic ones on the protein surface, preservation of existing stabilizing interactions, or decrease in the calculated free energy (ΔΔG) of the structure. In the first stage, 14 single and 4 double mutations were created based on these principles. Analysis of the results showed a correlation between the experimental data and the Foldit predictions: the substitutions that led to increased thermostability had the lowest (most favorable) calculated ΔΔG values. Based on these data, a second round of rational design was carried out, in which another 13 variants were created. In addition, 4 mutants were constructed that combined the most successful substitutions obtained from both the rational design method and the iterative saturation mutagenesis. The obtained results are presented in Table 1.

Table 1
Effects of rationally identified mutations on TdT activity
Substitution	Influence on activity	Substitution	Influence on activity	Substitution	Influence on activity
M313R	Increased	M269E	No Effect	D441K	No Effect
С302E	Increased	D170N	Decreased	D170K	Lost
C155V	Decreased	M470K+ L486E	Decreased	L189R	Lost
L251R	Lost	L327E+ Y440R	Decreased	L181R	Lost
D501E	Decreased	MR192EE	Decreased	R182N	Increased
M470Q	Lost	I150E+ V197Q	Lost	M288K	Lost
A510D	Decreased	L283Q	Decreased	M288R	Lost
R193E	Decreased	L167R	Increased	S207H	Decreased
M192E	Lost	G218F	Lost	M313K+ C302E	Decreased
M208K	Decreased	К370R	Lost	М313К+ L167R	Increased
M471K	Decreased	V320K	No Effect	Q287K+ C302E	No Effect
C188N	Decreased	G413R	No Effect	ED230AT + C302E	No Effect

Ultimately, only three of the proposed amino acid substitutions led to a noticeable increase in thermostability. Most of the constructed variants showed either a decrease in activity or a complete loss of catalytic activity. In some cases, this was due to a disruption of protein folding, which led to its aggregation and accumulation in inclusion bodies instead of being in the soluble fraction of the cell lysate.

Rational Design based on ProteinMPNN and CARBonAra Predictions

For further engineering, an approach based on the outputs of two neural network algorithms was applied. Using both ProteinMPNN [33] and CARBonAra [34]. The selection of target mutations for experimental evaluation was carried out according to several criteria: consensus of predictions, where positions were considered promising if both algorithms suggested making changes; coincidence of substitutions, where special attention was paid to cases, where the proposed amino acid substitutions were the same; and principles of rational design, taking into account general biophysical principles such as the formation of new stabilizing interactions. The amino acid substitutions predicted by the two neural network algorithms are visualized in Fig. 3. Based on this comprehensive analysis, 23 of what we considered the most promising variants were selected and created (Table S2).

However, screening showed that in most cases, the predicted mutations led to a significant decrease in catalytic activity or expression level, which did not allow for a reliable assessment of their effect on thermostability. The only successful variant was mutant 275, which was created by introducing the S275E/K276E double mutation into the backbone of mutant A4, the best variant obtained from saturation mutagenesis. Despite a decrease in the expression level, this new variant retained high activity comparable to that of its A4 parent. This allowed for a clear observation of a significant increase in its thermostability.

Characterization and Activity Analysis of Selected Mutants

The most promising mutants selected from the screening were expressed and purified to homogeneity. According to SDS-PAGE, the molecular weight of the purified proteins corresponded to the calculated value of ~ 57 kDa. Since the qualitative method used for activity assessment was characterized by high variability, all activity measurements were performed in eight replicates to obtain statistically reliable data. Figure 4 presents the results of a comparative analysis of the activity of the wild-type enzyme and the most stable mutant variant (A4, ED230AT + Q287K + S223K + RE182SR) obtained during iterative saturation mutagenesis.

From this figure, it is evident that the best variant from saturation mutagenesis has a much higher reaction rate than WT and reaches the maximum fluorescence plateau much earlier. Additional kinetic data for the wild-type enzyme and its mutant variants, obtained by HPLC with a fluorescence detector, are presented in Table S3.

To assess the thermostability of the mutant enzyme variants, an approach analogous to the activity analysis was used, employing a fluorescent probe. The key difference in the protocol was that before the main reaction, the samples were subjected to a preliminary incubation at an elevated temperature to evaluate their residual catalytic activity.

As shown in Fig. 5, after a 10-minute incubation at 43°C, the residual activity of mutant A4 and the wild-type (WT) enzyme are almost identical. However, at 46.1°C, a significant difference is observed: variant A4 retains about 80% of its activity, while WT is completely inactivated. At 50°C, variant A4 still shows slight residual activity, but it is evident that the protein is unstable at this temperature.

Peak deconvolution of the temperature dependence of excess heat capacity Δc_{p, exc} obtained from the results of DSC measurements and calculation of the mid-transition temperature Tₘ as deconvoluted peak maxima and specific enthalpies of protein denaturation at temperatures Tₘ (∆h) as the area under the deconvoluted peaks in Gaussian assumption was performed using OriginPro v.8.0 software (OriginLab Corp., Northampton, MA, USA).

Notably, according to differential scanning calorimetry (DSC) data (Fig. 6), the melting temperature (Tₘ) of variant A4 is only 1°C higher than that of the wild-type protein. Despite this, after a 10-minute incubation at 46.1°C, variant A4 demonstrates a significantly higher level of reaction product fluorescence compared to the wild-type. This observation indirectly suggests either an increased kinetic stability of the mutant or its increased catalytic activity.

Variant 275 was constructed by introducing the double mutation S275E/K276E, predicted by the ProteinMPNN algorithm, into the A4 mutant backbone. This combined mutant showed a significant improvement in thermostability: according to DSC and DSF data, its melting temperature (Tₘ) is 6.5°C higher than that of WT. Unexpectedly, in the hairpin-opening experiments, activity of the variant 275 after incubation at 50°C was repeatedly observed to be higher than the activity after incubation at 38 to 46°C. The nature of this phenomenon remains unclear, especially given that the residual activity for all samples was measured at a standard temperature of 37°C.

Attempt to delete Loop 2

A comparative structural analysis of TdT (PDB: 4i27) and Polymerase Mu (Pol µ, PDB: 6VEZ) showed that Pol µ lacks a domain homologous to Loop 2 in TdT, yet it retains weak terminal transferase activity [44]. In the TdT structure, this region is completely disordered and is characterized by high B-factor values. Based on these observations, we hypothesized that deleting this flexible loop could lead to an increase in the overall thermostability of the enzyme.

To test this hypothesis, a TdT variant with a deletion of Loop 2 was constructed (Fig. 7). However, contrary to expectations, the resulting mutant showed a decrease in thermostability: its melting temperature (Tₘ) was approximately 2°C lower than that of the wild-type protein. In addition, significant difficulties were encountered during protein purification due to its propensity for aggregation and precipitation. Nevertheless, the mutant retained its catalytic activity. Since the deletion strategy led to the destabilization of the enzyme, this approach was deemed unpromising. It was decided to change the approach and focus on stabilizing and structuring Loop 2 by performing saturation mutagenesis of key amino acid residues in this domain, as mentioned above.

TdT is a promising polymerase, important for several biotechnological applications. Here, we applied a number of semi-rational and rational protein engineering strategies, which yielded enhanced variants of Mus musculus TdT. These observations might be helpful for engineering of other TdTs as well as for engineering campaigns of other enzymes.

The approach of screening a limited part of the library (in this case, ~ 6%) has once again demonstrated its effectiveness, which is consistent with previous studies [45]. As early as the first stage of saturation mutagenesis, a double mutation was identified that simultaneously improved both catalytic activity and protein expression level.

Structural model analysis (Fig. 8) shows that the double substitution E230A/D231T leads to a significant rearrangement of the local interaction network at the junction of two alpha-helices. The ionic bond between the original residues D231 and E233 is broken. The new residue, T231, forms stable contacts with E233 and the backbone of G227. Furthermore, this substitution appears to induce a local bend in the alpha-helix, creating the conditions for an additional hydrogen bond to form between T231 and S235.

Thus, a single double mutation leads to the replacement of one interaction with a whole network of new ones, which, according to calculations in the Foldit program, makes a significant contribution to lowering the free energy of the structure (ΔΔG). Despite these substitutions being located far from the active site, they likely stabilize the overall conformation of the protein, preventing its local unfolding. Modeling in PyMOL also shows that this double mutation may cause a bend at the end of the alpha-helix, and in this context, the E230A substitution is preferable as it eliminates steric hindrance for such a conformational change.

The analysis of the Q287 site is of particular interest. According to X-ray structural data (PDB: 4i27), an ionic bond exists between residues Q287 and K290 in the wild-type protein. However, during the first two rounds of saturation mutagenesis, substitutions that disrupt this interaction were selected. This paradox is explained by structural relaxation in the Foldit program: in the optimized model, this ionic bond disappears. This is consistent with the high B-factor value for this site, which indicates its inherent flexibility and possibly the transient nature of this interaction in silico.

The nature of the stabilizing effect of the Q287K substitution remains not fully clear, as it does not lead to the formation of obvious new bonds. At the same time, the positive effect of the Q287P substitution, found in the first round, can be explained by the reduction in conformational flexibility of the main chain due to the introduction of the rigid proline structure. It can be assumed that the proline variant was accidentally missed in the sampling during the second round of mutagenesis, and therefore was not identified as the best mutant. It is noteworthy that when using the mutation prediction tool based on a neural network in Foldit, the program suggests a substitution to proline as the most promising for this position.

In the first round of mutagenesis at position S223, a substitution to arginine (S223R) was selected, which coincides with the results previously obtained by [23] and confirms the significance of this mutation. However, in the third round of iterative mutagenesis, which was carried out on a template with existing substitutions, arginine was replaced by lysine (S223K) at this same position.

The original residue S223 forms a polar interaction with E226 within the same alpha-helix. The S223R substitution likely either strengthens this existing interaction or promotes the formation of a new salt bridge with residue E236, located on an adjacent alpha-helix. The preference for lysine over arginine in the third round is apparently due to a synergistic effect with the mutations from the first round (E230A/D231T). Modeling in PyMOL shows that the combination of these substitutions leads to the formation of an extensive network of ionic bonds that connects the two adjacent alpha-helices, thereby making a significant contribution to the overall stabilization of the structure.

The analysis of positions R182 and E183 also revealed interesting features. The original residue R182, according to X-ray data, forms interactions that connect different parts of the protein globule (Fig. S4), which at first glance makes it an undesirable target for mutagenesis. Nevertheless, this residue is characterized by a high B-factor value, indicating its mobility. This observation is confirmed by modeling in Foldit, where after structural relaxation, most of the initial interactions of R182 disappear, leaving only a contact with N178 within the same alpha-helix. Despite the apparent importance of the original residues, the double mutation R182S/E183R led to a small increase in thermostability and, more importantly, to a significant increase in the protein's expression level.

When modeled in the Foldit program, the M313R substitution seemed promising as it was predicted to create a new intra-helical ionic bond with residue E317. Alternative modeling in PyMOL indicated an interaction of this arginine with the adjacent flexible Loop 2. Furthermore, another modeling scenario suggested the formation of a salt bridge network linking three residues: the engineered R313, E317, and R414 from Loop 2. Regardless of the exact nature of the interaction, the introduction of this mutation led to a slight increase in thermostability, which is likely due to the stabilizing effect of the new salt bridge. Similarly, for the C302E mutation, the formation of a new interaction with residue T340 was predicted during modeling. A key feature of this substitution is its location in close proximity to the active site: T340 is part of the loop involved in coordinating the Mg²⁺ ion. Contrary to expectations that changes in this critical area would lead to a loss of activity, the mutation led to a slight increase in the enzyme's thermostability. The effect of this substitution on specificity towards different nucleotides requires a separate detailed investigation.

The most significant contribution to the enzyme's stabilization was made by the double substitution S275E/K276E (Fig. 9), which was designed based on predictions from the ProteinMPNN and Carbonara algorithms. The best variant from the previous rounds of saturation mutagenesis, mutant A4, was used as the template for introducing these mutations. This change led to an increase in the protein's melting temperature by 5.5°C. Of particular interest is the fact that exactly the same substitutions were independently obtained in a large-scale directed evolution study carried out by [25]. Structural analysis shows that the K276E substitution helps to structure the disordered region between two alpha-helices by forming new interactions with the backbones of amino acids R272 and T273. Thus, an initially flexible region becomes more rigid. These substitutions are located in a domain that, although distant from the catalytic center, is involved in binding with single-stranded DNA. An increase in temperature likely leads to the local unfolding of this region, which interferes with substrate binding. The stabilization of this area prevents this process. The contribution of the S275E substitution to overall stability is less clear. Calculations in Foldit show that it lowers the overall Gibbs free energy, but its individual contribution to stabilization requires further experimental verification.

Unfortunately, this double substitution, along with stabilizing the enzyme without loss of activity, led to a significant decrease in its expression level. While the A4 mutant obtained by saturation mutagenesis showed a 26-fold increase in expression level compared to the wild-type protein, the introduction of this double mutation completely nullified this advantage. As a result, the expression level of the final variant returned to the initial values characteristic of the wild-type enzyme.

Thus, it can be concluded that the substitutions with the greatest stabilizing effect were those that allowed for the coordination of 3–4 amino acids together at the junction of secondary structures. The distance of such substitutions from the active site was not a decisive factor. All three successful substitutions predicted within the framework of classical rational design (without using neural networks) involved replacing a hydrophobic residue with a hydrophilic one, creating new polar bonds. Attempts to modify the existing system of ionic bonds using rational mutagenesis were essentially unsuccessful. Only the application of two neural network algorithms allowed for the discovery of a non-obvious substitution that made the most significant contribution to thermostability.

The direct screening methodology we employed, based on the analysis of mutant activity in crude cell lysates via polyacrylamide gel electrophoresis (PAGE), has several inherent limitations.

The primary drawback is its low sensitivity, which prevents the reliable detection of subtle improvements in thermostability. Consequently, mutations that confer a slight increase in stability but concurrently decrease catalytic activity or expression levels are likely to be falsely discarded. In such cases, a decrease in the product band intensity on the gel would be misinterpreted as a negative result, leading to the loss of potentially valuable variants.

In the work carried out to increase the thermostability of TdT, a comprehensive improvement of the key properties of the enzyme was achieved. Along with an increase in thermostability, the catalytic activity and, especially importantly, the protein expression level were significantly increased. For instance, while the yield of wild-type (WT) protein was 1.5 mg per liter of culture, after the first round of mutagenesis this figure reached 6 mg/L, and the final round mutant yielded 40 mg/L. The activity at 37°C also gradually increased because the more active a mutant was, the brighter the band it left on the electrophoresis gel, and the more likely it was to be selected as successful. In addition, we assessed the residual activity after a 10-minute heating period; naturally, the higher a mutant's activity and quantity, the more intense the band it will show after such heating, regardless of the increase in thermostability.

Analysis of various mutagenesis strategies showed that targeting residues with a high B-factor is an effective approach, while random mutations outside these flexible regions did not lead to success. Attempts at rational design aimed at "stitching" the enzyme's domains together to prevent its unfolding were also largely unsuccessful. This emphasizes that the use of modern computational tools, particularly neural networks, is critically important for narrowing the search space and identifying non-obvious but effective mutations.

Thus, the initial task of increasing only the thermostability led to the creation of an enzyme that was comprehensively improved in all its main parameters. The resulting mutant variant is an excellent scaffold for further engineering, particularly for improving its activity towards 3'-modified nucleoside triphosphates used in enzymatic DNA synthesis technologies.

A comparison of our final variant, 275, with recently engineered TdTs reveals the unique strengths of our approach. The thermostability enhancement achieved for mutant 275 (ΔTₘ = +6.5°C) is significant, although more profound increases were reported by (Niu et al.) (ΔTₘ = +10.3°C for M7-8) and (Forget et al.) (ΔTₘ = +20°C for TdT-33) .

However, our work highlights a critical parameter for the biotechnological viability of an enzyme: the recombinant expression level. Our intermediate variant, A4, demonstrated a remarkable 26-fold increase in protein yield compared to the wild-type. This dramatic improvement in manufacturability is a key advantage for any industrial application and was a central outcome of our engineering trajectory, a metric not quantitatively detailed in the aforementioned studies.

Furthermore, our stepwise evolution strategy (WT → A4 → 275) provides a clear and interpretable path of improvement, allowing for detailed structural analysis of the contribution of key mutations, as discussed above. This contrasts with approaches that introduce a large number of mutations simultaneously, where the effects of individual changes can be convoluted. Thus, our final variant 275 represents a highly balanced candidate, combining a significant increase in stability and a dramatically improved activity background inherited from its A4 parent. This makes it a valuable and technologically accessible enzyme for enzymatic DNA synthesis and a robust scaffold for further engineering.

Acknowledgements

Not applicable.

Author contributions

N.I.A. and F.O.V. conceived and designed the experiments. N.I.A., N.D.S., and Z.K.F. performed the experiments. K.J.E. synthesized the primers and molecular beacons. R.I.S. performed the Sanger sequencing. B.A.D. and G.I.Yu. predicted mutations using neural networks. S.M.P. and M.O.G. analyzed the proteins using DSC and DSF. N.I.A. wrote the manuscript. G.I.Yu. reviewed the manuscript. All authors have read and approved the final version of the manuscript.

Funding

The study is supported by the Russian state task number 125041005130-8.

G.I.Yu. acknowledges the support from the Ministry of Science and Higher Education of the Russian Federation (agreement 075-03-2025-662, project FSMG-2025-0003) for computational analysis with neural networks.

Availability of data and materials

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate

This article does not contain any studies with human participants or animals performed by any of the authors.

Consent for publication

Not applicable.

Competing interests

The authors declare they have no conflict of interests

Bollum, F. J. (1960). Calf thymus polymerase. The Journal of Biological Chemistry, 235, 2399–2403.
Landau, N. R., Schatz, D. G., Rosa, M., & Baltimore, D. (1987). Increased Frequency of N-Region Insertion in a Murine Pre-B-Cell Line Infected with a Terminal Deoxynucleotidyl Transferase Retroviral Expression Vector. Molecular and Cellular Biology, 7(9), 3237–3243.
Yang, B., Gathy, K. N., & Coleman, M. S. (1994). Mutational analysis of residues in the nucleotide binding domain of human terminal deoxynucleotidyl transferase. The Journal of Biological Chemistry, 269(16), 11859–11868.
Peterson, R. C., Cheung, L. C., Mattaliano, R. J., White, S. T., Chang, L. M., & Bollum, F. J. (1985). Expression of human terminal deoxynucleotidyl transferase in Escherichia coli. The Journal of Biological Chemistry, 260(19), 10495–10502.
Boulé, J. B., Johnson, E., Rougeon, F., & Papanicolaou, C. (1998). High-level expression of murine terminal deoxynucleotidyl transferase inEscherichia coli grown at low temperature and overexpressingargU tRNA. Molecular Biotechnology, 10(3), 199–208.
Bentolila, L. A., Fanton d’Andon, M., Nguyen, Q. T., Martinez, O., Rougeon, F., & Doyen, N. (1995). The two isoforms of mouse terminal deoxynucleotidyl transferase differ in both the ability to add N regions and subcellular localization. The EMBO Journal, 14(17), 4221–4229.
Delarue, M., Boulé, J. B., Lescar, J., Expert-Bezançon, N., Jourdan, N., Sukumar, N., & Papanicolaou, C. (2002). Crystal structures of a template-independent DNA polymerase: murine terminal deoxynucleotidyltransferase. The EMBO Journal, 21(3), 427–439.
Gouge, J., Rosario, S., Romain, F., Beguin, P., & Delarue, M. (2013). Structures of Intermediates along the Catalytic Cycle of Terminal Deoxynucleotidyltransferase: Dynamical Aspects of the Two-Metal Ion Mechanism. Journal of Molecular Biology, 425(22), 4334–4352.
Tu, C. P. D., & Cohen, S. N. (1980). 3′-End labeling of DNA with [α–32P] cordycepin–5′-triphosphate. Gene, 10(2), 177–183.
Deng, G. R., & Wu, R. (1983). [5] Terminal transferase: Use in the tailing of DNA and for in Vitro mutagenesis. Methods in Enzymology (Vol. 100, pp. 96–116). Elsevier.
Kumar, A., Tchen, P., Roullet, F., & Cohen, J. (1988). Nonradioactive labeling of synthetic oligonucleotide probes with terminal deoxynucleotidyl transferase. Analytical Biochemistry, 169(2), 376–382.
Tian, J., Ma, K., & Saaem, I. (2009). Advancing high-throughput gene synthesis technology. Molecular BioSystems, 5(7), 714.
Carr, P. A. (2004). Protein-mediated error correction for de novo DNA synthesis. Nucleic Acids Research, 32(20), e162–e162.
Minhaz Ud-Dean, S. M. (2008). A theoretical model for template-free synthesis of long DNA sequence. Systems and Synthetic Biology, 2(3–4), 67–73.
Hutter, D., Kim, M. J., Karalkar, N., Leal, N. A., Chen, F., Guggenheim, E., & Benner, S. A. (2010). Labeled Nucleoside Triphosphates with Reversibly Terminating Aminoalkoxyl Groups. Nucleosides Nucleotides & Nucleic Acids, 29(11–12), 879–895.
Mathews, A. S., Yang, H., & Montemagno, C. (2017). 3′- O ‐Caged 2′‐Deoxynucleoside Triphosphates for Light‐Mediated, Enzyme‐Catalyzed, Template‐Independent DNA Synthesis. Current Protocols in Nucleic Acid Chemistry, 71(1).
Koukhareva, I., & Lebedev, A. (2009). 3′-Protected 2′-Deoxynucleoside 5′-Triphosphates as a Tool for Heat-Triggered Activation of Polymerase Chain Reaction. Analytical Chemistry, 81(12), 4955–4962.
Efcavitch, J. W., & Sylvester, J. E. (2019). December 4). Modified template-independent enzymes for polydeoxynucleotide synthesis.
Flamme, M., Hanlon, S., Marzuoli, I., Püntener, K., Sladojevich, F., & Hollenstein, M. (2022). Evaluation of 3′-phosphate as a transient protecting group for controlled enzymatic synthesis of DNA and XNA oligonucleotides. Communications Chemistry, 5(1), 68.
Sabat, N., Katkevica, D., Pajuste, K., Flamme, M., Stämpfli, A., Katkevics, M., & Hollenstein, M. (2023). Towards the controlled enzymatic synthesis of LNA containing oligonucleotides. Frontiers in Chemistry, 11, 1161462.
Verardo, D., Adelizzi, B., Rodriguez-Pinzon, D. A., Moghaddam, N., Thomée, E., Loman, T., & Horgan, A. (2023). Multiplex enzymatic synthesis of DNA with single-base resolution. Science Advances, 9(27), eadi0263.
Chua, J. P. S., Go, M. K., Osothprarop, T., Mcdonald, S., Karabadzhak, A. G., Yew, W. S., & Nirantar, S. (2020). Evolving a Thermostable Terminal Deoxynucleotidyl Transferase. ACS Synthetic Biology, 9(7), 1725–1735.
Barthel, S., Palluk, S., Hillson, N. J., Keasling, J. D., & Arlow, D. H. (2020). Enhancing Terminal Deoxynucleotidyl Transferase Activity on Substrates with 3′ Terminal Structures for Enzymatic De Novo DNA Synthesis. Genes, 11(1), 102.
Palluk, S., Arlow, D. H., De Rond, T., Barthel, S., Kang, J. S., Bector, R., & Keasling, J. D. (2018). De novo DNA synthesis using polymerase-nucleotide conjugates. Nature Biotechnology, 36(7), 645–650.
Forget, S. M., Krawczyk, M. J., Knight, A. M., Ching, C., Copeland, R. A., Mahmoodi, N., & Lutz, S. (2025). Evolving a terminal deoxynucleotidyl transferase for commercial enzymatic DNA synthesis. Nucleic Acids Research, 53(4), gkaf115.
Niu, Y., Chen, B., Zhang, H., Zheng, W., Wu, J., Yang, L., & Yu, H. (2025). Computational Design of a Thermostable and Highly Active Terminal Deoxynucleotidyl Transferase for Synthesis of Long De Novo DNA Molecules. ACS Catalysis, 15(9), 7201–7216.
Xiong, A. S., Peng, R. H., Zhuang, J., Gao, F., Li, Y., Cheng, Z. M., & Yao, Q. H. (2008). Chemical gene synthesis: strategies, softwares, error corrections, and applications. FEMS Microbiology Reviews, 32(3), 522–540.
Fu, C., Donovan, W. P., Shikapwashya-Hasser, O., Ye, X., & Cole, R. H. (2014). Hot Fusion: An Efficient Method to Clone Multiple DNA Fragments as Well as Inverted Repeats without Ligase. Plos One, 9(12), e115318.
Zheng, L. (2004). An efficient one-step site-directed and site-saturation mutagenesis protocol. Nucleic Acids Research, 32(14), e115–e115.
Gouge, J., Rosario, S., Romain, F., Beguin, P., & Delarue, M. (2013). Structures of Intermediates along the Catalytic Cycle of Terminal Deoxynucleotidyltransferase: Dynamical Aspects of the Two-Metal Ion Mechanism. Journal of Molecular Biology, 425(22), 4334–4352.
Waterhouse, A., Bertoni, M., Bienert, S., Studer, G., Tauriello, G., Gumienny, R., & Schwede, T. (2018). SWISS-MODEL: homology modelling of protein structures and complexes. Nucleic Acids Research, 46(W1), W296–W303.
Duvaud, S., Gabella, C., Lisacek, F., Stockinger, H., Ioannidis, V., & Durinx, C. (2021). Expasy, the Swiss Bioinformatics Resource Portal, as designed by its users. Nucleic Acids Research, 49(W1), W216–W227.
Dauparas, J., Bennett, N., & Anishchenko, I. (2022). Robust deep learning–based protein sequence design using ProteinMPNN, Vol 378(Issue 6615), 49–56.
Krapp, L. F., Meireles, F. A., Abriata, L. A., Devillard, J., Vacle, S., Marcaida, M. J., & Peraro, D., M (2024). Context-aware geometric deep learning for protein sequence design. Nature Communications, 15(1), 6273.
Tareen, A., & Kinney, J. B. (2020). Logomaker: beautiful sequence logos in Python. Bioinformatics, 36(7), 2272–2274.
Permyakov, S. E., Vologzhannikova, A. A., Emelyanenko, V. I., Knyazeva, E. L., Kazakov, A. S., Lapteva, Y. S., & Permyakov, E. A. (2012). The impact of alpha-N-acetylation on structural and functional status of parvalbumin. Cell Calcium, 52(5), 366–376.
Uchiyama, Y., Takeuchi, R., Kodera, H., & Sakaguchi, K. (2009). Distribution and roles of X-family DNA polymerases in eukaryotes. Biochimie, 91(2), 165–170.
Nakane, S., Nakagawa, N., Kuramitsu, S., & Masui, R. (2009). Characterization of DNA polymerase X from Thermus thermophilus HB8 reveals the POLXc and PHP domains are both required for 3′–5′ exonuclease activity. Nucleic Acids Research, 37(6), 2037–2052.
Sun, Z., Liu, Q., Qu, G., Feng, Y., & Reetz, M. T. (2019). Utility of B-Factors in Protein Science: Interpreting Rigidity, Flexibility, and Internal Motion and Engineering Thermostability. Chemical Reviews, 119(3), 1626–1665.
Romain, F., Barbosa, I., Gouge, J., Rougeon, F., & Delarue, M. (2009). Conferring a template-dependent polymerase activity to terminal deoxynucleotidyltransferase by mutations in the Loop1 region. Nucleic Acids Research, 37(14), 4642–4656.
Musil, M., Stourac, J., Bendl, J., Brezovsky, J., Prokop, Z., Zendulka, J., & Damborsky, J. (2017). FireProt: web server for automated design of thermostable proteins. Nucleic Acids Research, 45(W1), W393–W399.
Reetz, M. T., Carballeira, J. D., & Vogel, A. (2006). Iterative Saturation Mutagenesis on the Basis of B Factors as a Strategy for Increasing Protein Thermostability. Angewandte Chemie International Edition, 45(46), 7745–7751.
Cooper, S., Khatib, F., Treuille, A., Barbero, J., Lee, J., Beenen, M., & Players, F. (2010). Predicting protein structures with a multiplayer online game. Nature, 466(7307), 756–760.
Andrade, P., Martín, M. J., Juárez, R., De López, F., & Blanco, L. (2009). Limited terminal transferase in human DNA polymerase µ defines the required balance between accuracy and efficiency in NHEJ. Proceedings of the National Academy of Sciences, 106(38), 16203–16208.
Parra, L. P., Agudo, R., & Reetz, M. T. (2013). Directed Evolution by Using Iterative Saturation Mutagenesis Based on Multiresidue Sites. Chembiochem, 14(17), 2301–2309.

Additionalfile1..docx
Supplementary Information Supplementary material available in Addition file 1, Calculated single amino acid substitution pmpnnres, Calculated single amino acid substitution carbres. Additional file 1. Table S1 List of primers used for saturation mutagenesis. Table S2Effects of substitutions predicted by neural networks on TdT. Table S3 Kinetic parameters. Fig. S1 Map of the pET-28a(+) plasmid containing the gene for SUMO-tagged TdT used for cloning and transformation. Fig. S2 An example of the activity screening for rationally designed mutants in cell supernatants using a fluorescently labeled primer. The image shows the residual activity at 37°C following a 10-minute incubation at the indicated temperature. Fig. S3 An example of the screening for the best mutants obtained via saturation mutagenesis at the ED230 site. The reaction was performed in cell supernatants using a fluorescently labeled primer. The image shows the residual activity at 37°C following a 10-minute incubation at the indicated temperature. Fig. S4 Interactions of the R182 residue. Fig. S5 Protein unfolding curves and peak temperature obtained using Tycho NT.6 Fig. S6 Temperature dependencies of the excess specific heat capacity and thermal denaturation parameters Tₘ and Δh of wild-type (WT) TdT protein, 275, A4, H11 and G2 mutant samples obtained by DSC measurements. Fig. S7 Measurement of residual activity of the wild-type (WT) enzyme. Samples were pre-incubated at 45°C for the times indicated in the figure. Residual activity was then measured at 37°C. Each curve represents the average of three independent replicates. Fig. S8 Measurement of residual activity of the mutant A4. Samples were pre-incubated at 45°C for the times indicated in the figure. Residual activity was then measured at 37°C. Each curve represents the average of four independent replicates.
Calculatedsingleaminoacidsubstitutioncarbres.csv
Calculatedsingleaminoacidsubstitutionpmpnnres.csv

Download PDF

Editorial decision: Resubmit revised form; Major revisions required
30 Nov, 2025
Reviewers agreed at journal
10 Oct, 2025
Reviewers invited by journal
21 Sep, 2025
Editor invited by journal
02 Sep, 2025
First submitted to journal
29 Aug, 2025

You are reading this latest preprint version

Optimization of Expression and Thermostability of Terminal Deoxynucleotidyl Transferase through Iterative Mutagenesis and Computational Design

Status:

Version 1

Abstract

Figures

Introduction

Material and methods

Plasmids construction

Protein Expression

Protein Purification

Site-Directed and Saturation Mutagenesis

Prediction of single mutations ProteinMPNN and CARBonAra

Analysis of Site-Directed Mutants

Screening of Saturation Mutagenesis Libraries

Thermostability and Activity Assays with Purified Protein

Protein Thermal Unfolding

Analysis of Kinetic Parameters

Results

Selection of residues for Saturation Mutagenesis

Rational Design and its Results

Rational Design based on ProteinMPNN and CARBonAra Predictions

Characterization and Activity Analysis of Selected Mutants

Attempt to delete Loop 2

Discussion

Declarations

References

Supplementary Files

Status:

Version 1