Directed evolution of mesophilic HNA polymerases providing insight into DNA polymerase mechanisms

Authors:

Ms. Paola Handal-Marquez^1,2, Dr. Leticia L. Torres¹ and Prof. Dr. Vitor B. Pinheiro^2*

Affiliations:

¹ University College London, Department of Structural and Molecular Biology, Gower Street, WC1E 6BT, London, UK

² KU Leuven, Rega Institute for Medical Research, Department of Pharmaceutical and Pharmacological Sciences, Herestraat, 49 – box 1041, 3000 Leuven, Belgium

Corresponding author information: ^* e-mail: v.pinheiro@kuleuven.be

Abstract

Detailed biochemical characterization of natural and mutant enzymes provides essential clues to understand their mechanisms. There are, however, limits to the throughput of such approaches and they are not without errors. DNA polymerases have benefited from over 50 years of detailed study and remain not fully understood. As such, methods that allow high-throughput interrogation of variants, and viable analysis pipelines to identify relevant variants, become an important tool to accelerate research. Using the DNA polymerase from B. subtilis Phi29 bacteriophage as a model, we demonstrate how coupling focused libraries, selection and deep sequencing can be combined to identify variants of interest for characterization. As selection parameters can be controlled, different areas of an enzyme’s mechanism can be explored. Focusing selection on faster HNA (1,5-anhydrohexitol nucleic acid) synthesis, we identified P562del as a variant of interest, enriching significantly between rounds. Characterization confirmed its faster HNA synthesis initiation but lower processivity and fidelity. P562 is a non-conserved residue, unlikely to be selected by more traditional approaches, but its deletion recapitulates knowledge on how Phi29 exonuclease, thumb and TPR2 subdomains regulate polymerase function. Our data further support the hypothesis that Phi29 shows a two-state binding to its template: a fast non-replicative complex that transitions to a replication-competent state.

Keywords:

Directed evolution; XNA synthase; Phi29 DNA polymerase; high-throughput biochemistry

Introduction

It is essential for all living organisms to store and replicate their genetic information. That process is carried out by specialist enzymes, termed template-dependent DNA (or RNA) polymerases. Despite their central role, the ancient origin of those proteins has given evolution enough time and opportunity to explore a vast range of characteristics and specializations, across processivity, fidelity, thermostability and multimerization^1,2. Those characteristics and specializations are the result of functional and sequence constraints imposed during the natural evolution of DNA polymerases, and they are not yet fully understood. Replicative DNA-dependent DNA polymerases are often multifunctional enzymes (harboring both polymerase and exonuclease activities) with complex dynamics and operating under multiple, and sometimes overlapping, layers of regulation. For instance, DNA polymerases can often discriminate between their natural substrate (deoxyribose nucleotide triphosphates) and other natural analogues present in the cellular milieu (ribonucleotides, partially phosphorylated nucleotides, and damaged nucleobases), even though the analogues are present at a higher concentration than the substrates³. This is done not only at incorporation in the active site of the enzyme^3,4, but also at multiple points during the elongation process^5,6.

Therefore, because of their complexity and as a result of their evolution, specific mutations can have different impacts on the function of homologous DNA polymerases. Mutations that in the Escherichia coli bacteriophage T4 result in 10-fold reduction of viability⁷, have no impact on the T4-related RB69 phage⁸. Similarly, a β hairpin mutation that decreases fidelity in both T4 and RB69, does not impact S. cerevisiae DNA polymerase δ (also a B-family polymerase).⁹

Detailed biochemical and mutational analysis has been very successful at identifying some of the sequence and structural determinants for the different DNA polymerase functions, but they are limited to the models we use to understand polymerases, to how those can be efficiently tested, and to how generalizable they are.

Deep mutational scanning (DMS) uses directed evolution strategies to assess the function of large numbers of protein mutants in parallel¹⁰. Mutants are partitioned based on function with high-throughput DNA sequencing used to map each population – thus generating a picture of the functional space of the protein. Such an approach bypasses gaps in our understanding of the sequence-structure-function relationships of the targeted proteins and has been successfully applied to engineer protein variants with novel properties. Nonetheless, DMS is usually focused on the isolation of desired variants (e.g. more efficient enzymes) rather than on the exploitation of the informationally-rich mutational datasets, which have the potential to improve our understanding of a protein’s sequence-structure-function relationships.

DNA polymerases can be engineered for the synthesis of xenobiotic nucleic acids (XNAs), and multiple novel mutations have been identified that can modulate enzymatic function and further contribute to our understanding of those enzymes^11,12. However, as with DMS approaches, improved functional understanding is a byproduct of the main research goal. By shifting the focus from “best” to “all” active mutants, we propose that the sequence landscape mapping in DMS and in directed evolution experiments (as summarized in Figure 1), can be harnessed to improving our understanding of protein function and to studying the impact of selection parameters.

XNAs such as HNA (1,5-anhydrohexitol nucleic acid) are of great interest for therapeutic and nanotechnology applications¹³ because of its enhanced chemical and biological stability¹⁴, which also make HNA a viable alternative genetic polymer for in vivo applications, such as novel platforms for biological containment¹⁵.

The hyperthermophilic DNA polymerase from Thermococcus gorgonarius was the first DNA polymerase to be engineered for HNA synthesis. From a starting variant that harboured 4 mutations, a further 14 mutations were acquired in directed evolution to enable efficient HNA synthesis – generating the enzyme named 6G12¹⁶. Nevertheless, the high temperatures required by 6G12 for HNA synthesis were incompatible with in vivo applications, prompting us to focus on mesophilic polymerases.

The high mutational load in 6G12 was hypothesised to reshape the enzyme thumb domain to accommodate the nascent HNA/DNA helix. However, our efforts on the B. subtilis bacteriophage Phi29, determined that reducing (N62D) or abolishing (D12A) exonuclease activity in this polymerase was sufficient to enable HNA synthesis¹⁷ – leading us to propose that XNA/DNA-polymerase complex stability was the key driving parameter for XNA synthesis, as also observed for engineered RNA polymerases¹⁸.

The high processivity of Phi29 DNA polymerase, which can synthesize up to 70 kb DNA per binding event¹⁹, is due in large part by the tight interaction between exonuclease, thumb and TPR2 domain, and therefore, a suitable focus of engineering for improving HNA synthesis and for gaining further understanding of how those regions coordinate enzyme processivity and fidelity.

Figure 1: Mesophilic XNA polymerase directed evolution pipeline. A library is initially constructed – represented (top left panel) by different color/lengths compared to the wild-type. The library is then subjected to a round(s) of selection through compartmentalized self-tagging (top right panel). The library pre- and post-selection is sequenced through NGS (bottom left panel). The proportion of highly active variants (e.g., green) should increase whereas that of variants with lower activity (e.g., orange, and red) should decrease. NGS data is filtered for quality and pre-processed data is then analyzed (bottom right). The frequency of each mutant corresponds to its abundance in the library pre- and post-selection. An enrichment score is used to evaluate the increase (or decrease) in abundance of specific mutants within the population and serves as a starting point for further analysis.

Here, we describe the identification of single residue in the DNA polymerase thumb (P562) implicated in Phi29 DNA polymerase processivity, demonstrating that high-throughput selections can be harnessed as a high-throughput biochemistry tool, allowing novel mechanistic insights in enzyme function. We characterise the isolated enzyme to demonstrate the impact of the mutation on DNA affinity, enzyme processivity, fidelity and strand displacement activities. Finally, we confirm that Phi29 DNA polymerase needs not only to bind DNA but also to switch to a stable conformation to achieve highly processive DNA synthesis.

Results

Library design: focused libraries sampling beyond biochemical knowledge.

Phi29 DNA polymerase processivity is closely associated with the terminal protein region 2 (TRP2) subdomain, which forms a structure akin to a sliding clamp around template and nascent DNA, and which harbours multiple contacts with the thumb (along residues V559-V566) and exonuclease (along residues T15-V24 and F65-N72) subdomains (see Figure 2A).

Deletion of the TRP2 subdomain, does not affect the ratio between exonuclease and polymerase activities of the polymerase but significantly decreases polymerase affinity for DNA and its processivity²⁰. This is analogous to what is observed in the T7 bacteriophage DNA polymerase (gp5), where high processivity is only observed when a clamp-like structure is formed around the DNA by the binding of the thioredoxin processivity factor²¹.

The close contacts between TRP2, thumb and exonuclease can, in principle, affect other polymerase processes, including exonuclease activity (given the packing against an alpha helix harbouring one of the catalytic exonuclease residues – D66), exonuclease sampling (DNA switching between polymerase and exonuclease catalytic sites, by modulating how exonuclease and thumb interact), strand displacing activity (through the interaction TRP2 with template DNA) and DNA binding.

Figure 2. Structure of the closed ternary complex of Phi29 DNA Polymerase and InDel library design. (A) There domains in the Phi29 DNA polymerase (PDB: 2PYJ) are involved in creating a clamp around the nascent DNA duplex: the polymerase thumb (green), the exonuclease domain (yellow) and the TPR2 domain (purple). The DNA duplex is shown in cyan. (B) InDel mutagenesis of the thumb loop through inverse PCR (iPCR). (B) A single reverse primer in combination with forward primers harbouring 1-3 NNS codons or with priming sites that skipped 1-4 codons were used to generate focused libraries or deletion mutants respectively. The same approach was used to target the TPR2 and exonuclease loops.

Although there are multiple possible contacts between TRP2, thumb and exonuclease subdomains, we hypothesised that contact between the three loops (TRP2: L406-L412, thumb: V559-V566, and exonuclease T15-V24) is likely functional, capable of providing the dynamic response needed for polymerase-exonuclease balance and enzyme processivity.

Traditionally, libraries for directed evolution focus on introducing amino acid substitutions without changes to sequence length²². Nonetheless, via the engineering of an XNA ligase based on the Chlorella virus DNA ligase²³, we recently determined that changes in length can be essential for the change in substrate specificity, allowing nucleic acids with different helical parameters to be accommodated into the enzyme’s active site. Given the tight packing of the loops and the nascent DNA in Phi29 DNA polymerase, we decided to investigate if optimization of sequence as well as length could create new molecular interactions to stabilize protein binding to the nascent heteroduplex and that could be limiting Phi29 to become a highly processive HNA polymerase.

Only limited diversity can be introduced in the 29 residues, given the rapid rise in the number of variants and the poor stability of the Phi29 DNA polymerase (half-life of 37 ± 4.5 min at 40°C, when in the presence of stabilizing DNA template²⁴). The wide range of available DNA library assembly strategies allows length and compositional variation to be introduced, randomly or in a targeted manner, across a gene of interest^25,26, such that technical bottlenecks in library generation are no longer relevant. Instead, design is constrained by the number of variants that can be generated and sampled (linking design to selection methodology), considering the likely destabilizing effect of a high mutational burden to the protein being engineered.

Starting from a previously reported thermostabilized Phi29 DNA polymerase²⁴ (t_1/2 > 16 h at 40°C in the presence of DNA), harboring the additional D12A mutation (which disables exonuclease activity) to allow limited HNA synthesis, we opted to concomitantly explore loop length in the three loops. Firstly, because insertions and deletions (indels) tend to be least disruptive to protein stability when they occur in loops. Secondly, we hypothesized that loop distortions, due to indels, would be a more sensitive tool to study loop-loop interaction than patterns of substitution, while also minimizing library diversity. That is further supported by sequence alignment with Phi29 DNA polymerase homologues (Figure S1), showing that these regions can accept indels.

The tips of the loops (residues D19, N409 and P562) were used as seed in design, with up to three insertions being introduced upstream of those seeds, or up to four deletions targeting the seed and downstream residues, as shown in Figure 2B. Exonuclease, TRP2 and thumb libraries had each a theoretical maximum diversity of approximately 8425 variants – well within the sampling range of the chosen selection platform (approximately 2 x 10⁷ per experiment) and analysis pipeline (between 5.0 and 7.0 x 10⁴ of sequences per run). Summary of the number of variants in the libraries pre-selection can be found on Table 1.

Although guided by the polymerase structures and phylogeny, the libraries are agnostic to the biochemical knowledge that TRP2^20,27 and the Phi29 DNA polymerase C-terminus^28,29 are crucial for processivity, with a single previous report targeting one of those residues²⁷.

InDel size	Exo R0	InDel Size	TPR2 R0	InDel size	Thumb R0
0	543	0	366	0	195
1	472	1	11520	1	107
2	14927	2	83	2	473
3	13238	3	5405	3	8104
-1	26	-1	57750	-1	143
-2	2421	-2	75	-2	2048
-3	195	-3	46	-3	11328
				-4	12944
Others	580		450		507
Total	32402		75695		35849
Theoretical library size	8424		8424		8425

Table 1. Sequence coverage of starting libraries across the different assembled lengths. Details of NGS processing and analysis are summarized in Figure S2 and Tables S3, S4 and S5.

Using enrichment to identify relevant mutations in selection

High-throughput selections rely on systems that link the phenotype being sought with efficient recovery of the responsible genotype, so that at each round of selection, the genotype of active enzymes becomes more frequent.

A variety of selection methodologies have been demonstrated for the selection of DNA and XNA polymerases, including in vivo³⁰, phage display³¹, and emulsion-based methods such as CSR (compartmentalised self-replication)³², CST (compartmentalised self-tagging)³³, and CBL (compartmentalised bead-labelling)³⁴. Technical characteristics of each method shape the range of polymerase activities that can be selected and what activities can emerge as selection parasites (active enzymes that co-enrich with the desired activity).

CST relies on polymerases capable of stabilizing the binding of the selection primer against their own gene through XNA synthesis (Figure 1). Given the diminishing return per XNA incorporation to the overall primer stabilization (i.e. analogous to oligonucleotide length effects on primer melting temperature), the pressure of CST selection is highest on the first few incorporations, with CST selections working akin to a high-pass filter (i.e. selection is unable to discriminate between polymerases that can synthesise beyond a certain number of incorporations). Expected parasites in this CST selection include polymerases that can make use of the low concentrations of natural cellular dNTPs still present, polymerases that express to significantly higher levels or that are more thermostable.

For the selection of HNA polymerases with higher processivity, extension time is a key parameter, sinceparameter since shorter extension times are likely to benefit highly processive enzymes as well as polymerases that can start XNA synthesis faster. The latter was not expected to be a significant possibility, considering Phi29 high affinity for DNA (K_D = 1 µM in conditions close to the selection conditions³⁵) and in view of our experience engineering thermostable XNA polymerases.

A single round of selection for HNA synthesis, with a short extension time of 10 min compared to the 1 h needed by the starting enzyme (Phi29 DNA polymerase: D12A) for 72 incorporations, was carried out for each library, followed by deep sequencing of naïve and selected libraries for analysis. Frequency counting post-selection can be a suitable route to identify best candidates but assumes that the starting library is devoid of biases, working best for selection methodologies that have low background (i.e. recovered genotypes that are randomly recovered) and for late-stage selection campaigns (i.e. libraries that have already been carried through multiple selection rounds and that are populated by few variants)³⁶. Given libraries were pooled (by site across all designed lengths) and since they require PCR amplification post-selection (two processes expected to introduce biases), we opted to use enrichment as a measure of selection, thus normalizing post-selection frequencies against pre-selection ones.

As expected, deletions were highly sampled in all three starting libraries while insertions showed partial coverage of the available sequence space. Based on the pre-selection NGS analysis, the diversity of the libraries was estimated to range between 1963 (TPR2 library) and 4570 variants (exonuclease library) (Figure S2).

We filtered the datasets to focus on variants that were present in both pre- and post-selection populations. We scored the sequences based on the change in frequency between rounds multiplied by the average number of observations, thus biasing detection towards well sampled sequences that were also enriching due to selection. Statistical significance of enrichment was tested (using only changes in sampling frequency) with an E-test³⁷, a sensitive test for detecting differences between two Poisson distributions. Significantly enriched sequences are summarised in Table 2.

Out of the identified mutants, P562del (Phi29 DNAP M8R, D12A, V51A, M97T, G197D, E221K, Q497P, K512E, F526L, ΔP562) and PG563del (Phi29 DNAP M8R, D12A, V51A, M97T, G197D, E221K, Q497P, K512E, F526L, ΔP562-G563) variants were notable, having enriched better than the parental enzyme (D12A-THR: Phi29 DNAP M8R, D12A, V51A, M97T, G197D, E221K, Q497P, K512E, F526L) in selection. In view that deletion of P562 alone was sufficient for the increased enrichment, we chose to focus on this variant for further characterisation.

Thumb loop variants	InDel	R0 Count (n= 35849)	R1 Count (n= 46907)	EC	E-test
KPKPVQV--GVVLVD	-2	1933	2821	9.7E+00	0.0E+00
KPKPVQV-GGVVLVD	-1	138	443	1.1E+00	0.0E+00
KPKPVQVPGGVVLVD	0	132	430	1.0E+00	0.0E+00
KPKPVQV-----LVD	-5	72	117	4.6E-02	1.4E-01
KPKPVQ--GGVVLVD	-2	96	132	1.6E-02	7.2E-01
TPR2 loop variants	InDel	R0 Count (n= 75695)	R1 Count (n= 73519)	EC	E-test
YLKE NGALGFRLGEE	0	123	730	3.5E+00	0.0E+00
YLKEGRGNGALGFRLGEE	3	56	94	4.0E-02	9.2E-04
YLK- -GALGFRLGEE	-2	49	79	2.7E-02	4.4E-03
YLKEAGGNGALGFRLGEE	3	48	73	2.2E-02	1.4E-02
YLKERGGNGALGFRLGEE	3	38	57	1.3E-02	3.5E-02
Exonuclease loop variants	InDel	R0 Count (n= 32402)	R1 Count (n= 4544)	EC	E-test
TKVE --RVWAYGYMN	-2	2329	406	2.4E+01	1.0E-01
TKVE DCRVWAYGYMN	0	259	111	3.0E+00	1.0E-06
TKVEARDCRVWAYGYMN	2	192	46	5.0E-01	1.6E-01
TKVEPGDCRVWAYGYMN	2	130	30	2.1E-01	3.3E-01
TKVERGDCRVWAYGYMN	2	224	38	1.9E-01	5.9E-01

Table 2. Most enriched variants isolated from individual libraries. Sequence, counts, complex enrichment score (EC) and statistical significance for the top 5 variants from each of the Phi29 DNAP Exonuclease, TRP2, and thumb InDel loop libraries. The EC is calculated as the difference in frequency between rounds, times the average number of counts. Wild-type sequences and their corresponding scores are shown in blue. Sequences with higher EC than the wild-type and significantly enriched (E-test p-value <0.05) are shown in red. Deletions are shown as hyphens and insertions are shown in regular font (i.e. not bold).

Deletion of P562 reduces processivity and fidelity

Although kinetic parameters of incorporation are often used to describe DNA polymerase activity and fidelity (when carried out with non-cognate substrates), they do not consider effects that emerge during elongation, which are known to be significant for XNA synthesis^31,38. For instance, hexitol nucleotide triphosphate (hNTP) incorporation kinetics is comparable to dNTP kinetics³⁹ in the B-family Thermococcus litoralis (Vent) DNA polymerase, yet the T. litoralis polymerase cannot incorporate more than 7 hNTPs in a row. Therefore, initial characterisation focused on confirming the increased activity of P562del for HNA synthesis by primer extension using concentration-normalised purified enzymes – confirming the results of selection and analysis. In short reactions (60 min extension), as shown in Figure 3, P562del indeed outperforms the starting D12A-THR polymerase, synthesising longer HNAs (~37 incorporations vs. ~18 from D12A-THR) and more efficiently (less visible un-extended primer). This difference in activity is observed at a range of protein concentrations (Figure 3E). Longer incubations (Figure S3A), however, allow D12A-THR to match and eventually outperform P562del – differences that cannot be seen with the natural nucleoside triphosphates. Notably, like the previously reported ΔTPR2 mutant²⁰, HNA synthesis by the P562del is distributive while D12A-THR is processive (with few visible accumulated intermediates between primer and full-length extension). Primer extension using alternative templates ruled out possible sequence-specific (Figure S3B-C).

Figure 3. P562 deletion in Phi29 DNAP enables faster initiation of HNA synthesis. Primer extension assays for HNA synthesis were carried out by incubating 10 pmol of single stranded DNA template pre-annealed to 1 pmol of fluorescently labelled DNA primer (100 nM in the reaction conditions) with 60 nM D12A-THR or P562del Phi29 DNAP over different amounts of time. (A) HNA primer extension products synthesized by D12A-THR and P562 mutants separated by denaturing PAGE. Un-extended primer (U) and extension products of ~37 incorporations (E) are highlighted. (B) Average depletion of un-extended primer (U) by D12A-THR (orange) and P562del (blue) normalized to the earliest time point (0 min). (C) Average extended products with ~37 incorporations (E) by D12A-THR and P562del normalized to the earliest time point (0 min). P562del shows a faster start to its hNTP incorporations than D12A-THR. (D) The effect of polymerase concentration on HNA primer extensions at a fixed time point of 15 min. Un-extended primer (U) and extension products of ~30 incorporations (E2) are highlighted. (E) Fraction of extended primer (E2) after 15 min incubations with each protein concentration used in (D). 3 biological replicates were carried out for (B-C) and 2 biological replicates were carried out for (E). In all cases, error bars show standard error of the mean.

Multiple polymerase mechanisms could account for the “early burst” phenotype observed. For instance, changes in binding affinity to the template (which begins as dsDNA but changes into an HNA/DNA heteroduplex as the polymerase extend the primer), changes in processivity (by increasing the dissociation rate of the enzyme from the template), or a drop in fidelity (misincorporations stall the polymerase and can also lead to more frequent sampling of the editing conformations). We therefore focused characterization on assays that could give some insight into those processes, always comparing P562del, to the parental enzymes, with (Exo+THR: Phi29 DNAP M8R, V51A, M97T, G197D, E221K, Q497P, K512E, F526L) or without (D12A-THR) active exonuclease function.

Figure 4. Phi29 DNAP P562del reduced DNA binding capacity. Commercial (NEB) Phi29 DNAP (Phi29(wt)), D12A-THR or P562del were allowed to bind a fluorescently labelled primer pre-annealed to a ssDNA template (A) EMSA reactions incubated for 0 min and 15 min with 13 nM of Phi29(wt), D12A-THR and P562del. (B) EMSA reactions incubated for 0 min and 15 min with 36 nM of D12A-THR and P562de and 18 nM of Phi29(wt) (capped at this concentration due to limit in reaction volume and commercial Phi29 DNA polymerase concentration). (C) Fraction of intermediate Pol-DNA complex (left), and the fraction of stable Pol-DNA complex by D12A-THR (orange) and P562del (blue) (right) over time. Two biological replicates were carried out and error bars represent the standard error of the mean between experiments. (D) Fraction of stable complex at the start of the reactions (0 min incubation - magenta) and after 15 min incubation (black) with different protein concentrations1, 13 and 36 nM of each mutant. These experiments were carried out only once.

Electrophoretic mobility shift assays (EMSAs) showed a significant difference between P562del and D12A-THR (Figure 4). The parental enzyme, as previously reported, shows complexes with two different mobilities: an unstable intermediate that transitions to a stable replication-competent complex²⁰. The deletion of P562 changes that equilibrium, with intermediate complexes visible for longer and at higher polymerase concentrations. This shows that the stability of the replication-competent complex is reduced. Reactions with longer incubation times were also performed and showed the same pattern (Figure S4).

Rolling circle amplification (RCA) confirmed the detrimental impact of D12A⁴⁰ and showed that deletion of P562 results in an enzyme unable to perform RCA (Figure 5A). Average extension lengths appear shorter than D12A-THR and no full-length product could be observed, supporting that P562 may be an important contact between thumb and TPR2 subdomains, and essential for TPR2’s role in strand displacement and high processivity. In the context of the previous primer extension and EMSA results, this supports a role for P562 in the dissociation rate of the polymerase from its DNA template.

DNA synthesis fidelity was investigated using an adaptation of our previously reported XNA fidelity assay, based on primer extension assays (Figure S3A). Given the very low error rates shown by wild-type Phi29 DNA polymerase (in the range between 3x10^-6 and 9.5x10^-6 errors per base ^41,42), errors introduced by PCR and deep sequencing steps significantly affect the data obtained. Deep sequencing errors alone are reported around 4.73 x10^-3 for MiSeq Illumina platforms⁴³.

Figure 5. Impaired rolling circle amplification (RCA) activity and reduced fidelity Phi29 DNAP P562del. (A) The RCA assay used 10 ng of plasmid template and 3 nM of each enzyme in a 3-hour reaction. A sample without enzyme (-ve control) was used to confirm the polymerase-dependent amplification. Restriction of the RCA yields full-length products. (B) Relative fidelity of the D12A-THR and p562del mutants to Exo+THR. The corrected total error rates from Table 3 of each mutant were divided by that of the Exo+THR, resulting in the relative fidelity scores.

We opted to use Exo+THR fidelity data to normalize for PCR and deep sequencing errors in the experiment, thus establishing a baseline against which the impact of D12A and P562 could be measured. As expected, disabling the exonuclease activity (D12A-THR) increased the rate of substitutions by approximately an order of magnitude, while showing marginal increases in deletions and insertions. Removal of P562 had a more pronounced effect, with substitution rates approximately 100-fold higher than Exo-THR and significant increases in deletion (7-fold) and insertion (6-fold) frequencies (Figure 5B and Table 3).

Mutant	Substitution ratea	Deletion ratea	Insertion ratea	Total error rate	Rb	Corrected total error rate	Total bases
Exo+THR	1.4x10-3	9.0x10-5	4.0x10-6	1.5x10-3	2.6x103	9.5x10-6	1,754,723
D12A-THR	1.4x10-3	1.7x10-4	6.1x10-6	1.6x10-3	2.7x103	1.5x10-4	1,808,769
P562del	1.8x10-3	6.1x10-4	2.3x10-5	2.5x10-3	2.7x103	9.9x10-4	1,834,463

Table 3. Quantification of insertions, deletions, and substitutions rates (errors/bp) of Phi29 DNAP mutants. ^aRaw error rates that do not consider PCR and NGS error rates. ^bR: PCR and Error estimate calculated from the number of errors expected to be introduced by the commercial Phi29 DNAP (using the published 9.5x10^-6 error rate ⁴¹) and subtracting this figure from the observed number of errors introduced by Exo+THR. The remainder was used to calculate the expected PCR and NGS error estimate for the total bases sequenced of the D12A-THR and p562del mutants.

Detailed analysis of error identity and position showed context-dependent mutational hot spots for the three tested enzymes but particularly salient for P562del (Figure S6-S7). P562del also showed elevated levels of transition substitutions (26.8%) compared to D12A-THR (21.2%) – shown in Figure S5C.

The phenotype shown by P562del bear similarities to the previously reported ΔTPR2, in that the variants form a less stable complex with their template, resulting in lower processivity and loss of strand displacement activity. That, however, is not sufficient to explain P562del enrichment in selection and its increased initial rate of HNA synthesis. Given the previous report of P562 interaction with TPR2²⁰, we postulate that P562 deletion removes a critical contact involved in regulating Phi29 polymerase binding to its template, making the process more dynamic – faster at assembling replication-competent complexes (explaining selection and primer extension), and faster at dissociating from DNA (explaining lower DNA binding stability and poor RCA performance). Dissociation from template becomes even more likely once it becomes an HNA/DNA heteroduplex, as it has been observed for other polymerases³⁹, explaining the “early burst” phenotype observed.

Discussion

DNA polymerases are essential for life and, logically, they have been extensively studied in the last 60 years. In addition to biochemical characterisation of natural enzymes, mutants – whether selected or targeted – have also significantly contributed to our current detailed mechanistic understanding of this family of enzymes¹².

Nonetheless, that understanding remains incomplete for polymerases. For instance, we still have no mechanistic understanding of how some natural DNA-dependent DNA polymerases display reverse transcriptase activity – such as the DNAP from Thermus thermophilus for RNA⁴⁴, or the DNAP from Geobacillus stearothermophilus for some XNAs⁴⁵. We do not know how to explain the different impact of a given modification between different polymerase chassis⁴⁶, and we still compare enzyme activities at standardized conditions, rather than at their individual reaction optimum⁴⁷. Such lack of understanding also explains the current generation of XNA polymerases, which lag severely behind natural DNA polymerases (using natural substrates) on processivity, fidelity and catalytic rates.

Numerous individual residues have been implicated in function, but they cannot act in isolation. How the different residues interact as a system to enable a polymerase to function remains elusive, with only a handful of examples demonstrating that such systems exist⁴⁸.

For less well-characterised enzymes, the prospect of biochemically characterising panels of mutants becomes increasingly difficult, and only accessible by technically challenging high-throughput screening platforms⁴⁹. Still, selection platforms have the potential to explore even higher throughputs (> 10⁷ variants) and at a lower technical entry barrier (e.g. emulsion or in vivo platforms).

Selection platforms are at the core of directed evolution, but here the emphasis is changed: rather than isolating a single best candidate, conditions are chosen to recover functional enzymes that perform best at the selected conditions. Consequently, there is greater need for robust statistical analysis of selection outputs, and for a more systematic exploration of how reaction conditions influence selection – areas that are increasingly being considered in directed evolution.

We implement an E-test to compare how individual sequences enrich between rounds, approximating individual sequence frequencies to Poisson distribution parameters. It assumes a degree of robustness in selection (so that frequencies can be good population estimates) that is only justifiable for heavily oversampled libraries – here, designed libraries were under 10⁴ variants, transformations recovered in excess of 10⁶ CFU and CST selections are carried out with over 10⁷ cells, fulfilling those criteria. Nonetheless, a natural progression would be to include stochastic measures that assume incomplete sampling, as commonly implemented for aptamer selection⁵⁰, and to extend analysis to include interacting residues^46,51.

Another crucial aspect of this approach is the library design. Substitutions are traditionally used to sample sequence space around wild-type enzymes, but they do not necessarily abolish backbone interactions. The P562 residue is not conserved among homologous Phi29-like enzymes, and a P562C mutation has been shown to have little functional impact in reducing conditions²⁷, suggesting that its interaction with TPR2 and exonuclease domains may be the result of backbone interactions. Indels, while abundant in nature, are usually less explored as a tool for protein engineering^52,53, because of their expected impact on protein structure and increased complexity in analysis.

While this work represents a proof-of-principle that selection can be used to gain biological insight into the workings of a well-characterised enzyme, it can be readily implemented to other enzymes and at larger scales, towards reconstructing the functional networks that define a protein function, such as obtained from statistical coupling analysis^54,55.

MATERIALS AND METHODS

Library design, selection, NGS and analysis leading to the identification of P562del. The exonuclease, TPR2 and thumb loop InDel libraries were generated through inverse PCR (iPCR) on the pET23-P2-D12A-THR (Table S1) with phosphorylated InDel mutagenic primers (Table S2) as previously described⁵⁶ and summarized in Figure 2B. Each InDel library was subjected to rounds of selection through CST³³ adapted for the mesophilic Phi29 DNAP as previously described⁵⁶. Briefly, 1x10⁸ cells from each library, post-induction (3 h at 30°C), were resuspended in 100μl of activity reaction mix, composed of 30 pmol of a short biotinylated oligo (CST_04(7)exoR, Table S2), 200 μM of hNTPs, 1x Phi29 reaction buffer (NEB), 1x Bovine Serum Albumin (BSA, Sigma Aldrich), 1 M betaine (Thermo Scientific), 2 μL NotI (NEB), 1 mg/mL lysozyme (Thermo Scientific) and 5 μg/mL polymyxin (Sigma-Aldrich) in 100 μL molecular grade water. Resuspended cells were emulsified in water-in-oil emulsions and subjected to freezing-thawing cycles for cell lysis and plasmid denaturation. HNA synthesis was then carried out for 10 min at 30°C and products were captured, post-emulsion disruption, via biotin-streptavidin pulldown using paramagnetic beads. Captured products were amplified through PCR using KOD Xtreme Hot Start DNA Polymerase (EMD Millipore), cloned into the pET23-P2-D12A-THR expression backbone, and transformed into fresh electrocompetent E. coli T7 Express cells (NEB). The libraries pre- and post-selection were then grown for 3 h at 37°C. Plasmids were then extracted and used as template in ~400 bp NGS amplicon generation PCR reactions using KOD Xtreme Hot Start DNA Polymerase (EMD Millipore) and NGS amplicon generation primers unique for each library (Table S2). The amplicons were gel-extracted using the Monarch DNA Gel Extraction kits (NEB). NGS-based amplicon sequencing was carried out for all libraries at Genewiz UK Ltd using the Amplicon-EZ service. Sequencing data, which can be found in the NCBI SRA database (BioProject: PRJNA883233), was pre-processed using the NGS_preprocessing.ga script (Supplementary Information S2) in the Galaxy⁵⁷ public server (usegalaxy.org). Total number of reads and the impact of the pre-processing workflow can be found on Tables S3-S5. Total number of reads and the impact of the pre-processing can be found on Tables S3-S5. InDel frequencies before and after selection, enrichment complex scores and the E-test for comparing two Poisson means³⁷ were calculated using the InDel_Quantification.ipynb Julia notebook (Supplementary Information S3). The complex enrichment score was calculated as:

$\left( {frequency}{R1} - \ {frequency}{R0} \right)\ \times \ \frac{({counts}{R1} + {counts}{R0})\ }{2}$ (1)

The calculated scores can be found in the InDel_EC.xlsx file (Supplementary Information S4).

Generation of Phi29 mutants. The P562del and Exo+THR mutants were generated on the pET23-P2-D12A-THR (Table S1) background through inverse Polymerase Chain Reactions (iPCR) using Q5® High-Fidelity DNA Polymerase (NEB) following the manufacturer’s instructions in 25 µl reactions with primers p2_thumb_loop_R/DEL1 (Table S2) and iPCR_P2_Exo+_F1/R1 (Table S2), respectively. Amplified products were treated with 0.4U/µL DpnI for 1 h at 37°C prior to the PCR purification with the GeneJET PCR Purification Kit (Thermo Scientific). 100 ng of DNA products were blunt-end ligated in 20 µL reactions with 40U/µL T4 DNA ligase in 1x T4 DNA ligase buffer overnight at room temperature. 5 µL of the ligation products were transformed in NEB® 5-alpha competent E. coli cells following the recommended High Efficiency Transformation Protocol (C2987, New England Biolabs) described by the commercial strain provider. Successful cloning was confirmed through Sanger Sequencing in all cases.

Expression, purification and protein quantification of mutants. The pET23 plasmid encoding the Phi29 DNA polymerase mutants were transformed into E. coli T7 Express cells (NEB). Isolated transformants were grown at 37°C until an OD₆₀₀ of 0.8 was reached. Protein expression was induced by adding IPTG to a final concentration of 1 mM and incubating cultures for 4 h at 30°C. Cells were then pelleted through centrifugation at 5000xg and the supernatant was discarded. Pellets were frozen for an hour at -20°C and resuspended in 4 mL of B-PER Reagent (Thermo Scientific) per gram of cell pellet with 50 mg/mL lysozyme and 250U/uL Benzoase (Sigma Aldrich NV). Cells were subsequently incubated at room temperature for 15 min. Protease inhibitor (1 mM Pefabloc) and reducing agents (1 mM DTT) were added, and cell debris removed through centrifugation at 5000xg. The supernatant was diluted 3-fold in 50 mM Tris-HCl pH 8.0, 1 mM DTT and 1 mM Pefabloc. Expressed protein was purified using the HisPur™ Ni-NTA Resin (Thermo Scientific) using 50 mM Tris-HCl pH 8.0, 1 M NaCl, 1 mM DTT and increasing concentrations of imidazole following the manufacturer’s recommendations. Eluted proteins were concentrated, and buffer-exchanged to 10 mM Tris-HCl, 100 mM KCl, 1 mM DTT, 0.1 mM EDTA, pH 7.4 at 25°C using Amicon Ultra-4 Centrifugal 30 kDa Filter Units (Sigma Aldrich NV). Glycerol (final concentration: 50% v/v), Tween 20 (0.5%), Nonidet P40 (0.5%) were then added for long term storage at -20°C.

To quantify protein concentration, a series of Bovine Serum Albumin (BSA) standards in the 100 μg/ml to 1,200 μg/ml range were run in parallel to the concentrated proteins in an SDS-PAGE gel to establish a BSA standard curve used to determine the concentration of the mutants and commercial Phi29 DNA polymerase.

Primer Extension Assay for HNA synthesis. Reactions were prepared by mixing in 10 µL: 1x NEB 10X Phi29 DNAP buffer, 0.1 mg BSA, 1 M Betaine, 0.2 mM hNTPs, 0.06 μM (unless stated otherwise) of each Phi29 DNAP mutant and 10 pmol of single stranded DNA template, TempN-exoR (Table S2) pre-annealed to 1 pmol of fluorescently labelled DNA primer, Tag01F3-exoR (Table S2). Pre-annealing was carried out by incubating primer and template at 95°C for 5 min and cooling them down to 4°C at a rate of 0.1°C/sec in a thermocycler. The primer extension assays were incubated at 30°C from 1 min to 16 h (depending on the experiment) but always inactivated by incubating it at 65°C for 20 min. Samples were diluted in an equal volume of 2x loading buffer (98% (v/v) formamide, 10 mM ethylenediaminetetraacetic acid (EDTA), 0.02% (w/v) Orange G) and boiled at 95°C for 5 min. Samples were loaded on 20% TBE-Urea polyacrylamide gels and run in 1X TBE buffer at 25 W for 2 hours. Gels were imaged using a Typhoon^TM FLA 9500 biomolecular imager (GE Healthcare) using recommended filter and detector settings based on the fluorophore being used and analysed with ImageJ⁵⁸. Three independent experiments (biological replicates, n = 3) were carried out for the quantification of the primer extension assays.

DNA Binding Assay – Electrophoretic mobility shift assay (EMSA). The EMSA protocol was adapted from Povilaitis et al.²⁴. Oligonucleotide Tag01F_ExoR (Table S2) was annealed to TemN-ExoR (Table S2) at a 1:1.2 molar ratio in the presence of 0.2 M NaCl and 60 mM Tris–HCl, pH 7.5 by incubating at 95°C for 5 min and cooling down to 4°C at a rate of 0.1°C/sec in a thermocycler. A 50 μL complex-forming reaction was prepared by mixing 33 mM Tris-acetate pH 7.9, 66 mM potassium acetate, 10% glycerol, 0.1 mg/ml BSA with the 2 pmol pre-annealed primer/template substrate (40 nM in the reaction conditions) and 1 ~ 60 nM of each corresponding enzyme. Reactions were incubated at 30°C for different time intervals (0, 15, 30 or 60 min). After incubation, samples were mixed with loading dye (10% glycerol and 0.03% orange G) and analysed by electrophoresis in 0.8% (w/v) agarose gel in 0.25x TBE buffer (22.25 mM Tris, 22.25 mM boric acid, and 0.5 mM EDTA pH 8.3). Electrophoresis was performed in 0.25x TBE buffer at room temperature for 2 h at 100 V. Gels were imaged using the Typhoon^TM bioimager and quantified using ImageQuant TL 8.2 (Cytiva). Four independent experiments (biological replicates, n = 2) were carried out for the quantification of the EMSA assays.

Rolling Circle Amplification (RCA). 100 pmol of P2_RCA_N8_ExoR (5’-NNNNNN*N*N-3’, Table S2) with two phosphorothioate-linked DNA bases at the 3’ end for exonuclease resistance was pre-annealed to 10 ng of pET23_P2_D12A_THR (Table S1) plasmid in 1x NEB 10X Phi29 DNAP buffer, per 100 µL RCA reaction, by incubating them at 95°C for 5 min and cooling them down to 4°C at a rate of 0.1°C/sec in a thermocycler. The pre-annealed template mix was supplemented with 0.2 mM dNTPs, 0.2 mg/mL BSA, 5 ng ETSSB (NEB), 4 nM of each Phi29 DNAP mutant in a 100 µL reaction and incubated for 3 h at 30°C. The RCA products were digested with 10U of NotI in 1x Cutsmart buffer for 1 h at 37°C. The digested RCA products were loaded on an agarose gel for visualisation.

Isothermal polymerase fidelity assay. Single-stranded template was generated by mixing 1.8 pmol of pET23_KOD_DA_Mut (Table S1) with 0.7U/μL Nb.BbvCI, 10U/μL of ExoIII and 1x Cutsmart buffer in a 30 μL reaction. 10 pmol of PH_pET23_DA_Biotin3 (Table S2) biotinylated oligonucleotide (4.9 μL) was pre-annealed to 0.2 pmol of single stranded template by incubating the mixture at 95°C for 5 min and cooling them down to 4°C at a rate of 0.1°C/sec in a thermocycler. The annealed primer-template was made up to 50 μL with deionized distilled water (ddH₂O). 5 μL of Dyanabeads MyOne C1 (ThermoFisher), per reaction, were washed twice with 100 μL of 2x BWBS-T (20 mM Tris-HCl pH 7.4, 2 M NaCl, 0.2% v/v Tween20, 2 mM EDTA) and blocked in 1 mL 2x BWBS-T on a rotator for 1 h at room temperature. Beads were captured, resuspended in 50 μL 2x BWBS-T, and mixed with the 50 μL reaction prior to being incubated for 3 h on a rotator at room temperature. Beads were washed twice with 200 μL 30 mM NaOH at 37°C, then with 200 μl EB-T (10 mM Tris-HCl pH 8.8, 0.1 mM EDTA, 0.01% Tween20). Beads were resuspended in 10 μL elution buffer (10 mM Tris-HCl, pH 8.5). Amplicons for deep sequencing were generated using KOD Xtreme™ Hot Start DNA Polymerase (EMD Millipore) following the manufacturer’s instructions in 50 μL reactions with 1 µL of the resuspended beads as template, and outnest_1 (Table S2) and P2_fidelity_inestR1 (Table S2) as primers. The cycling parameters used were 2 min at 95˚C, followed by 20 cycles of 15 s at 98˚C, 15 s at 65˚C and 10 s at 68˚C, with a final polishing step of 2 min at 72˚C. Primers were degraded by adding 0.8U/uL Exonuclease I to each 50 μL PCR reaction.

Next Generation Sequencing of fidelity assay data and analysis. The amplicons described above were purified with the Monarch PCR and DNA purification kit (New England Biolabs) and sent for NGS EZ Amplicon Sequencing by Genewiz. NGS data was pre-processed in the Galaxy public server (usegalaxy.org) using the NGS_preprocessing.ga script (Supplementary information S2) and trimmed using part 1 of the Julia (v1.7) Fidelity_Quantification.jl script (Supplementary Information S5). The reference sequence, ‘Fidelity_ref’ (Table S2), was added to the trimmed reads and the reads were then aligned using the random chain algorithm of MAFFT version 7⁵⁹ on the MAFFT server (mafft.cbrc.jp/alignment/server/large.html) with the following parameters: mafft --thread 8 --threadtb 5 --threadit 0 --inputorder --randomchain input > output. The number of insertions, deletions, and substitutions as well as the number of individual error types were calculated by comparing in a per-base manner the aligned reads to the reference ‘Fidelity_ref’ (Table S2) sequence within the alignment using part 2 of the Fidelity_Quantification.jl script (Supplementary Information S5). Raw error rates were obtained by dividing the sum of the errors by the total number of bases sequenced. The PCR and NGS error (R) estimate for Exo+THR was obtained by subtracting the expected number of errors of the commercial Phi29 DNAP (using the published error rate 9.5x10^-6) from the observed number of errors by Exo+THR as follows:

$$R_{Exo + THR}\ = {(Error\ rate}{Exo + THR} \times {Total\ bases}{Exo + THR)} - ({Error\ rate}{phi29} \times {Total\ bases}{Exo + THR}) $$(2)

The ${Err}_{Exo + THR}$ score was then used to calculate the PCR and NGS error of the D12A-THR and p562del mutants as follows:

$R_{mutant}\ = \frac{R_{Exo + THR}}{{Total\ bases}{Exo + THR} \times \ {Total\ bases}{mutant}}$ (3)

The corrected total error rates were then calculated by dividing the corrected error number over the total bases sequenced from each variant.

The location (start position, end position and length) and frequency of each Indel across the MSA were obtained using part2 of the Fidelity_Quantification.jl script (Supplementary Information S5), by comparing each sequence to the ‘Fidelity_ref’ sequence within the alignment. The frequency values were divided by the total number of insertions or deletions and multiplied by 100 to obtain a relative frequency distribution. Only sequences >5% relative frequency were plotted.

AVAILABILITY

All raw and processed data used in this manuscript are available in our GitHub repository (https://github.com/PinheiroLab/ ).

ACCESSION NUMBERS

Next generation sequencing data has been deposited on NCBI SRA under the following accession number: PRJNA883233.

ACKNOWLEDGEMENTS

VBP, LLT and PHM thank ERASynBio (grant BB/N01023X/1; invivoXNA). VBP and PHM thank FWO (grant G0H7618N). PHM thanks FWO (studentship 3M180645).

CONTRIBUTIONS

Paola Handal-Marquez: conceptualization, methodology, investigation, formal analysis, data curation, writing, visualization and funding acquisition. Leticia L. Torres: conceptualization, methodology, supervision. Vitor B. Pinheiro: conceptualization, methodology, formal analysis, writing, visualization, supervision, project administration and funding acquisition.

**
**

REFERENCES

(1) Garmendia, C.; Bernad, A.; Esteban, J. A.; Blanco, L.; Salas, M. The Bacteriophage Phi 29 DNA Polymerase, a Proofreading Enzyme. Journal of Biological Chemistry 1992, 267 (4), 2594–2599. https://doi.org/10.1016/S0021-9258(18)45922-4.

(2) Waters, L. S.; Minesinger, B. K.; Wiltrout, M. E.; D’Souza, S.; Woodruff, R. V.; Walker, G. C. Eukaryotic Translesion Polymerases and Their Roles and Regulation in DNA Damage Tolerance. Microbiology and Molecular Biology Reviews 2009, 73 (1), 134–154. https://doi.org/10.1128/MMBR.00034-08/ASSET/7C1A9D97-7040-4080-8EF7-1F7B91B23BD8/ASSETS/GRAPHIC/ZMR0010922090004.JPEG.

(3) Wang, W.; Wu, E. Y.; Hellinga, H. W.; Beese, L. S. Structural Factors That Determine Selectivity of a High Fidelity DNA Polymerase for Deoxy-, Dideoxy-, and Ribonucleotides. Journal of Biological Chemistry 2012, 287 (34), 28215–28226. https://doi.org/10.1074/jbc.M112.366609.

(4) Tsai, Y. C.; Johnson, K. A. A New Paradigm for DNA Polymerase Specificity. Biochemistry 2006, 45 (32), 9675–9687. https://doi.org/10.1021/bi060993z.

(5) Khare, V.; Eckert, K. A. The Proofreading 3′ → 5′ Exonuclease Activity of DNA Polymerases: A Kinetic Barrier to Translesion DNA Synthesis. Mutation Research - Fundamental and Molecular Mechanisms of Mutagenesis 2002, 510 (1–2), 45–54. https://doi.org/10.1016/S0027-5107(02)00251-8.

(6) Freund, N.; Taylor, A. I.; Arangundy-Franklin, S.; Subramanian, N.; Peak-Chew, S. Y.; Whitaker, A. M.; Freudenthal, B. D.; Abramov, M.; Herdewijn, P.; Holliger, P. A Two-Residue Nascent-Strand Steric Gate Controls Synthesis of 2′-O-Methyl- and 2′-O-(2-Methoxyethyl)-RNA. Nat Chem 2022. https://doi.org/10.1038/s41557-022-01050-8.

(7) Frey, M. W.; Nossal, N. G.; Capson, T. L.; Benkovic, S. J. Construction and Characterization of a Bacteriophage T4 DNA Polymerase Deficient in 3’ → 5’ Exonuclease Activity. Proc Natl Acad Sci U S A 1993, 90 (7), 2579–2583. https://doi.org/10.1073/PNAS.90.7.2579.

(8) Bebenek, A.; Dressman, H. K.; Carver, G. T.; Ng, S. S.; Petrov, V.; Yang, G.; Konigsberg, W. H.; Karam, J. D.; Drake, J. W. Interacting Fidelity Defects in the Replicative DNA Polymerase of Bacteriophage RB69. Journal of Biological Chemistry 2001, 276 (13), 10387–10397. https://doi.org/10.1074/JBC.M007707200.

(9) Darmawan, H.; Harrison, M.; Reha-Krantz, L. J. DNA Polymerase 3’→5’ Exonuclease Activity: Different Roles of the Beta Hairpin Structure in Family-B DNA Polymerases. DNA Repair (Amst) 2015, 29, 36–46. https://doi.org/10.1016/j.dnarep.2015.02.014.

(10) Fowler, D. M.; Fields, S. Deep Mutational Scanning : A New Style of Protein Science. 2014, 11 (8), 801–807. https://doi.org/10.1038/nmeth.3027.

(11) Laos, R.; Thomson, J. M.; Benner, S. A. DNA Polymerases Engineered by Directed Evolution to Incorporate Nonstandard Nucleotides. Front Microbiol 2014, 5 (OCT), 1–14. https://doi.org/10.3389/fmicb.2014.00565.

(12) Pinheiro, V. B. Engineering-Driven Biological Insights into DNA Polymerase Mechanism. Curr Opin Biotechnol 2019, 60, 9–16. https://doi.org/10.1016/j.copbio.2018.11.008.

(13) Pinheiro, V. B.; Holliger, P. Towards XNA Nanotechnology: New Materials from Synthetic Genetic Polymers. Trends Biotechnol 2014, 32 (6), 321–328. https://doi.org/10.1016/j.tibtech.2014.03.010.

(14) Pinheiro, V. B.; Taylor, A. I.; Cozens, C.; Abramov, M.; Renders, M.; Zhang, S.; Chaput, J. C.; Wengel, J.; Peak-Chew, S.-Y.; McLaughlin, S. H.; Herdewijn, P.; Holliger, P. SI Synthetic Genetic Polymers Capable of Heredity and Evolution. Science (1979) 2012, 336 (6079), 341–344. https://doi.org/10.1126/science.1217622.

(15) Torres, L.; Krüger, A.; Csibra, E.; Gianni, E.; Pinheiro, V. B. Synthetic Biology Approaches to Biological Containment: Pre-Emptively Tackling Potential Risks. Essays Biochem 2016, 60 (4), 393–410. https://doi.org/10.1042/EBC20160013.

(16) Pinheiro, V. B.; Taylor, A. I.; Cozens, C.; Abramov, M.; Renders, M.; Zhang, S.; Chaput, J. C.; Wengel, J.; Peak-Chew, S. Y.; McLaughlin, S. H.; Herdewijn, P.; Holliger, P. Synthetic Genetic Polymers Capable of Heredity and Evolution. Science (1979) 2012, 336 (6079), 341–344. https://doi.org/10.1126/science.1217622.

(17) Torres, L. L.; Pinheiro, V. B. Xenobiotic Nucleic Acid (XNA) Synthesis by Phi29 DNA Polymerase. Curr Protoc Chem Biol 2018, 10 (2), e41. https://doi.org/10.1002/cpch.41.

(18) Cozens, C.; Pinheiro, V. B.; Vaisman, A.; Woodgate, R.; Holliger, P. A Short Adaptive Path from DNA to RNA Polymerases. Proceedings of the National Academy of Sciences 2012, 109 (21), 8067–8072. https://doi.org/10.1073/pnas.1120964109.

(19) Lieberman, K. R.; Cherf, G. M.; Doody, M. J.; Olasagasti, F.; Kolodji, Y.; Akeson, M. Processive Replication of Single DNA Molecules in a Nanopore Catalyzed by Phi29 DNA Polymerase. J Am Chem Soc 2010, 132 (50), 17961–17972. https://doi.org/10.1021/ja1087612.

(20) Rodriguez, I.; Lazaro, J. M.; Blanco, L.; Kamtekar, S.; Berman, A. J.; Wang, J.; Steitz, T. A.; Salas, M.; Vega, M. de. A Specific Subdomain in Φ29 DNA Polymerase Confers Both Processivity and Strand-Displacement Capacity. Proceedings of the National Academy of Sciences 2005, 102 (18), 6407–6412.

(21) Tran, N. Q.; Lee, S. J.; Akabayov, B.; Johnson, D. E.; Richardson, C. C. Thioredoxin, the Processivity Factor, Sequesters an Exposed Cysteine in the Thumb Domain of Bacteriophage T7 DNA Polymerase. Journal of Biological Chemistry 2012, 287 (47), 39732–39741. https://doi.org/10.1074/jbc.M112.409235.

(22) Miton, C. M.; Tokuriki, N. Insertions and Deletions (Indels): A Missing Piece of the Protein Engineering Jigsaw. Biochemistry 2023, 62 (2), 148–157. https://doi.org/10.1021/ACS.BIOCHEM.2C00188/ASSET/IMAGES/LARGE/BI2C00188_0003.JPEG.

(23) Vanmeert, M.; Razzokov, J.; Mirza, M. U.; Weeks, S. D.; Schepers, G.; Bogaerts, A.; Rozenski, J.; Froeyen, M.; Herdewijn, P.; Pinheiro, V. B.; Lescrinier, E. Rational Design of an XNA Ligase through Docking of Unbound Nucleic Acids to Toroidal Proteins. Nucleic Acids Res 2019, 47 (13), 7130–7142. https://doi.org/10.1093/nar/gkz551.

(24) Povilaitis, T.; Alzbutas, G.; Sukackaite, R.; Siurkus, J.; Skirgaila, R. In Vitro Evolution of Phi29 DNA Polymerase Using Isothermal Compartmentalized Self Replication Technique. Protein Engineering, Design and Selection 2016, 29 (12), 617–628. https://doi.org/10.1093/protein/gzw052.

(25) Lutz, S. Beyond Directed Evolution - Semi-Rational Protein Engineering and Design. 2011, 21 (6), 734–743. https://doi.org/10.1016/j.copbio.2010.08.011.Beyond.

(26) Chica, R. A.; Doucet, N.; Pelletier, J. N. Semi-Rational Approaches to Engineering Enzyme Activity : Combining the Benefits of Directed Evolution and Rational Design. Curr Opin Biotechnol 2005, No. 16, 378–384. https://doi.org/10.1016/j.copbio.2005.06.004.

(27) Rodríguez, I.; Lázaro, J. M.; Salas, M.; de Vega, M. Involvement of the TPR2 Subdomain Movement in the Activities of Φ29 DNA Polymerase. Nucleic Acids Res 2009, 37 (1), 193–203. https://doi.org/10.1093/nar/gkn928.

(28) Kamtekar, S.; Berman, A. J.; Wang, J.; Lázaro, J. M.; de Vega, M.; Blanco, L.; Salas, M.; Steitz, T. A. Insights into Strand Displacement and Processivity from the Crystal Structure of the Protein-Primed DNA Polymerase of Bacteriophage Φ29. Mol Cell 2004, 16 (4), 609–618. https://doi.org/10.1016/j.molcel.2004.10.019.

(29) Truniger, V.; Lazaro, J. M.; Salas, M. Function of the C-Terminus of Phi29 DNA Polymerase in DNA and Terminal Protein Binding. Nucleic Acid Ther 2004, 32 (1), 361–370. https://doi.org/10.1093/nar/gkh184.

(30) Patel, P. H.; Loeb, L. A. Multiple Amino Acid Substitutions Allow DNA Polymerases to Synthesize RNA. Journal of Biological Chemistry 2000, 275 (51), 40266–40272. https://doi.org/10.1074/jbc.M005757200.

(31) Xia, G.; Chen, L.; Sera, T.; Fa, M.; Schultz, P. G.; Romesberg, F. E. Directed Evolution of Novel Polymerase Activities: Mutation of a DNA Polymerase into an Efficient RNA Polymerase. Proc Natl Acad Sci U S A 2002, 99 (10), 6597–6602. https://doi.org/10.1073/pnas.102577799.

(32) Ghadessy, F. J.; Ong, J. L.; Holliger, P. Directed Evolution of Polymerase Function by Compartmentalized Self-Replication. Proceedings of the National Academy of Sciences 2001, 98 (8), 4552–4557. https://doi.org/10.1073/pnas.071052198.

(33) Pinheiro, V. B.; Arangundy-franklin, S.; Holliger, P. Compartmentalized Self-Tagging for In Vitro-Directed Evolution of XNA Polymerases. Curr Protoc Nucleic Acid Chem 2014, 57, 9.9.1-9.9.18. https://doi.org/10.1002/0471142700.nc0909s57.

(34) Houlihan, G.; Arangundy-Franklin, S.; Porebski, B. T.; Subramanian, N.; Taylor, A. I.; Holliger, P. Discovery and Evolution of RNA and XNA Reverse Transcriptase Function and Fidelity. Nat Chem 2020, 12 (August), 1–8. https://doi.org/10.1038/s41557-020-0502-8.

(35) Lázaro, J. M.; Blanco, L.; Salas, M. Purification of Bacteriophage Φ29 DNA Polymerase. Methods Enzymol 1995, 262 (C), 42–49. https://doi.org/10.1016/0076-6879(95)62007-9.

(36) Wrenbeck, E. E.; Faber, M. S.; Whitehead, T. A. Deep Sequencing Methods for Protein Engineering and Design. Curr Opin Struct Biol 2017, 45, 36–44. https://doi.org/10.1016/j.sbi.2016.11.001.

(37) Krishnamoorthy, K.; Thomson, J. A More Powerful Test for Comparing Two Poisson Means. J Stat Plan Inference 2004, 119 (1), 23–35. https://doi.org/10.1016/S0378-3758(02)00408-1.

(38) Lutz, S.; Burgstaller, P.; Benner, S. A. An in Vitro Screening Technique for DNA Polymerases That Can Incorporate Modified Nucleotides. Pseudothymidine as a Substrate for Thermostable Polymerases. Nucleic Acids Res 1999, 27 (13), 2792–2798. https://doi.org/10.1093/nar/27.13.2792.

(39) Vastmans, K.; Pochet, S.; Peys, A.; Kerremans, L.; van Aerschot, A.; Hendrix, C.; Marliere, P.; Herdewijn, P. Enzymatic Incorporation in DNA of 1,5-Anhydrohexitol Nucleotides. Biochemistry 2000, 39 (42), 12757–12765. https://doi.org/10.1021/bi001297g.

(40) Bernad, A.; Blanco, L.; Lázaro, J. M.; Martín, G.; Salas, M. A Conserved 3′→5′ Exonuclease Active Site in Prokaryotic and Eukaryotic DNA Polymerases. Cell 1989, 59 (1), 219–228. https://doi.org/10.1016/0092-8674(89)90883-0.

(41) Paez, J. G.; Lin, M.; Beroukhim, R.; Lee, J. C.; Zhao, X.; Richter, D. J.; Gabriel, S.; Herman, P.; Sasaki, H.; Altshuler, D.; Li, C.; Meyerson, M.; Sellers, W. R. Genome Coverage and Sequence Fidelity of Phi29 Polymerase-Based Multiple Strand Displacement Whole Genome Amplification. Nucleic Acids Res 2004, 32 (9), 1–11. https://doi.org/10.1093/nar/gnh069.

(42) Nelson, J. R.; Cai, Y. C.; Giesler, T. L.; Farchaus, J. W.; Sundaram, S. T.; Ortiz-Rivera, M.; Hosta, L. P.; Hewitt, P. L.; Mamone, J. A.; Palaniappan, C.; Fuller, C. W. TempliPhi, ⌽29 DNA Polymerase Based Rolling Circle Amplification of Templates for DNA Sequencing. https://doi.org/10.2144/jun0208 2018, 32 (6 SUPPL.). https://doi.org/10.2144/JUN0208.

(43) Stoler, N.; Nekrutenko, A. Sequencing Error Profiles of Illumina Sequencing Instruments. NAR Genom Bioinform 2021, 3 (1), 1–9. https://doi.org/10.1093/NARGAB/LQAB019.

(44) Myers, T. W.; Gelfand, D. H. Reverse Transcription and DNA Amplification by a Thermus Thermophilus DNA Polymerase. Biochemistry 1991, 30 (31), 7661–7666. https://doi.org/10.1021/bi00245a001.

(45) Jackson, L. N.; Chim, N.; Shi, C.; Chaput, J. C. Crystal Structures of a Natural DNA Polymerase That Functions as an XNA Reverse Transcriptase. Nucleic Acids Res 2019, 47 (13), 6973–6983. https://doi.org/10.1093/nar/gkz513.

(46) Nikoomanzar, A.; Vallejo, D.; Chaput, J. C. Elucidating the Determinants of Polymerase Specificity by Microfluidic-Based Deep Mutational Scanning. ACS Synth Biol 2019, 8 (6), 1421–1429. https://doi.org/10.1021/acssynbio.9b00104.

(47) Nikoomanzar, A.; Dunn, M. R.; Chaput, J. C. Evaluating the Rate and Substrate Specificity of Laboratory Evolved XNA Polymerases. Anal Chem 2017, 89 (23), 12622–12625. https://doi.org/10.1021/acs.analchem.7b03807.

(48) Li, V.; Hogg, M.; J, R.-K. L. Identification of a New Motif in Family B DNA Polymerases by Mutational Analyses of the Bacteriophage T4 DNA Polymerase. J Mol Biol 2010, 400 (3), 295–308. https://doi.org/10.1016/j.jmb.2010.05.030.Identification.

(49) Gielen, F.; Hours, R.; Emond, S.; Fischlechner, M.; Schell, U.; Hollfelder, F. Ultrahigh-Throughput-Directed Enzyme Evolution by Absorbance-Activated Droplet Sorting (AADS). Proc Natl Acad Sci U S A 2016, 113 (47), E7383–E7389. https://doi.org/10.1073/pnas.1606927113.

(50) Hoinka, J.; Berezhnoy, A.; Dao, P.; Sauna, Z. E.; Gilboa, E.; Przytycka, T. M. Large Scale Analysis of the Mutational Landscape in HT-SELEX Improves Aptamer Discovery. Nucleic Acids Res 2015, 43 (12), 5699–5707. https://doi.org/10.1093/nar/gkv308.

(51) Salinas, V. H.; Ranganathan, R. Coevolution-Based Inference of Amino Acid Interactions Underlying Protein Function. Elife 2018, 7, 1–20.

(52) Tóth-Petróczy, Á.; Tawfik, D. S. Protein Insertions and Deletions Enabled by Neutral Roaming in Sequence Space. Mol Biol Evol 2013, 30 (4), 761–771. https://doi.org/10.1093/molbev/mst003.

(53) Tizei, P. A. G.; Harris, E.; Withanage, S.; Renders, M.; Pinheiro, V. B. A Novel Framework for Engineering Protein Loops Exploring Length and Compositional Variation. Sci Rep 2021, 11 (1), 1–13. https://doi.org/10.1038/s41598-021-88708-4.

(54) Lockless, S. W.; Ranganathan, R. Evolutionarily Conserved Pathways of Energetic Connectivity in Protein Families. Science (1979) 1999, 286 (5438), 295–299. https://doi.org/10.1126/SCIENCE.286.5438.295.

(55) Salinas, V. H.; Ranganathan, R. Coevolution-Based Inference of Amino Acid Interactions Underlying Protein Function. Elife 2018, 7. https://doi.org/10.7554/ELIFE.34300.

(56) Handal Marquez, P. Sampling the Functional Sequence Neighbourhood of Phi29 DNA Polymerase for XNA Synthesis. (MPhil) UCL (University College London). 2019.

(57) Afgan, E.; Baker, D.; van den Beek, M.; Blankenberg, D.; Bouvier, D.; Čech, M.; Chilton, J.; Clements, D.; Coraor, N.; Eberhard, C.; Grüning, B.; Guerler, A.; Hillman-Jackson, J.; Kuster, G. von; Rasche, E.; Soranzo, N.; Turaga, N.; Taylor, J.; Nekrutenko, A.; Goecks, J. The Galaxy Platform for Accessible, Reproducible and Collaborative Biomedical Analyses: 2016 Update. Nucleic Acids Res 2016, 44 (W1), W3–W10. https://doi.org/10.1093/nar/gkw343.

(58) Schneider, C. A.; Rasband, W. S.; Eliceiri, K. W. NIH Image to ImageJ: 25 Years of Image Analysis. Nat Methods 2012, 9 (7), 671–675. https://doi.org/10.1038/nmeth.2089.

(59) Katoh, K.; Standley, D. M. MAFFT Multiple Sequence Alignment Software Version 7: Improvements in Performance and Usability. Mol Biol Evol 2013, 30 (4), 772–780. https://doi.org/10.1093/molbev/mst010.