Engineering and Assay of Enzymes for Expanded Genetic Alphabet Replication

Chemicals and Reagents.

All reagents were of analytical grade and obtained from Sigma-Aldrich (St. Louis, MO, USA) or New England Biolabs (Ipswich, MA, USA) unless otherwise indicated. Synthetic oligonucleotides (Table S1) were custom synthesized by Integrated DNA Technologies (Coralville, IA, USA) or Firebird Biomolecular Sciences (Alachua, FL, USA) and resuspended in nuclease-free TE buffer (10 mM Tris (pH 8.0) and 0.1 mM EDTA pH 8.0). Stock solutions of oligonucleotide strand displacement (OSD) probes contained 1 μM fluorophore-labeled strand and 3 μM quencher-labeled strand in 1× isothermal buffer (20 mM Tris–HCl, 10 mM (NH₄)₂SO₄, 50 mM KCl, 2 mM MgSO₄, 0.1% Tween 20, pH 8.8 at 25 °C) from New England Biolabs (NEB; Ipswich, MA, USA), annealed by denaturing at 95 °C for 1 min and slow cooling at 0.1 °C/s to 25 °C. All OSD concentrations in the experiments are reported as the concentration of the fluorophore-labeled long strand preannealed with a 3-fold molar excess of the quencher-labeled complementary short strand. All oligonucleotides were stored at −20 °C. DNA polymerase coding sequences (Table S2) for Klenow (UniProt P00582), DNA Pol II (UniProt P21189), and DNA Pol III (UniProt P10443) cloned between the NdeI and XhoI restriction sites with N-terminal 6× histidine tag and expressed from the T7 RNA polymerase promoter in the pET28 plasmid were obtained from Twist Biosciences (South San Francisco, CA, USA) or created by site-directed mutagenesis using the Q5 Site-Directed Mutagenesis Kit (NEB). N-Terminal 6×-His-tagged Bst-LF (accession ID: 1L3S_A) or KlenTaq v5.9 DNA polymerases were expressed from the previously reported plasmids pAtetO_6xHis-Bst-LF (Addgene # 145799) and pAtetO_6xHis-v5.9 (Addgene #155259).^39,40

DNA Polymerase Purification.

DNA polymerase expression constructs were transformed into chemically competent E. coli BL21 or BL21(DE3) (NEB) and cultured overnight in 3 mL superior broth (Athena Enzyme Systems) containing 50 μg/mL kanamycin (pET28 plasmids) or 100 μg/mL carbenicillin (pAtetO_6xHis-Bst-LF and pAtetO_6x-His-KlenTaq v5.9) by incubating in a shaker maintained at 30 °C (DNA Pol II and IIIα) or 37 °C (KlenTaq v5.9, Klenow and Bst variants). Subcultures were prepared in 0.5 or 1 L flasks by inoculating 0.5 mL (KlenTaq v5.9, Klenow and Bst variants) or 3 mL (DNA Pol II and IIIα) of the overnight cultures into 100 or 200 mL, respectively, of fresh superior broth containing antibiotics. Cultures were grown in a shaker at 30 °C (DNA Pol II and IIIα) or 37 °C (KlenTaq v5.9, Klenow and Bst variants) until they reached an OD₆₀₀ of 0.45 to 0.8. Enzyme expression was induced either overnight (18 to 20 h) in an 18 °C shaker with 1 mM (KlenTaq v5.9 variants, Klenow variants, and Bst-SVV) or 0.05 mM (DNA Pol II and IIIα) IPTG or for 4 h at 37 °C with 200 ng/mL anhydrotetracycline (Bst-LF and KlenTaq v5.9). Subsequently, bacteria were pelleted at 5000g for 15 min at 4 °C and resuspended in 25 mL of ice-cold equilibration buffer (50 mM Tris–HCl, pH 8.0, 300 mM NaCl, and 20 mM imidazole) containing 2 tablets of cOmplete, Mini, EDTA-free Protease Inhibitor Cocktail (Roche, Basel, Switzerland). The bacterial suspensions were sonicated (1 s ON, 1 s OFF) on ice for a total time of 4 min with 40% amplitude. The lysates were centrifuged at 30,000g for 30 min at 4 °C. The His-tagged DNA polymerases were purified from the supernatant using metal affinity chromatography on a Ni-NTA column. Briefly, 1 mL of Ni-NTA agarose slurry was packed into a 5 to 10 mL disposable column, washed with 10 column volumes (CVs) of water, and equilibrated with 10 CVs of equilibration buffer. The clarified bacterial lysate was loaded onto the equilibrated column and allowed to flow through the column under gravity. The column was then washed with 10 CVs of equilibration buffer, followed by 10 CVs of wash buffer (50 mM Tris–HCl, pH 8.0, 300 mM NaCl, and 40 mM imidazole). DNA polymerases were eluted with 3 mL of elution buffer (50 mM Tris–HCl, pH 8.0, 300 mM NaCl, and 500 mM imidazole). The eluate was dialyzed twice, first for 4 h and second for 12–18 h, with 1 L each of fresh ice-cold Ni-NTA dialysis buffer (50 mM Tris–HCl, pH 8.0, 100 mM NaCl, and 1 mM DTT). The eluate was then dialyzed in 1 L of final dialysis buffer (50% glycerol, 50 mM Tris–HCl, pH 8.0, 100 mM NaCl, and 1 mM DTT). The purified polymerases were analyzed by UV spectrophotometry and by SDS-PAGE/Coomassie gel staining.

Nonenzymatic DNA Strand Exchange Reactions.

All strand exchange reactions were assembled at room temperature in 96-well PCR plates. Except where indicated, strand exchange reactions with individual invader oligonucleotides were prepared in 20 μL of 1× rCutSmart buffer (NEB; 50 mM potassium acetate, 20 mM Tris-acetate, 10 mM magnesium acetate, 100 μg/mL recombinant albumin, pH 7.9 at 25 °C) by mixing 500 nM of the invader with 75 nM of the indicated OSDs. Beginning immediately after addition of OSD probes to the invaders, the reactions were incubated in a LightCycler 96 real-time PCR machine (Roche, NC, USA) held at 37 °C, and the real-time measurement of OSD fluorescence was performed in the FAM channel every 1 to 3 min for a total of 60 to 90 min. Data are shown in triplicate with standard deviations. In some experiments, strand exchange reactions were performed in 10 μL volume by mixing 100 nM total concentration of indicated individual or mixtures of multiple invader oligonucleotide with 100 nM of OSD probes.

Strand Displacement Amplification Assays.

SDA assays for polymerase activity were assembled on ice in 20 μL reactions containing 1× rCutSmart buffer, 0.5 mM of deoxyribonucleotides (dNTP), 10 units of Nb.BbvCI nicking endonuclease (NEB), 100 nM of indicated OSD probes, and 0.5 μM of KlenTaq v5.9 variant, 1.9 μM of Bst variant, or 1.6 μM of Klenow variant. These reactions were seeded with equimolar amounts of SDA primers and templates at either 0.1 or 0.5 μM concentration. Reactions were incubated in a LightCycler 96 real-time PCR machine, held at either 50 °C (KlenTaq v5.9 and Bst) or 37 °C (Klenow), and amplification was performed in the absence or presence of 0.1 mM each of dZTP and dPTP deoxyribonucleotides (Firebird). OSD fluorescence was measured in real time in the FAM channel every 1 to 3 min for up to 2 h. Unless indicated, data are shown in at least triplicate with standard deviations.

Primer Extension Assays.

Standard DNA primer extension reactions were performed in 20 μL of 1× rCutSmart buffer (NEB) containing 0.5 mM dNTP and indicated combinations of 500 nM standard SDA template, 500 nM SDA primer, 25 ng/μL extreme thermostable single-stranded DNA binding protein (ET SSB) (NEB), and either wild-type DNA Pol IIIα (1.85 μM) or wild-type DNA Pol II (200 or 100 nM). Primer extensions were performed at 37 °C for 15 min (DNA Pol II) or 30 min (DNA Pol IIIα), followed by enzyme inactivation at 95 °C and cooling to 4 °C. Semisynthetic ZP DNA primer extension reactions were also assembled in 20 μL of 1× rCutSmart buffer (NEB) and contained 500 nM PZP.NSE.template, 0.5 mM dNTP, 25 ng/μL ET SSB, and one of the following DNA polymerases: 1.85 μM wild-type DNA Pol IIIα, 1.85 μM DNA Pol IIIα-Sso7d variant, 200 nM wild-type DNA Pol II, or 1.85 μM DNA Pol II Exo(−) variant. All ZP DNA primer extensions were performed at 37 °C for 15 min in the presence or absence of 500 nM SDA primer and 50 μM each of dZTP and dPTP followed by enzyme heat inactivation at 95 °C for 5 min and cooling to 4 °C. All primer extension products were visualized on native 10% polyacrylamide gels stained with SYBR-Gold, previously shown to be a cost-effective and efficient method for analyzing Taq and Klenow DNA polymerase activities.⁴¹

Nicking Strand Exchange Assays.

Nicking strand exchange (NSE) assays were performed by using the following three stepwise reactions. Step 1: primer extension reactions were assembled in 20 μL of 1× rCutSmart buffer containing 500 nM standard or semisynthetic ZP DNA templates, 500 nM SDA primers, 0.5 mM dNTP, 25 ng/μL ET SSB, and 1.85 μM DNA Pol II Exo(−) or DNA Pol IIIα. Some reactions containing the ZP DNA templates also received 50 μM each of dZTP and dPTP. Reactions were incubated for 15 min at 37 °C, followed by 5 min at 95 °C prior to cooling on ice. Step 2: nicking10 μL of the primer extension products was mixed with 10 μL of 1× rCutSmart buffer containing 10 units each of Nt.BsmAI, Nb.BbvCI, and Nb.BtsI nicking endonucleases. These nicking reactions were incubated for 15 min at 37 °C, followed by 5 min at 95 °C prior to being cooled on ice. Step 3: OSD readout10 μL of the nicked products was mixed with 10 μL of 1× rCutSmart buffer containing indicated OSD probes to a final concentration of 100 nM. These reactions were incubated at 55 °C for 30 min and then at 95 °C for 1 min, followed by 30 min at 37 °C for the measurement of OSD fluorescence.

Statistical Analysis.

All statistical analyses were performed using GraphPad Prism software. Nonlinear regression curves that fit data from three to six biological replicates were constructed using either logistic growth (SDA reactions) or one-phase exponential association (strand exchange reactions) models and least-squares fit. While these integrated kinetic models described curve progression reasonably well, the resulting best-fit values were not used to calculate initial rates because of significant differences in the timing of reaction completion, wherein some reactions had reached a plateau while others had not. To avoid bias and ensure valid and consistent comparison, average rates during comparable initial linear phase of the reactions were calculated using either simple linear regression of indicated three or more data points or the two-point method (Y₂ – Y₁/Δt), widely used for evaluating enzyme kinetics and toehold-mediated strand displacement,^27,42 from two indicated data points. All initial rate values are reported in relative fluorescence units per minute (RFU/min). The statistical significance of differences between reactions was assessed by using one-way ANOVA.

Impact of dZ:dP Base-Pairs on Toehold-Mediated DNA Strand Exchange.

Strand exchange reactions have proven to be excellent probes of the fidelity of base pairing.^27,43,44 To test if strand exchange reactions might also prove useful for monitoring non-natural nucleobases, we first evaluated the impact of dZ:dP pairings on strand exchange itself by designing two hemiduplex OSD probes to contain either three (PZP.OSD) or two (GZP.OSD) consecutive dZ:dP base pairs, followed by a standard single-stranded DNA toehold (Figure 2A). When incubated with fully complementary single-stranded DNA invaders (PZP.OSD with ZPZ.invader or GZP.OSD with CPZ.invader), both OSDs yielded rapid increases in fluorescence, suggesting that the dZ:dP base pairs were amenable to displacement by complementary nucleobases (Figure 2B,C), similar to the behavior of completely natural versions of the OSD and invaders (Figure S1).

Rapid accumulation of fluorescence was also observed upon incubating PZP.OSD with the ZGZ.invader, which presented a single dZ:dG mismatch, suggesting that a dG nucleobase could displace dP from an existing dZ:dP base pair in PZP.OSD, leading to dZ:dG base pairing and strand exchange (Figures 2B and S1). Conversely, when GZP.OSD was mixed with ZPZ.invader presenting a single dZ:dG mismatch, dZ could displace dC from an existing dG:dC base pair in the GZP.OSD and thereby form a dZ:dG base pair, again leading to strand exchange and fluorescence accumulation (Figures 2C and S1). These results validate the observation that G and Z can form a wobble-pairing.⁴⁵ However, unlike G:G wobble pairs^46,47 that severely restricted strand displacement of natural DNA (Figure S1), strand exchange using mismatched Z:G base pairs proved to be only moderately slower, with a 2- to 4-fold reduction in initial rates (Table S3), compared to the reaction between fully complementary partners. Indeed, when the formation of two consecutive dZ:dG base pairs was required for strand exchange (GZP.OSD and ZGZ.invader), strand exchange was highly inefficient (over a 35-fold decline in the initial rates; Figures 2C and S1 and Table S3). Single mismatches of dZ with any of the three other natural bases (dC, dT, and dA) also significantly inhibited strand exchange (Figure S1).

In contrast to Z pairing with either P or G, when either PZP.OSD or GZP.OSD was incubated with ZPC, ZPA, ZPG, or ZPT invader oligonucleotides, minimal fluorescence was observed, suggesting that a dP:dZ base pair is more stable than dP paired with any natural nucleobase (Figures 2B,C and S1 and Table S3).⁴⁸

To probe the impact of dZ and dP nucleobases on toehold function, the PZP.OSD sequence was modified by replacing three consecutive standard bases in its toehold with dZ and dP nucleobases. The resulting PZPPZZ.OSD displayed three adjacent non-natural base pairs in the branch migration domain, followed immediately by three consecutive non-natural nucleobases in the toehold. When incubated with its complementary strand exchange partner ZPZZPP.invader, PZPPZZ.OSD demonstrated a large increase in fluorescence indicating that both dZ and dP could serve as footholds for complementary toehold base pairing and facilitate robust strand exchange of six consecutive non-natural nucleobases (Figure 3).

Since the kinetics of strand exchange depends on toehold hybridization strength,^25,49 we hypothesized that mismatches of dZ and dP toeholds with standard nucleotides in the invader strands would lead to slower strand exchange. To evaluate this hypothesis, the strand exchange between PZPPZZ.OSD and several invader variants that would have presented one or more mismatches was analyzed via OSD fluorescence accumulation. The ZGZZGP.invader variant that introduced two dZ:dG “wobble” mismatches, one during toehold binding and the second during branch migration, led to OSD fluorescence accumulation similar to that observed during strand exchange with a fully complementary invader (Figure 3B and Table S4). In contrast, fluorescence accumulation was significantly slower with a 13-fold lower initial reaction rate when two adjacent dZ:dG wobble mismatches were present in the toehold (PZPPZZ.OSD and the ZPZZGG.invader; Figure 3C and Table S4). As with natural nucleotides, two consecutive wobble pairs significantly reduced the hybridization energy and thus toehold binding strength. When all three consecutive dZ and dP nucleobases in the PZPPZZ.OSD toehold were forced to mis-pair with standard bases in the ZPZ.invader, no fluorescence was generated by strand exchange (Figure 3C).

In contrast, dP:dC mismatches were much less well tolerated: strand exchange was completely blocked when a single dP:dC mismatch between the OSD and the invader was encountered near the beginning of the branch migration domain (Figure 2), and interactions between PZPPZZ.OSD and the ZPZCPP.invader variant, which would have resulted in only a single mismatch in the toehold, led to a 7-fold slower initial rate of fluorescence accumulation (Figure 3B and Table S4). These results are consistent with the fact that dP:dC mismatches are less stable than dZ:dG mismatches.⁴⁸

Overall, these results confirm that the base pairing stability of non-natural toeholds is a rate-limiting step for strand exchange, similar to results with natural base pairing. They also conclusively demonstrate that DNA strand exchange reactions can be programmed using six-letter genetic alphabets and can be reliably used to discriminate oligonucleotides with non-natural nucleobases, a feature we will further employ in the assessment of polymerases, below.

Applying Strand Exchange Reactions to Evaluate the Fidelity of Various DNA Polymerase I Enzymes with Six-Letter Genetic Alphabets.

E. coli DNA Pol I plays a significant role in lagging strand synthesis during DNA replication and has previously been the focus of attempts to incorporate noncanonical nucleotides in vivo.⁵⁰ Its large (Klenow) fragment can replicate many non-natural DNA base pairs in vitro and can also carry out nontemplated tailing of DNA with dZ and dP nucleobases.^20,51,52 We sought to determine how the major replicative machinery of E. coli might treat a six-letter genetic alphabet containing Z:P bases. While nanopore sequencing algorithms can be trained for directly sequencing Z and P,²⁰ these methods are not readily available for routine assay and high-throughput applications. We hypothesized that the specificity of DNA hybridization and strand displacement might provide a more convenient readout of incorporation and therefore designed an SDA assay for noncanonical DNA whose products would trigger dissociation of an OSD reporter. SDA uses a strand displacing DNA polymerase, a nicking endonuclease, a primer, and a DNA template with a strategically positioned nicking endonuclease recognition sequence to establish continuous isothermal amplification (Figure 4A). Displaced, single-stranded DNA amplicons should undergo toehold-mediated strand exchange with the complementary OSD reporter, ultimately leading to fluorescence accumulation.

The SDA-OSD assay was used to evaluate the E. coli Klenow DNA polymerase I D355A mutant (lacking 3′–5′ exonuclease activity)⁵³ for its ability to replicate single-stranded semisynthetic DNA containing two (ZZ.template), three (PZP.template), or six (PZPPZZ.template) consecutive dZ and dP nucleobases. Template designs were again based on our findings regarding the locations of single mismatches that could modulate the strand exchange. Single-stranded amplicons were measured via two OSD reporters: a universal OSD reporter (Uni.OSD) that contained only natural nucleotides and hybridized to an amplicon region that was downstream of the non-natural segment, or an OSD that contained Z:P nucleotides in its branch migration domain (ZZ.OSD, PZP.OSD, and GZP.OSD) or in both its toehold and branch migration domains (PZPPZZ.OSD) (Figure 4A).

Klenow produced a rapid increase in Uni.OSD fluorescence in SDA reactions containing dZ and dP nucleotides (Figures 4 and S2), suggesting that it could replicate non-natural DNA in the presence of dZ and dP nucleotides. However, about a 5-fold decrease in the initial reaction rate (Table S5) of the PZPPZZ.template as opposed to the PZP.template suggested that replication efficiency declined with longer stretches of non-natural base pairs. In SDA reactions lacking dZ or dP nucleotides, Klenow produced a comparable Uni.OSD signal when replicating the ZZ.template, suggesting that it could readily mis-incorporate natural nucleotides across from dZ (Figure S2). However, the absence of dZ and dP during the replication of PZP and PZPPZZ templates resulted in ∼10-fold and 600-fold decline, respectively, in the initial reaction rates, suggesting that natural nucleotide mis-incorporation across from dP was relatively inefficient (Figure 4 and Table S5). These observations were also true for SDA reactions seeded with 5-fold lower amounts of SDA templates and primers (Figure S3).

In parallel SDA reactions, when PZP, GZP, or ZZ.OSD OSD probes were used with amplicons generated in the presence of dZTP and dPTP, a rapid increase in fluorescence was again observed (Figures 4 and S2). Since we had previously observed that the presence of a single dP mismatch or two consecutive dZ mismatches within the first three branch migration steps blocks OSD strand exchange (Figures 2 and 3), these data strongly suggest that when dZ and dP nucleotides are available in the reaction, Klenow is faithfully generating amplicons via dZ and dP incorporation. Although some amplicons containing natural nucleotides across from Z or P may also still be produced, based on the lower rate of reaction with the PZP.OSD relative to the normalizing Uni.OSD probe, these are a small fraction of the overall signal, since PZP.OSD and GZP.OSD signals decline rapidly with increasing concentration of mutants (Table S5 and Figure S4). Klenow was also capable of replication of six consecutive non-natural nucleobases, with a similar reduction in initial reaction rates of PZPPZZ.OSD versus the Uni.OSD signal.

In order to determine whether these observations might apply to other Pol I polymerases, especially those of interest for biotechnology applications, we also carried out experiments with Bst DNA polymerase, a Family A Pol I homologue from G. stearothermophilus. Bst large fragment, routinely used for a variety of isothermal amplification assays, can reverse transcribe and replicate other xeno-nucleic acids and incorporate non-natural deoxyribonucleotides, such as dATPαS (2′-deoxyadenosine-5′-O-(1-thiotriphosphate)) and glycerol-nucleoside triphosphates, into DNA.^54–58 Bst has previously also been found to incorporate dZ and dP when replicating templates containing a single non-natural nucleotide.¹⁷ Interestingly, we have previously found that a chimeric KlenTaq DNA polymerase I containing a 14 amino acid region of Bst DNA polymerase among other point mutations had significantly improved strand displacing ability, allowing the mutant enzyme (termed v5.9) to perform loop-mediated isothermal amplification.³⁹ Similar to Klenow, both Bst and KlenTaq v5.9 generated comparable Uni.OSD signals when replicating ZZ.template in the absence or presence of dZ and dP nucleotides, suggesting that both v5.9 and Bst DNA polymerases could also mis-incorporate a natural nucleobase across from dZ with relative ease (Figures S5 and S6). Again, like Klenow, the absence of dZTP and dPTP during amplification of the PZP.template caused a decline (5-fold for Bst and 62-fold for KlenTaq v5.9) in the initial rate of normalizing Uni.OSD fluorescence accumulation (Table S5 and Figure S7), suggesting that both DNA polymerases inefficiently incorporate standard nucleotides across from dP. These observations were also true for SDA reactions containing 5-fold lower amounts of the PZP.template and the SDA primer (Figure S8). Interestingly, Bst was more efficient than both KlenTaq v5.9 and Klenow at replicating longer stretches of non-natural DNA in the presence of dZ and dP nucleotides, as indicated by the similar initial rates of its Uni.OSD signal with the PZPPZZ amplicon (Figure 4 and Table S5), while both other polymerases instead showed declines with the longer PZPPZZ.template (Table S5 and Figures S7 and S9). This enhanced efficiency might be a reflection of the greater processivity, better strand displacing ability, and higher k_cat of Bst DNA polymerase compared to both Klenow and Taq.⁵⁴

PZP, GZP, PZPPZZ, or ZZ.OSD OSD probes also demonstrated rapid increases in fluorescence when used to evaluate cognate SDA amplicons, suggesting that, like Klenow, Bst and v5.9 were capable of faithful dZ and dP incorporation (Figures 4 and S5–S9). Overall, while Bst, v5.9, and Klenow demonstrated similar trends toward semisynthetic DNA replication, in the presence of dZTP and dPTP, Bst and v5.9 may be more faithful than Klenow, based on inferred increased incorporation of ZGZ across from PZP (a Z:G mismatch), which would then fail to react with the (mismatch-detecting) GZP.OSD probe.

Evaluating E. coli DNA Pol II and IIIα Polymerases with Six-Letter Genetic Alphabets.

For full replication of genomic DNA, faithful incorporation by multiple polymerases will be necessary. While Pol I and other single-subunit DNA polymerases are capable of utilizing the Z:P base pair,^17,20 Z → C substitutions may remain a significant issue for other polymerases, especially large polymerase complexes such as Pol III. Since E. coli DNA polymerases II and IIIα do not perform strand displacement DNA synthesis, the polymerase activity of the purified enzymes was first verified using isothermal primer extension assays. In reactions containing a 3′-end-blocked single-stranded standard DNA template, primers, and DNA Pol IIIα subunit, which lacks exonuclease activity, the 3′-ends of hybridized primers were extended by the polymerase to yield duplex DNA that formed a distinctly larger band (Figure S10, lanes 1 and 2) than both the ssDNA template (lanes 6 and 8) and the template:primer hemiduplex (lane 4) on native polyacrylamide gels. In reactions lacking either DNA Pol IIIα, primer, or template, this larger DNA was absent (lanes 3–12), suggesting that it was produced upon template-dependent primer extension by DNA Pol IIIα. Inclusion of the extreme thermostable single-stranded DNA binding protein (ET SSB) (NEB) during primer extension also supported robust conversion of ssDNA templates into larger primer extension products (lane 1). Similar results were obtained in assays using DNA Pol II, where primer extension products were observed only in reactions containing templates, primers, and enzymes (Figure S11).

The abilities of DNA Pol II and IIIα to replicate non-natural DNA were evaluated in primer extension assays using single-stranded DNA templates containing three consecutive dZ and dP nucleobases. Either wild-type or Sso7d DNA binding domain-fused DNA Pol IIIα or 3′–5′ exonuclease-minus D335N DNA Pol II were used in the presence or absence of dZTP and dPTP nucleotides. Upon native polyacrylamide gel electrophoretic analysis of the reactions, primer extension products were observed only in reactions containing primers and dZTP and dPTP nucleotides (Figure 5). In reactions lacking non-natural nucleotides, hemiduplex products from incomplete primer extension were observed with slow migration patterns likely due to both partial self-dimerization and binding of extreme thermostable SSB. In contrast, the 3′– 5′ exonuclease-deficient DNA Pol II produced primer extension products even in the absence of dZTP and dPTP, albeit at about 0.6-fold of the amount compared to reactions containing the non-natural nucleotides (Figure 5). These results are the first indication that both DNA Pol III and DNA Pol II could replicate DNA containing dZ and dP nucleotides.

To evaluate the incorporation of dZ and dP nucleobases in Pol II and Pol IIIα amplicons using strand exchange probes, a variation of the primer extension assay, termed NSE, was designed (Figure S12A). Upon incubation with a nicking enzyme cocktail, the NSE template strand is cleaved at three sites, while the newly synthesized amplicon strand is cleaved at a single position immediately downstream of the primer. These multiple nick sites were included in order to minimize unintended hybridization between primer extension products/the OSD quencher strand and the template while still facilitating hybridization between the OSD fluorophore strand and full-length primer extension products. Upon heating and cooling the nicked DNA in the presence of a cognate OSD reporter, the newly synthesized amplicon strand should exchange the fragmented templates for the longer OSD fluorophore-labeled strand, and consequent separation of the OSD quencher-labeled strands would result in fluorescence accumulation. OSD reporters retained the ability to discriminate single-nucleotide differences between amplicons (CPZ.invader, ZPC.invader, and ZGZ.invader) even when subjected to heat denaturation (Figure S12B–D). This differentiation was similar to that observed during the interaction of OSD and invader sequences composed of only natural DNA (Figure S12E–G).

The NSE assay was validated using two standard DNA templates, wild-type (WT) or M1 containing a single point mutation (Figure S13), which were incubated with primers and either DNA Pol IIIα, DNA Pol II (Exo−), or no polymerase to facilitate primer extension. Subsequently, a portion of the amplicons was treated with nicking enzymes, while another portion was left intact. Upon adding WT OSD probes and heating the nicked amplicons, significant fluorescence increase was observed in reactions seeded with WT templates but not in reactions that had received the M1 template, since M1 extension products present a mismatch at the first step of branch migration with the OSD probe (Figure S13).

To further challenge the replicative polymerases, NSE assays containing DNA Pol IIIα (wild-type or fused with the Sso7d) or DNA Pol II were then seeded with single-stranded DNA templates containing three consecutive dZ and dP nucleobases, and primer extension was performed in the presence or absence of dZTP and dPTP (Figure 6). The template was designed based on our initial findings, above, regarding the fidelity of sensing, where single dP or two dZ mismatches could halt strand exchange. Nicked amplicons were probed with either fully complementary PZP.NSE.OSD or with PZG.NSE.OSD, which present single dZ:dG mismatches immediately downstream of the toehold. Both OSD reporters generated 2- to 3-fold higher end point fluorescence above background upon interacting with primer extension products produced by either DNA Pol IIIα or DNA Pol II. In contrast, OSD fluorescence in NSE reactions lacking templates remained largely quenched, again suggesting that DNA Pol III and DNA Pol II could accurately incorporate dZ and dP nucleobases.

Rationally Improving the Fidelity of DNA Polymerases for Six-Letter Genetic Alphabets.

While the full complement of E. coli polymerases is capable of Z:P incorporation, genomic stability with a six-letter Z:P alphabet may nonetheless require greater fidelity, especially considering potential fluctuations in cellular concentrations of synthetic nucleotides, which may be largely dependent on transport from the growth media. As we have shown, in the absence of the correct non-natural nucleotides, Klenow, Bst, and v5.9 DNA polymerases readily mis-incorporated natural nucleotides, a result that accords with the fact that natural nucleotide concentrations generally impact DNA polymerase fidelity.⁵⁹

Thus, we first attempted to improve DNA polymerase I fidelity, drawing from mutations in KlenTaq DNA polymerase variants that had been evolved for improved replication efficiency with non-natural nucleobases.³⁶ Although structure-guided mutation transplantation is not guaranteed to achieve functional transfer⁶⁰ and polymerase phenotypes are so complex that they often require directed evolution for modification,^61,62 the overall structures of many proteins are similar enough that they can sometimes be engineered by transplanting key amino acid residues between relatives.^63–65 For instance, replacing a single phenylalanine in Taq and Klenow DNA polymerases with a tyrosine found in the homologous location in bacteriophage T7 DNA polymerase allowed these polymerases to readily accept 2′,3′-dideoxynucleoside triphosphates.⁶⁶ We hypothesized that relocation of mutation sets (either SVV: N580S/L628V/E832V or VAEV: M444V/P527A/D551E/E832V) that improved Z:P replication fidelity in KlenTaq into corresponding positions in Klenow DNA polymerase, a Family A structural homologue of Taq,⁶⁷ would lead to similar improvements in fidelity of the resulting variants (Klenow-SVV and Klenow-VAEV) (Figure S14).^35,68 Similarities in key amino acid sequence motifs and three-dimensional structures of the two enzymes⁶⁹ were used to guide transplantation of the KlenTaq mutations to structurally corresponding, though not always sequence homologous, locations in Klenow (Figure S15).

While both mutants demonstrated similar rates of Uni.OSD fluorescence accumulation as the wild-type enzyme in SDA reactions containing dZ and dP nucleotides (suggesting faithful incorporation of natural nucleobases), in the absence of these nucleobases, Klenow-SVV demonstrated a 10-fold lower initial rate of amplifying the PZP.template compared to the wild-type enzyme (Figure 7 and Table S6). In a similar comparison, Klenow-VAEV demonstrated only a 2-fold reduction in the initial rate of amplification of non-natural DNA in the absence of corresponding synthetic nucleotides in the reaction. These results suggest that in the absence of dZ and dP nucleotides, the Klenow-SVV variant had the least propensity to mis-incorporate standard nucleotides, which was true even when replicating the PZPPZZ.template (Figure 7). Since all Klenow variants harbor the D355A mutation shown to diminish 3′–5′ exonuclease activity by 4 orders of magnitude, the observed difference in mis-incorporation is unlikely to be influenced by proofreading.⁵³ In general, the replication of PZP.templates was faster compared to the PZPPZZ.templates with or without dZ/dP nucleotides, suggesting that all Klenow polymerase variants are likely less efficient at replicating non-natural DNA compared to natural DNA.

When PZP.OSD or GZP.OSD probes were used for strand exchange readout of PZP.template amplicons, a rapid increase in fluorescence was observed in all reactions containing dZTP and dPTP with the wild-type enzyme showing a faster response especially with GZP.OSD (Figures 7 and S3). The Klenow-SVV variant produced the most PZPPZZ.OSD signal (1.5-fold improvement in rate) compared to both the wild-type and Klenow-VAEV variants (Figure 7 and Table S6). The SVV and VAEV mutations were found to reduce pausing of KlenTaq when incorporating dZTP opposite dP in a template, and also reduced the ability of the polymerase to incorporate dPTP opposite dZ in the template.³⁶ These results strongly suggest that the SVV mutations previously explored in the Taq polymerase can be readily transplanted to Klenow and allow the enzyme to incorporate dZ more efficiently. An additional mutation (R631K), corresponding to the R536K mutation in Taq that was previously shown to possess increased mismatch extension selectivity against standard mismatches,⁷⁰ did not improve fidelity (Figure 7). This might be due to the noncanonical nature of the Z and P mismatches or the sequence differences between the Taq and Klenow backbones.

Given the success of transplanting KlenTaq mutations to Klenow, we hypothesized that similar transfers might improve non-natural DNA replication for Bst DNA pol, another Family A polymerase structural homologue.⁵⁴ Adapting Bst to faithfully replicate multiple Z:P base pairs would open the way to using the AEGIS nucleotides in a variety of additional applications. In consequence, we synthesized, expressed, and purified Bst-SVV (Figure S16), harboring the set of substitutions previously shown to improve dZ incorporation by KlenTaq DNA polymerase³⁶ at structurally corresponding loci chosen using similarities in amino acid sequence motifs and structures of the two enzymes (Figure S17). When SDA reactions of PZP.templates were performed using Bst-SVV in the absence of dZ and dP nucleotides, the Uni.OSD signal was significantly lower, with a 18-fold reduction in the initial rate of amplification, than that observed with the wild-type enzyme suggesting greater selectivity of the polymerase variant against incorporating standard nucleotides (and in particular discrimination against P:C pairings since the evolved KlenTaq was identified to improve dZ incorporation opposite template dP) during non-natural DNA replication (Figure 8 and Table S7). In fact, when challenged to replicate a longer stretch of non-natural nucleotides in the PZPPZZ.template in the absence of dZ and dP nucleotides, the Bst-SVV variant did not generate any Uni.OSD signal over background, unlike the wild-type Bst polymerase (Figure 8 and Table S7). Since Bst DNA polymerase naturally lacks 3′–5′ exonuclease activity,⁷¹ the observed differences in mis-incorporation are not due to proofreading.

Taken together, these results all suggest that the transplantation of homologous substitutions across Family A polymerases (KlenTaq, Bst, and E. coli Pol I) can increase their fidelity with expanded genetic alphabets. We also attempted to transplant the SVV or VAEV mutations³⁶ into the KlenTaq-Bst chimera, v5.9³⁹ (Figures S18–S20). However, in contrast to augmentation of the wild-type polymerases, no significant improvements in the function of the chimera were noted.