Display of functional nucleic acid polymerase on Escherichia coli surface and its application in directed polymerase evolution

We report a first of its kind functional cell surface display of nucleic acid polymerase and its directed evolution to efficiently incorporate 2′‐O‐methyl nucleotide triphosphates (2′‐OMe‐NTPs). In the development of polymerase cell surface display, two autotransporter proteins (Escherichia coli adhesin involved in diffuse adherence and Pseudomonas aeruginosa esterase A [EstA]) were employed to transport and anchor the 68‐kDa Klenow fragment (KF) of E. coli DNA polymerase I on the surface of E. coli. The localization and function of the displayed KF were verified by analysis of cell outer membrane fractions, immunostaining, and fluorometric detection of synthesized DNA products. The EstA cell surface display system was applied to evolve KF for the incorporation of 2′‐OMe‐NTPs and a KF variant with a 50.7‐fold increased ability to successively incorporate 2′‐OMe‐NTPs was discovered. Expanding the scope of cell‐surface displayable proteins to the realm of polymerases provides a novel screening tool for tailoring polymerases to diverse application demands in a polymerase chain reaction and sequencing‐based biotechnological and medical applications. Especially, cell surface display enables novel polymerase screening strategies in which the heat‐lysis step is bypassed and thus allows the screening of mesophilic polymerases with broad application potentials ranging from diagnostics and DNA sequencing to replication of synthetic genetic polymers.

Phage-display-based strategies have often extraordinary throughput compared to plate-based directed evolution campaigns. However, phage display strategies enrich active variants by few nucleotide incorporations and do not usually ensure the identification of variants with high processivity, which is of interest when synthesizing modified NAs such as aptamers. Water-oil emulsion-based methods include compartmentalized self-replication (Ghadessy, Ong, & Holliger, 2002), compartmentalized self-tagging (Pinheiro, Arangundy-Franklin, & Holliger, 2014), and droplet-based optical polymerase sorting (Larsen et al., 2016). The three strategies also have often extraordinary throughput comparing to plate-based methods and have been proved powerful for thermostable polymerases, yet due to the requirement of a heat-lysis step in the three methods, they might be rather less suitable for evolving mesophilic polymerases.
Here, we report the first-time CSD of NA polymerase and demonstrated the stability and applicability of the CSD system by di- (California). The lysogeny broth (LB) medium used for cloning and precultures consisted of 10 g/L tryptone, 5 g/L yeast extract, and 10 g/L NaCl. In the case of LB agar plates, 20 g/L agar was added. The TBMg10 medium used for expressions comprised 12 g/L peptone, 24 g/L yeast extract, 4 g/L glycerol, 2.31 g/L KH 2 PO 4 , 12.5 g/L

| The AIDA-I-E-KF construct
An AIDA-I autotransporter construct (including a leading sequence, an E-epitope, a linker G 4 SGGS(G 4 S) 3 , a β1 extracellular domain, and a translocator domain of AIDA-I) was synthesized and cloned to the expression vector pALXtreme1a (Blanusa, Schenk, Sadeghi, Marienhagen, & Schwaneberg, 2010) using XbaI and HindIII restriction sites (Table S5). The gene encoding E. coli KF was amplified from the genome of E. coli DH5α using P_KF2 and P_KF3 primers (Table   S1). The PCR was carried out according to the general PCR protocol (Table S4; with a 5 min initial cell lysis and DNA denaturation step) and E. coli DH5α cells were used as a template. KF was then inserted between the leading sequence and the E-epitope of the AIDA-I transporter construct using NheI and BamHI restriction sites to generate pALX1a-AIDA-I-E-KF (see Supporting Information).

| The EstA-E-KF construct
An EstA autotransporter construct comprising a leading sequence, a passenger peptide, a 10-Alanine spacer, an E-epitope, an inactivated EstA extracellular domain, and an EstA translocator domain was kindly provided by Dr. Kristin Rübsam. A PCR using P_EstA1 and P_EstA2 primers (Table S1) was performed on the EstA autotransporter construct to replace the passenger peptide with a flexible linker (G 4 S) 3 . The modified EstA autotransporter construct was then amplified by PCR using P_EstA3 and P_EstA4 primers and then cloned into the expression vector pALXtreme1a (Blanusa et al., 2010) by NdeI and HindIII restriction sites to generate pALX1a-EstA-E (Table S5). Finally, the gene encoding E. coli KF was amplified from the genome of E. coli DH5α using P_KF1 and P_KF2 primers (Table   S1) and then inserted between the leading sequence and the linker of the EstA autotransporter construct using NheI and KpnI restriction sites to generate pALX1a-EstA-E-KF (see Supporting Information).

| OM fractionation
The cells expressing AIDA-I-E, AIDA-I-E-KF, EstA-E, and EstA-E-KF were resuspended in 20 mM Tris buffer with 100 mM NaCl (pH 8) and then sonicated at 40% amplitude for 5 min with 15 s on and 15 s off intervals. The sonicated cell lysate was subsequently centrifuged at 3,225g, 4°C for 1 hr, and the resulting supernatant was centrifuged at 100,000g, 4°C for 30 min. The resulting pellet was resuspended in PBS containing 0.01 mM MgCl 2 , 2% Triton X-100, and incubated at RT for 30 min. Finally, the suspension was centrifuged at 100,000g, 4°C for 30 min to precipitate the OM fraction. The resulting pellet was resuspended in PBS for further analysis (Grimm et al., 2018).

| Library generation
For each library, two fragments were amplified from pALX1a-EstA-E-KF by PCR (general PCR protocol, Tables S4 and S8) using primers listed in Table S3. The PCR-amplified fragments were then column purified. Fragments Nos. 1 and 2 were assembled to generate the EstA-E-KF I709E E710SSM library using NEBuilder DNA Assembly Kit (New England Biolabs; Table S6, Figure S9). The PCR-amplified Fragments Nos. 3 and 4 were assembled to generate the EstA-E-KF I709SSM E710G library using NEBuilder DNA Assembly Kit (New England Biolabs; Table S6, Figure S9). PolCSD buffer with gentle pipetting. The remaining procedure is identical as described in Section 2.5.

| RESULTS
Results on the development of the E. coli display system for directed polymerase (KF) evolution are summarized in three parts. In the first part, two autotransporter systems AIDA-I and EstA were assessed for their ability to display KF. CSD of active KF was achieved for the EstA autotransporter system, which was selected to establish a CSDbased screening system for KF-polymerase evolution. In the second part, the stability and applicability of the CSD-based screening system were validated in a semirational protein engineering campaign by screening KF libraries for synthesizing 2′-OMe modified DNA. The conceptual scheme of the screening system is shown in Figure 3. In Section 5, a variant that showed a 50.7-fold increased fluorescence signal for 2′-OMe-NTP incorporation was obtained.

| CSD of functional KF in E. coli
To display active KF on E. coli surface, we fused the C-terminal of KF to either AIDA-I or EstA autotransporter systems ( Figure 1a). The AIDA-I construct consists of a flexible G 4 SGGS(G 4 S) 3 linker, a β 1 extracellular domain, and a translocator domain. The β 1 extracellular domain was reported to assist the folding of the native passenger (Berthiaume, Rutherford, & Mourez, 2007), and a construct including the flexible G 4 SGGS(G 4 S) 3 linker and the β 1 extracellular domain was reported to successfully display a 54-kDa cytochrome P450 (Quehl et al., 2017). EstA consists of a (G 4 S) 3 A 10 linker, an inactivated native passenger, and a translocator domain (Grimm et al., 2018;Rübsam, Weber, Jakob, & Schwaneberg, 2018). An EstA construct including the inactivated native passenger domain was reported to successfully display a 62-kDa lipase (Becker et al., 2005). A (G 4 S) 3 A 10 linker was designed because an A 10 spacer was reported to facilitate the spatial separation of the E-epitope to its fusion protein partner (Rübsam et al., 2018). A flexible (G 4 S) 3 linker, which was reported to improve the activity of displayed P450 reductase (Quehl et al., 2017), was further added to facilitate a freer movement of the displayed KF.

| A CSD-based screening system for evolving KF for efficient incorporation of 2′-OMe-NTPs
Using the CSD of the polymerase, we established a screening system for mesophilic KF and applied it for the directed evolution of KF to efficiently incorporate 2′-OMe-NTPs in DNA synthesis (Figure 3).

F I G U R E 3
The scheme of a cell surface display (CSD)-based screening system for nucleic acid polymerases. A CSD polymerase library was generated, expressed in 96-well microtiter plates, and subjected to the screening system. In the system, the cells were mixed with a primed template, a cell-membrane impermeable DNA-intercalating fluorescent dye, dATP, dTTP, dCTP, and 2′-OMe-GTP, and incubated at 37°C for 15 min. The activity of each polymerase variant was detected by measuring the increase of fluorescence readout. Variants with higher fluorescence readout were taken for detailed analysis and may serve as a starting point for the next round of evolution. 2′-OMe-GTP, 2′-O-methyl nucleotide triphosphates; dATP, deoxyadenosine triphosphate; dCTP, deoxycytidine triphosphate; dTTP, deoxythymidine triphosphate; EstA, esterase A; KF, Klenow fragment [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 4 Measured fluorescence readouts of all wells in descending order of the DNA synthesis catalyzed by cell-surface-displayed KF in 96-well microtiter plates. The cells expressing EstA-E-KF in each well were mixed with primed template, DNA intercalating fluorescent dye, dATP, dTTP, dCTP, and 2′-OMe-GTP. The reaction mixtures were subsequently incubated at 37°C for 15 min and then the fluorescence intensity of each well was measured (ex./em. = 500/530 nm, gain = 100). The gray circles show the measured fluorescence readouts of all wells, the apparent coefficient of variation of the readouts is 9%. The white diamonds show the measured fluorescence readouts of all wells after subtracting the readout of the negative control (cells expressing EstA-E-KF*), the true coefficient of variation of the readouts is 14.3%. 2′-OMe-GTP, 2′-O-methyl nucleotide triphosphates; dATP, deoxyadenosine triphosphate; dCTP, deoxycytidine triphosphate; dTTP, deoxythymidine triphosphate; E, E-epitope; EstA, esterase A autotransporter; KF, Klenow fragment of Escherichia coli DNA polymerase I; KF*, mutant KF with abolished polymerase activity In the first library, we simultaneously mutated position I709 to glutamate and saturated position 710 with all 20 canonical amino acids using degenerate nucleotide codon "NNK" (N = A or T or C or G; K = G or T).
In the second library, we simultaneously mutated position E710 to glycine and saturated position 709 with all 20 canonical amino acids using degenerate nucleotide codon "NNK." To ensure that all 20 possible amino acid variants are present in randomly picked and screened clones, 168 clones per library were screened employing the CSD-based screening system (theoretical diversity coverage >94%; Firth & Patrick, 2008). In the first library, 44 clones showed a fluorescence readout of >150% of wild-type and 10 clones had a 190-250% improvement. After sequencing of 10 clones, glycine was found in the eight best clones (215-250% of wild-type); aspartate (191% of wildtype), as well as glutamate (203% of wild-type), were found once (Table 1). In the second library, only six clones showed a fluorescent readout of >150% of wild-type KF. Sequencing of the three best clones (>170% of wild-type KF) revealed a preferred substitution from isoleucine to glutamate (>200% of wild-type KF) and a substitution from isoleucine to aspartate (170% of wild-type KF; Table 1) at position 709.
In summary, KF I709E E710G was shown to be the most preferred variant found by CSD-based screening system in both libraries and the variant KF I709E E710G was subjected to further characterization.

| Characterization of KF I709E E710G
To characterize the polymerase's ability to successively incorporate 2′-OMe-NTPs, genes encoding KF* (mutant KF with abolished polymerase activity), wild-type KF, and KF I709E E710G were amplified from their CSD construct, cloned into an expression vector, expressed, and purified. A concentration of 0.35 µM of purified KF*, wild-type KF, or KF I709E E710G were supplied in a primer extension reaction comprising the DNA intercalating fluorescent dye, primed template, dNTPs, or 2′-OMe-NTPs at 37°C. Consistent with the criteria in the screening section, we defined the polymerase's ability to incorporate nucleotide successively as the measured fluorescence readout after the reaction reaches a plateau. When dNTPs were supplied in a primer extension reaction, purified wildtype KF and KF I709E E710G reached comparable fluorescence readout of (1.47 ± 0.03) × 10 4 RFU and (1.52 ± 0.02) × 10 4 RFU, respectively, after normalizing to the signal of KF* (Figure 5b). Comparable fluorescence readout suggested that wild-type KF and KF I709E E710G achieved similar completeness of primer extension, thus have similar ability to incorporate nucleotide successively when their natural substrates were supplied. When dTTP, dATP, dCTP, and 2′-OMe-GTP were supplied, wild-type KF and KF I709E E710G reached fluorescence readout of (8.76 ± 1.49) × 10 2 and (8.68 ± 0.19) × 10 3 RFU, respectively, after normalizing to the signal of KF* (Figure 5c). A 9.9-fold increase of fluorescence readout was achieved by KF I709E E710G compared to wild-type KF, suggesting that KF I709E E710G not only incorporated 2′-OMe-GTP but also continued the primer extension more efficiently than wild-type KF.
Despite the improvement, complete primer extension was not achieved by KF I709E E710G. The fluorescence readout detected when supplying dTTP, dATP, dCTP, and 2′-OMe-GTP was 43% less than supplying natural substrates (dNTPs). In addition, we analyzed the ability of purified KF I709E E710G to incorporate all four 2′-OMe-NTPs (Figure 5d). Direct comparisons are biased because complete 2′-OMe modification of one DNA strand reduces the measured fluorescence intensity by ∼25% ( Figure S7). Consequently, the measured fluorescence values were normalized to the signal of KF* and then multiplied by 1.33 before comparison. The fluorescence readout achieved by wild-type KF and KF I709E E710G after incorporating 2′-OMe-NTPs were (1.49 ± 0.56) × 10 2 RFU and (7.56 ± 0.36) × 10 3 RFU, respectively. Compared to wildtype KF, KF I709E E710G achieved a 50.7-fold increase of fluorescence readout when all four 2′-OMe-NTPs were used in primer extension experiments exclusively. The latter suggested that KF I709E E710G is not only able to incorporate 2′-OMe-GTP but also the other three 2′-OMe-dNTPs with much greater efficiency than wild-type KF. To directly evaluate the variant's ability to incorporate nucleotide successively, we repeated the primer extension reaction with fluorophore-conjugated primer and analyzed the product by denaturing polyacrylamide gel electrophoresis (urea PAGE; Figure 6). KF I709E E710G (Lane 12), or without polymerase (Lane 9). When KF* was present, the primers were partially digested by the enzyme's remaining 3′-5′ exonuclease activity. When dNTPs were supplied, both wild-type KF and KF I709E E710G achieved "fulllength synthesis." When dATP, dTTP, dCTP, and 2′-OMe-GTP were supplied, wild-type KF stopped primer extension while 2′-OMe-GTP was incorporated (see the sequence of primed template, Figure 6a) whereas KF I709E E710G continued to extend the primer after 2′-OMe-GTP incorporation. When 2′-OMe-NTPs were supplied, the exonuclease activity of wild-type KF outperformed its primerextension efficiency, and as a result, partially digested primers were obtained. KF I709E E710G extended the primers for at least two nucleotides without any observable primer degradation ( Figure 6 and Figure S10). The urea PAGE analysis served a direct confirmation of KF I709E E710G's ability to incorporate 2′-OMe-NTPs and achieve an efficient primer extension. The results also verified that the activity measured by the CSD-based screening system is transferable to purified enzymes.   Figure S5a). Since the 13-residue E-epitope is rather short and has not been reported to hinder membrane integrity, the ob-  (Patel, Suzuki et al., 2001;Sattar et al., 2004). The corresponding residue of position 709 in Taq has been reported to be involved in forming a hydrophobic pocket that binds to the base and the sugar part of the incorporating nucleotide (Patel, Kawate, Adman, Ashbach, & Loeb, 2001).
The bulky side chains of position 710 have been identified to act as a "steric gate" for sugar discrimination of incoming nucleotides (Astatke et al., 2002;Brown & Suo, 2011). The substitution from glutamate to glycine at position 710 likely reduced the "steric gate" and enlarged the active site for the insertion of 2′-OMe nucleotides (Astatke et al., 2002;Fa et al., 2004). Interestingly and in contrast to literature data, already I709E E710D and I709E alone enabled a significant degree of 2′-OMe-NTPs acceptance (186% and 182% of wt in the screening system).
E710D alone has a negligible effect on rNTP incorporation  which suggests that I709E alone facilitated the 2′-OMe-NTPs acceptance without altering the E710 steric gate in KF.
Polymerase CSD provides a novel tool for designing advanced polymerase screening strategies and enables efficient polymerase evolutions close to application conditions such as pH and buffer composition. Furthermore, the polymerase CSD system can potentially be linked to FACS readout by encapsulating single cells displaying polymerase variants in droplets together with its substrates using the water-in-oil-in-water double emulsion method reported by Larsen et al. (2016). In most emulsion-based polymerase screening or enrichment strategies, the cells expressing polymerase variants intracellularly are disrupted by heat lysis to allow the interaction be-

| CONCLUSION
A CSD platform based on the EstA autotransporter was found to be suitable to display a 68-kDa KF and can likely be applied to other polymerases. The applicability and stability of the platform were validated in a directed evolution campaign to significantly improve the incorporation of 2′-OMe modified nucleotides. 2′-OMe modified nucleotides are broadly applied in the design of nuclease resistant aptamers which are used as therapeutics and in diagnostics. In essence, this study provides a technology platform that expanded the scope of cell-surface displayable enzymes to the realm of (mesophilic) polymerases. The extracellular localization of target polymerases circumvents the otherwise inevitable heat-lysis step in numerous polymerases screening strategies, allowing a more efficient screening close to application conditions, especially for mesophilic polymerases.