Continuous directed evolution of proteins with improved soluble expression.

We report the development of soluble expression phage-assisted continuous evolution (SE-PACE), a system for rapidly evolving proteins with increased soluble expression. Through use of a PACE-compatible AND gate that uses a split-intein pIII, SE-PACE enables two simultaneous positive selections to evolve proteins with improved expression while maintaining their desired activities. In as little as three days, SE-PACE evolved several antibody fragments with >5-fold improvement in expression yield while retaining binding activity. We also developed an activity-independent form of SE-PACE to correct folding-defective variants of maltose-binding protein (MBP) and to evolve variants of the eukaryotic cytidine deaminase APOBEC1 with improved expression properties. These evolved APOBEC1 variants were found to improve the expression and apparent activity of Cas9-derived base editors when used in place of the wild-type cytidine deaminase. Together, these results suggest that SE-PACE can be applied to a wide variety of proteins to rapidly improve their soluble expression.

Introduction

Soluble protein expression is a critical requirement for the production and application of proteins. E. coli is generally the most convenient and commonly used organism for protein expression. It is estimated, however, that < 50% of bacterial and < 15% of non-bacterial proteins can be expressed in soluble form in E. coli; common barriers to soluble expression include proteolysis or misfolding into inclusion bodies[1]. Moreover, engineering or evolving proteins towards improved or novel function can often lead to reduced soluble expression, impeding the development and application of proteins with tailor-made functional properties[2]. While it is possible to optimize expression conditions[3], vary host strains[4], or fuse the protein of interest (POI) to solubility-enhancing tags[5], an attractive complementary approach is to engineer or evolve the POI itself to improve its soluble expression under conditions that preserve desirable preexisting activities.

To this end, researchers have successfully used directed evolution to identify protein variants that are resistant to high temperatures[6], denaturants[7], or proteases[8]. The resulting proteins possess enhanced thermodynamic stability, a property that often, but not always, leads to improved soluble expression[2]. Screening protein libraries in whole cells or cell lysates to identify highly active variants can also coincidentally yield proteins with superior expression[9], though this approach is often confounded by variants with higher activity and no improvements in expression. Finally, folding reporters measure soluble protein expression levels through fusion of the POI to a reporter protein such as GFP[10,11]. These reporters enable the selection of soluble variants outside of the context of protein activity, offering broad applicability but with potential loss in function.

In addition to the challenge associated with simultaneously yet independently selecting for soluble expression and protein function, traditional directed evolution approaches to improving protein expression introduce additional drawbacks, including substantial time and effort requirements[12]. Each round of traditional laboratory protein evolution can take a week or longer, rendering the prospect of evolving improvements in soluble expression in addition to new or improved functions an unattractive one. As a result, otherwise interesting targets for protein evolution are often not pursued due to the intractability of expression under standard laboratory conditions. A method that allows the rapid improvement of protein expression, coupled with the preservation or improvement of protein function, would therefore offer substantial benefits.

Here we report a phage-assisted continuous evolution (PACE) system for rapidly evolving proteins with improved soluble protein expression in E. coli, either in the presence or absence of a simultaneous selection for protein function. The system uses an AND logic gate that enables PACE under two simultaneous positive selections. We applied this system towards the evolution of antibody single-chain variable fragments (scFv) that canonically misfold and aggregate in the E. coli cytoplasm. In as little as three days, our system evolved several scFv variants with improved cytosolic expression and little to no loss of binding activity. Additionally, using disulfide-free scFvs, our PACE system evolved variants with enhanced thermodynamic stability in as little as five days. In an activity-independent mode, the system yielded APOBEC1 cytidine deaminase variants with improved solubility that enhanced the purification yields and editing activity of base editors in E. coli and mammalian cells. Together, these results establish soluble expression PACE (SE-PACE) as a rapid method to improve the expression and, in some cases, the stability, of a variety of proteins while preserving their function.

Results

A split T7 RNAP reporter for soluble expression PACE

In PACE (), a population of filamentous bacteriophage (selection phage, SP) is continuously diluted out of a fixed-volume vessel (the lagoon) by host E. coli cells. Engineered SPs carry the evolving gene of interest in place of gene III (gIII), which encodes the minor coat protein III (pIII) that is required to generate infectious progeny phage. SPs containing desired (active) variants of the gene of interest trigger gIII expression and pIII production from an accessory plasmid (AP) in the host cells. Progeny phage production scales with pIII levels[13]. As a result, SPs encoding desired target gene variants produce more infectious offspring capable of replicating faster than they are diluted from the lagoon. Diversity is generated through induction of a mutagenesis plasmid (MP) that dramatically increases SP mutation rates[14]. PACE has been successfully applied to a wide range of proteins, including polymerases[15-17], proteases[18,19], genome-editing proteins[20], insecticidal toxins[21], and aminoacyl-tRNA synthetases[22].

To link target protein expression to the production of pIII and phage propagation, we sought to render T7 RNA polymerase (T7 RNAP) activity dependent on the soluble expression of a target protein (), and then use the resulting T7 RNAP activity to drive gene III expression[15-19,22]. Although full-length T7 RNAP (883 amino acids) is too large to use as a folding reporter,[11] T7 RNAP can be split between amino acids 179 and 180 to produce two fragments that are individually inactive but spontaneously associate into functional T7 RNAP[23,24].

We reasoned that by fusing the smaller N-terminal fragment of this split T7 RNAP (T7n, amino acids 1–179) to the C-terminus of the protein of interest (POI) and providing the larger T7 RNAP fragment (T7c, amino acids 180–883) in the host cell, expression of a PT7-gIII transcript would be dependent on soluble POI expression. Previous folding reporter studies[11] suggest that the overall folded state of the POI–T7n fusion should be strongly determined by the degree to which the N-terminal POI is soluble and well-folded. Therefore, SP encoding a soluble, well-folded POI–T7n fusion should allow the T7n domain to associate with T7c, restoring T7 RNAP-mediated transcription of gIII from its T7 promoter (PT7) () and enabling propagation of SP encoding the POI during PACE (). In contrast, SP encoding POIs that express poorly, are unstable, or are prone to aggregation or proteolysis should result in less available POI–T7n and reduced T7 RNAP activity, and thus fewer progeny phage ().

To test this SE-PACE selection system, we performed transcriptional activation assays in which bacterial luciferase (rather than gene III) expression is controlled by PT7 (). We fused a variety of known maltose binding protein (MBP) mutants with variable soluble expression properties[25,26] to T7n. We observed a correlation between luminescence signal and soluble expression levels of the MBP mutants ( and ), as well as a weaker relationship between luminescence and the known thermodynamic stability of the MBP mutants (). These results demonstrate that POI–T7n fusions allow split T7 RNAP activity to be linked primarily to the soluble expression levels and, to a lesser extent, the thermodynamic stability of a model POI.

Next, we performed competitive phage propagation experiments using SP encoding wild-type MBP or a destabilized mutant (G32D+I33P) in a PACE format. Unlike wild-type MBP, the G32D+I33P mutant has been shown to express primarily in inclusion bodies[25]. SP encoding wild-type MBP-T7n selectively propagated at the expense of SP encoding MBP(G32D+I33P)-T7n on host E. coli cells containing an AP that constitutively expresses T7c and provides gIII from PT7. After 45 h of continuous propagation at a flow rate of 1 lagoon volume/hour, corresponding to 23 average generations of phage evolution, the SP encoding wild-type MBP were enriched at least 10,000-fold over the SP encoding the mutant MBP (). These findings establish the ability of SE-PACE to support the selective propagation of SPs encoding soluble proteins in a PACE format.

Finally, we applied this selection to evolve variants with improved solubility starting with the two most poorly expressed MBP mutants tested, Y283D and G32D+I33P, in PACE () using enhanced mutagenesis (mutagenesis plasmid MP6, ref. [14]) and a flow rate of 0.5–2 lagoon volumes/hour. After 72 h (80 average generations) of PACE, we observed complete reversion of the Y283D mutant to wild-type MBP (). The lagoon seeded with the MBP(G32D+I33P) mutant evolved a P33T substitution at the critical position 33, which improved expression levels by ~2-fold (). Together, these results suggest that the split T7 RNAP folding reporter can be applied to evolving model proteins with improved soluble expression in PACE.

Development of an AND gate for dual selection in PACE

Attempts to further evolve the MBP(G32D+I33T) mutant through SE-PACE with increased selection stringency, achieved through use of APs encoding deactivating mutations in T7c (), predominantly led to the evolution of frameshift mutations in the latter half of MBP. These frameshifts cause premature stop codons upstream of ATG (Met) codons present in the native protein sequence, enabling translational re-initiation at these sites to produce N-terminally truncated T7n fusions that allow the mutant SP to survive the selection in a manner that is not dependent on POI expression (). To overcome this challenge, we sought to add a simultaneous selection for POI function, as truncation products would likely experience defects in activity. Selecting both for soluble expression of T7n and for POI function should also reduce the likelihood of losing protein function while improving expression, a known disadvantage of activity-independent folding reporters[27].

To perform two simultaneous positive selections during PACE, we used a trans-splicing split intein to create an AND gate for pIII-dependent phage propagation. Although a split intein T7 RNAP has been previously reported[28], we reasoned that an AND gate constructed using pIII itself would be compatible with a wider scope of PACE selections. To implement this strategy, we divided pIII into two fragments, each fused to one half of the split intein. Each PACE positive selection leads to the production of one of the two split-intein–pIII halves. If the POI passes only one selection, the other split-intein–pIII half will not be produced, and trans-splicing to generate pIII therefore cannot take place. In contrast, a POI capable of passing both selections results in production of both halves of the split-intein–pIII, allowing the trans-splicing reaction to generate full-length pIII necessary for progeny phage production ().

Because pIII is normally directed to the periplasm after translation[29], we chose to split pIII within its signal peptide sequence to generate translocation-competent pIII only upon trans-splicing. To determine the optimal site, we inserted a Cys-Phe-Asn sequence (CFN; the native C-extein scar left behind by the fast-splicing cyanobacteria DnaE split inteins[30]) at various positions in the hydrophobic region of the signal peptide and evaluated its effect on phage propagation. Insertion of CFN after Ile8 or Leu10 was well-tolerated (), suggesting that pIII might be dividable at these positions. Next, we split pIII between amino acids 10–11 of its signal peptide and fused each half to the Nostoc punctiforme (Npu) DnaE split intein[31] to create split-intein–pIII halves. We expressed both halves of split-intein–pIII from the phage shock protein promoter (PPSP)[15,32] on two separate APs (AP1 and AP2) and tested whether both APs in combination could support the propagation of SP encoding a kanamycin resistance gene in place of gIII (). Host cells co-transformed with both AP1 and AP2 supported robust phage propagation at a level comparable to host cells transformed with a single AP containing full-length gIII. Providing only one half of the split-intein pIII (AP1 only or AP2 only), or mutating residues critical to trans-splicing, resulted in negligible phage propagation, demonstrating the AND gate-like behavior of the split-intein–pIII system ().

Next, we tested if the split-intein–pIII could link two different selections with two separate APs, each expressing one half of split-intein–pIII under control of a selection for protein binding activity or protein expression (). For protein binding activity, we used the HA4 monobody–ABL1 kinase SH2 domain interaction[33], as we previously showed that SP encoding HA4 propagate robustly on an AP containing a 434cI–SH2 domain fusion[21]. Because monobodies are typically highly soluble, we expected that HA4 would also pass the protein expression selection. Indeed, only SP encoding a T7n–HA4 fusion, which passes both the protein binding and soluble expression selections, form plaques on cells harboring the two split-intein–pIII APs (). Conversely, SPs that pass only one selection (encoding either HA4 or T7n alone) exhibit no plaque activity. These results suggest that the split-intein–pIII system should enable two distinct, simultaneous positive selections in PACE. Finally, we used split-intein–pIII to evolve the binding-impaired HA4 Y87A mutant[33] and successfully rescued its ABL1 SH2 domain binding activity (), demonstrating the ability of our system to restore targeted protein binding activity in PACE.

PACE with the split-intein–pIII dual-selection system

Single-chain variable fragments (scFvs) are fusion proteins that consist of the VH and VL variable domains of antibodies joined with a flexible linker[34]. While they retain the epitope binding affinity and specificity of their parent antibodies, scFvs offer several advantages including substantially smaller size and compatibility with expression in E. coli. Typically, the VH and VL domains each contain a stabilizing disulfide bond, which is unable to form in the reducing environment of the E. coli cytosol, leading to very poor expression. Because of this limitation, engineering and evolving these proteins to improve their cytosolic expression has been the focus of major effort[35].

We hypothesized that scFv soluble expression could be improved while preserving binding activity using the dual-selection PACE system (). We began with an scFv targeting the yeast GCN4 transcription factor leucine zipper[36]. Initially isolated through ribosome display, this scFv was extensively engineered for improved expression to ultimately yield the Ω graft (Ωg) variant[37]. Despite these efforts, the majority of Ωg remains insoluble in the E. coli cytosol[38]. We sought to improve the soluble expression of this scFv with SE-PACE.

We evolved SP encoding Ωg in SE-PACE selecting for both protein binding activity and expression. The protein binding selection used a GCN4 leucine zipper peptide fused to 434cI, while the SP carried Ωg fused to rpoZ (RNA polymerase omega subunit) as well as a PACE-optimized version of T7n (eT7n; ) to minimize fitness gains that could be evolved in T7n independent of mutations in the POI. After 72 h (31 average generations), we observed enrichment of phage carrying several mutations in Ωg that enhanced protein expression to a level comparable to that of m3, a recently identified Ωg variant with greatly improved cytosolic expression ( and ). The best PACE-evolved clones, 29.1.2 (S38L+V40I+F87S+N190D) and 29.1.5 (F87S+G103S+L224V) improved expression > 2.5-fold by gel densitometry () and increased expression yields up to 5-fold (with a range of 2- to 17-fold) (). The individual mutations comprising 29.1.5 were found to have a positive effect on expression levels when introduced into Ωg; however, none of these single point mutants expressed as well as the triple F87S+G103S+L224V mutant ( and ). The binding affinities of Ωg and 29.1.5 for the GCN4 leucine zipper sequence were comparable by ELISA, while 29.1.2 exhibited approximately five-fold lower affinity ( and ).

Our PACE-evolved variants showed slightly decreased (~2 °C lower) melting temperatures (Tm) compared to Ωg ( and ). This may result from the discrepancy between evolving in an environment that requires folding without disulfide formation and conducting Tm measurements on purified protein under oxidizing conditions. To probe the disulfide dependence of scFvs evolved in SE-PACE, we mutated the disulfide-forming cysteines in m3 and 29.1.5 to serines (denoted m3-noCys and 29.1.5-noCys, respectively), decreasing the Tm of each scFv by > 10 °C (). PACE of these disulfide-free scFvs selecting for both protein binding and expression resulted in variants with improved thermodynamic properties ( and ). For 29.1.5-noCys, this effect was dramatic, with one clone (58.3.1) possessing an improvement in Tm of > 10 °C. Additionally, variants 58.1.1 and 58.1.2 showed a + 5 °C increase in Tm when compared to m3-noCys (). These results suggest that, despite selecting primarily for protein expression, our system is capable of producing variants with improved thermodynamic stability when other factors influencing stability such as disulfide bond formation are removed.

Finally, we used the dual PACE selection system to evolve a second scFv, C4, which recognizes the first 17 amino acids of the huntington (Htt) protein and has been shown to reduce aggregation of Htt fragments with large polyglutamine expansions[39]. 120 h (70 average generations) of dual selection PACE with a protein binding AP encoding the first 17 amino acids of Htt and a moderately stringent soluble expression AP enriched for two different mutations that each increased soluble expression ( and ). An additional 114 h (94 average generations) of PACE using the same protein binding AP and a more stringent protein expression AP produced genotypes that further increased soluble expression by > 2-fold compared to C4 by gel densitometry ( ) and increased expression yields up to 6-fold (). Similar to the PACE-evolved Ωg variants, these mutants did not possess higher stability (). Two evolved anti-Htt scFvs (34.1.2 and 34.2.3) exhibited similar binding as C4 to a peptide containing the first 17 amino acids of htt, while the third characterized variant (34.2.6) showed slightly reduced affinity ( and ). Together, these results suggest that dual selection SE-PACE is capable of producing improvements in protein expression with little or no loss of binding activity.

Effect of activity selection on expression improvements

In some directed evolution experiments[9], including ones employing PACE[19], an activity-based selection alone can be sufficient to produce gains in expression. To investigate the contribution of the activity selection to improving protein expression, we evolved MBP(G32D+I33T) for binding to the anti-MBP monobody YSX1[40]. We performed two parallel PACE experiments: one simultaneously selecting for protein binding and protein expression (P26), and one selecting for protein binding alone (using a protein expression AP with zero stringency; P28). Most of the mutations fixed in both experiments mapped to the YSX1-MBP binding interface (), suggesting strong selection pressure for protein binding was present. However, the phage pool that experienced selection for protein expression enriched an additional mutation at Gly32 that was not present in the phage pool selected on protein binding alone (). Variants from P28 exhibited improved soluble expression when compared to MBP(G32D+I33T) (). However, variants from P26 containing the G32D mutation showed even greater increases in expression (); in the case of 26.5, soluble expression was comparable to that of wild-type MBP. These results confirm previous observations[2] that selections for activity can increase protein soluble expression, but also suggest that a dedicated selection for soluble expression can further improve this property.

Activity-independent soluble expression PACE

Although the simultaneous positive selection approach possesses the advantage of preserving protein activity while improving expression, it requires an additional PACE selection for activity. An activity-independent selection for improving protein expression might provide a useful complement to the dual selection system. To achieve a useful activity-independent SE-PACE selection, we sought to address the primary source of “cheating” observed with the split T7 RNAP folding reporter—the formation of premature truncation products upstream of translation-initiating Met residues. We hypothesized that this problem could be circumvented through judicious application of the protein binding selection, which requires the translation of rpoZ (the omega subunit of E. coli RNA polymerase) fused to the target protein. If we added a small affinity tag at the N-terminus of the target protein in the POI–T7n fusion and fused rpoZ to the C-terminus, then the resulting tag–POI–T7n–rpoZ construct must be translated in its entirety in order to maintain covalent linkage of the tag and rpoZ. In this system, truncations in the middle of the POI–T7n fusion would separate the tag and rpoZ, impairing the ability of the POI to recruit rpoZ to a tag-binding protein localized to the gene III promoter ().

We used the GCN4 leucine zipper peptide epitope as our affinity tag and the anti-GCN4 scFv clone m3 as the tag-binding protein fused to 434cI, thereby localizing tag-linked proteins to the gene III promoter (). To validate this activity-independent, cheater-resistant selection strategy, we returned to the evolution of the G32D+I33P mutant of MBP, which was originally frustrated by enrichment of truncating genotypes. An SP encoding the GCN4 peptide–[MBP(G32D+I33P)]–eT7n–rpoZ quadruple fusion was subjected to PACE with host cells transformed with split-intein pIII APs selecting for both GCN4 peptide binding and soluble protein expression. After 96 h, we obtained enrichment of SPs bearing mutations fully reverting Asp32 and substituting either Thr (50.1.5) or Leu (50.2.2) for Pro33 (). A third genotype, G24V+P33S (50.2.1), was also observed. All three genotypes showed greatly enhanced (> 10-fold) expression relative to the starting G32D+I33P mutant ( ). 50.1.5 and 50.2.2 also possessed higher thermodynamic stability than G32D+I33P (). We did not observe full reversion to the wild-type genotype, likely because a Pro to Ile substitution requires changing three adjacent nucleotides, while substitution to Thr or Leu requires mutation of only a single nucleotide and is more frequently accessed during random mutagenesis. Notably, no truncating genotypes were observed among SP surviving this PACE, suggesting that the peptide affinity tag strategy enabled successful evolution of protein expression in the absence of a selection for protein activity without evolving cheaters.

Finally, we applied this cheater-resistant activity-independent protein expression evolution to improving the soluble expression of rat apolipoprotein B mRNA editing catalytic subunit 1 (rAPOBEC1). APOBEC1 is a potent cytidine deaminase that can act on both RNA and DNA. Like many eukaryotic proteins, rAPOBEC1 expresses poorly in E. coli and localizes almost exclusively to the insoluble fraction (). We evolved SP encoding rAPOBEC1 for a total of 370 h (260 average generations) in PACE sequentially using three protein expression APs of increasing stringency (). After the first 74 h, SPs containing the two dominant genotypes were pooled and this mixture was used to infect two parallel, identically-run lagoons. At 186 h, both lagoons converged on mutations (F113C+A123E+F205S) that yielded the 36.1 genotype (); however, lagoon 1 had also collected an inactivating mutation (E63A). The final 184 h of PACE produced two variants (43.1 and 43.2) that enhanced expression > 4-fold ( and ). As 43.2 was identified from lagoon 1, it also carried the E63A mutation, which was reverted to obtain 43.2-rev (). When tested in a rifampicin resistance assay[41], neither 36.1 nor 43.1 showed defects in apparent deaminase activity. Evolved APOBEC clone 43.2-rev also showed comparable activity to wild-type rAPOBEC1, while 43.2 was completely inactive as expected (). Clones 36.1, 43.1, and 43.2-rev all had a negative impact on cell growth rate when expressed, while wild-type rAPOBEC1 and 43.2 exhibited no such effect, suggesting that the observed growth inhibition may be due to increased expression of active deaminase ().

These SE-PACE-evolved APOBEC1 variants also increased soluble expression levels when incorporated into base editors, engineered genome editing proteins consisting of a fusion of rAPOBEC1, a catalytically impaired Cas9, and an uracil glycosylase inhibitor that enable targeted conversion of individual DNA base pairs in living cells[42-45]. Substitution of wild-type rAPOBEC1 with 36.1, 43.1, and 43.2-rev in the base editor BE3 improved soluble expression yields by 2–3 fold ( and ), despite the fact that rAPOBEC1 only accounts for ~15% of the total BE3 protein by molecular weight. Finally, base editors incorporating evolved APOBEC variants 36.1, 43.1, and 43.2-rev exhibited higher apparent editing activity in E. coli, as measured by the ability of BE2 (ref. [42]), which contains dCas9 in place of Cas9 nickase, to rescue an active site mutation (H193R) in chloramphenicol acetyl transferase[44] (). Increased cytidine deamination was also observed in HEK293T cells transfected with BE3 variants 36.1 and 43.1 (), albeit with modestly decreased product purities (). Since these APOBECs were not selected for deaminase activity during PACE, we attribute deaminase efficiency increases primarily to improvements in base editor expression. Together, these findings establish that activity-independent SE-PACE can improve the soluble expression of potentially growth-inhibiting proteins, in some cases without loss of activity.

Discussion

We have developed methods for the continuous evolution of proteins with improved soluble expression. Our approach quickly generates improved variants without the need for library cloning or discrete, time-intensive steps usually required to implement protein directed evolution methods. Through the use of a PACE-compatible split-intein pIII which acts as an AND gate to bridge two orthogonal positive selections, SE-PACE is capable of evolving proteins in either the presence or absence of a simultaneous selection for protein activity.

The split-intein pIII AND gate allows for two positive selections to take place in the same PACE experiment. In this work, we demonstrated its ability to bridge selections for target protein binding and soluble protein expression. In theory, however, this strategy can link any two orthogonal PACE selections. In addition, we show that the split-intein pIII can be used to constrain evolving SP populations to disfavor undesirable outcomes such as the formation of truncation products that cheat an activity-independent soluble expression selection.

The SE-PACE-evolved proteins did not always show improvements in Tm, and in some cases exhibited small decreases. As PACE is conducted at 37 °C and the starting proteins melted well above this temperature (Tm > 55 °C), no direct selection pressure was present for improved thermodynamic stability, and increases in Tm likely resulted from mutations that improved soluble expression. Since thermodynamic stability and soluble expression are not always correlated, SE-PACE’s selection for the latter property may not necessarily produce improvements in the former. With respect to expression yields, the evolved anti-GCN4 and anti-Htt scFvs reported here exhibited increases in the 2–6 fold range. In comparison, previous directed evolution efforts toward improving expression of scFv and related antibody-derived proteins in E. coli have produced variants that increase yields by 2 to >10 fold[38,46,47].

We have demonstrated that simultaneous selection for two different traits (here, soluble expression and binding activity) is possible in PACE through the evolution of scFvs with enhanced soluble expression in the E. coli cytosol. All dual-selection evolved scFvs tested showed binding to their target epitopes similar to that of the starting ScFv. In contrast, the activity-independent SE-PACE of rAPOBEC1 yielded some variants that lost enzymatic activity, as expected.

We have shown that single- and dual-selection SE-PACE can improve the expression of a diverse variety of proteins, including proteins of bacterial origin (MBP), engineered antibodies (scFvs), and eukaryotic enzymes (rAPOBEC1). It may also be interesting to apply SE-PACE to proteins that have emerged from long directed evolution campaigns, which often result in proteins with diminished stability[21,48]. By enforcing selection for protein activity using the split-intein pIII AND gate, it should be possible to preserve newly acquired protein function while improving soluble expression. Therefore, SE-PACE may also provide a way to rapidly reverse the deleterious effects of extensive mutation to improve the utility of previously engineered or evolved proteins.

Online Methods

General methods.

Antibiotics (Gold Biotechnology) were used at the following working concentrations: carbenicillin 50 μg/mL, spectinomycin 50 μg/mL, chloramphenicol 25 μg/mL, kanamycin 50 μg/mL, tetracycline 10 μg/mL, streptomycin 50 μg/mL. HyClone water (GE Healthcare Life Sciences) was used for PCR reactions and cloning. For all other experiments, water was purified using a MilliQ purification system (Millipore). Phusion U Hot Start DNA polymerase (Thermo Fisher Scientific) was used for all PCRs. Plasmids and SPs were cloned by USER assembly[21]. Genes were obtained as synthesized gBlock gene fragments from Integrated DNA Technologies or PCR amplified directly from E. coli genomic DNA. Plasmids were cloned and amplified using either Mach1 (Thermo Fisher Scientific) or Turbo (New England BioLabs) cells. Unless otherwise noted, plasmid or SP DNA was amplified using the Illustra Templiphi 100 Amplification Kit (GE Healthcare Life Sciences) prior to Sanger sequencing. A full list of plasmids used in this work is given in . A full list of reagents and equipment used in this work is given in .

Preparation and transformation of chemically competent cells.

Strain S206020 was used in all luciferase, phage propagation, and plaque assays, and in all PACE experiments. To prepare competent cells, an overnight culture was diluted 1000-fold into 50 mL of 2xYT media (United States Biologicals) supplemented with tetracycline and streptomycin and grown at 37˚C with shaking at 230 RPM to OD600 ~ 0.4–0.6. Cells were pelleted by centrifugation at 4000 g for 10 minutes at 4˚C. The cell pellet was then resuspended by gentle stirring in 2 mL of ice-cold LB media (United States Biologicals) 2 mL of 2x TSS (LB media supplemented with 5% v/v DMSO, 10% w/v PEG 3350, and 20 mM MgCl2) was added. The cell suspension was stirred to mix completely, aliquoted and frozen on dry ice, and stored at −80˚C until use.

To transform cells, 100 μL of competent cells thawed on ice was added to a pre-chilled mixture of plasmid (2 μL each; up to 3 plasmids per transformation) in 95 μL KCM solution (100 mM KCl, 30 mM CaCl2, and 50 mM MgCl2 in H2O) and stirred gently with a pipette tip. The mixture was incubated on ice for 10 min and heat shocked at 42˚C for 75 s before 600 μL of SOC media (New England BioLabs) was added. Cells were allowed to recover at 37˚C with shaking at 230 RPM for 1.5 h, streaked on 2xYT media + 1.5% agar (United States Biologicals) plates containing the appropriate antibiotics, and incubated at 37˚C for 16–18 h.

Luciferase assay.

S2060 cells were transformed with the AP(s) and CP(s) of interest as described above. Overnight cultures of single colonies grown in 2xYT media supplemented with maintenance antibiotics were diluted 500-fold into DRM media[17] with maintenance antibiotics in a 96-well deep well plate (Axygen). The plate was sealed with a porous sealing film and grown at 37˚C with shaking at 230 RPM for 4 h, whereupon the culture reached OD600 ~ 0.4–0.6. The cells were then induced with 100 ng/mL anhydrotetracycline (aTc; Fluka) and 5 μM arabinose (Gold Biotechnology) before incubation for an additional 1 h at 37˚C with shaking at 230 RPM. 120 μL of cells was transferred to a 96-well black-walled clear-bottomed plate (Costar), then 600 nm absorbance and luminescence were read using an Infinite M1000 Pro microplate reader (Tecan). OD600-normalized luminescence values were obtained by dividing raw luminescence by background-subtracted 600 nm absorbance. The background value was set to the 600 nm absorbance of wells containing DRM only.

Phage propagation assay.

S2060 cells were transformed with the AP(s) of interest as described above. Overnight cultures of single colonies grown in 2xYT media supplemented with maintenance antibiotics were diluted 1000-fold into DRM media with maintenance antibiotics and grown at 37˚C with shaking at 230 RPM to OD600 ~ 0.4–0.6. Cells were then infected with SP at a starting titer of 5 × 104 pfu/mL. Cells were incubated for another 16–18 h at 37˚C with shaking at 230 RPM, then centrifuged at 8000 g for 2 min. The supernatant containing phage was removed and stored at 4˚C until use.

Plaque assay.

S2060 cells were transformed with the AP(s) of interest as described above. Overnight cultures of single colonies grown in 2xYT media supplemented with maintenance antibiotics were diluted 1000-fold into fresh 2xYT media with maintenance antibiotics and grown at 37˚C with shaking at 230 RPM to OD600 ~ 0.6–0.8 before use. SP were serially diluted 100-fold (4 dilutions total) in H2O. 150 μL of cells was added to 10 μL of each phage dilution and to this 1 mL of liquid (55˚C) top agar (2xYT media + 0.6% agar) supplemented with 2% Bluo-gal (Gold Biotechnology) was added and mixed by pipetting up and down once. This mixture was then immediately pipetted onto one quadrant of a quartered Petri dish already containing 2 mL of solidified bottom agar (2xYT media + 1.5% agar, no antibiotics). After solidification of the top agar, plates were incubated at 37˚C for 16–18 h.

Phage-assisted continuous evolution.

Unless otherwise noted, PACE apparatus, including host cell strains, lagoons, chemostats, and media, were all used as previously described[17,21]. To reduce the likelihood of contamination with gIII-encoding recombined SP, phage stocks were purified as previously described[21].

Chemically competent S2060s were transformed with AP(s) and MP6 or DP6 as described above, plated on 2xYT media + 1.5% agar supplemented with 25 mM glucose (to prevent induction of mutagenesis) in addition to maintenance antibiotics, and grown at 37˚C for 18–20 h. Four colonies were picked into 1 mL DRM each in a 96-well deep well plate, and this was diluted 5-fold eight times serially into DRM. The plate was sealed with a porous sealing film and grown at 37˚C with shaking at 230 RPM for 16–18 h. Dilutions with OD600 ~ 0.4–0.8 were then used to inoculate a chemostat containing 80 mL DRM. The chemostat was grown to OD600 ~ 0.8–1.0, then continuously diluted with fresh DRM at a rate of ~1.5 chemostat volumes/h as previously described[17]. The chemostat was maintained at a volume of 60–80 mL.

Prior to SP infection, lagoons were continuously diluted with culture from the chemostat at 1 lagoon volume/h and pre-induced with 10 mM arabinose for at least 2 h. If DP6 was used, the lagoons were also pre-induced with aTc. Lagoons were infected with SP at a starting titer of 106 pfu/mL and maintained at a volume of 15 mL. Samples (500 μL) of the SP population were taken at indicated times from lagoon waste lines. These were centrifuged at 8000 g for 2 min, and the supernatant was passed through a 0.22 μm PVDF Ultrafree centrifugal filter (Millipore) and stored at 4˚C. Lagoon titers were determined by plaque assays using S2060 cells transformed with pJC175e[15]. For Sanger sequencing of lagoons, single plaques were PCR amplified using primers AB1793 (5’-TAATGGAAACTTCCTCATGAAAAAGTCTTTAG) and AB1396 (5’- ACAGAGAGAATAACATAAAAACAGGGAAGC), both of which anneal to regions of the phage backbone flanking the evolving gene of interest. Generally, eight plaques were picked and sequenced per lagoon.

Continuous flow competition of MBP and G32D+I33P.

Host cells transformed with pTW006aP1a were maintained in a 40 mL chemostat. Lagoons were infected with a 1000:1 ratio of SP02a (G32D+I33P) to SP01a (MBP). No inducers were used in this continuous flow experiment. The SP pool was monitored by PCR using primers AB1793 (5’-TAATGGAAACTTCCTCATGAAAAAGTCTTTAG), which anneals to the phage backbone 5’ of the MBP–T7n insert, and TW0023 (5’-ACCACCAGATCCACCCUTGGTGATACGAGTCTGCG), a USER primer which anneals to the 3’ end of MBP. The purified PCR product (200 ng) was digested with BsaWI (New England BioLabs), which cuts a ACC|GGA motif in MBP that is absent in the G32D+I33P mutant, according to manufacturer’s protocol.

Evolution of MBP Y283D and G32D+I33P.

Host cells transformed with pTW006aP1a and MP6 were maintained in a 40 mL chemostat. Lagoons were infected with SP02a (G32D+I33P) or SP02b (Y293D). Upon infection, lagoon dilution rates were decreased to 0.5 volume/h. For G32D+I33P, the lagoon dilution rate was maintained at 0.5 volume/h throughout the experiment. For Y283D, the lagoon dilution rate was increased to 1 volume/h at 22 h. Both experiments ended at 72 h.

Evolution of HA4 Y87A monobody.

Host cells transformed with pTW048a3, pTW051b2, and DP6 were maintained in a 40 mL chemostat. The lagoon was cycled at 1 volume/h with 10 mM arabinose and 20 ng/mL aTc for 3 h prior to infection with SP10b2. Upon infection, lagoon dilution rates were maintained at 1 volume/h. The lagoon dilution rate was increased to 1.5 volume/h at 93 h and 2 volumes/h at 119 h. The experiment ended at 122 h.

Evolution of anti-GCN4 scFv Ωg.

Host cells transformed with pTW055a3, pTW051b4, and MP6 were maintained in a 40 mL chemostat. Lagoons were infected with SP16c3. Upon infection, lagoon dilution rates were decreased to 0.5 volume/h. After 19 h, the lagoon dilution rate was increased to 1 volume/h. The experiment ended at 72 h.

Evolution of disulfide-free anti-GCN4 scFvs.

Host cells transformed with pTW055a3, pTW051b4, and MP6 were maintained in a 40 mL chemostat. Lagoons were infected with SP39b or SP39d. Upon infection, lagoon dilution rates were decreased to 0.5 volume/h. For SP39b, the lagoon dilution rate was increased to 1 volume/h after 19 h, 1.5 volumes/h after 44 h, 2 volumes/h after 68 h, and 3 volumes/h after 90 h. For SP39d, the lagoon dilution rate was increased to 1 volume/h after 44 h, 1.5 volumes/h after 68 h, and 2 volumes/h after 90 h. The experiment ended at 120 h.

Evolution of anti-htt scFv C4.

Host cells transformed with pTW074c, pTW051b4, and MP6 were maintained in an 80 mL chemostat. Lagoons were infected with SP24a3. Upon infection, lagoon dilution rates were decreased to 0.5 volume/h. The lagoon dilution rate was increased to 1 volume/h at 42 h, 1.5 volumes/h at 70 h, 2 volumes/h at 90 h, and 3 volumes/h at 114 h. The experiment ended at 120 h.

Further evolution of anti-htt scFv C4.

Host cells transformed with pTW074c, pTW051d2, and MP6 were maintained in a 40 mL chemostat. Lagoons were infected with the 120 h SP pool from the previous PACE. Upon infection, lagoon dilution rates were decreased to 0.5 volume/h. The lagoon dilution rate was increased to 1 volume/h at 18 h, 1.5 volumes/h at 42 h, 2 volumes/h at 69 h, and 3 volumes/h at 90 h. The experiment ended at 114 h.

Evolution of MBP G32D+I33P with affinity tag strategy.

Host cells transformed with pTW084b, pTW051b4, and MP6 were maintained in an 80 mL chemostat. Lagoons were infected with SP27b. Upon infection, lagoon dilution rates were decreased to 0.5 volume/h. For lagoon 1, the dilution rate was increased to 1 volume/h at 44 h and 2 volumes/h at 69 h. For lagoon 2, the dilution rate was increased to 1 volume/h at 44 h, 1.5 volumes/h at 69 h, and 2 volumes/h at 88 h. The experiment ended at 112 h.

Evolution of rAPOBEC1 with affinity tag strategy.

Host cells transformed with pTW084b, pTW051d, and DP6 were maintained in a 40 mL chemostat. The lagoon was cycled at 1 volume/h with 10 mM arabinose and 20 ng/mL aTc for 4 h prior to infection with SP30. Upon infection, lagoon dilution rates were decreased to 0.5 volume/h. At 18 h, the aTc concentration was decreased to 0 ng/mL. The lagoon dilution rate was increased to 1 volume/h at 28 h and 1.5 volumes/h at 66 h. The experiment ended at 74 h.

Evolution was continued on host cells transformed with pTW084b, pTW051d, and MP6 in an 80 mL chemostat. Lagoons were infected with a 1:1 ratio of 35.1 and 35.2, SP clones isolated from the previous experiment. Upon infection, lagoon dilution rates were maintained at 1 volume/h. The lagoon dilution rate was increased to 1.5 volume/h at 18 h, 2 volumes/h at 66 h, and 3 volumes/h at 90 h. The experiment ended at 112 h.

Evolution was continued on host cells transformed with pTW084b, pTW051b4, and MP6 in an 80 mL chemostat. Lagoons were infected with either 36.1 or 36.2, SP clones isolated from the previous experiment. Upon infection, lagoon dilution rates were decreased to 0.5 volume/h. The lagoon dilution rate was increased to 1 volume/h at 17 h, 1.5 volumes/h at 41 h, and 2 volumes/h at 66 h. At 96 h, the chemostat was replaced with a fresh 80 mL chemostat containing host cells transformed with pTW084b, pTW051d2, and MP6, and the lagoon dilution rate was decreased to 1 volume/h. The lagoon dilution rate was increased to 1.5 volumes/h at 115h and 2 volumes/h at 168 h. The experiment ended at 184 h.

Small-scale protein expression.

BL21 DE3 cells (New England BioLabs) were transformed with the expression plasmids (EPs) of interest according to manufacturer protocol. Overnight cultures of single colonies grown in 2xYT media supplemented with maintenance antibiotics were diluted 1000-fold into fresh 2xYT media (2 mL) with maintenance antibiotics and grown at 37˚C with shaking at 230 RPM to OD600 ~ 0.4–0.6 before induction with 1 mM isopropyl-β-D-thiogalactoside (IPTG; Gold Biotechnology) or rhamnose (Gold Biotechnology). Cells were grown for a further 3 h at 37˚C with shaking at 230 RPM. Cells from 1.4 mL of culture were isolated by centrifugation at 8000 g for 2 min. The resulting pellet was resuspended in 150 μL B-per reagent (Thermo Fisher Scientific) supplemented with protease inhibitor cocktail (Roche) and incubated on ice for 10 min before centrifugation at 16,000 g for 2 min. The supernatant was collected as the soluble fraction and the resulting pellet was resuspended in an additional 150 μL B-per reagent to obtain the insoluble fraction. To 37.5 μL of each fraction was added 12.5 μL 4x Laemmli sample loading buffer (Bio-Rad) containing 2 mM Dithiothreitol (DTT; Sigma Aldrich). After vortexing, the fractions were incubated at 95˚C for 10 min. 12 μL of each soluble fraction and 6 μL of each insoluble fraction was loaded into the wells of a Bolt 4–12% Bis-Tris Plus (Thermo Fisher Scientific) pre-cast gel. 6 μL of Precision Plus Protein Dual Color Standard (Bio-Rad) was used as a reference. Samples were separated by electrophoresis at 180 V for 35 min in Bolt MES SDS running buffer (Thermo Fisher Scientific). Gels were stained with InstantBlue reagent (Expedeon) for 1 h to overnight, then washed several times with H2O before imaging with a G:Box Chemi XRQ (Syngene). Band densities were quantified using ImageJ and normalized to reference bands to control for protein loading.

Western blot analysis.

SDS-PAGE was performed as described above. Transfer to a PVDF membrane was performed using an iBlot 2 Gel Transfer Device (Thermo Fisher Scientific) according to manufacturer’s protocols. The membrane was blocked in Odyssey Blocking Buffer (LI-COR) in PBS for 1 h at room temperature, then incubated with mouse anti-6xHis (abcam ab18184; 1:2000 dilution) and rabbit anti-GroEL (Sigma-Aldrich G6532; 1:20,000 dilution) in SuperBlock Blocking Buffer (Thermo Fisher Scientific) overnight at 4˚C. The membrane was washed 3x with TBST (TBS + 0.5% Tween-20) for 10 min each at room temperature, then incubated with IRDye-labeled secondary antibodies goat anti-mouse 680RD (LI-COR 926–68070) and donkey anti-rabbit 800CW (LI-COR 926–32213) diluted 1:5000 in SuperBlock for 1 h at room temperature. The membrane was washed as before, then imaged using an Odyssey Imaging System (LI-COR).

Medium-scale protein expression and purification.

BL21 DE3 cells were transformed with the EPs of interest according to manufacturer protocol. Overnight cultures of single colonies grown in LB or 2xYT media supplemented with maintenance antibiotics were diluted 1000-fold into fresh LB or 2xYT media (250 mL) with maintenance antibiotics and grown at 37˚C with shaking at 230 RPM to OD600 ~ 0.4–0.6. Cells were chilled on ice for 1 h, then induced with 1 mM IPTG and grown for a further 16–18 h at 16˚C with shaking at 200 RPM. For 37˚C post-induction growth, the cold shock step was omitted and the cells were grown for a further 3 h at 37˚C with shaking at 200 RPM after induction with 1 mM IPTG. Cells were isolated by centrifugation at 8000 g for 10 min. The resulting pellet was resuspended in 4 mL B-per reagent supplemented with EDTA-free protease inhibitor cocktail (Roche) and incubated on ice for 20 min before centrifugation at 12,000 g for 15 min. The supernatant was decanted into a 15 mL conical tube and incubated with 250 μL of TALON Cobalt (Clontech) resin at 4˚C with constant agitation for 1–2 h, after which the resin was isolated by centrifugation at 500 g for 5 min. The supernatant was decanted, and the resin resuspended in 2 mL binding buffer (50 mM NaH2PO4, 300 mM NaCl, 20 mM imidazole, pH 7.8) and transferred to a column. The resin was washed 4x with 1 mL binding buffer before protein was eluted with 1 mL of binding buffer containing increasing concentrations of imidazole (50–250 mM in 50 mM increments). The fractions were analyzed by SDS-PAGE for purity. Combined fractions were buffer-exchanged with TBS (20 mM Tris-Cl, 500 mM NaCl, pH 7.5) and concentrated using an Amicon Ultra-15 centrifugal filter unit (10,000 molecular weight cutoff; Millipore), then stored at 4˚C until further use. Proteins were quantified using Quick Start Bradford reagent (Bio-Rad) using BSA standards (Bio-Rad).

Medium-scale BE3 expression and purification.

BE3 variants were expressed and purified as described previously[45] with a few modifications. BL21 DE3 cells were transformed with the BE3-expressing EPs of interest according to manufacturer protocol. Overnight cultures of single colonies grown in 2xYT media supplemented with maintenance antibiotics were diluted 1000-fold into fresh 2xYT media (200 mL) with maintenance antibiotics and grown at 37˚C with shaking at 230 RPM to OD600 ~ 0.7–0.8. Cells were chilled on ice for 1 h, then induced IPTG and grown for a further 16–18 h at 16˚C with shaking at 200 RPM. Cells were isolated by centrifugation at 8000 g for 15 min. The resulting pellet was resuspended in 8 mL high salt buffer (100 mM Tris-Cl, 1 M NaCl, 5 mM tris(2-carboxyethyl)phosphine (TCEP; Gold Biotechnology), 20% glycerol, pH 8.0) supplemented with EDTA-free protease inhibitor cocktail and 1 mM phenylmethane sulfonyl fluoride (PMSF; Sigma-Aldrich). Cells were sonicated on ice (3 s on/3 s off; 6 min total) and the lysate centrifuged at 16,000 g for 15 min. The supernatant was decanted into a 15 mL conical tube and incubated with 500 μL of TALON Cobalt resin at 4˚C with constant agitation for 1 h, after which the resin was isolated by centrifugation at 500 g for 5 min. The resin was washed 5x with 1 mL high salt buffer, then eluted with 1 mL of elution buffer (100 mM Tris-Cl, 500 mM NaCl, 5 mM TCEP, 200 mM imidazole, 20% glycerol, pH 8.0). The isolated protein was then buffer-exchanged with medium salt buffer (100 mM Tris-Cl, 500 mM NaCl, 5 mM TCEP, 20% glycerol, pH 8.0) and concentrated using an Amicon Ultra-15 centrifugal filter unit (100,000 molecular weight cutoff). Proteins were quantified using Quick Start Bradford reagent using BSA standards.

Protein melt temperature assay.

Protein melt temperatures were determined using a Protein Thermal Shift Dye Kit (Life Technologies) according to manufacturer’s protocols. Fluorescence was monitoring using a CFX96 Real-Time PCR Detection System (Bio-Rad).

ELISA.

Anti-GCN4 ELISA.

A MaxiSorp 96-well plate (Nunc) was coated with a 10 μg/mL solution of purified antibody fragments in TBS (20 mM Tris-Cl, 500 mM NaCl, pH 7.5), 50 μL per well, and incubated overnight at 4˚C. The wells were washed 3x with TBST (TBS + 0.5% Tween-20) and blocked with 1% w/v Bovine Serum Albumin (BSA; Sigma-Aldrich) in TBS for 1–2 h. This and all subsequent incubations were performed at room temperature. The blocking solution was removed and the wells were incubated with serial dilutions of purified MBP-TEV-GCN4 in TBS for 1 h. The wells were washed 3x with TBST, then incubated with a 1:5000 dilution of HRP-conjugated anti-MBP (abcam ab49923) in TBS for 1 h. The wells were washed 3x with TBST, then developed with 1-Step Ultra TMB-ELISA Substrate (Thermo Fisher Scientific) for 5 min. The reaction was quenched with addition of 2 M H2SO4 before 450 nm absorbance was read using a Tecan Infinite M1000 Pro microplate reader.

Anti-htt ELISA.

A MaxiSorp 96-well plate was coated with a 50 μg/mL solution of purified antibody fragments in TBS, 50 μL per well, and incubated overnight at 4˚C. The wells were washed and blocked as above. The blocking solution was removed and the wells were incubated with serial dilutions of biotinylated htt peptide (MATLEKLMKAFESLKSFK(Biotin)-NH2; New England Peptide) in TBS for 1 h. The wells were washed 3x with TBST, then incubated with a 1:10,000 dilution of HRP-conjugated streptavidin (BioLegend) in TBS for 1 h. The wells were washed 3x with TBST, then developed, quenched, and absorbance read as above.

Rifampicin resistance assay.

BL21 DE3 cells were transformed with the EPs of interest according to manufacturer protocol. Overnight cultures of single colonies grown in DRM media supplemented with maintenance antibiotics were diluted 1000-fold into DRM media with maintenance antibiotics in a 96-well deep well plate. The plate was sealed with a porous sealing film and grown at 37˚C with shaking at 230 RPM for until the culture reached OD600 ~ 0.4. The cells were then either induced with 5 mM rhamnose or repressed with 5 mM glucose before incubation for an additional 16–18 h at 37˚C with shaking at 230 RPM. Cultures were serially diluted on 2xYT + 1.5% agar plates supplemented with 50 μg/mL spectinomycin, 100 μg/mL rifampin (Alfa Aesar), and 25 mM glucose. The total number of colony-forming units (cfus) was determined by serially diluting the same cultures on 2xYT + 1.5% agar plates supplemented with 50 μg/mL spectinomycin and 25 mM glucose. Plates were grown at 37˚C for 16–18 h. The surviving colonies on the plates containing rifampin were counted and this number was normalized to the total cfu count.

Chloramphenicol resistance assay.

Chemically competent S2060s were transformed with selection plasmid[44] and the base editor EPs of interest as described above and streaked on 1.5% agar plates supplemented with 50 μg/mL carbenicillin, 50 μg/mL spectinomycin, and 25 mM glucose. Overnight cultures of single colonies grown in DRM media supplemented with maintenance antibiotics were diluted 500-fold into DRM media with maintenance antibiotics and grown at 37˚C with shaking at 230 RPM for until the culture reached OD600 ~ 0.5. The cells were then induced with 10 mM arabinose and incubated for an additional 16–18 h at 37˚C with shaking at 230 RPM. Cultures were serially diluted on 2xYT + 1.5% agar plates supplemented with 32 or 64 μg/mL chloramphenicol and 25 mM glucose (to repress further expression). The total number of cfus was determined by serially diluting the same cultures on 2xYT + 1.5% agar plates supplemented with 50 μg/mL carbenicillin, 50 μg/mL spectinomycin, and 25 mM glucose. Plates were grown at 37˚C for 16–18 h. The surviving colonies on the plates containing chloramphenicol were counted and this number was normalized to the total cfu count.

HEK293T transfection and genomic DNA extraction.

HEK293T cells (ATCC CRL-3216) maintained in Dulbecco’s Modified Eagle’s Medium plus GlutaMax (Thermo Fisher Scientific) supplemented with 10% (v/v) FBS at 37 °C with 5% CO2 were seeded on 48-well poly-D-lysine coated plates (Corning). 16–20 h after seeding, cells were transfected at approximately 80–85% confluency as previously described[44]. Cells were cultured for 3 days post-transfection, before genomic DNA was extracted as previously described[44].

High-throughput sequencing of genomic DNA.

Genomic sites were amplified with primers targeting the region of interest and the appropriate universal Illumina forward and reverse adapters. Primer pairs for the first round of PCR (PCR 1) were the same as reported previously[42]. 25 μL scale PCR 1 reactions used 1.25 μL each of 10 μM forward and reverse primers and 0.5 μL genomic DNA extract, all diluted to 12.5 μL with nuclease-free water, and 12.5 μL Phusion U Green Multiplex PCR MasterMix (Thermo Fisher Scientific). PCR 1 conditions: 98 °C for 2 min, then 30 cycles of (98 °C for 15 s, 61 °C for 20 s, 72 °C for 15 s), followed by a final 72 °C extension for 2 min. PCR products were verified by comparison with DNA standards (Quick-Load 2-Log Ladder; New England BioLabs) on a 2% agarose gel supplemented with ethidium bromide. Unique Illumina barcoding primers which anneal to the universal Illumina adapter region were subsequently appended to each PCR 1 sample in a second PCR reaction (PCR 2). PCR 2 reactions used 1.25 μL each of 10 μM forward and reverse Illumina barcoding primers and 1 μL of unpurified PCR 1 reaction product, all diluted to 12.5 μL with nuclease-free water, and 12.5 μL Phusion U Green Multiplex PCR MasterMix (Thermo Fisher Scientific). PCR 2 conditions: 98 °C for 2 min, then 12 cycles of (98 °C for 15 s, 61 °C for 20 s, 72 °C for 20 s), followed by a final 72 °C extension for 2 min. PCR products were pooled and purified by electrophoresis with a 2% agarose gel using a Monarch DNA Gel Extraction Kit (New England BioLabs) eluting with 30 μL H2O. DNA concentration was quantified with the KAPA Library Quantification Kit-Illumina (KAPA Biosystems) and sequenced on an Illumina MiSeq instrument (paired-end read – R1: 220 cycles, R2: 0 cycles) according to the manufacturer’s protocols.

General HTS data analysis.

Sequencing reads were demultiplexed in MiSeq Reporter (Illumina). Alignment of amplicon sequences to a reference sequence was performed as previously described using a custom Matlab script[42]. In brief, the Smith-Waterman algorithm was used to align sequences without indels to a reference sequence; bases with a quality score of less than 30 were converted to ‘N’ to prevent base miscalling as a result of sequencing error. Indels were quantified separately using a modified version of a previously described Matlab script in which sequencing reads with more than half the base calls below a quality score of Q30 were filtered out[44]. Indels were counted as reads which contained insertions or deletions of greater than or equal to 1 bp within a 30 bp window surrounding the predicted Cas9 cleavage site.

Base editing values are representative of N = 4 independent biological replicates collected over different days, with the mean ± s.e.m shown. Base editing values are reported as a percentage of the number of reads with cytidine mutagenesis over the combined number of aligned reads and indel-containing reads.

Stastics.

Information regarding the number and type of replicates, along with the type of error depicted, is provided in the figure legends.

Data availability.

High-throughput sequencing data for Figures 5f and SI Figure 19 have been deposited in the NCBI Sequence Read Archive database under accession code SRP152041. Selection plasmids used in this study will be available through Addgene. Other data and materials are available upon reasonable request.

Figure 5.

Activity-independent selection for soluble proteins in PACE.

(a) SE-PACE of the MBP(G32D+I33P) mutant produces partial reversions to the wild-type sequence after four days. (b) PACE-evolved MBP variants show greatly improved soluble expression in E. coli at 37 °C when compared to starting MBP(G32D+I33P). The full gel is provided in . This experiment was not repeated. (c) PACE of cytidine deaminase rAPOBEC1. The stringency of the protein expression selection was modulated through increasing the rate of lagoon dilution (flow rate; dashed lines) as well as the use of activity-diminishing T7 RNAP and T7 promoter mutants in the AP. (d) rAPOBEC1 variants obtained after 186 h and 370 h show greatly improved soluble expression at 37˚C when compared to wild-type. The full blot is provided in . This expression experiment was repeated once with similar results. (e) Base editor (BE2) variants employing evolved rAPOBEC1s show enhanced apparent activity in E. coli, as measured by cells surviving selection on increasing concentrations of chloramphenicol due to BE2-dependent reversion of an active site mutation in chloramphenicol acetyl transferase. Data reflects the mean and s.d. of three technical replicates (unique clones). (f) Base editor (BE3) variants using evolved rAPOBEC1 36.1 and 43.1 variants show enhanced editing efficiency in HEK293T cells, as measured by the percentage of total DNA sequencing reads with Cs in the indicated target positions converted to another base. The sequences of the six genomic loci tested (EMX-RNF2) are provided in . Data reflects the mean and s.e.m. of four biological replicates (experiments performed on different days).

Code availability.

The custom Matlab script used in this work can be found in the Supplementary Information of reference 39.