3.1. Overview of Kunkel mutagenesis
One method frequently used to construct phage-display libraries is Kunkel mutagenesis (). First, a phagemid genome is introduced into E. coli CJ236 cells, which lack functional dUTPase and uracil-N glycosylase. With the aid of a helper virus, M13-K07, the transformed CJ236 cells secrete phage particles, from which single-stranded, circular DNA (ssDNA), containing uracil in place of thymine (i.e., uracilated), is recovered. A pair of mutagenic oligonucleotides are phosphorylated and annealed to the ssDNA template for in vitro synthesis of heteroduplex, double-stranded DNA (dsDNA) with T7 DNA polymerase and T4 DNA ligase. The resulting product is purified and evaluated by agarose gel electrophoresis; as seen in , the ssDNA can be quantitatively converted to the larger dsDNA species by this technique. The dsDNA is electroporated into E. coli TG1 cells, where the parental uracilated strand is cleaved in vivo by uracil-N glycosylase, so that only the recombinant strand survives.
Synthesis of heteroduplex dsDNA in vitro
3.2. Phage replication at lower temperature improved yields of ssDNA
Infected CJ236 cells often have low yields of ssDNA, and of low purity [7
]. To find conditions leading to higher yields and better purity, multiple settings (i.e., growing temperature, shaking speed, different type of flasks) were evaluated. Incubation at 25°C for 22 h consistently yielded 2- to 7-fold more ssDNA than cultures grown at 37°C (). A similar increase in ssDNA yields was also observed for cultures of TG1 cells carrying the M13 bacteriophage genome (data not shown). However, if cells were resuspended in a much higher volume (i.e., 30 fold) of fresh medium after infection, the same growing conditions did not lead to higher yields.
Single-stranded DNA extracted from bacterial cultures grown at two different temperatures
The effects of shaking speed and flask shape on phage yields were also examined. High shaking speed (i.e., 280 rpm), with 37°C incubation, led to lower yields of ssDNA due to excessive degradation of DNA (data not shown). As both greater aeration and higher temperatures lead to rapid cell growth and potential cell lysis [29
], these conditions are not recommended for phage replication. When cultures were grown in baffled and non-baffled flasks, there was no significant change in the yields or purity of the ssDNA (data not shown). In summary, keeping the culture volume the same after infection and incubating it for 22 h at 25°C, with a shaking speed of 200 rpm, contributed to the best results, among all the conditions tested.
3.3. Different reaction conditions for in vitro dsDNA synthesis exhibited minimal influences on mutation rate
As mentioned above, it is beneficial to reduce or eliminate phage particles displaying the ‘wild-type’ form of a scaffold protein, thereby increasing the diversity of a library and enhancing its effectiveness in yielding binders. One way to reduce the number of wild-type clones in a phage library is to identify conditions that improve the efficiency of in vitro synthesis of dsDNA during Kunkel mutagenesis. To learn what modifications of the extension step might lead to enhanced yields, nine different conditions, ranging from different annealing ratios of oligonucleotide/ssDNA (3, 20, and 100) to different extension times (30 min, 3 h, and 16 h), were tested. While, none of these conditions appeared to significantly increase the mutation rate, there are two points worth noting. First, the average mutation rate (i.e., replacement of both loops with degenerate sequences) for all nine conditions was 38 ± 3%. Second, for reactions of all three annealing ratios, as the extension time went from 30 min, 3 h to 16 h, there was a 1-8% increase in the mutation rate from 30 min to 3 h, and then a 3-6% decrease from 3 h to 16 h (data not shown). A 16 h extension not only led to a lower mutation rate, but also a significantly lower yield of dsDNA (data not shown). Based on our results, an annealing ratio of 3:1 (oligonucleotide/ssDNA template), with an extension time of 3 h, is recommended for the synthesis of heteroduplex dsDNA in vitro.
Curiously, even in the absence of any synthesized oligonucleotide primer, ssDNA is still able to self-prime and generate dsDNA [6
] during the extension reaction, which can be how the non-recombinant dsDNA is synthesized. To test the possibility that one could outcompete self-priming with longer oligonucleotide primers to reduce the synthesis of non-recombinant dsDNA, primers of different lengths were tested. No significant difference in the mutation rate was observed, indicating that once a minimum length requirement was met (Tm
=50-55°C), longer primers offered no advantage in boosting the mutation rate compared to shorter ones. We also hypothesized that degraded small DNA segments, which might be present in preparations of the ssDNA template, could be responsible for the self-priming. To test this notion, ssDNA was resolved by agarose gel electrophoresis and extracted from any contaminating DNA segments. However, this extra purification step only slightly raised the mutation rate (i.e., from 40% to 50%), implying that the persistence of non-recombinants was somehow inherent to the ssDNA template itself.
3.4 Removal of non-recombinant clones by restriction enzyme digestion
To enhance the efficiency of Kunkel mutagenesis, a set of three phagemid vectors were devised with stop codons and restriction endonuclease recognition sites inserted in the BC and FG loop coding regions. The stop codons are intended to prevent the display of wild-type FN3 domain and its N-terminal fused Flag epitope, whereas inclusion of restriction endonuclease cleavage sites is to permit differential destruction of non-recombinant DNA by digestion with restriction endonucleases.
Three restriction endonucleases were chosen because their recognition sequences were absent from the original phagemid genome and their activities were robust. Vectors were designed with SacII (5′-CCGC↓
GG-3′), SmaI (5′-CCC↓
GGG-3′), or StuI (5′-AGG↓
CCT-3′) sites in the BC and FG coding regions. (Note that each vector carried two recognition sites for the same enzyme). For StuI recognition site, even though it contains thymine, which can be replaced by uracil when ssDNA genome is propagated in the CJ236 E. coli
strain, and thus can make non-recombinant dsDNA resistant to StuI cleavage, we rationalized that the chance that one or both thymines in the StuI site are replaced by uracil is low (i.e., 3-4%), given that only 20-30 uracils are inserted in each ssDNA genome when propagated in CJ236 cells [30
When the three vectors were used for in vitro synthesis of dsDNA, digestion of the dsDNA with the cognate restriction enzyme prior to electroporation led to only a 10% increase in the frequency of recombinants (51% to 61%) obtained. At the moment, we cannot account for this result, even though it is reproducible. However, we noted that the phagemid carrying the two StuI sites consistently yielded larger transformation outputs, compared to the phagemids with SacII or SmaI sites; consequently, we used the phagemid DNA with the two StuI sites throughout this study.
Since we could not selectively degrade the in vitro synthesized, non-recombinant dsDNA molecules by restriction enzyme cleavage, we decided to first electroporate the dsDNA into TG1 cells, and then digest the purified, replicated DNA. After re-electroporating the digested DNA into bacteria, the resulting transformants were observed to be entirely recombinant (). By this approach, a phage-display library, with 1010 members and >99% recombinant, has been recently constructed in the lab by performing 140 electroporations (i.e., 100 electroporations of the in vitro synthesized dsDNA, followed by 40 electroporations of the digested DNA, into bacteria).
Removal of non-recombinant clones by digestion with StuI
3.5. Kunkel mutagenesis with DNA segments generated by error-prone and asymmetric PCR
To bypass the time-consuming steps of restriction enzyme digestion, gel purification, and ligation in the conventional approach for generating a secondary mutagenic library, a protocol was devised that utilizes DNA segments amplified from PCR in lieu of oligonucleotide primers. First, an error-prone PCR was carried out with Mutazyme II, a mixture of two mutant DNA polymerases that introduce point mutations at high frequency in the targeted gene [31
]. The amplified DNA segments were resolved by agarose gel electrophoresis, gel purified, denatured, and annealed to the ssDNA template. After DNA synthesis in vitro
, the resulting dsDNA was electroporated into TG1 cells. But as revealed by phage ELISA and DNA sequencing, only 3-8% of the transformants were recombinants. With this method, a secondary library with a diversity of 1.0×108
was constructed (). To investigate whether digestion by StuI could elevate the incorporation of the mutated strand, the dsDNA was subjected to digestion with StuI before transformation. Although digestion did increase the mutation rate to 17-28%, the transformation efficiency decreased 5- to 7-fold.
Construction of two secondary mutagenic libraries with DNA segments amplified by error-prone and asymmetric PCR
As Kunkel mutagenesis is usually conducted with one to several 30-60 nucleotide long oligonucleotide primers, we rationalized that a double-stranded 300-bp DNA segment might not anneal efficiently to the ssDNA template for in vitro
DNA synthesis because of the equimolar presence of its complementary DNA strand. Therefore, we decided to generate DNA segments that were predominately single-stranded through asymmetric PCR [27
], and use the resulting mixture to prime DNA synthesis on the template strand (). DNA segments that were predominantly single-stranded were generated through 40 cycles of DNA synthesis with Mutazyme II. They were subsequently annealed to the uracilated ssDNA template, converted to dsDNA in vitro
, and electroporated into TG1 cells. A library consisting of 4×108
recombinants was created, with a transformation efficiency of 7.4×108
cfu/μg dsDNA (), which is comparable to the transformation efficiency, 5-6×108
cfu/μg, obtained with dsDNA synthesized with oligonucleotide primers ordered from commercial vendor (data not shown). Upon sequencing 24 recombinants, we observed equal numbers of transition mutations and transversion mutations, 18/35 compared to 17/35, respectively, and an overall point mutation rate of 5.3 mutations per kilobase (i.e., 0.5%).
While conducting our experiments, we witnessed that asymmetric PCR amplification was very sensitive to many factors (i.e., type of DNA polymerase, buffer, thermal cycling conditions). Even with a third primer to perform a nested asymmetric PCR [32
], amplification results were not consistent (i.e., the yields of single-stranded product varied). Thus, when asymmetric PCR fails to generate high-quality single-stranded DNA, the double-stranded segment from the error-prone PCR can be used instead for constructing a secondary library.
To elevate the mutation rate, error-prone PCR was performed with Mutazyme II under conditions reported to increase mutation rate of Taq
DNA polymerase [28
]. The resulting DNA segment was used directly as the primer for in vitro
dsDNA synthesis, as described above (). Electroporation of the dsDNA yielded 2.8×109
transformants, with a transformation efficiency of 5.3×108
cfu/μg dsDNA. Among the transformants, 3.5% were recombinant, and on the average each recombinant clone had 6.5 mutations, which is equal to a point mutation rate of 2.5% (i.e., 25 mutations/per kb) (). While the fraction of recombinants among the transformants is low, the diversity of the resulting library is still sufficient (i.e., 1.0×108
) for generating enhanced binders.
3.6. Identification of stronger binders through affinity selection of a mutagenic library
The mutagenic library containing 4×108 variants was screened through three rounds of affinity selection under stringent conditions. After the third round of selection, eighty-eight clones were evaluated by phage ELISA: ~50% of the clones showed tighter binding to Pak1 kinase than the original clone (data not shown). Three clones with the greatest ELISA signals were chosen for further testing, and two variants, C2 and D3, were observed to bind 4-fold stronger to Pak1 kinase than the original clone, while a third variant, E10, bound 2-fold better (). Compared to the primary structure of the original binder, each of the three FN3 variants had three amino acid substitutions (data not shown). Thus, with this example, we demonstrate the utility of our modifications to the Kunkel mutagenesis protocol in generating high quality recombinant affinity reagents.
Comparing the original clone and its three variants for binding to Pak1 kinase