In our pipeline, we combined a two-fragment PCR approach, in which mutagenesis primers are used in two separate PCR reactions, with an in vitro assembly of the resulting fragments.

The mutagenesis primers are combined with another pair of primers, annealing approximately on the opposite side of the plasmid (Fig. 1). The two vector fragments generated in separate PCRs are combined in the simple and efficient Gibson assembly¹⁴, merits of which were extensively discussed by Benoit et al.¹⁵. Because shorter fragments are amplified, the PCR is more robust and efficient. This is clearly an advantage over other techniques in which a full-length plasmid is amplified^{11, 12, 16, 17}, including the one we have used earlier for scanning mutagenesis of the human G_ai1 (354 residues, UniProtKB: P63096) and bovine arrestin-1 (404 residues, P08168) proteins¹³. After the initial success of the two-fragment approach in generating several “difficult-to-make” arrestin-1 mutants, we decided to use this strategy for high-throughput scanning mutagenesis of two human GPCRs – cannabinoid CB2 receptor (CB2; 360 residues, P34972) and vasopressin V2 receptor (V2R; 371 residues, P30518).

Here, each vector fragment was amplified by PCR using one mutagenic primer and the appropriate ColE1 primer annealing to the complementary DNA strand within the vector’s bacterial origin of replication (Fig. 1). Although one could design a non-mutagenic primer that anneals anywhere on the vector further away from the mutation site, the origin of replication offers several advantages: (1) many different plasmids share the same origin of replication, so the same pair of the reverse-complementary ColE1 primers can be used in all such cases, (2) only plasmid DNA with a properly assembled replication origin can be replicated in bacterial cells and produce colonies on plates with selection antibiotic and (3) an origin of replication is usually situated on a plasmid approximately “opposite” the gene of interest, which results in more or less balanced sizes of the two vector fragments that are to be assembled.

The mutagenic primers were designed in a high-throughput way using the program AAscan¹³. The amino acids of the target protein were changed to alanine, while alanines were exchanged for glycines (V2R) or valines (CB2). The alanine and glycine substitutions remove the side chain of the amino acid and are used for probing its role in protein function. Valines are one of the prevalent types of amino acids in membrane proteins and promote stability of transmembrane helices. These substitutions are generally well tolerated and used for thermostabilisation of GPCRs (reviewed in refs 3, 18 and 19). For both receptors, we have used the mammalian expression vector pcDNA4/TO. In preliminary experiments, we tested different DNA polymerases and optimized PCR conditions for the preparation of 30 randomly chosen CB2 mutants, analysing the PCR products on agarose gels (Fig. 2). Success of mutagenesis was dependent on DNA polymerase and PCR conditions (see Materials and Methods). The same PCR conditions used for CB2 were also successfully applied in high-throughput mutagenesis of V2R without any further optimization. The PCR elongation time was calculated based on the longest fragment to be made in the whole alanine scan. We used a step-down PCR protocol²⁰ with a decrease of annealing temperature of 0.5 °C per cycle to accommodate the variation of primer melting temperatures inherent to a large high-throughput screen.

After PCR, the two matching fragments carrying one mutation were combined without checking PCR products on an agarose gel (Fig. 3). The methylated template DNA in a mixture of fragments was digested with DpnI at 37 °C for 18 h. This resulted in a very low background of wild-type template plasmid (Fig. 4). DpnI digestion was followed by DNA clean-up in order to remove remaining primers, enzymes and deoxynucleotides, crucial for the success of the following enzymatic reaction. A Gibson assembly at 50 °C for 10 min followed by 1 h at 37 °C with 1 μL cleaned-up DNA and 3 μL 1.33× Gibson assembly mix (see Materials and Methods) proved to be sufficient for the high-throughput applications. The ratio of the fragments used in Gibson assembly was not adjusted for the high-throughput mutagenesis. Instead, both fragments were combined directly after the PCR reaction to save time and reagents. A separate clean-up of fragments and adjustment of fragment ratios might increase success rates in low-throughput applications. Transformations were done with 2 μL assembly product and 20 μL chemically competent Escherichia coli XL1-Blue cells. As shown for V2R mutants, transformed cells could be plated on a quarter of selective lysogeny broth (LB) agar plates to save space without losing efficiency of mutagenesis. Alternatively, bacteria could be plated on 24-well plates using an expanding pipette²¹. It should be noted that 68 CB2 transformants gave no visible colonies on the selection plates. We hypothesized that the corresponding PCRs were not successful, so we repeated them using another DNA polymerase and the corresponding PCR conditions (see Materials and Methods).

Colonies were sent for sequencing on 96-well LB agar plates with the appropriate selection antibiotic. Initially, we sent out only one colony per mutant to reduce sequencing costs. If the first sequencing reaction was not successful, we sequenced DNA from one to three additional colonies. Difficult CB2 mutants (14, 4%) were obtained by repeating PCR, analysing products on agarose gels and cutting the right bands out in case of significant side-products. Alternatively, the rest of the missing V2R mutants (3, 0.8%) were obtained by repeating the PCR with two mutagenic primers in the same reaction mixture (Fig. 1), followed by in vivo recombination in the E. coli Mach1 strain²², according to our earlier one-fragment approach^{5, 13}. When few mutants remain to be cloned, we tend to try different approaches in parallel and/or send more clones at once. Therefore, the standard single-fragment approach was used alongside the two-fragment method.

Once PCR conditions have been optimized for the particular vector used, the mutagenesis is straightforward. Most steps are done in 96-well plates, contributing to the protocol’s simplicity. Most mutations not obtained in the first round of mutagenesis can be obtained by sending additional clones for sequencing, while others can be obtained by repeating Gibson assembly or PCR, depending on the cause of failures (see Fig. 5).

We defined cloning efficiency as a percentage of the expected sequences obtained in all successful sequencing reactions (excluding sequencing failures and non-informative sequencing data). We use this term when discussing occurrence of primer tandem repeats, which were the main cause of failure in the one-fragment approach used in arrestin-1 high-throughput mutagenesis we use as a comparison. However, it should be noted that sequencing failures could also be caused by unsuccessful cloning, e.g. when a part of plasmid to which sequencing primer should anneal is missing. Because we cannot distinguish a true sequencing failure from a cloning failure in our data, we decided to use an overall success rate of the method as well. It takes into account all encountered practical difficulties and it is defined as a percentage of the expected sequences obtained in all sequencing attempts (including failed sequencing and non-informative sequencing data). In our opinion, this information is valuable in the context of high-throughput site-directed mutagenesis with direct repercussions on expected costs.

Taken together, we successfully generated complete alanine scanning libraries of the two GPCRs. For V2R, 467 out of 565 sequencing reactions confirmed the designed single amino acid mutations, which correspond to an overall success rate of 83%. For CB2, we obtained the right sequences for 390 out of 537 colonies sent for sequencing (73%). The lower efficiency in this case was mostly due to failed sequencing or lower quality of sequencing data not allowing their reliable interpretation as well as primer misannealing resulting in large deletions (Fig. 4). In general, the most common problems encountered in these projects were unsuccessful sequencing attempts, too short sequence reads and deletions of a part of the vector. We observed very few cases of vectors bearing no mutation (unsuccessful DpnI digestion reaction) or an additional mutation (PCR error or mutation in a primer oligonucleotide).

We also used the same two-fragment cloning approach to prepare twenty N- and/or C-terminally truncated constructs of Arabidopsis thaliana and tomato ethylene receptor 1 (ETR1) – two plant membrane proteins with an extensive cytoplasmic domain. Out of 50 successfully sequenced DNA samples, 45 had the correct sequence. Although the remaining 5 had the correct truncation, they also contained an additional point mutation, presumably introduced as a DNA polymerase error in PCR. While the Gibson method is an established cloning method, these results suggest that it is very well suited for general high-throughput applications.

The success of mutagenesis for a given template would, naturally, depend on a number of parameters in addition to the mutagenesis method chosen. For example, the length of the plasmid to be amplified needs to be considered, as well as extremely high or low GC content¹³. In our case, the pCDNA4/TO vector encoding for fusions of the CB2 and V2R constructs used in this work have the same length of 7 kbp and an average GC content of 55 and 54%, respectively (Fig. 6). The length of the plasmid encoding for arrestin-1 fusion to mCherry⁴ is 7.6 kbp and it is also similar in GC content (50.4%). The constructs encoding for the ETR1 were in the range of 6.2–7.5 kbp. Based on the two parameters that can dramatically affect the PCR efficiency, the length and GC content, these samples are relatively similar. The plasmid size is very critical for co-transformation cloning, where overlapping fragments are simply used to co-transform E. coli, without in-vitro treatment with In-Fusion or assembly mix¹⁵. However, the assembly mix is efficient at combining DNA fragments up to several hundred kilobases, a size that is never reached with plasmids¹⁴. If insert-plasmid mixtures are combined in vitro using assembly mix, as in our method, the limiting factor is the PCR. Because we use two fragments and 6 kb can be robustly amplified today, the method works for plasmids up to at least 12 kb. The larger plasmid size also has a slight negative effect on transformation efficiency, but this is not a problem with competent cells of reasonably high transformation efficiency. In comparison to the method previously used in our lab with both mutagenic primers in one PCR and recombination in Mach1 cells¹³, we found that a number of instances with tandem repeats of the primer sequence before or after the mutation site was drastically reduced (Fig. 4), resulting in increased cloning efficiency (86% and 90% compared to 54% of successfully sequenced samples). The main improvements clearly came from performing two separate PCR reactions which amplify two fragments of the target vector.

In our opinion, even higher mutagenesis success rates can be achieved using a novel, not expensive and high-fidelity DNA polymerase (such as Q5), longer elongation times (at least 30 s/kbp) and less PCR cycles in combination with annealing temperatures not lower than 60 °C (for the primers designed as described in Materials and Methods). Furthermore, the Hot Fusion method^{15, 23} can be used for assembling two fragments instead of the Gibson reaction, which further reduces the cost of mutagenesis by omitting a DNA ligase.

The two-fragment approach presented here can be further extended to a three-fragment approach where the whole plasmid backbone is prepared by restriction digestion, mutagenesis is then done with mutagenesis primers and primers flanking the gene of interest, followed by PCR clean-up and Gibson assembly. This could be a very efficient way of making mutants, especially in the GC rich regions that are challenging for the PCR. An additional benefit is that the backbone, since it is only produced once, can be verified for the absence of mutations and deletions, something that is not practical with the two-fragment approach. It needs to be noted that, from our experience, it is more difficult to assemble three fragments successfully, especially when the molar ratio of the assembling fragments has not been chosen carefully. In our two-fragment high-throughput approach, we do not measure or adjust fragment concentration before the assembly reactions.

Although there are many general cloning methods developed and kits commercially available, it is of crucial importance to have a highly-efficient, simple, fast and cost-effective method when it comes to generating libraries comprised of hundreds of mutants with 100% coverage. Even though the presented high-throughput mutagenesis and cloning pipeline is slightly more laborious than other established mutagenesis protocols, we found this to be a convenient and efficient alternative.