The budding yeast Saccharomyces cerevisiae has been a model eukaryote for decades, in part because of the ease with which the genome can be precisely altered. For example, gene function can be studied in a number of ways by deleting the open reading frame (ORF), changing the coding sequence, inserting an epitope or fluorescent tag, or altering expression. Two properties that make yeast especially amenable to genome manipulation are the ease at which it can be transformed with exogenous DNA and the high fidelity of integrating the DNA into the genome by homologous recombination (Rothstein, 1991).
Although the methods used to alter the yeast genome have changed as new technologies have been developed, the underlying mechanism based on transformation and homologous recombination remains the same. Twenty years ago, if one wanted to disrupt a gene, she would first clone the gene and surrounding DNA onto a plasmid. Using in vitro manipulations, all or part of the gene would be replaced with a selectable marker, such as the URA3 gene. The plasmid would then be digested with restriction enzymes to release a linear fragment used to transform yeast, with the result of replacing the wild-type gene with the disruption marked by URA3 (Rothstein, 1991). The advent of PCR-based, one-step gene replacement made the method much simpler (Lorenz et al., 1995; Baudin et al., 1993). The linear DNA used for yeast transformation is now created by PCR, using synthetic oligonucleotides (oligos) designed with yeast sequences up- and downstream of the deletion end points and sequences to amplify a selectable marker. This PCR product is used to transform yeast with the same result as the cloning-based method above.
Some experiments require seamless genome modification without a selectable marker or extraneous DNA remaining, for example, to study the effect of a single codon change on gene function, or to build a new strain requiring multiple changes that would make the accumulation of selectable markers impractical. Seamless genome modification requires more work than one-step gene replacement. The original method requires two steps and is referred to as 'pop-in/pop-out' gene replacement (Scherer and Davis, 1979). A plasmid is assembled in vitro that carries a segment of yeast DNA containing the desired alteration and the selectable/counterselectable URA3 marker. The first step is transformation of yeast with the plasmid, which has been cut at a restriction site within the yeast DNA on one side or the other of the alteration to direct 'pop-in' integration. Transformation with a gapped plasmid produces on the chromosome nearly a tandem duplication of the cloned DNA separated by the URA3 plasmid, with the altered copy on one side and wild-type on the other. The second step is initiated by growing the transformed cells in the absence of uracil selection. At a low frequency, cells arise in the population that have undergone recombination between the repeats that flank the URA3 plasmid. These rare 'pop-out' recombinants can be selected for on 5-fluoro-orotic acid (5-FOA) medium, which is toxic to URA3 cells (Boeke et al., 1984). Whether a pop-out strain retains altered or wild-type DNA depends on where the crossover takes place. If the crossover occurs in the yeast DNA between the alteration and the URA3 plasmid, then the alteration will remain while the wild-type is evicted. Likewise, if the crossover occurs outside the alteration relative to the URA3 plasmid, then wild-type DNA remains.
There are cloning-free methods for marker-free genome modification, but they are either not seamless or they require multiple oligos, PCRs and transformations. One method for marker-free gene deletion is easy to perform, but it leaves behind extraneous DNA, such as a fragment of bacterial hisG or a loxP site (Güldener et al., 1996; Schneider et al., 1996). Strictly speaking, the products of these techniques are marker-free, but they are not seamless. The DNA remnants preclude use of the methods for applications such as in-frame deletions. Moreover, repeated use of the techniques in the same strain, so-called 'marker recycling', can lead to problems with subsequent transformation (Davidson and Schiestl, 2000) and genome rearrangements (Delneri et al., 2000). Other PCR-based techniques are available that result in truly marker-free, seamless genome modification. Some require multiple PCR primers and PCR steps to assemble a DNA molecule for yeast transformation that carries sequences to target integration, the alteration, a repeat sequence for pop-out recombination and a URA3 marker (Erdeniz et al., 1997; Akada et al., 2006). Another method uses four long DNA oligos and two yeast transformations to integrate URA3 at the target locus and subsequently to evict it with complementary oligos containing the alteration (Storici et al., 2001).
Here we describe the 50:50 method for marker-free, seamless genome editing that is almost as simple as PCR-based, one-step gene replacement. The technique requires only two primers, one PCR with a URA3 cassette, and a single transformation. It is a two-step method, with pop-in of the URA3 cassette, followed by selection with FOA for pop-out recombinants. The key is one of the two PCR primers: the 50:50 primer, a hybrid containing 50% pop-in sequences, the alteration, and 50% pop-out sequences. Together with a standard reverse primer and the URA3 cassette, the 50:50 primer provides all that is needed for PCR-based, seamless genome editing in yeast.
Materials and methods
Standard techniques were used for DNA manipulations (Ausubel et al., 1995). DNA oligos (Eurofins MWG Operon, AL, USA) are listed in Table 1. Yeast media and culture conditions were as described by Amberg et al. (2005). PCR cassettes for yeast transformation were amplified using Takara ExTaq polymerase, which gives a consistently high yield. Assays for α-factor and mating were as described by Sprague (1991). A PCR-based, one-step gene replacement of MATα1 (α1::kanMX4) was created using primers al1.52.1 and al1.52.2 with plasmid pFA6–kanMX4 (Wach et al., 1994).
Table 1. PCR primers
MATα1 –50 to −1, double underline; 50 nt following stop codon, dotted underline; U2, single underline.
MATα1 –41 to +9, double underline; 2 nt deletion and XhoI site, italics; +14 to +63, dotted underline; U2, single underline.
Reverse complement of MATα1 +14 to +63, dotted underline; D2, single underline.
MATα1 –60 to −1, double underline; U2, single underline.
Reverse complement of MATα1 60 nt following stop codon, double underline; D2, single underline.
U2, single underline.
D2, single underline.
Plasmid pJH136 (Addgene 47554) carries URA3 flanked by PCR priming sites U2 and D2 (Chu and Davis, 2008). It was constructed by using Phusion HSII (Thermo Scientific) to PCR-amplify URA3 DNA from JHY222 (S288c background; Lardenois et al., 2011) with primers URA3.54.1 and URA3.54.3. The 1160 bp blunt-end PCR product was cloned into the SrfI site of pCR-Script Amp SK+ (Agilent Technologies) and verified by DNA sequencing. The URA3 sequences in pJH136 extend from 243 bp upstream of the start codon to 79 bp downstream of the stop codon. Because U2 and D2 sequences flank the common heterologous markers (Goldstein and McCusker, 1999; Wach et al., 1994), pJH136 can also be used to make standard PCR-based, one-step gene replacements.
The two PCR primers used to amplify URA3 for yeast transformation are the 50:50 primer and a standard reverse primer. Design of the 50:50 primer is straightforward: (5′ to 3′) 50 nt directly upstream of the alteration (pop-in sequence to target integration), the alteration, 50 nt directly downstream of the alteration (pop-out sequence) and a sequence for priming URA3 PCR (e.g. U2; Table 1). In our examples we used 50 nt for each of the pop-in and pop-out sequences, but they can be shorter or longer. The length of DNA inserted or changed is limited by oligo synthesis technology, which presently allows for ca. 125 nt total and therefore a maximum of 7 bp inserted or changed, although this can be increased if the 50:50 sequences are decreased. To create a deletion (Figures 2A, 3), the primer is 50 nt upstream of the deletion, followed by 50 nt downstream of the deletion plus U2. For an insertion (Figure 2B), the primer is 50 nt upstream of the insertion, the insertion (e.g. GCA), followed by 50 nt downstream plus U2. For a single nucleotide change (Figure 2C), the primer is 50 nt upstream of the change, the single nucleotide change (e.g. C instead of G), followed by 50 nt downstream plus U2. Finally, in our example of deleting two non-consecutive base pairs (Figure 2D, and see below), the primer is 50 nt upstream of the altered sequence, the alteration (CG instead of TCGA), followed by 50 nt downstream plus U2.
The standard reverse primer is composed of 50 nt yeast sequence followed by a sequence for priming URA3 PCR (e.g. D2; Table 1). One function of the reverse primer's yeast sequence is to target pop-in integration of the URA3 cassette. Depending on the alteration being created, the reverse primer's yeast sequence can also serve as a direct repeat downstream of URA3 of the pop-out sequence present in the 50:50 primer. Thus, for alterations that are not a deletion, the reverse primer should be the reverse complement of the 50:50 primer's pop-out sequence plus D2. Beware that PCR with 50 bp direct repeats flanking URA3 can be problematic for some polymerases (see below). For deleting an ORF, the yeast sequences in the reverse primer can be the reverse complement of any 50 nt in the ORF, or even the 50 nt immediately downstream of the ORF, which is equivalent to the 50:50 primer's pop-out sequence.
We used the 50:50 method to create two MATα1 mutations in JHY337 (MATα ura3∆0 leu2∆0 lys2∆0, a derivative of JHY222). Precise deletion of the ORF (α1∆) was accomplished with primers al1.46.1 and al1.46.3. A −2 bp frameshift mutation after the third codon (α1-2x) was accomplished with primers al1.46.2 and al1.46.3. For each mutation, primer pairs were used to amplify URA3 by PCR from pJH136. PCR products were either used directly to transform yeast (Gietz and Schiestl, 2007) or first concentrated by ethanol precipitation and resuspended in 0.1× TE buffer. We plated one-fifth of the transformation to SC – Ura plates and then replica-plated to fresh plates after 2 days to eliminate the background lawn that is often present with PCR transformations. Several dozen colonies were obtained for each transformation. Correct transformants (18 of 20 scored) were identified by genomic PCR, using primers specific to both sides of the integrated URA3 cassette. Four independent transformants for each mutation were streak-purified on YPD plates and a single colony of each was used to inoculate 3 ml YPEG broth (2.5% ethanol, 2% glycerol; used to prevent growth of petites). Cultures were grown to saturation (~2 × 108 cells/ml) for 2 days at 30°C; 50 µl of the culture (~1 × 107 cells) was spread onto a plate of synthetic complete medium containing 0.8 mg/ml FOA (US Biological). An average of 74 colonies arose on each plate (range 32 to ~200, eight plates scored). Correct URA3 pop-out recombinants were identified by genomic PCR and confirmed by DNA sequencing.
PCR amplification of DNA segments that contain repeated sequences can be problematic, with results being dependent on length of the repeat and PCR conditions. PCR amplification of the URA3 cassette using primers al1.46.2 and al1.46.3 produces a product with a 50 bp direct repeat on each end. We found that Takara ExTaq polymerase can efficiently generate this product in a single reaction (Figure 4A). However, the same primers and template used with Phusion HSII polymerase failed (data not shown). A simple and robust solution to problematic PCRs with repeat-containing products is to perform split-URA3 PCRs, which separate the repeats into two tubes. Primers URA3.for and URA3.rev can be used with the long D2- and U2-tailed primers, respectively, to set up two PCRs that are later pooled and used for yeast transformation (Figures 1, 4). The URA3 fragments share 280 bp overlap and recombine upon transformation to restore URA3. Single and split-URA3 PCRs using ExTaq polymerase are shown in Figure 4A. We have used both single and split-URA3 PCRs for transformation with similar results.
We wished to develop a simple method for seamless genome editing in yeast that satisfied the following criteria: cloning-free (no in vitro plasmid construction); based on synthetic DNA primers; requires only one PCR and one yeast transformation; and has no special strain requirements other than the ura3 genotype. The original and elegant two-step gene replacement technique (Scherer and Davis, 1979) (see Introduction) provided the framework for us to convert a plasmid-based method to a PCR-based one that satisfied our criteria.
What are the important features of the original two-step method, and can they be more-simply accomplished, and even improved, without cloning yeast DNA and introducing the alteration on a plasmid? One important feature is the selectable/counterselectable marker used to select for the pop-in and pop-out recombinants. URA3 is commonly used and has an advantage over another such marker, LYS2, because of its smaller ORF length (804 bp vs 4179 bp). In a cloning-free method, only URA3 is required; plasmid sequences are not. For the pop-in step, linear DNA is used for transformation that has yeast sequences flanking URA3. In the original method, cutting within the yeast sequences cloned into a URA3 plasmid produces the linear DNA. The pop-in step can just as well be accomplished using linear DNA, with yeast ends created by synthetic DNA primers and PCR amplification of a URA3 cassette. Indeed, the components of a cloning-free pop-in step are essentially PCR-based, one-step gene replacement using the URA3 marker.
This leaves two features that require adapting: the alteration and the duplicated yeast DNA that allows for pop-out recombination. Because of advances in DNA synthesis technology, we found that both can be incorporated into one of the two primers used to PCR-amplify the URA3 cassette (Figures 2, 3). This primer is a hybrid composed of two segments of yeast DNA flanking the alteration, followed by a sequence for priming URA3 PCR. The first segment serves as a pop-in sequence to target integration of URA3 into the genome. The second serves as a pop-out sequence, with homology to a segment located on the other side of the integrated URA3 marker. In between the pop-in and pop-out segments is the alteration. Because DNA primer length is limited by synthetic chemistry, the types of alterations are limited to deletions, short insertions or substitutions. In principle, the hybrid primer contains 50% pop-in sequence, the alteration, and 50% pop-out sequence. Thus, it is called the 50:50 primer, and it is used in the 50:50 method.
An additional advantage of 50:50 over the original two-step method is that none of the transformed sequences are duplicated upon integration, because the PCR cassette is linear DNA and not a gapped plasmid. The only sequences present in duplicate are the pop-out sequences designed into the 50:50 primer between the alteration and URA3, and the cognate sequences on the other side of URA3 (Figure 3C). Thus, unlike the original method, where sequences up- and downstream of the alteration are duplicated and give rise to both wild-type and altered pop-out recombinants, the 50:50 method gives rise only to altered recombinants.
To demonstrate the technique, we used the 50:50 method to introduce two mutations in MATα1. The α1 transcription factor is expressed only in MATα cells and functions in a complex with Mcm1 to activate α-specific genes, such as those encoding α mating pheromone and the a-factor receptor (Sprague, 2005). Mutants lacking α1 are sterile and cannot mate to MATa cells. An α1::kanMX4 deletion strain yielded an unexpected and new result: although haploid α1::kanMX4 cells had the expected phenotypes of not producing α-factor and not mating to MATa cells, we found that, in contrast to wild-type MATα cells, α1::kanMX4 cells mated as MATa cells, albeit with low efficiency (Figure 5, and data not shown). One explanation for the ability of α1::kanMX4 cells to mate as MATa cells is that the kanMX4 cassette, with its strong A. gossypii TEF1 promoter, interferes with transcription of the adjacent MATα2 gene, which encodes a repressor of a-specific genes (Sprague, 2005). Reduced MATα2 transcription would lead to a defect in a-specific gene repression, which in turn would allow mating as MATa cells. To test this hypothesis, we used the 50:50 method to create two MATα1 mutations without the kanMX4 marker: one a precise ORF deletion and the other a deletion of two non-consecutive base pairs near the start of the ORF to introduce a frameshift and a diagnostic XhoI site.
To create the precise MATα1 deletion (α1∆), we designed the 50:50 primer (al1.46.1) with 50 nt upstream of the MATα1 start codon, the deletion (lack of the MATα1 ORF), 50 nt downstream of the stop codon, and the 18 nt U2 sequence. The reverse primer (al1.46.3) for URA3 cassette PCR contained the reverse complement of the MATα1 ORF from +14 to +63, followed by the 19 nt D2 sequence (Figure 3A). For deletions larger than 50 bp, such as the MATα1 ORF, the URA3 cassette can be inserted at, or upstream of, the genomic pop-out homology. In this case, the 50:50 URA3 cassette was inserted in the first part of MATα1, a site we chose so that we could recycle the reverse primer when making the −2 bp deletion (see below). The transformed α1∆::URA3 yeast strain had the configuration 50 bp upstream of MATα1 ATG, 50 bp downstream of MATα1 stop codon, URA3, and then the MATα1 ORF from +14 continuing into the wild-type genomic sequences (Figure 3C). Culture in the absence of uracil selection allowed for growth of FOA-resistant recombinants that had undergone a crossover between the 50 bp MATα1 ORF pop-out sequence incorporated into the 50:50 primer and its cognate sequence downstream of URA3 on the chromosome (Figure 3C). The resulting strain carries a precise deletion of the MATα1 ORF without URA3 or any other DNA (Figures 3D, 4B; DNA sequence data not shown). We tested the α1∆ strain and found it to be α-specific sterile, as expected (Figure 5A, B). We also found that α1∆ cells mated with low efficiency as MATa cells (Figure 5C). Thus, it is not the kanMX4 cassette per se that causes the MATa mating phenotype. It is either the lack of α1 activity or an off-target effect of deleting 528 bp DNA near MATα2.
To discriminate between these two possibilities, we used the 50:50 method to introduce a frameshift mutation at the start of the MATα1 ORF, a change that likely would not have an off-target effect MATα2 transcription. The α1-2x mutation deletes two non-consecutive base pairs after the third codon and creates a diagnostic XhoI site (CTCGAG). Thus, the wild-type ORF begins ATGTTTACTTCGAAG, whereas α1-2x begins ATGTTTACTCGAG. We designed the 50:50 primer (al1.46.2) with sequences −41 to +9 relative to the wild-type ORF, the alteration (CG instead of TCGA following +9), sequences +14 to +63, and the 18 nt U2 sequence. The reverse primer (al1.46.3) for URA3 cassette PCR was the same as that used for α1∆. Note that the pop-out sequence in the second half of the α1-2x 50:50 primer is identical to the pop-in sequence in the reverse primer. Non-selective growth of α1-2x::URA3 cells followed by selection on FOA medium yielded pop-out recombinants that were verified as α1-2x based on a correct size genomic PCR product that could be digested with XhoI (Figure 4B) and DNA sequencing (data not shown). We found the α1-2x strain to be α-specific sterile (Figure 5A, B). As with α1::kanMX4 and α1∆, α1-2x mutants mated with low efficiency as MATa cells (Figure 5C). Given that α1-2x alters the MATα locus by deletion of only 2 bp at the start of the MATα1 ORF, it is highly unlikely that the unexpected mating phenotype we observed is the result of an off-target effect. From these observations, we conclude that α1 has a function outside its known role as a positive regulator of α-specific gene expression. For example, α1 might positively regulate MATα2 transcription. Another possibility is based on the fact that both α1 and α2 function by binding the Mcm1 transcription factor: absence of α1 might upset the balance of Mcm1 and α2 association, leading to a defect in a-specific gene repression. Whatever the mechanism, it is intriguing that a transcription factor that had only been assigned a role in activation of α-specific genes also seems to have a role in repression of a-specific genes.
Here we have described the 50:50 method for simplified, marker-free, seamless genome editing in yeast. The utility of the method is that it requires only two primers, one PCR, and a single yeast transformation. The key components are the 50:50 primer, which provides the pop-in sequence, the alteration, and the pop-out sequence. Because the alteration is engineered upstream of the pop-out sequence relative to URA3, recombination between the 50:50 primer pop-out sequence and the cognate sequence downstream of URA3 only leaves the altered sequence on the chromosome. This is a significant advantage over the original, plasmid-based method that can produce both wild-type and altered recombinants. We used the S. cerevisiae URA3 marker, but if the host strain is not ura3∆0 the method can also be used with a heterologous URA3 marker, such as CaURA3MX4 (Goldstein et al., 1999). One advantage of using S. cerevisiae URA3 is its smaller size (1160 vs 1509 bp cassette). Another is that it does not contain the flanking A. gossypii TEF1 promoter and terminator sequences that are present in most heterologous markers. If a yeast strain already carries a heterologous marker, such as a kanMX4 gene replacement, the common sequences present in CaURA3MX4 will cause problems with obtaining correct integrants at the new locus and with obtaining correct pop-out recombinants that have not become FOA-resistant by way of gene conversion from kanMX4.
There are PCR-based methods that accomplish the same goal as the 50:50 method, although none are as efficient and economical. One method most similar to ours uses PCR primers that create a roughly 60 bp direct repeat flanking the Kl URA3 marker (Längle-Rouault and Jacobs, 1995). The repeats contain 30 bp upstream of the desired alteration, the alteration itself, and 30 bp downstream. The targeting homology is short, leading to inefficient transformation, even with the heterologous marker. Moreover, because the upstream–alteration–downstream sequences are present on both sides of the Kl URA3 cassette, there is the potential for undesired integration events at either the upstream sequence or the downstream sequence, pop-out of either of which would return wild-type. The 50:50 method reduces the possibility of incorrect integration because the upstream pop-in sequence is confined to one primer, which also makes the method more economical. Also, the stretches of pop-in homology on the 50:50 and reverse primers are longer. Together, these features contribute to nearly all 50:50 Ura+ transformants being correct (see Materials and Methods).
Other methods have been described that require only one yeast transformation, but they require multiple PCR primers and PCRs (Erdeniz et al., 1997; Akada et al., 2006). The Akada method is limited to seamless gene deletion and requires four PCR primers and three PCRs, divided into two steps (Akada et al., 2006). The Erdeniz method adapted for de novo mutations likewise requires seven PCR primers and five PCRs, divided into two steps (Erdeniz et al., 1997). In both methods, one of the first-step PCRs uses yeast genomic DNA as a template to amplify several hundred base pairs of target locus DNA, which in a subsequent PCR is fused to either URA3 or Kl URA3. The long stretch of yeast DNA on one (Akada) or both (Erdeniz) sides of the marker can increase transformation efficiency. In the Erdeniz method, the long stretch of DNA is repeated on both sides of Kl URA3, which can also increase pop-out efficiency. Certainly, one advantage of our method over these is that it requires only two PCR primers and one PCR; but another important advantage is that the length of chromosomal DNA affected by the 50:50 method is limited to about 50 bp up- and downstream of the alteration. When making a seamless genome alteration, we believe it is important to use DNA sequencing to confirm the change. In our experience, we have found a significant number of mutations that could be attributed to oligo synthesis or PCR. Genome-editing methods that introduce longer than necessary stretches of synthetic DNA increase the chance of undesired mutations and the work required for DNA sequence confirmation. Regarding short flanking yeast homology and transformation efficiency, we have never encountered a problem obtaining transformants with the 50:50 method, or with any PCR-based gene deletion or modification method for that matter. Likewise, the 50 bp pop-out sequences in the 50:50 primer are sufficient to yield many FOA-resistant recombinants (see Materials and methods).
Finally, another variation on cloning-free seamless genome editing is delitto perfetto, which requires two sets of long DNA oligos and two transformations (Storici et al., 2001). Yeast is first transformed with a 3.2 kb kanMX4-Kl URA3 double heterologous marker (the CORE cassette) that is amplified by PCR with primers that target integration at the desired genomic locus. Correct integrants are then transformed a second time, typically with a pair of long, complementary oligos that have ends homologous to sequences up- and downstream of the integrated CORE cassette, plus the desired change near the centre. There is no direct selection with the second transformation; rather, cells are cultured non-selectively at first and then later plated to FOA medium, with the goal of identifying ura3 cells that have replaced the CORE cassette with the altered oligo sequence. Recovering correct recombinants can be a challenge, because transformation and homologous recombination of oligos is inefficient. To increase the frequency of obtaining recombinants in the second transformation, a 4.7 kb Kl URA3–kanMX4–GAL1-I–SceI endonuclease cassette has been introduced (Storici and Resnick, 2006). In contrast to delitto perfetto, every cell with an integrated 50:50 URA3 cassette has the 50 bp pop-out sequences repeated on the chromosome up- and downstream of URA3. Besides not requiring a second transformation, having the pop-out repeats in every cell means that the efficiency of obtaining recombinants is not dependent on the efficiency of transformation with exogenous oligos.
In summary, we have described the 50:50 method for PCR-based, seamless genome editing in yeast. There have been other methods described that achieve the same goal, but none are as simple or economical.
We thank Angela Chu for helpful discussions; thanks to Angela, Rishi Rakhit and Eric Foss for comments on the manuscript. This study was supported by the National Institutes of Health (Grant No. 5P01HG000205, to R.W.D).