Simple-to-use CRISPR-SpCas9/SaCas9/AsCas12a vector series for genome editing in Saccharomyces cerevisiae

Abstract Genome editing using the CRISPR/Cas system has been implemented for various organisms and becomes increasingly popular even in the genetically tractable budding yeast Saccharomyces cerevisiae. Because each CRISPR/Cas system recognizes only the sequences flanked by its unique protospacer adjacent motif (PAM), a certain single system often fails to target a region of interest due to the lack of PAM, thus necessitating the use of another system with a different PAM. Three CRISPR/Cas systems with distinct PAMs, namely SpCas9, SaCas9, and AsCas12a, have been successfully used in yeast genome editing. Their combined use should expand the repertoire of editable targets. However, currently available plasmids for these systems were individually developed under different design principles, thus hampering their seamless use in the practice of genome editing. Here, we report a series of Golden Gate Assembly-compatible backbone vectors designed under a unified principle to exploit the three CRISPR/Cas systems in yeast genome editing. We also created a program to assist the design of genome-editing plasmids for individual target sequences using the backbone vectors. Genome editing with these plasmids demonstrated practically sufficient efficiency in the insertion of gene fragments to essential genes (median 52.1%), the complete deletion of an open reading frame (median 78.9%), and the introduction of single amino acid substitutions (median 79.2%). The backbone vectors with the program would provide a versatile toolbox to facilitate the seamless use of SpCas9, SaCas9, and AsCas12a in various types of genome manipulation, especially those that are difficult to perform with conventional techniques in yeast genetics.


Introduction
Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) is an adaptive immune system in eubacteria and archaebacteria that functions to counteract foreign nucleic acids, such as those of invading bacteriophages (Jinek et al. 2012). The CRISPR array encodes guide RNAs (gRNAs) that form complexes with Cas (CRISPR-associated) proteins. In a number of CRISPR/ Cas systems, the Cas-gRNA complexes function as DNA endonuclease to cleave double-stranded DNA. The sequence of gRNA specifies the cleavage target. Accordingly, co-expression of a Cas protein and its cognate gRNA can introduce a DNA double-strand break (DSB) at a specific position in the genome. Cleaving the genome at a specific site is a critical process of genome editing. Because of the ease of specifying the target sequence, the CRISPR/Cas systems are being widely used for genome editing in a variety of organisms (Hsu et al. 2014).
A protospacer adjacent motif (PAM) is a short DNA sequence that flanks the target sequence defined by the gRNA and is required for the Cas-gRNA complex to recognize its target sequence (Sternberg et al. 2014). The nucleotide sequence of PAM is different from one CRISPR/Cas system to another. SpCas9 from Streptococcus pyogenes has a G-rich PAM (NGG) that follows the 3 0 end of the target sequence (Jinek et al. 2012). AsCas12a from Acidaminococcus sp. has a T-rich PAM (TTTV) that precedes the 5 0 end of the target sequence (Zetsche et al. 2015). SaCas9 from Staphylococcus aureus has a PAM with an intermediate GC content (NNGRRT) that flanks the 3 0 end of the target sequence (Ran et al. 2015).
An absolute prerequisite for genome editing to insert a gene fragment, a single amino acid substitution, and a small indel to a specific site is a PAM in the vicinity of the target site. It, however, often happens that no appropriate PAM for a single CRISPR/Cas system is found in the region of interest. If multiple systems with different PAMs are available, it would theoretically become much easier to find a PAM and hence the target sequence to introduce a DNA DSB for gene fragment insertion, the introduction of single amino acid substitutions, and the introduction of small indels. Indeed, a simple calculation indicates the power of the combined use of SpCas9, SaCas9, and AsCas12a in genome editing of budding yeast Saccharomyces cerevisiae (see "Expansion of editable fraction of the yeast genome by combining three CRISPR/Cas systems" in the Results section and Figure 1).
These three CRISPR/Cas systems have already been implemented for the budding yeast (DiCarlo et al. 2013;Laughery et al. 2015;Generoso et al. 2016;Jessop-Fabre et al. 2016;Swiat et al. 2017;Degreif et al. 2018;Verwaal et al. 2018). However, the vectors for these systems were independently developed in different laboratories. Consequently, the methods and design principles for genome-editing plasmid construction (e.g., copy number, selection marker, promoter, cloning sites, and so on) are different from one to another. In practice, such differences often hamper seamless, stress-free use of the most suitable system to a given target site of interest. If all three systems can be used in a unified manner, then the genome-editing processes will be substantially accelerated.
Based on these theoretical and practical needs, we developed in this study a series of four backbone vectors under a unified design principle to seamlessly exploit SpCas9, SaCas9, and enAsCas12a in yeast genome editing. A single highly efficient method, Golden Gate Assembly, is applicable to construct genome-editing plasmids on these backbones. To facilitate the design of synthetic oligodeoxyribonucleotides (ODNs) required for the Golden Gate Assembly process, we developed a simple program that automatically calculates the ODN sequences corresponding to a given target sequence. We demonstrated that genome-editing plasmids thus constructed were efficient enough for routine use in gene knock-in at essential genes (median 52.1%), complete deletion of open reading frames (ORFs) (median 78.9%), and the introduction of single amino acid substitutions (median 79.2%).

Yeast strains
Yeast strains used in this study are listed in Supplementary  Table S1. All strains are derived from S. cerevisiae BY4741 or BY4742 (Brachmann et al. 1998). Standard culture media were used in this study (Guthrie and Fink 1991). Conventional gene deletion was performed using a PCR-based method (Longtine et al. 1998). Plasmids used for yeast strain construction are listed in Supplementary Table S2.

Construction of backbone vectors for genome editing
ODNs used in this study are listed in Supplementary Table S3. All ODNs for plasmid construction were purchased from Sigma-Aldrich Japan (Tokyo, Japan) and Eurofins Genomics K. K. (Tokyo, Japan). The four backbone vectors for genome editing were constructed using the seamless cloning with HiFi DNA Assembly (E2621, New England Biolabs, Ipswich, MA, USA). Restriction enzymes used for plasmid construction were purchased from New England Biolabs. PCR fragments used for plasmid construction were amplified by Q5 DNA polymerase (M0491, New England Biolabs) according to manufacturer's instruction. Escherichia coli competent cells NEB 5-alpha (C2987, New England Biolabs), NEB Stable (C3040, New England Biolabs), or Champion DH5a high (CC5202, SMOBIO Technology, Hsinchu City, Taiwan) were used for transformation to amplify and extract plasmids. Plasmids were extracted by FastGene Plasmid Mini Kit Nippon Genetics,Tokyo,Japan). Plasmids used in this study are listed in Supplementary Table S2. The DNA sequence files of the backbone vectors for genome editing are available on our repository at GitHub (https://github.com/poccopen/Genome_editing_plas mid_for_budding_yeast).

Selection of target sequences for genome editing
For the insertion of mNeonGreen-encoding sequence into the CSE4 gene, we selected target sequences from the region encoding the unstructured N-terminal loop of Cse4 protein (Zhou et al. 2011;Yan et al. 2019). For the insertion of mScarlet-I-encoding sequence into the CDC3 gene, we first performed secondary structure prediction by JPred4 (Drozdetskiy et al. 2015) of Cdc3 protein N -20 NNNNNNNNN -11 N -10 N -9 N -8 N -7 N -6 N -5 N -4 N -3 N -2 N -1 NGG N -20 NNNNNNNNN -11 N -10 N -9 N -8 N -7 N -6 N -5 N -4 N -3 N -2 N -1 NNGRRT  Figure 1 Estimation of editable nucleotides in the yeast genome. (A-C) Definition of candidate nucleotides for editing. PAM sequence for each CRISPR/ Cas system is underlined with magenta. The target sequence in sgRNA/crRNA is underlined with blue. Candidate nucleotides for editing with each CRISPR/Cas system are highlighted in orange. Green arrowheads indicate cleavage sites. (D) Fraction of nucleotides in the reference genome editable with each and all possible combinations of the three CRISPR/Cas systems. and then selected target sequences from the region encoding the N-terminal region with no predicted secondary structure.
For designing single guide RNAs (sgRNAs) for SpCas9 and SaCas9, CRISPRdirect (Naito et al. 2015) was used to select target sequences. For designing CRISPR RNAs (crRNAs) for enAsCas12a, CRISPOR (Concordet and Haeussler 2018) was used to select target sequences. Target sequences for genome editing used in this study are listed in Supplementary Table S4.

Construction of genome-editing plasmids
All genome-editing plasmids were constructed using the seamless cloning with Golden Gate Assembly using NEB Golden Gate Assembly Kit (BsaI-HF v2) (E1601, New England Biolabs). The ODNs for Golden Gate Assembly were automatically designed with an in-house program.

Plasmid extraction from yeast cells
Plasmids were extracted from yeast cells by Easy Yeast Plasmid Isolation Kit (#630467, Takara Bio) and transformed into E. coli competent cells (Champion DH5a high, SMOBIO Technology).

Fluorescence microscopy and image processing
Image acquisitions of yeast cells were performed on a microscope (Ti-E, Nikon, Tokyo, Japan) with a 100Â objective lens (CFI Apo TIRF 100XC Oil, MRD01991, Nikon), a sCMOS camera (ORCA-Fusion BT, C15440-20UP, Hamamatsu photonics, Hamamatsu, Japan), and a solid-state illumination light source (SOLA SE II, Lumencor, Beaverton, OR, USA). Image acquisition was controlled by NIS-Elements version 5.3 (Nikon). The binning mode of the camera was set at 2 Â 2 (0.13 lm/pixel). Z-stacks were 13 Â 0.3 lm. For imaging of Cse4-mNeonGreen, a filter set (LED-YFP-A, Semrock, Rochester, NY, USA) was used with excitation light power set at 20% and the exposure time set at 200 msec/ frame. For imaging of Cdc3-mScarlet-I, a filter set (LED-TRITC-A, Semrock) was used with excitation light power set at 7% and exposure time set at 70 msec/frame. For DIC (differential interference contrast) image acquisition, the exposure time was set at 20 ms/frame. DIC images were captured only at the middle position of the Z-stacks.
Image processing and analysis were performed using Fiji (Schindelin et al. 2012). To generate 2-dimensional images of fluorescence channel from Z-stacks, background subtraction (sliding paraboloid radius set at 10 pixels with disabled smoothing) and maximum projection using 13 Z-slices were performed. Maximum projected fluorescence images and corresponding smoothed DIC images were superimposed. After global adjusting of brightness and contrast and cropping of the images, sequences of representative images were generated.
Editable fraction of yeast genome with three CRISPR/Cas systems S. cerevisiae reference genome sequence available at Saccharomyces genome database (SGD) (S288C strain, version R64-2-1, http://sgd-archive.yeastgenome.org/sequence/S288C_ reference/genome_releases/S288C_reference_genome_R64-2-1 _20150113.tgz) without mitochondrial genome and plasmid sequences were searched for PAMs (NGG for SpCas9, NNGRRT for SaCas9, and TTTV for AsCas12a). Both strands were included in the PAM search. After identifying the PAM sequence, nucleotides in a defined distance from the PAM were assigned as candidate nucleotides for editing. In genome editing, it is critical for a successfully edited target sequence in the genome not to be cleaved again by the Cas-gRNA complex bearing the gRNA corresponding to the original, unedited target sequence. We thus defined a nucleotide as a candidate for editing if its substitutions leading to mismatches with the gRNA can significantly reduce the efficiency of re-cleavage by the Cas-gRNA complex. For SpCas9 and SaCas9, the nucleotides at one to eleven nt away from the PAM and the nucleotides consisting the PAM were defined as the candidates based on previous reports ( Figure 1C). Degenerate nucleotides in the PAMs (i.e., N, R, and V) were excluded from the calculation (Figure 1, A-C). The total number of candidate nucleotides for editing is summarized in Supplementary Table S5 and Figure 1D.

Results
Expansion of editable fraction of the yeast genome by combining three CRISPR/Cas systems We evaluated the potentials of three well-established CRISPR/Cas systems with distinct PAMs, namely SpCas9 from S. pyogenes (Jinek et al. 2012), SaCas9 from S. aureus (Ran et al. 2015), and enhanced AsCas12a (enAsCas12a) derived from Acidaminococcus sp. (Kleinstiver et al. 2019), in editing the budding yeast genome. For each system, we calculated the number of "editable" nucleotides in the reference genome sequence of the budding yeast strain, S288C (see "Editable fraction of yeast genome with three CRISPR/ Cas systems" in the Materials and methods section). This simple simulation indicated that while SpCas9, SaCas9, and AsCas12a can target 59.9%, 31.4%, and 61.2% of the genome, respectively, their combined use can cover as much as 89.5% ( Figure 1D, Supplementary Table S5). Based on this simulation, we decided to develop a backbone vector series sharing a single design principle to enable the seamless use of the three CRISPR/Cas systems to expand the repertoire of editable genes.

Design of backbone vectors for yeast genome editing
In the design of the vector series, we defined the following three requirements: (1) both Cas protein and sgRNA/crRNA are encoded on a single plasmid, (2) expression of Cas protein and/or sgRNA/crRNA can be artificially induced, and (3) target sequence of sgRNA/crRNA can be incorporated using the Golden Gate Assembly (Engler et al. 2008). Fulfilling these three requirements, we developed four backbone centromeric plasmid vectors marked with URA3 for the three CRISPR/Cas systems ( Figure 2). Note that enAsCas12a was used instead of AsCas12a because of its improved activity at lower temperatures suitable to grow budding yeast cells (Kleinstiver et al. 2019).
As an inducible promoter, we use the well-characterized GAL1 promoter because it is actively repressed by glucose and strongly activated by galactose in the absence of glucose. Cas-encoding genes on the four vectors are placed under the control of the GAL1 promoter. Similarly, sgRNA/crRNA precursors on three vectors are controlled by the GAL1 promoter.
The Golden Gate Assembly uses type IIS restriction enzymes such as BsaI and BbsI (Engler et al. 2008). Our vector series harbors two BsaI recognition sites for Golden Gate Assembly. Because the target sequence lies at the 5 0 terminal side of the sgRNA scaffold for SpCas9 and SaCas9, one BsaI site is placed just downstream of the GAL1 or SNR52 promoter, and the other site is placed just upstream of the sgRNA scaffold sequence (Figure 2, A-C). In the case of enAsCas12a, the target sequence is located at the 3 0 terminal side of the crRNA scaffold. Accordingly, one BsaI site is placed just downstream of the crRNA scaffold, and the other site is placed further downstream ( Figure 2D). Extra sequences at the 5 0 and 3 0 ends of sgRNA often compromise the efficiency of genome editing and hence should be adequately trimmed. To remove the 5 0 extra sequence, a hammerhead ribozyme is inserted at the beginning of the sgRNA-containing transcript from the three Cas9 vectors (Figure 2, A-C).
To remove the 3 0 extra sequence in the transcripts driven by GAL1 promoter, the hepatitis delta virus (HDV) ribozyme is inserted to the 3 0 side of the sgRNA scaffold ( Figure 2, A and C). For the other vector using SNR52 promoter driven by RNA polymerase III, SUP4 terminator is used to define the 3 0 end of transcript ( Figure 2B). In the enAsCas12a vector, the crRNA is preceded and followed by tRNA(Gly) ) and the HDV ribozyme, respectively, for the removal of extra sequences ( Figure 2D). Furthermore, a sequence encoding "UUUUAUUUU" is inserted between the second BsaI site (i.e., 3 0 end of the crRNA) and the HDV ribozyme, because this 9-mer sequence was demonstrated to increase the efficiency of genome editing ( Figure 2 Schematic representation of backbone vectors for genome editing. (A) SpCas9 þ pGAL1-sgRNA system. (B) SpCas9 þ pSNR52-sgRNA system. (C) SaCas9 system. (D) enAsCas12a system. pGAL1, GAL1 promoter; pSNR52, SNR52 promoter; tADH1, ADH1 terminator; tCYC1, CYC1 terminator; tSUP4, SUP4 terminator; HDV, HDV ribozyme; U 4 AU 4 , 9-mer encoding "UUUUAUU UU" for improvement of genome-editing efficiency; URA3, URA3 marker cassette; CEN/ARS, centromere and autonomously replicating sequence A program to design ODNs for Golden Gate Assembly To construct a genome-editing plasmid, a pair of ODNs corresponding to its target sequence must be synthesized so that they include 4-nt sequences compatible with the backbone vectors. Furthermore, a hammerhead ribozyme compatible with each target sequence must be designed and included in the ODNs (Supplementary Figure S1A). To facilitate this complicated process without the risk of human errors, we created a simple program that automatically calculates the ODNs for a given target sequence (Supplementary Figure S1B). Upon entering a target sequence with its name followed by the selection of a backbone vector, the program readily provides ODN sequences to be synthesized for the Golden Gate Assembly of a genome-editing plasmid on the selected backbone vector. The program is available from our repository on GitHub (https://github.com/poccopen/ Genome_editing_plasmid_for_budding_yeast).

Application example 1: gene insertion by SpCas9 and enAsCas12a systems
In budding yeast, while the C-terminal tagging of an essential gene can be performed easily in a single-step procedure (Longtine et al. 1998), the insertion of a DNA fragment to an essential gene at a location other than its C-terminus requires multiple steps and more extended time. Genome-editing using CRISPR/Cas systems can simplify the gene fragment insertion processes. As an application example of the CRISPR/Cas systems, we attempted to insert a fluorescent protein gene into an internal portion of essential genes. We chose the CSE4 gene and the CDC3 gene as our targets of fluorescent protein gene insertion. The CSE4 gene encodes a centromere-specific histone H3 variant, Cse4 (Stoler et al. 1995). It was reported that when a fluorescent protein is fused at the C-terminus of Cse4, the cells show temperature sensitivity (Wisniewski et al. 2014). In contrast, when the fluorescent protein is inserted into the unstructured N-terminal loop of Cse4 (Zhou et al. 2011;Yan et al. 2019), the cells grow normally at a higher temperature (Wisniewski et al. 2014). The CDC3 gene encodes one of the septin proteins, which form a ring structure along the bud neck (Caviston et al. 2003). It was shown that when a fluorescent protein is fused to the C-terminus of Cdc3, the localization of Cdc3 protein becomes abnormal, leading to a morphological defect (Huh et al. 2003;Dubreuil et al. 2019). In contrast, when a fluorescent protein is inserted into an N-terminal loop of Cdc3, the tagged Cdc3 protein correctly localizes at the bud neck, and the cells grow normally without showing any morphological defect (Caviston et al. 2003).
As an application example of the SpCas9 þ pGAL1-sgRNA system, we attempted to insert a gene fragment encoding mNeonGreen, a bright yellow-green fluorescent protein (Shaner et al. 2013), into the N-terminal loop of Cse4. We designed 8 target sites in the genic region encoding the N-terminal loop of Cse4. To insert the mNeonGreen gene fragment into these sites, we prepared donor PCR fragments harboring 45-bp homology arms at both termini ( Figure 3A). For each of the 8 target sites, the yeast cells were co-transformed with the corresponding genomeediting plasmid and the donor PCR fragment. The transformants formed a mixture of large and small colonies (Supplementary Figure S2A). The gene fragment insertion efficiency of the small and large colonies was evaluated by a PCR assay using primer sets flanking the insertion position. For the target sequence CSE4-1, small colonies showed a significantly higher insertion efficiency (87.5%, n ¼ 24, or 8 colonies in each of 3 biological replicates) than large colonies (4.2%, n ¼ 24, or 8 colonies in each of 3 biological replicates) (Supplementary Figure S2B).
Observing the heterogeneity in colony size, we hypothesized that loss of genome-editing function results in loss of cell cycle arrest induced by DNA DSB and its repair and leads to the formation of large colonies. It was likely that intramolecular recombination between the two GAL1 promoters led to the loss of SpCas9 expression cassette (Supplementary Figure S2C). To test this hypothesis, we analyzed the structure of the genome-editing plasmids in the cells forming large colonies by restriction enzyme digestion and PCR. All the 24 plasmids derived from large colonies showed the structural change consistent with the predicted deletion caused by recombination between the two GAL1 promoters (Supplementary Figure S2, D and E). Based on these results, we picked only small colonies in the subsequent genomeediting experiments with plasmids harboring two GAL1 promoters.
Colony size and the number of colonies were analyzed for the plates shown in Supplementary Figure S2 using image analysis (Supplementary Figure S3). While the number of small colonies is highly variable depending on the target sequence, the frequency of appearance of large colonies is almost constant among the eight plates (Supplementary Figure S3, B and C). This result shows that it is likely that the frequency of intramolecular recombination through two GAL1 promoter sequences is stable. There is a strong negative correlation between the number of small colonies and the insertion efficiency (Supplementary Figure S3F). This result suggests that the gRNA efficacy, which is reflected in the viability of the cells, is a major determinant of the success rate of the genome-editing. In contrast, a significant correlation was not observed between the number of large colonies and the insertion efficiency (Supplementary Figure S3E). This result suggests that the intramolecular recombination takes place independently of the genome-editing processes. Based on the colony numbers of the cells (CSE4-3, showing the highest viability among the eight conditions), the frequency of intramolecular recombination is estimated to be no more than 4.6% (65/1407) (Supplementary Figure S3, A and B).
Using CSE4-1 as a model target sequence, we also investigated the relationship between the length of homology arms and the insertion efficiency. We used PCR fragments harboring 4 different homology arm lengths (15-, 25-, 35-, and 45-bp) for genome editing. There was a positive correlation between homology arm length and insertion efficiency (Supplementary Figure S4). A similar positive correlation has been reported between homology arm length and genome-editing efficiency in fission yeast (Hayashi and Tanaka 2019). When a donor PCR fragment harboring 15-bp homology arms was used, no gene insertion was observed (Supplementary Figure S4). Based on these results, PCR fragments harboring 45-bp homology arms were used in the subsequent genome-editing experiments for gene insertion.
Among the 8 target sequences (Supplementary Table S4), the gene insertion efficiency varied from 0% to 91.7% (n ¼ 24 for each target sequence) ( Figure 3B, Supplementary Table S4). Clones with successful gene insertions were obtained for 7 out of the 8 target sequences.
We investigated the phenotype of the successfully genomeedited cells. In all the genome-edited clones tested, the mNeonGreen fluorescence signal was localized as a single spot or a pair of spots in each cell ( Figure 3C and Supplementary Figure  S5A). The localization pattern of the mNeonGreen signal was indistinguishable between the genome-edited cells and the cells generated using the conventional method to harbor mNeonGreen

CSE4-mNeonGreen (internal) (enAsCas12a genome-edit)
As an example of the use of SpCas9 þ pSNR52-sgRNA system, we attempted to insert a gene fragment encoding a bright red fluorescent protein, mScarlet-I (Bindels et al. 2017), into an N-terminal region of Cdc3 predicted to lack any secondary structure. We designed 4 target sequences in the genic region corresponding to the N terminal region. To insert the mScarlet-I gene fragment, we prepared donor PCR fragments harboring 45-bp homology arms at both termini ( Figure 3A). The transformation of yeast cells with a genome-editing plasmid and a corresponding donor PCR fragment resulted in a mixture of large and small colonies (Supplementary Figure S6A). However, for all the 4 target sequences, there was no statistically significant difference in insertion efficiency between the small and large colonies (Supplementary Figure S6B).
The gene insertion efficiency varied from 8.3% to 66.7% among the 4 target sequences (n ¼ 48 for each target sequence) ( Figure 3D). In all the genome-edited clones examined, mScarlet-I fluorescent signal was localized as a ring structure at the bud neck ( Figure 3E and Supplementary Figure S6C). None of the 12 genome-edited clones (3 clones for each target sequence) exhibited morphological defect (Supplementary Figure S6C). All the genome-edited clones (4 clones for each target sequence) showed comparable growth at 37 C with the wild-type cells (Supplementary Figure S6D).
As an application example of enAsCas12a þ pGAL1-crRNA system, we attempted to insert a gene fragment encoding mNeonGreen to the genic regions encoding the N-terminal loop of Cse4, as we did above (Figure 3, B and C). We designed 8 target sequences and prepared donor PCR fragments harboring 45-bp homology arms at both termini ( Figure 3A). The gene insertion efficiency varied from 0% to 87.5% among the 8 target sequences (n ¼ 24 for each target sequence) ( Figure 3F). Clones with successful gene insertions were obtained for 6 out of the 8 target sequences ( Figure 3F). In all genome-edited clones, the mNeonGreen fluorescence signal was localized as a single spot or a pair of spots in each cell ( Figure 3G and Supplementary Figure S7A), and the growth at 37 C was comparable to the wild-type cells (Supplementary Figure S7B). These results were consistent with those obtained for the cells generated using the conventional approach and the SpCas9 þ pGAL1-sgRNA system.

Application example 2: complete ORF deletion by SaCas9 system
As an example of the use of SaCas9 þ pGAL1-sgRNA system, we attempted to delete an entire ORF. We chose the ADE3 gene as a target of complete ORF deletion. The ADE3 gene encodes C1tetrahydrofolate synthase, an enzyme required for adenine biosynthesis (McKenzie and Jones 1977). The cells lacking the ADE2 gene encoding phosphoribosylaminoimidazole carboxylase, another enzyme in the adenine biosynthesis pathway, form red colonies by accumulating intermediate metabolites of red color (Hieter et al. 1985). When the ADE3 gene is deleted in the cells lacking ADE2, colony color returns to white because of the loss of accumulation of the red metabolites (Koshland et al. 1985). We attempted to delete the ADE3 ORF in an ade2D strain and convert colony color from red to white ( Figure 4A). We selected 4 target sequences in the ORF, constructed the corresponding genomeediting plasmids, and used them to transform the ade2D cells with or without a 100-bp donor PCR fragment composed of the 5 0 -and 3 0 -flanking sequences of the ORF ( Figure 4A). The transformation with the genome-editing plasmids resulted in the formation of white colonies on galactose-containing plates ( Figure 4B, top row). Transformation with a control plasmid, YCplac33 vector with no expression of Cas protein and sgRNA, failed to form white colonies ( Figure 4B). Even when the genome-editing plasmids were used, white colonies did not appear among the transformants on glucose-containing plates ( Figure 4B, bottom row). These results indicated that the galactose-inducible SaCas9 system worked as we expected.
When the cells were transformed with a genome-editing plasmid and the donor PCR fragment for the ORF deletion, the proportion of white colonies on galactose-containing plates was in the range of 73.7-81.6% ( Figure 4C). The formation of white colonies does not necessarily indicate complete deletion of ADE3 ORF, as small insertion or deletion (indel) could also result in the loss of Ade3 function. We performed a PCR assay to distinguish the complete deletion of ADE3 ORF from small indels (Supplementary Figure S8A). PCR products consistent with complete deletion of ADE3 ORF were obtained in all the 32 white colonies examined (8 colonies for each target sequence) (Supplementary Figure S8B, top). We then checked the sequences of these PCR products and confirmed the complete loss of ADE3 ORF (Supplementary Figure S9).
We also attempted to knock out the ADE3 gene through nonhomologous end joining. For this purpose, we transformed the yeast cells solely with the genome-editing plasmids. In this case, loss of the ADE3 gene function should be attributable to frameshift mutations caused by indels in the vicinity of SaCas9 cleavage site. The proportion of white colonies on galactosecontaining plates ranged from 32.2% to 66.1%, which is lower than that of the cells co-transformed with the genome-editing plasmids and the donor PCR fragment (73.7%-81.6%) ( Figure 4C). We performed a PCR assay to exclude the possibility of large deletions (Supplementary Figure S8, A and B, bottom). Sequencing of these PCR products confirmed the presence of small indels (1-2 bp) in the vicinity of the expected SaCas9 cleavage sites (Supplementary Figure S10).
Application example 3: introduction of single amino acid substitutions by SpCas9, SaCas9, and enAsCas12a systems As another example of the use of the CRISPR/Cas systems, we attempted to introduce single amino acid substitutions and chose three target genes, PIF1, ORC6, and TAO3.
PIF1 encodes a DNA helicase which facilitates DNA synthesis during break-induced replication (Chung et al. 2010). Pif1 functions in the nucleus and mitochondria (Schulz and Zakian 1994). The nuclear and mitochondrial functions of PIF1 can be separated by mutations (Schulz and Zakian 1994). PIF1 has two initiation codons which determine localization of the Pif1 protein: translation starting from the first initiation codon (M1) and the second one (M40) leads to the production of Pif1 protein bearing and lacking the N-terminal mitochondrial targeting signal, respectively. Accordingly, pif1-m1 (M1A) allele encodes only the protein isoform that starts from M40 to lack the targeting signal and is hence incapable of mitochondrial localization. In contrast, pif1-m2 (M40A) allele encodes only the isoform that starts from M1 to bear the targeting signal and is hence imported to mitochondria. In this study, we attempted to generate the pif1-m2 (M40A) allele using the CRISPR/Cas systems. For the introduction of the M40A substitution, only a single sgRNA/crRNA can be designed for each of the CRISPR/Cas ( Figure 5B). We thus  Figure 4 Complete deletion of ADE3 ORF using the SaCas9 system. (A) Process for deleting the entire ADE3 ORF and converting colony color. The dark magenta arrows represent the positions of the target sequences for SaCas9. The 50-bp regions used as the 5 0 -and 3 0 -homology arms of the 100-bp donor PCR fragment are colored blue and orange, respectively (schematic not proportional to actual size). While ade2D cells accumulate red pigments on adenine-limited medium, ade2Dade3D cells fail to do so and form white colonies. (B) Representative images of colonies after transformation of the genome-editing plasmids with or without the donor PCR products on adenine-limited galactose-containing medium (top, Gal) or adenine-limited glucose-containing medium (bottom, Dex). The target sequences used for genome editing are shown above the panels. (C) The efficiency of colony color conversion. Each white bar indicates the percentage of white colonies in each experimental condition shown at the bottom. The value is the average of 3 biological replicates indicated by black dots.  constructed a total of four genome-editing plasmids using the four backbone vectors (Figure 2). We prepared a donor PCR fragment containing mutations relevant to M40A substitution ( Figure 5A). For each of the target sites, the yeast cells were cotransformed with the corresponding genome-editing plasmid and the donor PCR fragment. The knock-in efficiency was evaluated by a pair of PCR assays using primer sets specific for the M40A mutant allele and the wild-type M40M allele (Supplementary Figure S11). We also checked the sequence of the region and confirmed the introduction of the M40A substitution (Supplementary Figure S12). Among the four systems, the knockin efficiency varied from 25.0% to 100% (n ¼ 24 for each target sequence) ( Figure 5C). Compared to the systems using the pGAL1-sgRNA/crRNA systems (83.3%-100%), the SpCas9 þ pSNR52-sgRNA system showed a lower efficiency (25.0%). It is possible that the T 4 stretch in the target sequence ( Figure 5B) can lead to a pause or premature termination of transcription of the sgRNA by RNA polymerase III (Arimbasseri and Maraia 2015).

5'-ACATTTGCTGTTGATCTACTTTCAGCTGTTGAAACGAAATTACATAATTCTTCTTATACCAAG-3' 3'-TGTAAACGACAACTAGATGAAAGTCGACAACTTTGCTTTAATGTATTAAGAAGAATATGGTTC-5' T F A V D L L S A V E T K L H N S S Y T K
ORC6 encodes a subunit of the origin recognition complex required to start DNA replication (Li and Herskowitz 1993). It has been reported that mutations causing substitution of Ala for Ser-116 of Orc6, a target site of cyclin-dependent kinases, increases the frequency of re-replication under a specific genetic background (Richardson and Li 2014). We attempted to generate orc6 (S116A) allele using CRISPR/Cas systems. For the introduction of the S116A substitution, one sgRNA, no sgRNA, and three crRNA can be designed for SpCas9, SaCas9, and AsCas12a, respectively ( Figure 5D). We obtained the co-transformants of the four pairs of the genome-editing plasmid and the donor PCR fragment and analyzed them with PCR and sequencing (Supplementary Figures  S13 and S14). The knock-in efficiency varied from 0% to 79.2% (n ¼ 24 for each target sequence) ( Figure 5E). Clones with successful gene insertions were obtained for 3 out of the 4 target sequences.
TAO3 encodes a protein involved in apical bud growth and cell morphogenesis. Intriguingly, TAO3 (E1493Q) allele was identified as one of the quantitative trait loci affecting the sporulation efficiency (Deutschbauer and Davis 2005). While the S288C strain and its derivative BY4741 strain have the TAO3 (E1493E) allele and show a very low sporulation efficiency, the SK1 strain has the TAO3 (E1493Q) allele and shows a very high sporulation efficiency. The introduction of the TAO3 (E1493Q) allele in the S288C cells increased the sporulation efficiency. In this study, we attempted to generate the TAO3 (E1493Q) allele using the CRISPR/Cas systems. Notably, the E1493Q substitution was found to be achievable only with AsCas12a ( Figure 5F). For each of the three target sites, the yeast cells were co-transformed with the corresponding genome-editing plasmid and the donor PCR fragment containing the relevant mutations, and the cotransformants were examined with PCR and sequencing (Supplementary Figures S15 and S16). Among the three crRNAs, the knock-in efficiency varied from 79.2% to 91.7% (n ¼ 24 for each target sequence) ( Figure 5G). Clones with successful gene insertions were obtained for 3 out of the 3 target sequences.

Discussion
Here, we reported a series of vectors for yeast genome editing using three different CRISPR/Cas systems, namely SpCas9, SaCas9, and enAsCas12a (Figure 2). Because the three systems have distinct PAMs, their combined use expands the editable fraction of the yeast genome, as indicated by our simulation (Figure 1). To facilitate the seamless use of these systems, we constructed a vector series under a unified design principle. First, all the vectors harbor URA3 marker and GAL1 promoter, thus sharing the media required for their use. Accordingly, if a single certain system fails to edit a region of interest, one can readily switch to another system without preparing any additional medium. Second, all the vectors are compatible with the highly efficient Golden Gate Assembly, thus making the construction step virtually free from failure. Furthermore, a dedicated program is developed to design ODNs for Golden Gate Assembly of individual genome-editing plasmids on these backbone vectors. Target search with CRISPRdirect (Naito et al. 2015) and CRISPOR (Concordet and Haeussler 2018) followed by ODN design with this program would thus streamline the entire process to design genome-editing plasmids.
The realistic schedule of genome editing described in this study is summarized in Table 1. The entire process from designing a genome-editing plasmid to obtaining genome-edited strains can be completed within 2 weeks. This period is substantially shorter than the one required for the traditional yeast genetics approach, especially, in the case of inserting a gene fragment into an essential gene. For instance, when using the own promoter of an essential gene, the traditional approach includes the construction of a cover plasmid carrying the wild-type allele of the essential gene, transformation of the cover plasmid, disruption of the genomic copy of the essential gene, introduction of an adequately modified allele, and the curing of the cover plasmid, thus taking at least 17 days or, more realistically, >20 days.
It is intriguing to note that the period for genome editing can be further shortened with these plasmids. In this study, we used galactose to strongly activate GAL1 promoter at the expense of substantially compromised growth compared to that in the presence of glucose, the ideal carbon source for S. cerevisiae. Notably, the artificial transcription factor GEV (i.e., a fusion protein composed of Gal4 DNA-binding domain, estrogen receptor, and VP16) can activate the GAL1 promoter upon estradiol addition in glucose media (Hickman et al. 2011). We thus expect that a GEVbearing strain reconciles efficient induction and rapid growth, thereby further shortening the period required for genome editing using these vectors.
We examined the performance of genome-editing plasmids using these backbone vectors in our attempts for the insertion of gene fragments to essential genes (Figure 3), the complete deletion of an ORF (Figure 4), and the introduction of single amino acid substitutions ( Figure 5). In the case of gene fragment Pick up yeast colonies 10 Perform yeast colony PCR; inoculate yeast cells in YPAD 11 Spread yeast cells for single colony isolation 12 13 Pick up single yeast colonies 14 Perform yeast colony PCR insertion, successfully genome-edited cells were obtained for 17 out of 20 target sequences examined (Figure 3, B, D, and F), and the insertion efficiency exceeded 50% for 11 target sequences. In the case of complete ORF deletion, the efficiency was larger than 70% for all the 4 target sequences tested ( Figure 4C). We also examined the growth of genome-edited clones at 30 C and 37 C (Supplementary Figures S5B, S6D, and S7B). None of the 55 clones showed temperature-sensitive growth. In the case of the introduction of single amino acid substitutions, successfully genomeedited cells were obtained for 10 out of 11 sgRNAs/crRNAs examined ( Figure 5, C, E, and G), and the knock-in efficiency exceeded 50% for 9 target sequences. These results proved the practical utility of the vector series developed in this study. For one of the target genes for the introduction of single amino acid substitutions [TAO3 (E1493Q)], it is impossible to design sgRNAs for SpCas9 and SaCas9 due to the lack of the cognate PAMs in the vicinity of the target site ( Figure 5F). Accordingly, the genome-editing was only possible and successfully attained with the enAsCas12a system ( Figure 5G). This result demonstrates the value of having all three systems available as a vector series constructed in a unified design principle.
We should refer to a practical rule of thumb for successful genome editing using the three backbone vectors bearing two GAL1 promoters. When using these vectors, intramolecular recombination between the two promoters tends to lead to the formation of large colonies with low efficiency of genome editing (Supplementary Figure S2). We thus recommend the users of these vectors to simply discard large colonies and select small ones for further analyses because the latter showed significantly higher genome-editing efficiency than the former (Supplementary Figure S2). While the single GAL1 promoter plasmid also led to heterogeneity in colony size, no difference in genome-editing efficiency was observed between large and small colonies (Supplementary Figure S6).
Our application examples included the insertion of fluorescent proteins into positions that are neither N-nor C-end of the essential proteins Cse4 and Cdc3 (Figure 3). Tagging at inappropriate sites of these proteins was reported to induce temperaturesensitive growth and/or morphological defects. To avoid the adverse effects of inserting a fluorescent protein on the recipient protein folding, we took a strategy to select an insertion site from regions demonstrated or predicted to have no secondary structure. All the proteins thus fluorescently tagged, including those using previously unvalidated sites, showed physiological localization, and the cells thus modified exhibited neither temperaturesensitive growth nor abnormal morphology. These results suggest the general utility of our strategy.
Taken together, the backbone vectors and the program developed in this study would provide a versatile toolbox to facilitate various types of genome manipulation in S. cerevisiae, including those challenging to perform with conventional techniques in yeast genetics.

Data availability
The four backbone vectors are available from NBRP Yeast Resource Center (https://yeast.nig.ac.jp/yeast/). The source codes of programs for ODN design and PAM search are available from our repository at GitHub (https://github.com/poccopen/Genome_ editing_plasmid_for_budding_yeast). Other strains and plasmids are available upon request. The authors state that all data necessary for confirming the conclusions presented here are represented fully within the article.
Supplementary material is available at G3 online.