Construction and evaluation of gRNA arrays for multiplex CRISPR‐Cas9

Abstract

Endonuclease system CRISPR‐Cas9 represents a powerful toolbox for the budding yeast's Saccharomyces cerevisiae genome perturbation. The resulting double‐strand breaks are preferentially repaired via highly efficient homologous recombination, which subsequently leads to marker‐free genome editing. The goal of this study was to evaluate precise targeting of multiple loci simultaneously. To construct an array of independently expressing guide RNAs (gRNAs), the genes encoding them were assembled through a BioBrick construction procedure. We designed a multiplex CRISPR‐Cas9 system for targeting 6 marker genes, whereby the gRNA array was expressed from a single plasmid. To evaluate the performance of the gRNA array, the activity of the designed system was assessed by the success rate of the introduction of perturbations within the target loci: successful gRNA expression, followed by target DNA double‐strand breaks formation and their repair by homologous recombination led to premature termination of the coding sequence of the marker genes, resulting in the prevention of growth of the transformants on the corresponding selection media. In conclusion, we successfully introduced up to five simultaneous perturbations within single cells of yeast S. cerevisiae using the multiplex CRISPR‐Cas9 system. While this has been done before, we here present an alternative sequential BioBrick assembly with the capability to accommodate many highly similar gRNA‐expression cassettes, and an exhaustive evaluation of their performance.

Constructs to express guide RNAs (gRNAs) individually were assembled as BioBricks. The evaluation of single and multiplex CRISPR‐Cas9 gRNA systems was performed. Up to five successful simultaneous perturbations were achieved and assessed.

Take‐aways

• Constructs to express guide RNAs (gRNAs) individually were assembled as BioBricks.

• The evaluation of single and multiplex CRISPR‐Cas9 gRNA systems was performed.

• Up to five successful simultaneous perturbations were achieved and assessed.

1. INTRODUCTION

Double‐strand breaks (DSBs) that are created by the Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)‐Cas9 endonuclease system in the budding yeast Saccharomyces cerevisiae are preferentially repaired via homologous recombination (HR) (Bao et al., 2015; DiCarlo et al., 2013; Jakobson et al., 2019; Jakočiūnas, Bonde, et al., 2015; Stovicek et al., 2017), as opposed to the preference for nonhomologous end‐joining (NHEJ) in mammalian cells (Cong et al., 2013). Thus, providing to the yeast DNA repair mechanism a designated DNA matrix with adequate homology‐like regions usually results in precise genome editing. Transforming S. cerevisiae with a Cas9‐expression vector, followed by a guide RNA (gRNA)‐expression vector and excess of linear DNA matrix for HR therefore typically results in successfully modified genomes (DiCarlo et al., 2013; Malcı et al., 2020; Stovicek et al., 2017). However, to introduce multiple perturbations within the genome, sequential CRISPR‐Cas9 systems, whereby plasmid loss of the preceding gRNA‐expression vector is required (Bao et al., 2015; Mans et al., 2015, 2018; Stovicek et al., 2017), have often been used. The goal of this study was to improve and evaluate precise genome editing in S. cerevisiae from a single locus to multiple loci simultaneously, by using a single vector for an array of gRNAs. In nature, such multiplexing of type II bacterial CRISPR‐Cas system as an adaptive immune response is achieved through RNAse III‐mediated cleavage of the pre‐crRNA (i.e., direct repeat‐spacer‐direct repeat) array (Ciurkot et al., 2019; Cong et al., 2013; Liu et al., 2022; Stovicek et al., 2017). Each of the processed crRNAs then forms a duplex with the tracrRNA, which in using CRISPR as a genome editing tool can be mimicked by a synthetic chimeric gRNA that functions as a search unit of the ribonucleoprotein complex (Bao et al., 2015; Cong et al., 2013; DiCarlo et al., 2013; Liu et al., 2022; Stovicek et al., 2017).

S. cerevisiae is well‐suited for genetic engineering and synthetic biology applications because of the high performance of its endogenous HR machinery (Lee et al., 2015). With adequate homologous‐like flanking sequences, linear parts of DNA can be assembled in vivo (Shao et al., 2009) or incorporated within a target locus, to e.g. construct a complete gene knockout collection (Giaever & Nislow, 2014). By the introduction of DSBs with a CRISPR‐Cas9 system, the efficiency of such desired site‐specific perturbations is severalfold increased (DiCarlo et al., 2013; Lee et al., 2015; Mans et al., 2015; Ryan et al., 2014). Notably, while multiplexed CRISPR‐Cas9 can be used for high‐throughput genome‐wide DNA editing approach that within a pool of yeast cells perturbs each cell once, but in a cell‐specific manner (Bao et al., 2018; Lian et al., 2017, 2019; McCarty et al., 2020; Meng et al., 2020; Roy et al., 2018; Sharon et al., 2018), the goal of this study was to simultaneously perturb the genome of single cells at multiple loci. Multiplex CRISPR‐Cas9 systems that rely on the co‐transformation of up to three expression vectors for one or two gRNAs have been described (Adiego‐Pérez et al., 2019; Malcı et al., 2020; Mans et al., 2015, 2018; Stovicek et al., 2017; Wijsman et al., 2019), but their drawback is the limited number of selection markers and the plasmid cell burden. Another approach to multiplex CRISPR‐Cas9 system that has been developed is the fusion of a gRNA to the 3′ end of the self‐cleaving hepatitis delta virus (HDV) ribozyme. The HDV ribozyme sequence protects the 5′ end of the gRNA from degradation and cleaves the sequence at its 5′ end. A construct [tRNA‐HDV ribozyme‐gRNA‐SNR52t] _n , where n can be between 1 and 4, has been constructed, whereby tRNA acts as RNA Pol III promoter, and HDV ribozyme‐gRNA‐SNR52t RNA is released to form a ribonucleoprotein complex with Cas9 (Jakočiūnas, Bonde, et al., 2015; Lee et al., 2015; Liu et al., 2022; Malcı et al., 2020; Ryan et al., 2014; Stovicek et al., 2017; Y. Zhang et al., 2019). Such an approach requires relatively long constructs, which in addition to the tRNA promoter, gRNA, and the transcription terminator contain also the HDV sequence. Notably, the activity of tRNA as an RNA Pol III promoter has been shown to be strain‐specific (Ryan et al., 2014). To avoid the HDV ribozyme sequence, arrays of SNR52p‐tRNA‐gRNA constructs have been designed, and shown that they could be processed with the host RNase P and RNase Z to some extent (McCarty et al., 2020; Meng et al., 2020; Y. Zhang et al., 2019), but this RNase activity is insufficient to process longer RNA arrays (Adiego‐Pérez et al., 2019; Bao et al., 2015; Meng et al., 2020). In addition, expression of recombinant endonuclease Csy4, which can excise specific supplementary repeats between gRNAs within an array, has been reported (Adiego‐Pérez et al., 2019; Ciurkot et al., 2019; Ferreira et al., 2018; Liu et al., 2022; Malcı et al., 2020; McCarty et al., 2020; Meng et al., 2020; Y. Zhang et al., 2019).

To avoid the need for the expression of supplementary components, such as HDV ribozyme or Csy4 endonuclease, and to bypass the suboptimal efficiency of the S. cerevisiae endogenous endonucleases, we designed DNA constructs to express each gRNA independently. We decided to use the SNR52 promoter because of the reported dependence of the activity of tRNAs as RNA Pol III promoters on the strain background. The resulting gRNA arrays contain a relatively high amount of sequence repetition. To assess the efficiency of the selected design and genetic elements, a thorough evaluation of the efficiency of the designed constructs was performed.

2. METHODS

2.1. Strains and media

A prototrophic CEN.PK113‐7D MATa haploid S. cerevisiae strain (Daranlapujade et al., 2003; Nijkamp et al., 2012) was used as the starting strain.

Yeast extract, peptone, and glucose for the YPD medium (1% yeast extract, 2% peptone, 2% glucose), agar (2% for solid media plates), complete supplement mixtures (CSM) with single dropout of amino acids or uracil (CSM [790 mg/L], CSM‐His [770 mg/L], CSM‐ura [770 mg/L], CSM‐Arg [740 mg/L], CSM‐Lys [740 mg/L], CSM‐Leu [690 mg/L], CSM‐Trp [740 mg/L]), and yeast nitrogen base (YNB) without amino acids (6.9 g/L) were purchased from Formedium. The antibiotic nourseothricin (clonNAT; Jena Bioscience) was added to the final concentration of 0.1 mg/ml, and the antibiotic G‐418 disulphate (Formedium) to the final concentration of 0.5 mg/ml. To assess the activity of the CAN1 gene, the arginine toxic analog L‐canavanine sulfate (Sigma‐Aldrich) was added to the CSM‐Arg medium to the final concentration of 60 μg/ml.

Plasmids were amplified by Escherichia coli DH5α strain, for which transformation by a standard heat‐shock protocol (Froger & Hall, 2007) was used. The bacteria were grown in the LBA medium (1% tryptone, 1% NaCl, 0.5% yeast extract, 100 μg/ml ampicillin).

2.2. Plasmid construction

Single gRNA‐expressing plasmids were constructed by Gibson assembly (Gibson et al., 2009) of two fragments. The respective fragments were PCR‐amplified (Q5 Hot Start High‐Fidelity DNA Polymerase, NEB) from the backbone of a pre‐existing p426 plasmid (DiCarlo et al., 2013; Addgene #43803), using a pair of universal primers and gRNA‐specific pairs which incorporated distinct 20 bp crRNA sequences within their overhangs (Supporting Information: Table 1).

Linear donor DNA molecules used as matrices for HR were either PCR‐amplified (Taq DNA Polymerase) from the genomic DNA of the CEN.PK S. cerevisiae strain within target loci with one of the primers inserting an in‐frame stop codon by point mutation in the place of the protospacer adjacent motif (PAM) sequence (DiCarlo et al., 2013), or prepared by annealing two long oligonucleotides into dsDNA to delete the open reading frame entirely (Supporting Information: Table 1). Regardless of the method used, at least one arm of the donor DNA had a 45 bp‐long sequence identical to the target genome locus.

Synthetic DNA fragments for arrayed gRNAs were purchased from IDT DNA as gBlocks. Their structure to express each gRNA independently was arranged as spacer (58 bp)‐SNR52 promoter (269 bp)‐specific crRNA part (20 bp)‐tracrRNA part (75 bp) of chimeric gRNA‐spacer (4 bp)‐SUP4 terminator (20 bp)‐spacer (5 bp), with the total length of 451 bp (Supporting Information: Table 2).

For the multiple gRNA arrays we implemented BioBrick assembly (BglBricks, BB RFC 21 [Røkke et al., 2014]) procedure. A detailed protocol is provided in the Results section. Restriction enzymes were purchased from ThermoFisher Scientific as FastDigest derivatives and T4 DNA Ligase was purchased from NEB.

2.3. Transformation

We performed CRISPR‐Cas9 endonuclease experiment as described in DiCarlo et al. (2013). Briefly, Cas9 with the SV40 nuclear localization sequence was expressed from a low‐copy CEN6/ARS4 p414 plasmid with the clonNAT selection marker (Addgene #43802) under a strong constitutive TEF1 promoter. Next, we co‐transformed the transformants with linear PCR matrices for HR and high‐copy 2μ gRNA expression plasmid p426 with the G‐418 selection marker (Addgene #43803).

Yeast transformation was performed with the lithium acetate‐ssDNA‐PEG method as described in Gietz and Schiestl (2007). Special caution was taken during the second transformation, whereby cells expressing Cas9 from the p414 plasmid were grown for four generation times (OD₆₀₀ 0.05–0.5) after the dilution of the overnight culture into fresh liquid YPD medium with clonNAT. Cells from 10 ml of such freshly grown culture were harvested and washed with 0.1 M lithium acetate. They were co‐transformed with 200 ng of one of the gRNA‐expressing plasmids and 5 μg of each linear matrix with prolonged 1^st incubation time at 30°C for 1 h, followed by a heat shock at 42°C for 30 min. Recovery time of the final incubation at 30°C was prolonged to 4 h.

Selection of double‐transformants was achieved on solid agar YPD clonNAT G‐418 plates after 3 days of incubation at 30°C.

2.4. Phenotyping

Final selected double‐transformants were transferred to fresh YPD plates (60 colonies per plate) and replica plated onto six test dropout solid media plates (CSM‐His, CSM‐ura, CSM‐Arg+canavanine, CSM‐Lys, CSM‐Leu, CSM‐Trp) and two control plates (CSM, YNB). The growth of each colony was assessed on all media plates. Editing efficiencies were calculated as the percentage of transformants with the expected phenotype, that is, the transformants in which all the respective loci had been successfully targeted and modified, over all transformants.

3. RESULTS

The aim of this study was to design and evaluate a multiplex CRISPR‐Cas9 system for precise targeting of multiple yeast genomics loci simultaneously. The first step of the experiment was to construct gRNA‐expressing plasmids, either individually or as an array of independently expressing gRNAs, in a manner that could be used for any combination of yeast genomic loci.

The target sequences for gRNAs were selected using the Benchling tool (Doench et al., 2016; Hsu et al., 2014) as shown in Table 1. For adequate activity, the on‐target score >60, indirectly based on RNA secondary structures formation (Doench et al., 2016), was required, and for specificity, the off‐target score of 99.9%–100% was required (Hsu et al., 2014). For each of the six single gRNA‐expressing plasmids, 20 bp long crRNA‐encoding sequences were inserted by PCR amplification and Gibson assembly (Gibson et al., 2009).

Table 1.

Properties of the CRISPR‐Cas9 genomic targets

Marker gene	Protospacer sequence	On‐target score	DSB position	Chr.
HIS3	CCG‐TAGTGAGAGTGCGTTCAAGG	60.6	566	XV
URA3	GGGTCAACAGTATAGAACCG‐TGG	91.9	448	V
CAN1	CTAAGGATAAAAACGAAGGG‐AGG	83.3	840	V
LYS2	CCC‐CTCAGTTGTTCCGTTTGGCC	70.8	406	II
LEU2	CCG‐CCATGATCCTAGTTAAGAAC	72.0	686	III
TRP1	CCC‐TTGTTTGATTCAGAAGCAGG	92.7	429	IV

Note: The 20 bp protospacer sequences are shown in the 2^nd column in 5′−3′ orientation of the marker gene, whereby the PAM sequence is written in italics and separated by a hyphen from the remaining sequence. The target sequences for gRNAs within the marker genes were selected according to their high specificity and activity (3^rd column). The corresponding double‐strand break position within the coding sequence of the gene is written in the 4^th column, followed by the chromosome on which the gene is located.

To construct a single plasmid expressing multiple gRNAs, BglBricks BioBrick assembly was used (Figure 1). Note that any existing type of the BioBrick assembly could be implemented for such a purpose, as long as a mixed site could be formed after the ligation of two different restricted overhangs and no other restriction site is found within the plasmid or the synthetic DNA fragment. For multiplex CRISPR‐Cas9 purpose, numerous gRNA constructs could be assembled this way—up to six were attempted in this study, but most probably more could be combined as well.

Figure 1.

Graphical overview of the assembly of the multiple gRNAs‐expression vector. First, the synthetic gRNA fragment (gray dashed line) and the cloning vector (solid black line) were PCR‐amplified to add the restriction sites. The cloning vector already contained the first gRNA that was inserted by the Gibson assembly, and/or an array of additional gRNAs from the previous assembly cycles (black dashed line). The overhangs of both linear fragments to be assembled were processed by restriction endonucleases (encircled by dashed lines). Ligation was performed and as a result of the formed BamHI‐BglII mixed restriction site, a new gRNA fragment (gray dashed line) was inserted at the 5′ end of the assembled array (black dashed line). The BioBrick assembly cycle can be repeated by the EcoRI‐BglII restriction and linearization of the cloning vector, since the formed mixed site is unrecognized by any of the restriction endonucleases used. Finally, after the construction process has been completed, the array of gRNAs (dashed line) was transferred from the cloning vector (solid black line) to p426 gRNA expression vector (solid gray line) by restriction‐ligation process within the EcoRI and BamHI sites to avoid additional recognition by restriction endonucleases during the assembly. The NotI restriction sites are used throughout the process for on‐site control before the final verification by Sanger sequencing.

To sequentially insert the gRNA constructs into a plasmid vector, on Day 1, a cloning vector containing the bacterial replication (ori) and antibiotic resistance (AmpR for ampicillin resistance) elements (backbone from Addgene #43803) was PCR‐amplified and linearized to introduce the EcoRI restriction site at the 3′ overhang, and the BglII‐BamHI‐NotI‐XhoI restriction sites at the 5′ overhang, denoted as EcoRI//BglII‐BamHI‐NotI‐XhoI. A synthetic DNA fragment of gRNA construct was PCR‐amplified to introduce the EcoRI‐NotI‐BglII restriction sites at the 5′ overhang and the BamHI restriction site at the 3′overhang (Supporting Information: Table 3). Both PCR products were purified and overhangs of the gRNA fragment were digested with EcoRI and BamHI restriction endonucleases, whereas the overhangs of the linearized cloning vector were digested with EcoRI and BglII. Ligation within the EcoRI sites and the BamHI + BglII sites was performed and the construct transformed into E. coli cells.

On Day 2, single bacterial transformants with cloning plasmid were inoculated into liquid LBA medium.

On Day 3, the assembled cloning plasmid with gRNA construct was isolated from the bacterial culture. As a result of the BamHI‐BglII mixed site formed after the ligation, which is unrecognizable to either enzyme, BioBricks could further be sequentially assembled. The cloning plasmid with gRNA construct was linearized with EcoRI and BglII restriction endonucleases and another gRNA fragment inserted at the 5′ end of the preceding array. In this way, numerous highly repetitive gRNA constructs can be assembled. Finally, the arrays of multiple gRNAs, assembled as BioBricks, were transferred into a linearized p426 gRNA‐expression vector through EcoRI//BamHI restriction sites (Supporting Information: Table 3). This vector already contained the first gRNA construct, previously inserted by the Gibson assembly. The array was excised from the cloning plasmid with EcoRI and BamHI restriction endonucleases. The p426 gRNA‐expression plasmid (full length of the Addgene #43803) was PCR‐amplified and linearized to introduce the EcoRI restriction site at the 3′ overhang, and the BamHI restriction site at the 5′ overhang; note that this plasmid already contains another BglII restriction site. The gRNA array was ligated into the p426 gRNA‐expression plasmid that already contained the first gRNA‐coding sequence (see the Plasmid construction chapter of the Materials and methods section) and transformed into E. coli.

On Day 4, or Day (2n − 1) + 1, where n is the number of gRNA constructs, single bacterial transformants of the p426 gRNA‐expression plasmid were inoculated into liquid LBA medium.

On Day 5, or Day (2n − 1) + 2, where n is the number of gRNA constructs, the p426 gRNA‐expression plasmid was finally isolated and verified by Sanger sequencing.

The efficiency of the constructed CRISPR‐Cas9 systems was evaluated by phenotyping the resulting transformants in which 6 marker genes were targeted with the respective CRISPR‐Cas9 systems: HIS3, URA3, CAN1, LYS2, LEU2, and TRP1. To inactivate the marker genes that were targeted with the gRNAs, their protein‐coding sequences were completely deleted, or in‐frame stop codons inserted. The stop codons were designed such that they disrupted the PAM sequence. The donor DNA molecules that served as templates for HR possessed an in‐frame stop mutation within the PAM sequence and at least 45 bp long homologous‐like sequence on each side of the genome target locus. The donor DNA molecules were either PCR‐amplified and ranged from 471 bp for URA3 to 891 bp for CAN1 or, in the case of HIS3 and LEU2 genes, to delete the protein‐coding sequence entirely, generated by annealing 90 bp long double‐stranded oligonucleotides (Supporting Information: Table 1).

Individually expressing one gRNA at a time resulted in a nearly complete loss of function of the targeted marker genes, in between 94% and 100% strains, with an average of 98.6% (Figure 2a), indicating the high level of performance of the constructed CRISPR‐Cas9 systems.

Figure 2.

Evaluation results of multiplex CRISPR‐Cas9 constructs. (a) Performance of single gRNAs. Genetic modification of each gRNA with the corresponding matrix for HR (denoted as M) was verified at least twice, and each time at least 60 transformants were investigated. In the last column, the average value of all 6 different single gRNAs is shown. Note that the scale on the y‐axis starts at 0.80, and the error bars correspond to 95% confidence intervals for the population value. (b) Efficiency of multiple gRNA CRISPR‐Cas9 systems. Labeled construct compositions and donor DNAs (denoted as M) are defined in the table below the graph. To depict the scale of the study, K represents the number of biological replicates, and N the total number of the tested transformants, which for 2–6 gRNAs equals their total number. The mean value is calculated over all biological replicates and the error bars represent the corresponding 95% confidence intervals for the population value. (c) Dependence of average multiplex CRISPR‐Cas9 success rate on the average number of the recovered transformants per transformation reaction. The transformation procedure was consistently performed as described in the Methods section. Note that the number of transformants for single gRNAs was not determined due to near confluence of colonies on the plates.

To target multiple loci simultaneously, we created a single plasmid expressing multiple gRNAs, each independently from its own SNR52 promoter, so that each gRNA was expressed by RNA Pol III, as described above. We assigned a random order to the gRNA cassettes, such that the first gRNA targeted the HIS3 gene, and at the 5′ end of the previous gRNA construct, the next one was inserted in a random sequential manner: URA3, CAN1, LYS2, LEU2, and TRP1 (Figure 1b). The synthetic dsDNA Bio‐Bricks, consisting of snoRNA SNR52 promoter (SNR52p), a distinctive 20 bp crRNA part linked to the tracrRNA part of the chimeric gRNA and the SUP4 terminator (SUP4t), corresponded to each targeted gene. The design of the gRNA constructs on their expression plasmids was [SNR52p‐gRNA‐SUP4t] _n ‐CYC1t, where n ranged from 1 to 6. The array was completed by the CYC1 terminator, a remnant from the original p426 plasmid backbone (DiCarlo et al., 2013), which could in fact be functionally redundant. Throughout the study, we observed no indices of position‐dependent efficiency of gRNAs within the arrays for multiplex CRISPR‐Cas9. Namely, the observed failures to inactivate the target gene of the individual gRNAs within the array did not correlate with the position in the array (Supporting Information: Figure 1).

Finally, to test the efficiency of the generated multiplex CRISPR‐Cas9 system, gRNAs targeting marker genes HIS3, URA3, CAN1, LYS2, LEU2, and TRP1 were expressed from a single plasmid. Targeting up to four genes simultaneously resulted in a gradual decline of the system's efficiency to 89.19%, 84.07%, and 50.27%, respectively, whereas targeting of the 5th gene resulted in a drastically lower efficiency of only 1.96% (Figure 2b). Notably, URA3 and CAN1 genes are both localized on chromosome V. We found no evidence from the results for 2 and 3 gRNAs (Figure 2b) that this linkage affects the efficiency of genome editing. Generally, while the number of genomic targets increased, the number of recovered transformants, and thus, the efficiency of the multiplex CRISPR‐Cas9 system decreased significantly (Figure 2c). Because of the limitation in the number of transformants which we could analyze in our assay, a successfully sextuple‐modified transformant, if it exists, lies below the level of detection.

To thoroughly evaluate the engineered CRISPR‐Cas9 systems, several additional control experiments were conducted (Supporting Information: Table 4). First, we performed transformations with gRNA plasmids alone, and observed that the transformation efficiency was high, as expected for neutral plasmids. Second, we attempted to express multiple gRNAs together with Cas9, but not providing the linear DNA matrix for HR. The resulting transformants were all prototrophic or L‐canavanine intolerant, indicating that no perturbations of the selected marker genes were present. Notably, when in the later experiments the matrix for HR was provided for targeting a single locus, the number of transformants was very high (several hundred colonies) and their growth on the selective media plates almost confluent. In contrast, up to 200‐times less transformants were obtained when matrix DNA was not provided. Third, co‐transformation of multiple gRNA‐expressing vectors with only a subset of the possible corresponding species of matrices for HR resulted strictly in auxotrophs related to the marker genes targeted by the supplemented matrix species (Supporting Information: Table 4). This indicates that the selected gRNAs were of high specificity and that they acted independently. Decreasing the number of the possible donor DNA species co‐transformed with the corresponding multiple gRNA‐expressing vector to 0 was typically followed by a decrease in the number of transformants. Generally, the greatest reduction step was observed when one matrix species less than the number of all expressed gRNAs was present (Supporting Information: Table 4).

4. DISCUSSION

Here, we present an alternative sequential BioBrick assembly of highly similar SNR52p‐gRNA‐SUP4t constructs that commonly raise limitations in synthesis or assembly. To assemble gRNA arrays, previously reported studies have relied on Golden‐Gate assembly or construction in yeast cells by HR, but these approaches have limitations due to high number and similarity of the fragments to be assembled. We, therefore, applied a complementary sequential BioBrick assembly of highly similar gRNA constructs, which is especially practical when concatenating a large number of almost identical elements due to the absence of HR or the requirement of element‐unique PCR, or when subsequently assembled products are already useful. Commercially synthesizing such arrays remains difficult due to extensive recurring sequences.

In contrast to previous studies, an exhaustive evaluation of their efficiency for multiplex CRISPR‐Cas9 marker‐free genome editing purpose was performed. To increase the performance of genome editing, Cas9 endonuclease was expressed separately from its own plasmid, as opposed to a one‐plasmid system (DiCarlo et al., 2013; Lee et al., 2015), which allows the protein component to be synthesized in advance. Since freshly transformed cells grow slower, the recovery time of the final incubation step of the yeast transformation was prolonged compared to the commonly performed protocol (Gietz & Schiestl, 2007). Another reason for prolongation was that the repair mechanisms for DSBs caused by the Cas9 endonuclease are active in the G2 phase and it is reasoned that it takes additional time for a thorough repair.

For targeting one locus at a time, the efficiency of our CRISPR‐Cas9 systems was close to or even 100%, as has been reported for other gRNA expression constructs (Bao et al., 2018; DiCarlo et al., 2013; Lee et al., 2015; Malcı et al., 2020). There was no significant difference in the performance of each individual gRNA and its linear matrix to incorporate stop codon within the PAM sequence or to entirely delete the marker genes through HR (Figure 2a). The fact that some donor DNAs were PCR‐amplified and ranged from 471 to 891 bp, or were double‐stranded 90 bp long oligonucleotides (Supporting Information: Table 1), is therefore of negligible impact.

Overall, with our constructed systems of ribonucleoprotein complexes, we succeeded to simultaneously introduce up to four perturbations with adequate efficiency within single cells in yeast S. cerevisiae using the CRISPR‐Cas9 technique, where each gRNA was expressed independently, whereas five simultaneous modifications were achieved with a drastically lower efficiency. It has previously been reported that targeting two loci at a time has produced 76% (Lee et al., 2015) or 85% (Ryan et al., 2014), three loci 66% (Lee et al., 2015), 76.9% (Y. Zhang et al., 2019) or 80% (Ryan et al., 2014), four loci 18.2% (Y. Zhang et al., 2019) or 21% (Lee et al., 2015), and five loci 6.7% (Y. Zhang et al., 2019) of successfully modified haploid genomes when expressing each gRNA separately in some manner. Our experimental results (Figure 2b) overlap fairly well with the literature data, considering the success rates as well as the failure to produce a sextuple mutant. Since we preliminarily showed that our selected gRNAs and corresponding matrices were of high performance when functioning alone (Figure 2a), we succeeded to improve the efficiency of targeting four loci simultaneously on average by 30% compared to previously reported success rates (Lee et al., 2015; Y. Zhang et al., 2019). On the other hand, performance of targeting two or three loci was not altered significantly. It should be noted that previous studies have not evaluated multiplex gRNA arrays in such a comprehensive manner, and therefore a direct comparison of our CRISPR‐Cas9 systems with those previously described is not possible for all the parameters. Nevertheless, to partially compare our approach with different multiplexing techniques: it has been reported that targeting three different loci via yeast's own RNAse III mediated cleavage of pre‐crRNA array has resulted in 87% of successful triple mutants, but such an approach has failed at an array of spacers of size 5, and possibly already 4 (Bao et al., 2015). Targeting three loci with a tRNA‐gRNA array has been reported to be 93.0% effective, four loci 100%, five loci 88.9%, six loci 80%, seven loci 70%, and eight loci 86.7% effective (Y. Zhang et al., 2019). The superiority of the tRNA promoter compared to the SNR52 promoter, proposed due to its higher transcription rate (Y. Zhang et al., 2019), has been suggested to result in such relatively high efficiency; however, we argue this explanation is inadequate. Namely, the observed decrease in the efficiency of multiplex CRISPR‐Cas9 and the number of recovered transformants with the growing number of genomic targets (Figure 2c) could be explained in four ways: (a) the presented genome editing is a result of multiple independent events, none of which is 100% efficient; (b) the yeast cells may not provide enough protein machinery for gRNA or Cas9 expression; (c) the DNA repair mechanisms fail to resolve DSBs; and (d) a global genome catastrophic event, such as chromoanagenesis, could affect genome stability.

Regarding the option (a), we can calculate from the efficiency of independent events that multiplex CRISPR‐Cas9 system's activity is not composed of independent events only, since the product of the efficiency of its components exceeds the actual result. Specifically, experimental efficiency of targeting with two gRNAs is diminished by 5%, three gRNAs by 11%, four gRNAs by 46%, and five gRNAs by 98% compared to the theoretically calculated efficiency. Moreover, this explanation alone cannot explain the sharp decrease in the efficiency from the quadruple to the quintuple CRISPR‐Cas9 system. In addition, such high amounts of donor DNA were used that it is highly unlikely that DNA matrices were the limiting factor, especially given that up to 44 different DNA species have been shown to be efficiently introduced into target cells (Postma et al., 2021).

Regarding the option (b), the level of Cas9 or gRNA expression could at least in theory limit ribonucleoprotein complex formation and consequently the genome editing efficiency. The reported superior performance of tRNA‐gRNA array to the SNR52 promoter system has been interpreted as RNA Pol III insufficiency to transcribe all the gRNAs and further by a higher transcription rate of the tRNA promoter (Y. Zhang et al., 2019). However, too low rate of gRNA transcription or diminished Cas9 abundance would result in a limited number of DSBs. Consequently, the number of transformants harboring the plasmids for Cas9 and gRNA, and the HR matrix, would not be altered, since the number of harmful DSBs would be limited. However, while increasing the number of individually expressed gRNAs, we observed diminished number of the transformants (Figure 2c). This has also been reported elsewhere (DiCarlo et al., 2013; Jakočiūnas, Bonde, et al., 2015; Lee et al., 2015) as caused by increased toxicity of Cas9‐mediated DNA cleavage. Moreover, it has been reported that in applications with dCas9, where no DSBs are formed, up to 24 different gRNAs could be expressed simultaneously (Shaw et al., 2022), therefore the transcription rate is probably not the limiting factor in our case.

Regarding the option (c), before the DNA repair takes place, DSBs are sensed by protein kinases Tel1 and Mec1 that arrest the cell cycle and amplify the signal to establish and activate the DNA repair mechanisms (Jackson, 2002). In the presence of the CRISPR‐Cas9 system, target DSBs in yeast are repaired mainly through HR when donor DNA is provided (DiCarlo et al., 2013; Fleiss et al., 2019), whereas nontarget DSBs, which occur due to gRNA tolerance toward mismatches or partial promiscuity of Cas9 toward the PAM sequence, are repaired through NHEJ (Boutin et al., 2022). The observed phenomena of the decreased number of the transformants and the drop in the multiplexing efficiency (Figure 2c) could be explained with the failure to sense all Cas9‐mediated DSBs and resolve them mainly through HR, but also through NHEJ.

Regarding the option (d), chromosomal rearrangements via DSBs caused by CRISPR‐Cas9 systems have already been described (Boutin et al., 2022; Fleiss et al., 2019; Liu et al., 2022; McCarty et al., 2020). Moreover, DSBs within multiple loci have been reported to be excessively more harmful compared to a single DSB. This is presumably due to the triggering of extensive chromosomal rearrangements (Boutin et al., 2022; Fleiss et al., 2019), that is, translocations ranging up to the level of chromoplexy (C.‐Z. Zhang et al., 2013), and accompanying deletions (Boutin et al., 2022) ranging up to the level of chromotripsis (Boutin et al., 2022; C.‐Z. Zhang et al., 2013), or duplications ranging up to the level of chromoansynthesis (Fleiss et al., 2019; C.‐Z. Zhang et al., 2013), altogether named chromoanagenesis (C.‐Z. Zhang et al., 2013). Such single complex catastrophic event dismantles the chromosome and assembles it through error‐prone NHEJ into its new derivative (Boutin et al., 2022; C.‐Z. Zhang et al., 2013). The observed decreased performance of multiplex CRISPR‐Cas9 (Figure 2c) could be interpreted via chromoanagenesis, which arrests the cell cycle. Additionally, diminished number of detected transformants might result from the toxicity of rearrangements within chromoanagenic chromosomes.

Taken together, our results show that multiplexing more than four gRNAs independently for genome editing is limited (Figure 2c), plausibly by the DNA repair machinery failure and/or catastrophic events, that is, chromoanagenesis, that destabilize the genome. Further investigations are required to underlie the mechanistic explanation.

CRISPR‐Cas9 is a powerful genetic tool, and multiplexing it enables unprecedented applications. To demonstrate the proof of concept, only marker genes were selected in this study, yet presumably almost any part of the yeast genome could be targeted in such a way. In basic research, it could be used in functional genomics for genotype‐phenotype mapping via high‐throughput construction and dissection of natural variants (Bao et al., 2018), yet also to simultaneously test numerous predicted QTLs of variable contributions by variant allele swapping that was previously done sequentially (De Chiara et al., 2022; Pačnik et al., 2021). Multiplexing CRISPR‐Cas9 when dissecting variants would be especially advantageous for the detection of genetic interactions. In biotechnology, multiplexing CRISPR‐Cas9 could speed up genetic engineering with introductions of recombinant genes to construct new metabolic pathways for yeast cell factories (Ciurkot et al., 2019; Jakočiu̅nas, Rajkumar, et al., 2015; Lian et al., 2018; Malcı et al., 2020; Mans et al., 2015; Meng et al., 2020; Stovicek et al., 2017; Wang et al., 2018) or by rewiring cell metabolism (Adiego‐Pérez et al., 2019; Jakočiūnas, Bonde, et al., 2015; Jakočiu̅nas, Rajkumar, et al., 2015; Lian et al., 2017, 2018, 2019; Malcı et al., 2020; Meng et al., 2020; Wang et al., 2018; Wijsman et al., 2019; Y. Zhang et al., 2019) to improve the existing pathways with genome‐scale engineering by applying new genetic variants (Lian et al., 2017, 2019; Liu et al., 2022; Roy et al., 2018; Ryan et al., 2014). Our study provides additional aspects to the CRISPR‐Cas toolbox, where alternative cloning strategies could play an advantageous role. Multiplexing CRISPR‐Cas9 in the yeast S. cerevisiae for marker‐free genome editing is greatly effective, yet, as comprehensively evaluated here, it has its limitations presumably due to DNA cleavage and following repair mechanisms and/or genome instability. Therefore, applying either the hereby described approach or previously published techniques, but without the step including generation of DSBs, that is, with nicking and/or fusions of dCas9 with base editors or transcriptional effectors (Adiego‐Pérez et al., 2019; Lian et al., 2017; Liu et al., 2022), possess a considerable multiplexing potential.

AUTHOR CONTRIBUTIONS

Gašper Žun and Uroš Petrovič designed the study. Gašper Žun and Katja Doberšek performed the experiments. Gašper Žun, Katja Doberšek and Uroš Petrovič analyzed the results and wrote the manuscript.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Supporting information

Supporting information.

Žun, G. , Doberšek, K. , & Petrovič, U. (2023). Construction and evaluation of gRNA arrays for multiplex CRISPR‐Cas9. Yeast, 40, 32–41. 10.1002/yea.3833