Dramatic Improvement of CRISPR/Cas9 Editing in Candida albicans by Increased Single Guide RNA Expression

Candida albicans is an important human fungal pathogen. An understanding of fungal virulence factors has been slow because C. albicans is genetically intractable. The recent development of CRISPR/Cas in C. albicans (V. K. Vyas, M. I. Barrasa, G. R. Fink, Sci Adv 1:e1500248, 2015, https://doi.org/10.1126/sciadv.1500248) has the potential to circumvent this problem. However, as has been found in other organisms, CRISPR/Cas mutagenesis efficiency can be frustratingly variable. Here, we systematically examined parameters hypothesized to alter sgRNA intracellular levels in order to optimize CRISPR/Cas in C. albicans. Our most important conclusion is that increased sgRNA expression and maturation dramatically improve efficiency of CRISPR/Cas mutagenesis in C. albicans by ~10-fold. Thus, we anticipate that the modifications described here will further advance the application of CRISPR/Cas for genome editing in C. albicans.

clustered regularly interspaced short palindromic repeat system with CRISPR-associated protein (CRISPR/Cas) in C. albicans (1) is an important advance that promises to accelerate the pace of progress in Candida albicans biology research.
CRISPR/Cas is a system that provides adaptive immunity in bacteria (2)(3)(4). Adaptation of type II CRISPR/Cas for eukaryotic gene editing requires RNA-guided recruitment of the Cas9 nuclease to specific DNA sequences that are adjacent to an "NGG" protospacer adjacent motif (PAM) (5)(6)(7). Cas9 is recruited to the target site by a single guide RNA (sgRNA) comprised of a 20-nucleotide guide RNA (gRNA), which directs cleavage specificity by base-pairing complementarity to the target site, and an~80 nucleotide trans-activating CRISPR-targeting RNA (tracrRNA), which binds Cas9. The sgRNA/Cas9 ribonucleoprotein is recruited to the target site, where Cas9 creates a double-strand break (DSB) 3 bases upstream of the PAM site (3). Since chromosomal breaks are lethal, there is a strong selection for their repair, either by nonhomologous end joining (NHEJ) or by homology-directed repair. The introduction of a DSB can increase the rate of recombination by repair with linear homologous DNA by several thousand-fold (5,8,9). Thus, donor repair fragments containing homology to regions flanking the DSB can be designed to introduce deletions or other alterations with single-base-pair precision.
Evidence has supported the idea that intracellular levels of correctly folded, nuclear localized sgRNA limit the rate of Cas9-dependent cleavage (10,11). Up until now, C. albicans CRISPR/Cas studies have used the RNA polymerase III (Pol III) SNR52 promoter (P SNR52 ) to drive sgRNA expression (1,12). Unlike most genes transcribed by Pol III that contain internal promoters within the transcribed region, SNR52 has an upstream promoter (13). Thus, P SNR52 Pol III transcripts are initiated downstream of the promoter but are not confounded by RNA polymerase II (Pol II)-associated 5= cap and 3= poly(A) additions that are predicted to influence sgRNA specificity and nuclear retention. P SNR52 has been used successfully for sgRNA expression in both Saccharomyces cerevisiae and C. albicans (1,14), but its relative efficiency in C. albicans has not been compared to those of other sgRNA delivery schemes.
Our attempts to use the P SNR52 CRISPR system described by Vyas et al. (1) did not lead to high-efficiency mutagenesis of several C. albicans target genes, despite the use of different gRNAs that target different sites within each gene. Therefore, to improve CRISPR mutagenesis, we sought to optimize sgRNA expression and CRISPR/Cas efficiency in C. albicans by comparing different promoters and RNA processing mechanisms. Our results demonstrate that gRNA expression can indeed be rate limiting for efficient DSB cleavage in C. albicans, and we describe new modifications that can increase the efficiency of gene editing in C. albicans by 10-fold. While our major goal was to optimize gRNA expression, our experiments also revealed that in addition to homology-directed repair, under certain conditions, Cas9-dependent DSB can be repaired by nonhomologous end joining at low frequency.

RESULTS AND DISCUSSION
Assay system. To examine gRNA-dependent Cas9 nuclease activity, the C. albicans gene encoding codon-optimized enhanced monomeric red fluorescent protein (yEmRFP [henceforth referred to as RFP]) was targeted (15). C. albicans strains expressing RFP display a distinct pink colony color phenotype as well as bright red fluorescence (16). Inactivation of RFP by mutation results in a reversal of the pink colony color, red fluorescence phenotype, so colonies derived from rfp mutants could be easily distinguished by their white color and absence of fluorescence (Fig. 1A).
We constructed RFP hemizygous strains in which HIS1-marked RFP was integrated at single copy at the RPS1 locus (Fig. 1B). As a control for Cas9-dependent rfp mutations, we constructed RFP strains that do or do not also express CAS9. C. albicans uses a noncanonical genetic code that reads the leucine CTG codon as serine. Thus, a synthetic Candida-optimized CAS9 gene (CaCAS9) was designed to alter each CTG serine codon to an AT-rich leucine codon. In addition, the codon bias of the entire gene was maximized toward AT-rich codons (see Materials and Methods). These RFP/CaCAS9 strains were transformed with URA3-marked sgRNA plasmids (described below) that targeted RFP cleavage proximal to the PAM site at position ϩ132 of the RFP open reading frame (ORF) (Fig. 1B). The efficiency of Cas9/sgRNA-dependent cleavage was measured by determining the percentage of white, nonfluorescent rfp mutant colonies in the population of transformants.
These transformations were performed in the presence or absence of a donor repair fragment designed to delete 370 bp of the RFP ORF after homology-directed repair. This repair fragment was made using overlap extension PCR and contained 285-bp arms of upstream and 308-bp downstream sequence homology to regions flanking the DSB (see Materials and Methods) (Fig. 2). To confirm that RFP in white colonies was correctly deleted by recombination with the repair fragment, RFP was amplified by PCR, analyzed by gel electrophoresis for the presence of the deleted allele, and also and are fluorescent and can be easily distinguished from colonies that arise through CRISPR-mediated deletion of RFP, which are white and nonfluorescent. Panel B depicts the strategy for quantitating Cas9and sgRNA-dependent cleavage of RFP. RFP (EPC1) or RFP CaCAS9 (EPC2) hemizygous strains were constructed as described in Materials and Methods. These strains were transformed with or without a donor healing fragment and a series of URA3-marked plasmids that differ in expression of an RFP sgRNA (Fig. 3). The gRNA targets a 20-bp DNA sequence proximal to the RFP PAM site at position 132. The number of white and red colonies in each transformation was counted, and cleavage efficiency was calculated as the percentage of white colonies in the population.
Optimizing CRISPR/Cas in Candida albicans sequenced (see below). This screen allowed a simple and accurate way to measure the efficiency of Cas9-dependent cleavage and repair of RFP. sgRNA delivery schemes. To optimize sgRNA expression, several parameters were varied, including the promoters used to drive transcription, as well as removable flanking sequences designed to be posttranscriptionally cleaved to produce transcripts with precise 5= and 3= ends. A schematic diagram of the different constructs and the predicted RNA transcripts is shown in Fig. 3, and their DNA sequences are listed in Fig. S1 in the supplemental material. Each construct had the same sgRNA sequence but differed in the promoters used to drive sgRNA transcription and/or flanking sequences that encode ribozymes or tRNAs that are posttranscriptionally cleaved.
The first construct used the Pol III P SNR52 to drive sgRNA expression (P SNR52 -sgRNA). The C. albicans P SNR52 promoter sequence and sgRNA junction were the same as those used in previous studies (Fig. 3A) (1). P SNR52 Pol III transcripts are initiated at a purine, G or A, and are terminated by a poly(U) stretch (six in yeast) (13).
The second delivery scheme used the strong Pol II ADH1 promoter (P ADH1 ) to drive sgRNA transcription (Fig. 3B). Because P ADH1 is a strong constitutive promoter, we reasoned that P ADH1 may increase the levels of sgRNA. To prevent potential interference of sgRNA specificity by the 5= cap and 3= poly(A) tail associated with Pol II transcripts, self-cleaving hammerhead (HH) and hepatitis delta virus (HDV) ribozymes were added to the 5= and 3= ends of sgRNA. The presence of these ribozymes at the sgRNA ends may also protect it from nucleases and therefore further increase its intracellular levels.
In the design of this cassette, the first six nucleotides of the HH ribozyme are comple- The donor repair fragment used for repair of the Cas9-induced DSB was constructed by overlap PCR. Two fragments were amplified by PCR, one of which contains a 3= end that is homologous to the 5= end of the other. After the two fragments are mixed, these homologous regions anneal to one another, and then a round of DNA replication produces a single fragment with arms of homology to the 5= and 3= regions of RFP, but which is deleted for 370 bp of the ORF, including the PAM site at position 132. Homologous recombination of this rfp⌬33-403 fragment with chromosomal RFP can easily be detected by PCR amplification (Fig. 4).
mentary to the first six nucleotides of the gRNA, thus allowing the formation of the stem structure required for HH self-cleavage at the gRNA-HH junction (17). The sites of the HH and HDV predicted cleavage sites are indicated by the arrowheads in Fig. 3B.
To compare the efficiency of sgRNA maturation, the third delivery scheme used the same strong P ADH1 to transcribe sgRNA, but its 5= end was flanked with C. albicans tRNA Ala instead of the HH ribozyme (Fig. 3C)  ) ribozymes. HH-sgRNA-HDV transcription is driven by the strong ADH1 promoter. The stem structure formed between the 5= HH and 5= gRNA forms the structural motif recognized for autocleavage (depicted by arrowhead). The 3= cleavage relies on the autonomous pseudoknot structure of HDV (depicted by arrowhead). (C) P ADH -tRNA(ϪHDV). This construct is exactly like that shown in panel B, but the hammerhead sequence is replaced by the 75-bp C. albicans alanine tRNA gene. In this scheme, the mature 5= end of the sgRNA is generated by endogenous RNase Z, which recognizes the stem structure formed by tRNA 5= and 3= sequences. 3= cleavage relies on the autonomous stem-loop structure of HDV (depicted by arrowhead). Transcription of the tRNA-sgRNA-HDV is driven by the strong ADH1 promoter. (D) P tRNA (ϪHDV). This construct is exactly like that of panel C, but the P ADH1 sequence is deleted. In this cassette, RNA Pol III transcription is driven by the internal A and B box elements of the tDNA promoter, and the mature 5= end of the sgRNA is generated by endogenous RNase Z, while 3= cleavage relies on the autonomous pseudoknot of HDV (depicted by arrowhead).
Optimizing CRISPR/Cas in Candida albicans endogenous RNase P and Z mediate the robust posttranscriptional cleavage of 5= and 3= ends of pre-tRNAs (18,19). These endogenous RNase activities may be more efficient than HH autocatalysis and therefore lead to improved sgRNA maturation. Second, cleavage by RNase Z at the tRNA-sgRNA junction requires recognition of sequences that are only within the tRNA. Therefore, this architecture, if efficient, would provide a more portable platform for swapping out the 20-bp gRNA coding sequence with any another gRNA. This is not possible using HH ribozyme, whose unique 5= sequence is constrained by each different gRNA. The sequence of the sgRNA-tRNA junction was designed to yield the precise sgRNA 5= end after cleavage with RNase Z (11,20). The sites of the tRNA and HDV predicted cleavage sites are indicated by the arrowheads in Fig. 3C. As tRNAs are very abundant, we considered the possibility that the tRNA promoter itself may be efficient in directing sgRNA expression (11,20). To test this, the fourth delivery scheme used the tRNA Ala fused to sgRNA-HDV (Fig. 3D). Cleavage of tRNA at its 3= trailer by endogenous RNase Z (see arrows for cleavage site) allows maturation of sgRNA with a precise 5= end. The sites of the tRNA and HDV predicted cleavage sites are indicated by the arrowheads in Fig. 3D.
Measuring RFP mutation as a function of sgRNA delivery. To compare their relative activities, each of the URA3-marked CIp gRNA plasmids described above, or a negative (Ϫ) gRNA vector alone (ϪgRNA) was transformed side by side into RFP strains that did (EPC2) or did not (EPC1) express Cas9 (see Materials and Methods). Each transformation was performed in the presence or absence of a repair fragment (see Materials and Methods). After 2 days, colonies from each transformation were counted by fluorescence microscopy to determine the number of white and red colonies for each transformation. Unexpectedly, in addition to pure red and white colonies, under certain conditions, we observed red colonies with white sectors (Fig. 4). White or white-sectored colonies were seen at a frequency of less than 10 Ϫ4 in control experiments, in which the isogenic Cas9 Ϫ EPC1 strain was transformed with gRNA, vector, and/or the donor repair fragment (data not shown) ( Table 1). White or white-sectored colonies were also not observed in Cas9 ϩ EPC2 cells transformed with vector lacking gRNA with or without the donor repair fragment. These control experiments demonstrated that the high frequency of white or white-sectored colonies was not due to spontaneous looping out of RFP, which is flanked by RPS1 repeats, or by replacement of RFP with the CIp-gRNA plasmid during integration at RPS1. Instead, white and white-sectored colonies were dependent on the expression of both Cas9 and sgRNA ( Fig. 5C; Table 1). Each white sector represented a Cas9-mediated RFP cleavage event that occurred in a cell at the vertex of the white sector. The sizes and numbers of white sectors ranged from those that were 1/2 the colony, in which RFP cleavage likely occurred in one daughter cell of the first division, to those that appear mostly white but had a remaining sliver of red cells in a colony ( Fig. 4A and B). The rfp mutation that gave rise to white sectors was stable, since restreaking a sectored colony on -Ura plates resulted in colonies that were almost completely white (not shown). This sectoring phenotype could explain why PCR analysis of RFP amplified from genomic DNA from what appeared to be visually "white/healed" colonies occasionally yielded both fulllength and deleted RFP products (Fig. 4C).
Thus, to more accurately quantitate the relative efficiency of sgRNA delivery, we counted the number of white, red, and sectored colonies produced after transformation with each of the different plasmids, in the presence or absence of healing fragment. These results, shown graphically in Fig. 5A and B, demonstrated a clear hierarchy of cleavage efficiency as measured by the percentage of pure white rfp⌬ mutant colonies in the population, in the following order: P ADH1 -tRNA (~94% white) Ͼ P ADH -HH (~22% white) Ͼ P SNR52 (9% white) ϭ P tRNA (~8% white). When sgRNA transcription was driven by the Pol II ADH1 promoter but flanked by the 5= hammerhead ribozyme instead of the tRNA, the Cas9-directed mutation frequency was reduced bỹ 3.5-fold. When sgRNA transcription was driven by the Pol III SNR52 or tRNA promoters, the mutation frequency was reduced by~10-fold. While the experiments shown in Fig. 5 were performed using CIp-based integration plasmids, the same trend was observed using CaARS-containing plasmids (data not shown). Therefore, we inferred that the high efficiency observed when sgRNA was delivered using the P ADH1 -tRNA plasmid is due to the combination of (i) increased Pol II P ADH1 promoter strength,    Fig. 3. Note the near absence of sectored colonies when sgRNA is efficiently delivered (P ADH -tRNA-gRFP-HDV) compared to when poorly delivered (P tRNA or P SNR52 ).

A P A D H 1 -H H -g R F P -H D V ( -) C A S 9 t R N A A g R F P -H D V P A D H 1 -t R N A -g R F P -H D V P
compared to the Pol III P SNR52 or P tRNA , and (ii) the presence of the tRNA flanking sequence. These results support the idea that the presence of the tRNA may promote sgRNA nuclear retention or stabilization and in addition is processed by endogenous RNase Z more efficiently than HH autocatalysis. An additional factor that may explain the reduced efficiency of transcripts driven by P SNR52 is that the sgRNA is constrained by a 5= purine, which is the favored nucleotide for transcription initiation by P SNR52 . A non-base-paired 5= G, which is built into the cloning site of P SNR52 CRISPR vectors, may result in 5= overhang in the sgRNA-target DNA complex that could impact sgRNA efficiency, although this has not been formally tested. Thus, of the sgRNA delivery schemes we tested, the P ADH1 -tRNA-driven sgRNA expression was the most efficient, by almost 10-fold. Quantitation of sectored colonies demonstrated a strong correlation between the abundance of sectored colonies and the inefficiency of the sgRNA delivery. In a population of P ADH1 -tRNA transformants, over 90% of colonies were white, while less than 1% were sectored. Conversely, in a population of P tRNA transformants, less than 10% of colonies were white, while more than 50% were sectored (Fig. 5C). These results are shown quantitatively in Fig. 5. One interpretation of these results is that efficient Cas9-dependent cleavage requires a threshold level of sgRNA that is not met in founder transformants when sgRNA is poorly expressed. After one or more cell divisions, daughter cells may eventually accumulate sufficient levels of sgRNA to allow Cas9 DSB and repair.
RFP DSB repair. Surprisingly, we observed that the relative numbers of white, red, and sectored colonies for each of the different sgRNA plasmids were similar whether or not a donor repair fragment was included in the transformation (compare Fig. 5A and B). These results raised the question of how breaks in RFP were repaired in the absence of homologous recombination with the repair fragment. The strain used for these experiments is hemizygous for RFP. Since this sequence is not repeated anywhere else in the genome, this DSB cannot be repaired by gene conversion. However, RFP is flanked by a duplication of RPS1 loci (see Materials and Methods) ( Fig. 1 and Fig. 6A), and there are additional duplicated RPS1 loci that were introduced by integration of the gRNA plasmid, as well as CAS9. Thus, Cas9-mediated DSB in RFP could stimulate its deletion by a variety of homology-directed pathways that involve recombination between adjacent RPS1 loci (i.e., by flip-out or single-strand annealing as depicted in Fig. 6A) (21) or even RPS1 on the homologous chromosome (Fig. 6A). These types of homology-directed repair pathways would lead to deletion of RFP sequence, which includes RFP and HIS1. An alternative means of DSB repair is nonhomologous end joining (NHEJ), in which the breaks are directly ligated, although this is usually accompanied by short insertions or deletions (Fig. 6A) (22,23).
To investigate how these DSB were repaired in the absence of donor healing fragment, RFP from these white colonies (~30) was amplified by PCR, analyzed by gel electrophoresis, and subjected to DNA sequencing. If DSBs at RFP were repaired by homologous recombination between adjacent RPS1 loci, then both the RFP and HIS1 genes should be completely deleted. To test this, these colonies were also replica plated on ϪHis plates to determine the presence or absence of HIS1. The result of this experiment demonstrated that by PCR analysis, 70% of these white colonies lacked the RFP gene entirely and were also histidine auxotrophs, suggesting that in these colonies, RFP DSB repair occurred via a homology-directed repair pathway (Fig. 6A).
The remaining 30% of these white colonies contained RFP of a similar electrophoretic mobility as the full-length gene. As exemplified in Fig. 6B, the alignment of wild-type RFP with one such mutant showed the presence of a 21-bp deletion that removes the PAM site and sgRNA targeting sequence (Fig. 6B [GGTGAAGGTAGACCAT ATGAA represents the gRNA sequence underlined in red]). The mutant DNA sequence (shown in Fig. 6C and Fig. S2 in the supplemental material) indicates that the Cas9 cut site occurred 3 bases from the PAM (AGG at position 132, depicted by blue asterisks in Fig. 6B). This result suggested that repair occurred via nonhomologous end joining, resulting in a short deletion. This was a surprising result because several studies have demonstrated that, in C. albicans as in S. cerevisiae, homologous recombination is a vastly preferred mechanism for repair of DSB since mutations that knock out homologous recombination affect the sensitivity of strains to agents that produce DSB, while mutations in NHEJ do not (24,25 RFP. (B) Alignment of the wild-type RFP sequence with that of RFP amplified from a white colony isolated after CRISPR mutagenesis in the absence of a donor repair fragment. The sequence targeted by the RFP gRNA is underlined in red and precedes the PAM site (AGG, denoted with blue asterisks) at n ϭ 132. Note that the deletion begins 3 bases from the PAM, precisely at the predicted Cas9 cleavage site.
Extension of system to endogenous genes. The results described above demonstrated a system for efficient targeting and repair of RFP. However, RFP was present in the chromosome as single copy and flanked by direct repeats, which may affect its mutability. Therefore, to confirm the utility of the P ADH1 -tRNA sgRNA delivery scheme for deletion of endogenous diploid genes, we constructed URA3-or recyclable ura3-dpl200 (26)-marked P ADH -tRNA vectors (pND494 and pND501) to allow optimized expression of any gRNA. These plasmids allow ligation of short (23-bp) annealed oligonucleotides into a cloning cassette containing two nonpalindromic SapI sites at the tRNA-gRNA junction (see Materials and Methods and Fig. S1). This cassette was designed to maintain the RNase Z cleavage recognition site at the tRNA-gRNA junction as well as the seamless fusion of gRNA to tracrRNA after ligation of annealed gRNA oligonucleotides (Fig. S1).
As proof of principle, we targeted LEU2 with a gRNA complementary to the region proximal to the PAM at position ϩ123 of the LEU2 ORF (Fig. 7). A Cas9 ϩ strain (HNY30) was transformed with a LEU2-specific gRNA plasmid marked with URA3 or vector lacking gLEU2 (Table 2). These transformations were performed in the absence or presence of a donor repair fragment designed to replace a 434-bp fragment within the LEU2 ORF with a unique EcoRI site (see Materials and Methods) (Fig. 7A). This repair fragment contained 47-bp arms of homology to regions flanking the break. As noted by Vyas et al. (1), we found that repair fragments with arms of longer homology (80, 250, or 500 bp) did not improve the efficiency of mutagenesis (not shown). Controls included the same transformations but of a Cas9 Ϫ isogenic strain (BWP17). To quantify mutations in LEU2, Ura ϩ transformants were patched on plates containing synthetic dropout medium without uracil [SD(ϪUra)] and then replica plated on SD(ϪUra) and SD(ϪLeu) to identify leucine auxotrophs ( Table 2). LEU2 from both Ura ϩ and Leu Ϫ transformants was genotyped by PCR ( Fig. 7B; Table 2).
The results of these experiments revealed some important differences between LEU2 and RFP mutagenesis. First, in contrast to what was observed with RFP, the number of Ura ϩ gLEU2 transformants that were obtained in the absence of healing fragment was~25-fold lower than in its presence. This reduction in transformation was dependent on the presence of both Cas9 and LEU2-specific gRNA, suggesting it was caused by DSBs (Table 2). Second, in the absence of healing fragment, none of the viable Ura ϩ transformants were leucine auxotrophs (Table 2). Thus, a plausible explanation for this reduction is that donor repair fragment is required to repair breaks in LEU2, and that unrepaired breaks are lethal. This implies that other non-homologydirected repair pathways, including NHEJ, occur very rarely. Third, in the presence of healing fragment,~80% of Ura ϩ transformants were Leu Ϫ . PCR analyses of these Leu Ϫ mutants suggest that all of them arose through homologous recombination with the donor repair fragment (Table 2; Fig. 7B). Finally, from random PCR analysis of Ura3 ϩ transformants, we found no evidence for heterozygosity, since PCR amplification products from single colonies yielded either wild-type or truncated (EcoRI-sensitive) LEU2, but not both ( Table 2; data not shown). The lack of heterozygosity was not specific for LEU2 since similar results were obtained for other nonessential endogenous genes that were targeted for mutagenesis (data not shown), including ANP1, VAN1, and BMT1, whose further characterization will be described elsewhere.
These experiments demonstrate the applicability of this expression system for creating DSBs in endogenous diploid genes with high efficiency. This system has several advantages compared to others that have been described (1,27). First, the higher mutation frequency, through use of the ADH1 promoter and tRNA instead of the SNR52 Pol III promoter, means that fewer yeast colonies require screening in order to identify a desired mutation. Second, while the transient system described by Min et al. (12) is very simple, we found that even under optimal gRNA expression, the mutation frequency is too low for mutagenesis without selection of a marker-linked donor repair fragment. The expression system described in the present study facilitates "scarless" mutations, although it requires selection of marker-linked gRNA. gRNA can be linked to URA3 as described here or nourseothricin resistance as described by Vyas et al. (1). The advantage of uracil selection is that it is faster and cheaper than Nat r . Also, certain mutations-for instance those that strongly interfere with cell wall biosynthesisdisplay altered nourseothricin sensitivities (our unpublished observation), which can complicate their isolation using Nat r selection. On the other hand, recycling ura-dpl200 requires counterselection with 5-fluoroorotic acid (FOA), which may bias toward selection of aneuploidy (28). A basic protocol for the day-to-day timeline for a single-gene knockout using the vectors described in this study is outlined in Fig. S3 in the supplemental material.
Concluding remarks. Genetic analysis of C. albicans has been complicated because it is a diploid that does not readily undergo sexual reproduction. Without CRISPR, genetic modifications, including knockouts, must be applied to both chromosomes, requiring sequential modification of each locus. The application of the CRISPR/Cas system as described by Vyas et al.
(1) has enormous potential because it can circumvent these problems. However, as has been found in other systems, the efficiency of CRISPR/Cas can be frustratingly variable. Studies from other systems suggest that several factors influence efficacy of CRISPR/Cas, including (i) location and accessibility of gRNA target site, (ii) gRNA sequence, and (iii) sgRNA intracellular levels. Here, we systematically examined parameters hypothesized to alter sgRNA intracellular levels in order to optimize CRISPR/Cas in C. albicans. Our most important conclusion is that increased sgRNA expression and maturation dramatically improve efficiency of CRISPR/ Cas mutagenesis in C. albicans. Large-scale analyses of the sgRNA target site effects, chromatin structure, and gRNA sequence preferences have led to an increasing knowledge base, as well as online tools that help design gRNAs. Features of a "good" gRNA include guanines at the Ϫ1 and Ϫ2 positions (i.e., a 3= GG) and cytosine at the Ϫ3 DNA cleavage site and at ϩ1 relative to the N 0 G 0 G 0 PAM site (29)(30)(31)(32). Features of a good target position appear to be nucleosome-free locations upstream of transcriptional start sites (29-31, 33, 34). It is notable that neither the gRFP nor its target location used in the present study conforms to any of these predictive guides, yet when optimized for expression, it nevertheless resulted in an almost 100% mutation frequency. These results suggest that in C. albicans, sgRNA levels may in part compensate for a less than optimal gRNA design. Thus, we anticipate that the modifications described here will further advance the application of CRISPR/Cas for genome editing in C. albicans.

MATERIALS AND METHODS
Plasmid construction. The plasmids and their relevant features are listed in Table 3, and the sequence of each relevant cassette and the oligonucleotides used for the construction are listed in Table S1 in the supplemental material. All DNA generated by PCR was verified by DNA sequence analysis.
YPB-P ADH1 tA gRFP HDV (pND468) contains the 75-bp C. albicans tRNA Ala gene between the ADH1 promoter and the sgRNA in YPB1-Adhp. The tRNA gene was amplified by PCR from genomic DNA. Gibson assembly (38) was used to assemble the tRNA-sgRNA-HDV fragments and the YPB-ADHp vector.  The number of Ura ϩ transformants that were leucine auxotrophs. c The percentage of LEU2/leu2⌬ heterozygotes was based on PCR analysis of LEU2 using genomic DNA isolated from 40 random Ura ϩ /Leu ϩ transformants. d The percentage of leu2⌬/leu2⌬ homozygotes was based on PCR analysis of LEU2 amplified from genomic DNA isolated from 60 random Ura ϩ /Leu Ϫ transformants.
YPB-P tRNA gRFP HDV (pND479) was generated by deleting the ADH1 promoter of YPB-P ADH1 tA gRFP HDV (pND468) by digestion with NotI and BglII, filling in overhangs with Klenow DNA polymerase and ligating.
CIp10-P ADH1 tA SapI HDV (pND494) is a vector that allows cloning and expression of any gRNA such that the sgRNA transcript is flanked with 5= tRNA and 3= HDV and transcribed by P ADH1 . It was constructed by using site-directed mutagenesis to replace the RFP gRNA segment in CIp10-P ADH1 tA gRFP HDV with a cassette containing two SapI sites and 22 bp of intervening sequence, including a ClaI site (Fig. S2). In addition, site-directed mutagenesis replaced the unique SapI site at position 4381 of CIp10-P ADH1 tA gRFP HDV with an NsiI site (Fig. S2). It should be noted that while constructing this plasmid, we discovered that the CIp10 sequence (GenBank accession no. AF181970) between KpnI and SacI (containing pBluescript KSϩ sequence) was incorrectly annotated and actually flipped, which places the T7 promoter adjacent to RPS1 and the T3 promoter adjacent to CaURA3. To allow sequential introduction of additional gRNAs into the same strain, we also constructed a plasmid that is identical to pND494 but contains the recyclable ura3-dpl200 allele (26). This URA3 is flanked by 200-bp repeats that facilitate its recycling by FOA selection.
The C. albicans strains used in this study are listed in Table 4 and were derived from BWP17. EPC1 contains a single integrated copy of the RFP gene (16) and was constructed by targeting StuI-linearized CIp-HIS-P ADH -RFP to RPS1. This integration results in a duplication of RPS1 flanking P ADH1 -RFP, HIS1 and the intervening plasmid sequence. Single integration of P ADH1 -RFP at RPS1 was confirmed by Southern blotting (not shown). EPC2, which expresses both RFP and CaCAS9, was constructed by targeting CIp-dpl-P ADH1 tA SapI HDV pND494 but contains ura3-dpl200 instead of URA3 This study  ), and resuspended in 500 l 100 mM LiOAc. The concentration of plasmid DNA was titrated to yield~200 colonies per plate. Typically, transformations included 50 l of the cell suspension (2.5 OD 600 units),~2 to 4 g of linearized sgRNA plasmid (1 pmol), 1 to 2 g of RFP donor repair DNA fragment (15 pmol), or annealed, filled-in repair oligonucleotides (200 pmol). After transformation, plates were incubated at 30°C for 2 days before being viewed by fluorescence microscopy with a Zeiss microscope equipped with low-power magnification (1.5 to ϫ10). Visual detection of pink (RFP) and white (rfp) colonies required 3 days of incubation. The number of red, white, and sectored colonies per plate was counted to determine the efficiency of Cas9-mediated cleavage. Cleavage efficiency was calculated as the no. of white colonies/total no. of colonies per plate.
PCR analysis of Cas9-mediated cleavage. For analyses of RFP mutagenesis, white colonies were patched and replica plated on selective media to test for prototrophy of selective markers (URA3, HIS1, and ARG4). Genomic DNA from 20 white colonies per experiment was prepared and used as the template for PCR amplification using primers specific for RFP (Fig. S1). Colonies were inoculated into 0.5 ml medium and incubated at 30°C with shaking for 2 to 3 h. After harvesting of cells, pellets were suspended in 30 l 0.2% SDS, heated for 3 min at 95°C, and centrifuged for 1 min at 14,000 ϫ g. Three microliters of the supernatant was used as genomic DNA template in a standard 25-l PCR.
For the analyses of LEU2 deletions, after transformation with various sgLEU2 gRNA plasmids or control vectors, and with or without donor repair fragments, colonies were patched onto to SD(ϪUra) and replica plated onto SD(ϪUra) and SD(ϪLeu) plates. PCR amplification of Leu2 ϩ and Leu2 Ϫ colonies was performed, followed by digestion with EcoRI to determine the percentage of colonies that were heterozygous or homozygous for the leu2⌬ allele.
Construction of donor healing fragments. The rfp⌬ repair fragment targeted deletion of nucleotides 55 to 402 of the RFP ORF, including the PAM site located at ϩ132. It was generated by fusion PCR (40). PCR was used to amplify two fragments: one homologous to the 5= region of RFP and the other homologous to the 3= region. The 5= fragment also contained a 3= 20-bp tail that was homologous to the 5= end of the second fragment ( Fig. 2A). These two fragments were mixed, annealed, and then extended. After extension, the full-length "fused" fragment was amplified by PCR. The resulting 593-bp fragment contains a 285-bp arm of upstream sequence homology and a 308-bp arm of downstream sequence homology to regions flanking the DSB. Approximately 1 to 2 g of this DNA (~5 pmol) was used for transformation of yeast. Recombination with RFP results in deletion of an internal 370 bp within the RFP ORF, including the PAM site to produce the rfpΔ33-403 allele.
The LEU2 donor repair fragment was made by annealing and filling in two 60-mer nucleotides that contained 20 bases of sequence complementarity at their 3= ends. This complementary region included an EcoRI recognition sequence (highlighted in blue in Fig. 7A). The resulting fragment contained 47 bp of homology to sequences flanking the DSB. Three microliters of each oligonucleotide (300 pmol) was annealed and then extended in a 25-l reaction mixture containing 0.2 mM deoxynucleoside triphosphate (dNTP), buffer, and Taq DNA polymerase (Denville Scientific, Inc., United States) and subjected to 25 to 30 cycles of PCR with an extension time of 30 s. Twenty microliters of this reaction mixture was used per yeast transformation as a repair donor fragment (~240 pmol of repair fragment). Recombination with LEU2 results in replacement of an internal 434-bp fragment within the LEU2 ORF, including the PAM site, with an EcoRI site to produce the leu2Δ71-505 allele.