CRISPR-Cas9 ribonucleoprotein-mediated co-editing and counterselection in the rice blast fungus

The rice blast fungus Magnaporthe oryzae is the most serious pathogen of cultivated rice and a significant threat to global food security. To accelerate targeted mutation and specific genome editing in this species, we have developed a rapid plasmid-free CRISPR-Cas9-based genome editing method. We show that stable expression of Cas9 is highly toxic to M. oryzae. However efficient gene editing can be achieved by transient introduction of purified Cas9 pre-complexed to RNA guides to form ribonucleoproteins (RNPs). When used in combination with oligonucleotide or PCR-generated donor DNAs, generation of strains with specific base pair edits, in-locus gene replacements, or multiple gene edits, is very rapid and straightforward. We demonstrate a co-editing strategy for the creation of single nucleotide changes at specific loci. Additionally, we report a novel counterselection strategy which allows creation of precisely edited fungal strains that contain no foreign DNA and are completely isogenic to the wild type. Together, these developments represent a scalable improvement in the precision and speed of genetic manipulation in M. oryzae and are likely to be broadly applicable to other fungal species.

In recent years, the use of the clustered regularly interspaced short palindromic repeats (CRISPR)-associated RNA-guided Cas9 endonuclease, has facilitated genome editing technologies and become the leading tool used to generate specific changes to DNA sequences in a wide range of species 1 . To generate double stranded breaks (DSBs) in the genome of a target organism, the CRISPR-Cas9 system requires Cas9 endonuclease, which cleaves target DNA at a genomic target sequence [2][3][4] , and a single RNA molecule, which in the CRISPR-Cas9 system, uses a linker sequence to join the nuclease-binding tracrRNA and the target specific crRNA molecules found in naturally occurring complexes in the source organism, Streptococcus pyogenes 5 . The sgRNA associates with the nuclease and directs it to a genomic target sequence 5 . The DSB created by the nuclease can then be repaired by non-homologous DNA-end joining (NHEJ) or using homologous recombination (HR), by introduction of donor DNA homologous to the sequence around the break, which allows very specific edits to the DNA sequence, or very precise insertions or deletions 6 . The only target sequence requirement necessary for CRISPR-Cas9 genome editing is the presence of the protospacer adjacent motif (PAM), a triplet NGG located immediately 3′ of the genomic target sequence 7,8 . Because HR-based repair can be used to introduce modifications at some distance to the DSB, for example up to 30 bp in human stem cells 9 , the majority of fungal genomes are accessible to manipulations using CRISPR-Cas9 editing.
CRISPR-Cas9 genome editing offers huge potential to accelerate the pace of research in key fungal research areas, such as biotechnology, medical mycology and plant pathology, by dramatically reducing the time required to undertake common objectives, such as targeted gene deletion, overexpression, single nucleotide changes, or tagging the products of genes of interest with fluorescent proteins 10 . The technique also permits targeting of gene families, making multiple mutations 11 , and generating mutants in dikaryotic, or polyploid fungi 12 . Moreover, the potential exists to carry out 'selectable marker-free' manipulations for precise genetic changes, a prerequisite for

Results
Evidence of toxicity of Cas9 in Magnaporthe oryzae. We made multiple attempts to generate a strain stably expressing a Cas9 gene, together with small guide RNAs (sgRNA) targeting the melanin biosynthetic polyketide synthase-encoding genes ALB1 (MGG_07219) or RSY1 (MGG_05059), mutations which give rise to easily identifiable white (albino) or orange-red (rosy) colour phenotypes, respectively, to fungal colonies 16 . Although Arazoe and co-workers reported successful generation of rsy1 mutants in M. oryzae using CRISPR-Cas9, it is likely that their constructs (and hence Cas9) were only transiently present in the cell (although this was not tested in their work) 14 . Despite many attempts, and using three different versions of Cas9 under control of two different promoters (see methods for details), we were never able to generate mutants showing altered pigmentation, among the few transformants which resulted from transformation with either vector. Importantly, we were not able to reproduce the generation of mutants reported previously, even when the same vectors were used 14 . We did, however, observe that the transformation of all constructs containing Cas9-encoding sequences, either with or without a sgRNA present, always gave rise to far fewer transformants than empty vector controls ( Fig. 1 and Table S1.). Based on our observations and on the previous report 14 , it would seem that stable expression of Cas9 is very toxic to M. oryzae, but that genome editing using Cas9 might be possible by transient expression of the nuclease. CRISPR-Cas9 genome editing using purified Cas9 and sgRNAs. In view of the likely toxicity of Cas9 nuclease to M. oryzae cells, we set out to assess whether CRISPR-Cas9 genome editing might instead be accomplished by transient introduction of purified Cas9 protein along with in vitro synthesised sgRNAs, into protoplasts of M. oryzae. To this end, we purchased nuclear-localised Cas9 (Cas9-NLS) from commercial suppliers (see Methods for details) and complexed this to sgRNAs capable of directing Cas9 to the ALB1 locus (Fig. 2a). We independently tested a second RNP that targets the RSY1 locus. We introduced these Cas9-NLS-sgRNA ribonucleoprotein complexes (RNPs) independently into a wild type M. oryzae strain Guy11, together with donor DNAs which would introduce an insertion containing a selectable marker (HPH -the hygromycin phosphotransferase gene cassette) near the 5′ end of the coding regions of these genes, by repair of the DSB by homologous Figure 1. Toxicity of stably expressed Cas9. Binary vectors containing the gene encoding Cas9-NLS under the control of the TrpC promoter and terminator were introduced into Guy 11 using Agrobacterium-mediated transformation. Transformant numbers were assessed after 7 days on selective medium (transformants were subsequently sub-cultured for assessment of pigmentation after growth on CM). recombination with donor DNAs containing homologous regions on either side of the selectable marker gene (Fig. 2b). In both cases the selectable marker gene was expected to integrate close to the DSB, normally 3-4 bp from the PAM site. We were able to demonstrate very efficient targeting of both genes, as shown in Fig. 2c,d, and Table S2. Remarkably, mutation of RSY1 was near to 100% efficient in multiple experiments (see for example Fig. 2d) and the efficiency of targeting ALB1 was also greater than 50% in every test, with typically 70-80% albino transformants generated (Table S2). As expected, the rates of mutation using donor-only controls were more typical of rates reported for gene deletion using conventional gene deletion strategies (see Table S2). These observations provide evidence that RNP-CRISPR-Cas9 generated DSBs strongly induce HR repair and can be exploited for efficient gene manipulation in M. oryzae.
Highly efficient mutation of genes using CRISPR-Cas9 induced recombination of micro-homologous donor DNAs. The efficient mutation of ALB1 by CRISPR-Cas9 induced homologous recombination of a donor DNA, prompted us to define the minimum length of homologous DNA that would facilitate efficient genome editing. Recent reports in some fungi suggest that micro-homologous regions are sufficient to allow repair by short donor DNAs of CRISPR-induced DSBs with high efficiency [17][18][19] . To test whether this was possible in M. oryzae, we amplified the BAR gene which confers resistance to the herbicide glufosinate ammonium 20 , with 30 bp and 40 bp flanking regions on either side of the selectable marker. In this way we were able to demonstrate that a 30 bp region of homology was sufficient to induce repair by HR of the CRISPR-Cas9 generated DSB and result in mutation of ALB1, with efficiencies approaching those achieved with the much longer donor DNAs (see Table S3). The increased rates of mutation, compared to those observed with donor-only controls, provided further evidence that RNP-CRISPR-dependent gene replacement is efficient, and also demonstrated that RNP-CRISPR-Cas9 gene inactivation can be generated without the need for laborious cloning strategies.
Direct demonstration of marker free mutation of the ALB1 melanin biosynthesis gene using donor free CRISPR-Cas9 RNPs. Our observation of efficient gene replacement when RNPs were introduced into M. oryzae with donor DNA fragments, suggested that the Cas9-NLS enzyme creates DSBs and, in so doing, induces repair by the homologous recombination pathway. However, because these experiments used donor DNAs to repair the breaks using a selectable marker gene, they were genotypically indistinguishable from alb1 or rsy1 strains that would be generated from an experiment using the donor DNA only to disrupt each geneas in 'traditional' gene disruption approaches. We therefore set out to examine CRISPR events more directly by introduction of the RNP complex targeting ALB1 only, without donor DNAs to direct repair by homologous recombination. In the absence of donor DNA, the resultant DSB can be repaired by the non-homologous DNA end-joining (NHEJ) pathway which, because it is frequently inaccurate, should result in alb1 (albino) mutants. Although such events were found to be rare under the conditions tested, we were able to observe albino mutants among a large background (showing confluent growth) of normally pigmented regenerants (Fig. 3a). Control transformations without RNPs yielded no such white patches. Unfortunately, three attempts to dilute the transformants to a concentration where 10-20,000 individuals could be isolated, failed to yield any albino colonies, suggesting that much less than 0.01% of regenerated protoplasts harbour alb1 mutations. These observations suggest that the efficiency of the delivery system used would make marker-less gene targeting impracticable in the absence of an easily identifiable phenotype, because this would necessitate analysis of more than 20,000 individuals. Nevertheless, the albino mutants generated allowed us to directly demonstrate that RNPs are functional in vivo and to understand the nature of mutations generated through NHEJ. To this end, we purified albino mutants by several rounds of subculture followed by single spore isolation (Fig. 3b,c) and in a few cases, the insertion or deletion was large enough to be apparent by gel electrophoresis of the amplicons (Figs 3d and S1). In one case, no amplification was possible, indicating that a larger deletion had removed the sites where one or both primers, anneal. Sequencing of DNA around the genomic target site of the RNP in these albino mutants revealed true CRISPR mutations, showing a range of insertions or deletions close to the PAM site, as shown in Fig. 3e. These results demonstrated that the RNP complex functions by creating a DSB at the expected site and that marker-less single mutations are feasible, but not at a frequency which would be useful in the absence of an easily identifiable phenotype.
Specific single nucleotide gene edits can be very efficiently accomplished using short oligonucleotide donor DNA. One of the attractive features of CRISPR-induced genome editing is the ability to make highly specific changes to the coding sequence of a gene that could, for example, give rise to a single amino acid change in a protein product. To test whether single nucleotide edits were feasible using RNP-CRISPR-Cas9 in M. oryzae, we attempted to edit the gene SDI1, which encodes a subunit of the succinate dehydrogenase enzyme. We designed a SDI1-targeting RNP, to introduce a mutation that leads to an amino acid change in the enzyme Development of a gene co-editing strategy in M. oryzae. We observed that mutants generated by NHEJ using RNPs without a selectable marker did not arise at a frequency that would make gene targeting practicable. We therefore set out to develop a different method to enrich for gene-edited transformants. We decided to adopt a co-targeting approach in which two independent RNPs are transformed together into M. oryzae. We reasoned that a proportion of cells would take up both complexes and be edited at both loci. If one of the genes had an easily scorable phenotype that could be used as a selectable marker, we should be able to select transformants more easily and then determine whether the second locus had also been edited. As a proof of principle, we introduced RNPs targeting the succinate dehydrogenase subunit-encoding gene SDI1 and ALB1, simultaneously, together with the 80 bp oligonucleotide donor dsDNA, which we had already established was able to convert the SDI1 gene to an allele bestowing carboxin resistance. We employed two different ALB1-targeting donor DNAs, both of which introduce a premature stop codon in the gene, close to its 5′ end (Fig. S2). One of the donor DNAs introduces the edit within the genomic target sequence. The other donor is predicted to generate an edit 40 bp from the PAM allowing us to assess whether we can create edits at some distance from the DSB. We found that when Guy 11 was used as a recipient, these donor DNAs integrated in approximately 50% of the albino transformants, which represented 1-2% of all the carboxin resistant transformants, as shown in Table S4. Furthermore, the donor DNA that introduces an edit outside the genomic target sequence was as efficient at editing as the other donor. Albino mutants that lacked the integrated donor DNAs exhibited indels, typically 3-4 bp 5′ from the PAM. These are indicative of mutations generated by NHEJ of the CRISPR-generated DSB. Surprisingly, in one instance an albino mutant arose by integration of the SDI1-targeting donor DNA at the ALB1 locus. By contrast, when we employed a Δku70 mutant 22 that lacks the NHEJ pathway, all albino mutants generated showed precise integration of the ALB1 targeting donor DNA. Moreover, the efficiency of co-editing both loci increased, although the number of overall transformants was reduced (Table S4). To determine whether this approach, which we henceforth refer to as co-editing, was applicable to other genes, we set out to co-edit both the ILV2 and TUB2 genes, which encode acetolactate synthase and β-tubulin, respectively. These genes can be edited at a single nucleotide to give rise to alleles encoding sulfonylurea and benomyl resistant mutants, respectively (see Fig. S3; refs 23,24 ). RNPs were first created to introduce these edits and then tested individually (Fig. S3). We then conducted a co-editing experiment by transforming the ILV2 and TUB2 targeting RNPs, together with the two corresponding oligonucleotide donors. We selected for sulfonylurea resistance and then calculated the proportion of sulfonylurea resistant transformants that were also benomyl resistant. Consistently, we observed ~1% efficiency of co-editing (Table S4). We were able to confirm the edits that had occurred in these transformants by direct sequencing of amplicons containing the target sequence. Mutations generated by NHEJ repair would in most conceivable instances, not be selected for by these experiments. Together, these experiments demonstrated that we are able to generate marker-less mutations in M. oryzae at target loci by employing a straightforward co-editing strategy.
Co-editing allows generation of in-locus GFP-tagged gene fusions and conditional mutant alleles in M. oryzae without a selectable marker. To demonstrate that co-editing could be employed to generate novel genotypes in any gene of interest, we decided to tag the SEP5 septin-encoding gene 25 with GFP, using CRISPR co-editing at the native locus. We also tested whether we could exploit co-editing to introduce a two nucleotide edit into the SEP6 septin-encoding gene to create a temperature-sensitive allele. We generated a G234E substitution into SEP6, which corresponds to a mutation (G247E) that in the Sep6 orthologue Cdc12 in Saccharomyces cerevisiae, gives rise to a temperature sensitive (ts) form of the septin (Fig. 5a) 26 . Using CRISPR-mediated co-editing we were able to generate both a sep6 G234D allele (Fig. 5b,c) and a SEP5-GFP strain, as shown in Fig. 5d,e. To confirm that the correct genotype had been created at the corresponding loci, we sequenced amplicons of both genomic regions (for the sep6 G234D strain's edit see Fig. 5c). The SEP5-GFP strain was identified by examination of 200 transformants, while the sep6 G234D mutation was identified among 79 transformants. These experiments confirmed that co-editing can be employed to rapidly and precisely manipulate genes in M. oryzae. We confirmed that the sep6 G234D mutation leads to a temperature sensitive loss of virulence and is unable to cause rice blast symptoms at the non-permissive temperature of 30 °C, as shown in Fig. 5b. Replacement of SEP5 with a SEP5-GFP gene fusion at the native locus meanwhile leads to visualisation of a GFP-tagged septin ring at the M. oryzae appressorium pore (Fig. 5e), identical to that previously reported for an ectopically integrated gene fusion 25 .
A novel selection strategy allowing construction of isogenic gene edited mutants in M. oryzae. It is likely that expression of antibiotic resistance genes has unknown consequences in fungi. Ultimately, it would therefore be desirable to generate a mutant that contains a specific edit, but that is in all other respects, isogenic to the progenitor strain. We reasoned that the reported negative cross resistance of certain benomyl resistant mutations to the compound diethofencarb 27,28 , might provide a novel counterselection strategy that would allow us to generate isogenic CRISPR mutants in M. oryzae. We were able to confirm in plate growth tests that the wild type M. oryzae strain, Guy11, can exhibit normal growth in the presence of 10 μg mL −1 diethofencarb, whereas a benomyl-resistant strain harbouring a TUB2 allele with a E198A mutation cannot grow under the same conditions (Fig. 6a). We therefore used a TUB2-targeting RNP and an 80 bp oligo donor that restores the TUB2 sequence to wild type (Fig. 6b), and introduced this into a benomyl resistant transformant, previously created using RNP-CRISPR-Cas9. This led to the generation of diethofencarb-resistant transformants at a very high frequency (Fig. 6c). The diethofencarb-resistant transformants were as sensitive to benomyl as Guy 11. No diethofencarb-resistant transformants were generated in the absence of the TUB2-targeting RNP. Furthermore, we observed that the counterselection is very tight, because no wild type TUB2 strains can grow on 10 µg mL −1 benomyl, and no benomyl-resistant TUB2 E198A strains are able to grow at all on 10 µg mL −1 diethofencarb as shown in Fig. 6a. We conclude that CRISPR-Cas9-RNP mediated generation of TUB2 E198A benomyl resistant strains of M. oryzae, followed by a second round of CRISPR-Cas9-RNP to restore a wild type TUB2 sequence, bestowing diethofencarb resistance, provides a means of generating isogenic, markerless, genome-edited mutants. The counter selection strategy also represents an excellent way by which to build multiple mutations in a single strain of the fungus, as it is straightforward to cycle between the two states, and thereby introduce further specific edits or other manipulations each time by co-editing.

Assessment of off target effects in albino mutants generated by RNP-CRISPR without donor DNA.
One potential constraint on the use of CRISPR-Cas9 genome editing is the possibility of off-target mutations at sites showing significant similarity to the genomic target sequence 29 . Although no mismatch is tolerated in the seed sequence proximal to the PAM, in some species up the 5 bp mismatches are tolerated at the PAM-distal end of the genomic target sequence, resulting in potential for off-target cleavage 30 . For this reason, most programmes used for automated protospacer selection search for a protospacer with minimal potential for off target cleavage. The degree of off target cleavage, however, varies considerably between organisms and is, of course, also a function of genome size. However, careful sgRNA design and limited longevity of the Cas9-SgRNA complex in a cell, seem to be major means to maintain editing specificity 30 . An advantage of using RNP-CRISPR genome editing is that there is evidence that the transient presence of the CRISPR machinery in the cell may also limit off target effects 31 .
In order to assess whether off target mutations had occurred in CRISPR generated mutants, we randomly selected two alb1 mutants generated by CRISPR-Cas9-RNP and, as a control, two strains that had been through the transformation procedure and regenerated but not exposed to RNP complexes. The genomes of these M. oryzae strains were sequenced (Table S5)

Discussion
In this report, we have demonstrated the efficient generation of CRISPR induced gene edits in the rice blast fungus using purified CRISPR machinery components. We were motivated to develop this method because stable expression of Cas9 appears to be very toxic to M. oryzae in the same way as reported in some other species, including fission yeast [32][33][34] . The transient nature of CRISPR-Cas9-ribonucleoprotein expression provides an excellent means of circumventing the problems associated with Cas9 toxicity and rapidly generating gene edited mutants in M. oryzae. This contrasts with the previously published method requiring expression of the Cas9-encoding gene 14 , which we were unable to reproduce in spite of extensive efforts. The key advantages of RNP-mediated CRISPR-Cas9 editing are its efficiency, accuracy, and especially its speed. Use of oligonucleotide-based or PCR-amplified donor DNAs obviates the need for labour-intensive DNA cloning and thereby dramatically reduces the time and cost required to make precise gene manipulations in this species. Additionally, we demonstrated that RNP-CRISPR editing is highly specific, because we saw no evidence of off-target mutations in the genomes of two CRISPR generated mutants. RNP-CRISPR therefore is an extremely useful and adaptable addition to the Magnaporthe gene manipulation toolbox as a simple amendment to existing transformation protocols. Furthermore, as the price of the necessary components of the RNP complex falls over time, we predict that RNP-CRISPR-Cas9 editing will become a standard manipulation in a very short time in M. oryzae.
We observed that although mutations resulting from RNP-CRISPR and NHEJ-dependent repair were possible without a selectable marker, these occurred at a frequency too low to be practically exploited. One interpretation of this result is that NHEJ may be highly accurate in M. oryzae, but a more likely explanation is that a very small proportion of fungal protoplasts actually take up the RNP. It is therefore possible that RNP complexes might be more efficiently delivered by other means, such as electroporation or biolistic delivery. We were, however, able to overcome this potential limitation by developing a gene co-targeting strategy that we termed co-editing. This co-editing approach significantly enriches for specific edits in a marker-less fashion without introduction of a further selectable marker sequence at the locus of interest. Co-editing was found to reproducibly occur in 0.5-2% of transformants that were also edited to bestow carboxin, benomyl or sulfonylurea resistance, respectively. This required a second RNP and donor DNA targeting a gene of interest, in addition to the RNP donor pair generating the mutation to confer antibiotic resistance. Using oligonucleotide donor DNAs also had the additional advantage that it generated very large numbers of antibiotic-resistant transformants, from which it was straightforward to select co-edited mutants. When the gene co-editing method is combined with the benomyl/diethofencarb-based counterselection strategy, we have provided a mechanism to generate gene edited mutants in M. oryzae that are truly isogenic to a progenitor wild type. This is very advantageous, not only for basic research (where studying a single mutation in isolation from any other genome perturbations is the best possible method), but also in fungal biotechnological applications where the lack of a resistance gene marker is important from a regulatory perspective. Although the efficiency of gene co-editing that we report here is rather low, it is likely that optimisation by adjustment of the ratio of the RNPs and/or donor DNAs may be possible in future. Furthermore, in the current report we show that 1 or 2 base changes are possible with an 80 bp donor DNA, but preliminary results in our laboratory indicate that similarly sized donor DNAs can efficiently and precisely delete small sections of genes of 40-50 bp, that would facilitate simple PCR-based screens for gene-inactivated mutants. Additionally, the generation of small deletions may be a more attractive method for high-throughput gene functional analysis because they may be more stable than changes to a single nucleotide. The time saved by not having to construct vectors for gene manipulation makes co-editing an attractive option, even if it necessitates sequencing 100 or more transformants to identify the specific mutant required.
Our study does, however, raise some important questions too, that we will address in future. It is apparent, for instance, that some RNPs work more efficiently than others as has been reported in other species 35,36 . In making the most efficient use of co-editing, it is important that we better understand which protospacers are likely to generate the best results. Additionally, it has been suggested that protoplast-mediated transformation is in itself, mildly mutagenic 37 , and in the long term it would be worth investing time to explore other means to deliver RNP complexes to M. oryzae. Electroporation, for example, might represent a less disruptive means of delivering the RNP-CRIPSR-Cas9 complex. If RNP complexes can be delivered at the same time as DNA donors via electroporation, it would, for instance, be possible to directly compare SNPs and indel frequencies in the genomes of fungal strains manipulated by these different methods to test whether electroporation or other delivery systems really are less mutagenic than protoplast generation.
In summary, the current study demonstrates that RNP-CRISPR-Cas9 genome editing is a simple, precise, reproducible, and rapid means by which gene manipulation can be carried out in M. oryzae. It is our hope that researchers investigating Magnaporthe species and related fungi will adopt the procedures described here and that this will empower researchers and accelerate discovery towards understanding and ultimately combatting the devastating rice blast fungus M. oryzae and similarly important pathogenic fungal species.

Materials and Methods
Strains and Culture conditions and infection assays. The wild type strain Guy 11 and the NHEJ deficient mutant Δku70 were grown in a controlled temperature room at 25 °C with a 12 h light/dark cycle. For tests of temperature sensitivity an incubator at 30 °C with a 12 h light/dark cycle was used to conduct cut leaf infection assays. Infection assays used two cm long leaf sections cut from 3 week old leaves of the dwarf indica rice cultivar CO-39 and were assessed after 5 days incubation at either 25 °C or 30 °C with a 12 h light/dark cycle. For glufosinate and sulfonylurea selection, we used Basta defined complex medium (BDCM) 38 . For standard growth, we used CM 39 . For selection using hygromycin, benomyl, carboxin or diethofencarb we used OCM which is CM osmotically stabilised with 0.8 M Sucrose. Selective agents were used at a final concentration of 200 µg mL −1 for hygromycin or 150 µg mL −1 for sulfonylurea (chlorimuronethyl) 40 µg mL −1 glufosinate or 10 µg mL −1 benomyl 50 µg mL −1 carboxin or 10 µg mL −1 diethofencarb. SgRNA synthesis and RNP formation. SgRNA for complexing with Cas9-NLS was synthesised using the EnGen sgRNA synthesis kit available from New England Biolabs (NEB #E3322), according to the instructions provided, and purified before complexing to Cas9 using the RNA Clean & Concentrator-25 kit (Zymo Research). Cas9-NLS was purchased from New England Biolabs (NEB Catalog #: M0646M). Cas9-NLS was complexed with the sgRNA by a 10 minute incubation at room temperature. The sgRNA selection was carried out using the E-CRISP program at http://www.e-crisp.org/E-CRISP/ 40 .

Fungal transformation and introduction of RNPs and donor DNAs. PEG-mediated fungal trans-
formation of Glucanex-generated protoplasts was achieved, as previously described 39 . The RNP complexes and donor DNAs were introduced together before the step in the standard transformation procedure, where 50% PEG is added and the mixture then incubated for 25 min. During optimisation for reproducibility of the method, sgRNAs were always freshly synthesised and purified on the day of transformation and kept on ice following formation of the RNP complex. All experiments except co-targeting experiments using the SDI1-targeting RNP, used 2 μg of donor DNA and 6 μg of Cas9 mixed at a 1:1 molar ratio with the respective sgRNA (pre-complexed together for 10 min at RT) and protoplasts, in a volume of 150 μl at a concentration of 1.5 × 10 8 protoplasts/ml. For the co-targeting experiments using the SDI1-targeting RNP, we used 1 μg of Cas9 precomplexed with 0.2 μg of sgRNA. For Agrobacterium-mediated transformation of Cas9-containing vectors, conidia of Guy11 were transformed as previously described 41 .
Oligonucleotide design and source. Oligonucleotides used in this study were designed with the aid of the Primer3 program (http://primer3.ut.ee/) or as templates for sgRNA synthesis were designed with the aid of NEB's online design tool (http://nebiocalculator.neb.com/#!/sgrna). All the oligonucleotides used in this study were purchased desalted from Eurofins Genomics and are listed in Table S6. For generation of PCR products to recombine in vivo in S. cerevisiae 40-45 bp regions of homology to the recipient vector were added to the 5′ end of the primer sequence.
Cloning of ALB1 and RSY1 targeting donor DNAs. To generate cloned donor DNA for repair of RNP-CRISPR generated DSBs, in the case of ALB1 we amplified a 1.4 kb ALB1 fragment using the primers Scientific RepoRts | (2018) 8:14355 | DOI:10.1038/s41598-018-32702-w ALB1-for-EcoRI and ALB1-rev-SpeI. The resultant PCR product was digested with EcoRI and XhoI gel purified and ligated to EcoRI + SalI digested pCAMBIA0380 thus destroying the SalI site in this vector's multi-cloning site. The resultant vector pCAMB-ALB1 was linearized using a SalI site in the middle of the ALB1 fragment and ligated to the HPH cassette, conferring hygromycin resistance excised from plasmid pCB1636 38 using SalI to create pCAMB-ALB-HPT. A donor DNA was generated by PCR amplification of the HPH interrupted ALB1 fragment from pCAMB-ALB-HPT using the primers ALB1-for and ALB1-rev. In the case of RSY1, RSY-donor-f and RSY-donor-r were used to amplify a 1.8 kb RSY1 fragment which was cloned into the vector pGEMT-easy (Promega) to give pGEM-RSY. pGEM-RSY was digested with XhoI and ligated to a SalI HPH fragment to give pGEM-RSY-HPT which could then be used as a template for amplification of the required donor DNA (HPT interrupted RSY1) using RSY-donor-f and RSY-donor-r.

Cloning of Agrobacterium tumefaciens compatible Cas9 expressing vectors. To generate an
Agrobacterium compatible vector to introduce Cas9 and an ALB1 targeting sgRNA transcribing sequence into M. oryzae, we used the nuclear localised, Cas9 codon-optimised for Neurospora crassa in vector p415-PtrpC-Ca s9-TtrpC-CYC1t 42 as a template for high fidelity PCR amplification of Cas9-NLS under control of the Aspergillus nidulans TrpC promoter and terminator sequences, using the primers Cas9-recom-f and Cas9-recom-r. We then used yeast recombination in Saccharomyces cerevisiae strain DS94 (MATα, ura3-52, trp1-1, leu2-3, his3-111, and lys2-801 43 ) to recombine this fragment into a 12,869 bp XhoI-BamHI fragment of the "soft-landing" vector pS315 pMMag_Cbx-mCherry, which will integrate at the SDI1 locus as a single copy to confer carboxin resistance (pS315-pMMag-Cbx-mCherry was a kind gift from Prof. Gero Steinberg, University of Exeter). The resultant vector p315-Cas9-csr-1 was then further modified to introduce sgRNA transcribing sequences targeting ALB1 or RSY1. This was achieved by generating a PCR-amplified fragment containing the desired guide RNA encoding sequence created by using the plasmid p426-SNR52p-gRNA.csr-1.Y-SUP4t as a template, with primers PKS1gRNA-f and PKS1gRNA-r for ALB1 and in the case of SDH1 the PKS1gRNA-r primer was replaced with the primer SDH1-sg-r. This PCR results in the replacement of the N. crassa csr-1 targeting guide with the M. oryzae ALB1 or RSY1 targeting guide under the control of the SNR52 promoter. The resultant amplicons were then used as a template to generate recombination competent fragments using the primers gRNAto315Cas-R and sgRNA-rec-R. The products of this amplification were recombined separately in yeast into XbaI digested p315-Cas9-csr-1. Correct assembly of the vectors in yeast was confirmed following extraction from yeast strains, transformation of, and purification from E. coli followed by analysis by restriction digests and partial sequencing. The vectors p415-PtrpC-Cas9-TtrpC-CYC1t and p426-SNR52p-gRNA.csr-1.Y-SUP4t were gifts from Christian Hong (Addgene plasmid numbers 68059 and 68060 respectively). Additionally, to test the Cas9-NLS promoter gene derived from p415-PtrpC-Cas9-TtrpC-CYC1t, a TEF1 promoter-Cas9 with a glucoamylase-encoding gene terminator sequence 14 was used. Note that the codon optimised Cas9 allele used by these authors differs from the native Cas9 amino acid sequence at two amino acids, but is reported as functional despite these changes. We also tested our own TrpCp-Cas9-TrpC terminator construct (M. oryzae codon optimised Cas9-NLS) generated by gene synthesis. These constructs were generated by a yeast recombination cloning strategy using the vector pMMag_Cbx-mCherry as detailed above and replace the mCherry ORF with the appropriate Cas9 ORF or replace the promoter and gene in the case of the TEF1 promoter-Cas9 construct. The sgRNA expressing constructs tested used the S. cerevisiae SNR52 promoter 42 or the U6-1 or U6-2 promoters of M. oryzae 14 . Bioinformatic analysis for detection of potential off target mutations. Genomic DNA from two regenerant colonies (not exposed to RNP but which have been through the transformation procedure) and two hygromycin resistant alb mutants was extracted using the Qiagen Plant DNAeasy mini kit and was sequenced on Illumina HiSeq2500 using standard reagents and protocols producing 125 base paired-end reads by the University of Exeter Sequencing Service. Reads were filtered using the fastq-mcf program from the ea-utils package, ea-utils 44 : "Command-line tools for processing biological sequencing data"; https://github.com/ ExpressionAnalysis/ea-utils). Genomes were assembled 'de novo' using SPAdes 3.11.0 45 . For SNP and INDEL calling reads were aligned against the M. oryzae reference genome 46 using BWA 47 . SNPs and INDEL were identified with a bespoke pipeline using bcftools and vcfutils from the SAMtools package as well as SnpSift 48,49 . SNPs called from the two alb1 mutants were compared to two regenerated strains that had been subjected to the transformation protocol but not exposed to RNPs, based on analysis made by Schuster and co-workers 38 .

Data Availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files) or are available from the corresponding author on reasonable request.