The CRISPR/Cas9 technology is increasingly popular in plant research to generate mutants in target genes. In CRISPR/Cas9, the gRNA mediates a double-stranded DNA break, which when repaired often leaves small mutations. These mutations can knock out the function of the target gene, but only when all alleles of the gene are mutated. Because the repair is imprecise, the exact mutations will differ between cells. In a genetic chimera such as a hairy root, it is often difficult to separate a multitude of small mutations from the WT sequence. Without the ability to rule out any remaining WT alleles in the transformed tissue, one cannot unequivocally link genotype with phenotype.
Our approach is to use a pair of gRNAs to target different parts of the same gene, so that they induce two double stranded DNA breaks simultaneously. When the cell’s DNA repair systems are engaged, the repair can bring together two distal DNA breakpoints caused by different gRNAs, resulting in a deletion (Fig. 2). When the two gRNA target sites are sufficiently far from each other, the resulting deletion can be large enough to be detectable on the gel following PCR, and is likely a loss-of-function mutation. A transformed root where a WT-sized band can no longer be detected will have deletions in both alleles. Such a cell could be homozygous for the deletion, or contain two different deletions and thus be bi-allelic. In either case, one can reasonably assume it is a functional null. The entire root may still be a genetic mosaic containing combinations of mutations, but is devoid of a functional copy of the target gene. If desired, the PCR products can then be cloned and sequenced to reveal the exact mutations. One can then be confident that the phenotype observed on such roots is due to disruptions in the candidate gene.
Roots on different plants are separately transformed and mutagenized; the exact mutations are often different; and thus the phenotype can be independently verified through an allelic series.
Causing large deletions in a target gene requires delivering two gRNAs into the same plant cell. The most efficient strategy is to clone both gRNAs in the same vector. Several platforms have been developed to clone two or more gRNAs (Lowder et al., 2015; Ma et al., 2015; Xing et al., 2014). Furthermore, the two double-stranded DNA breaks should be induced simultaneously; otherwise, each break will be repaired locally, resulting in two small mutations that may not be detectable by simple PCR. To improve synchrony, the two gRNAs should be expressed at comparable levels. This rules out systems that use different promoters to drive gRNA expression; even when the same promoter is used twice to express two gRNAs separately, the result can still be suboptimal, because multiple copies of the same promoter can trigger gene silencing. The ideal system is one that produces two or more gRNAs from a polycistronic transcript, as the gRNAs are generated in identical quantities. Cloning is also easier when all gRNAs share one common set of promoter and terminator sequences.
We decided to use the system developed by Dr. Dan Voytas at the University of Minnesota (?ermák et al., 2017). The vector allows the expression of a polycistronic transcript, which is then processed by the Csy4 enzyme to produce several gRNAs. One great versatility enabled by this vector is that the gRNA sequences are incorporated via Golden Gate cloning. This cloning method means there is no theoretical upper limit to the number of gRNAs that can be incorporated in one construct: the same cloning protocol can accommodate one to many gRNAs, requires only three enzymes, and is completed in one step. The authors demonstrated the feasibility to incorporate a maximum of eight gRNAs. To incorporate two gRNAs, the procedure involves three PCR products (Fig. 3); any additional gRNA simply requires one more PCR product. All products will then ligate with the vector in a Golden Gate reaction.
Genotyping for mutants
The resulting CRISPR/Cas9 construct is delivered using regular Agrobacterium rhizogenesis-mediated hairy root transformation. Because different roots will contain distinct mutations, we keep only one long root per plant to better track each individual. The transformation protocol includes a 2-week period to select for transgenic roots with antibiotic resistance. Our experience shows that often mutations can already be detected after this stage, if not sooner. Often we pool root tissues at this stage for one DNA extraction, to test the efficacy and to estimate the efficiency of the construct. Once the predicted deletion is detected, we proceed to genotype each root.
At this point, a section of each antibiotic-resistant root is excised for genotyping, as the rest of the plant is transferred to soil to be inoculated with rhizobia. Using an easy DNA extraction method (Satya et al., 2013), one can quickly genotype dozens of individuals to identify deletion mutants lacking WT-sized PCR products. Concentrating on these individuals significantly saves labor further downstream, especially when the efficiency of the gRNAs is low.
Once we identify homozygous/biallelic mutants, we will be able to examine nodule phenotype to determine the requirement of the target gene in the symbiosis. If the correlation between phenotype and genotype is unambiguous, in additional batches of hairy root transformation we can simply focus on roots with mutant nodules. In a proof-of-concept, we were able to knock out a component (SPC18) of the nodule-specific signal peptidase complex (Wang et al., 2010), and observed defective nodules similar to the dnf1 mutant (Fig. 4). We also generated a construct targeting the gene encoding the nodule-specific small peptide NCR247, which would be difficult to recover from the available insertional or deletion mutant collections due to its small size. For reasons that are not well understood, not all gRNAs perform equally well. However, the cloning and transformation steps are sufficiently easy and fast, so that new gRNAs can be tested easily.
Beyond individual genes
At present, we have employed this CRISPR/Cas9 strategy to disrupt individual genes, for example SPC18, NCR247, etc. However, the construct can also be used to induce larger deletions. The authors of the original vector have demonstrated a maximum deletion of 58 kb. Such scales are suitable for removing tandemly duplicated genes. We expect that the larger the deletion, the lower the rate of recovery. Therefore, the selection of gRNAs may need to be optimized through trial and error.
Our paired gRNAs are designed to cause one deletion. However, the approach also allows the construct to carry more than one pair of gRNAs, which can generate higher order mutations. This will be useful when multiple unlinked genes are suspected of redundant functions. Knocking out multiple genes at once will be superior to crossing single mutants even when they are available, or complicated RNAi strategies. Making such constructs requires ligating five or more DNA fragments, which may make it necessary to optimize the cloning.
We have also gained the expertise to generate fully transgenic plants. This allows us to make stable mutants when desired, for example when the mutagenesis is complex. This can include very large deletions and multiple deletions, where it may not be feasible to recover the desired homozygous mutants or mutant combinations. We can then first produce full transgenic plants, recover heterozygous individuals, and then screen for homozygotes in the progeny.
Finally, CRISPR/Cas9 technology has been increasingly utilized in animal systems as a forward genetic tool to discover novel genes (Yu and Yusa, 2019): rather than disrupting a candidate gene to test a hypothesis regarding its function, a CRISPR screen uses a gRNA library to knock out many genes, sometimes at genome scale, to discover genes with previously unknown roles in a process. This approach is beginning to be adopted by the plant research community (Gaillochet et al., 2021). Our platform, because of its multiplex nature and rapid generation of transgenic material, can be a powerful basis for a gRNA library. We will probe the feasibility of a screen. We will focus on mutations or their combination that is hard to fully cover using existing M. t. genetic tools, such as small genes and gene families.