CRISPR/Cas-directed evolution platform
Applications of CRISPR/Cas9 in plants are mainly focused on the generation of functional knockouts via targeted DSB formation and efficient NHEJ, because HDR is quite inefficient in plants. To develop an efficient directed evolution platform in plants, we used a combination of targeted mutagenesis and selection. To this end, we designed a CRISPR/Cas-directed evolution (CDE) platform employing CRISPR/Cas9 to generate DSBs at all possible coding sequence sites in a specific gene of interest. CRISPR/Cas9 requires a protospacer adjacent motif (PAM) of NGG for cleavage; therefore, only sequences adjacent to a PAM can be targeted.
For CDE, a targeted library of single guide RNAs (sgRNAs) is designed, cloned into a binary vector, and then transformed into Agrobacterium tumefaciens for stable plant transformation. Agrobacteria harboring the sgRNA library are used for plant transformation, and plants are regenerated under selective pressure to force accelerated evolution. This enables the recovery of protein variants that give the plant the ability to survive under selective pressure. Plant regenerants that survive the selective pressure are genotyped and their progeny phenotyped to link genotype to phenotype and examine mutant variants of the protein (Fig. 1). The ability to segregate out the CRISPR-Cas9 machinery and recover progeny plants harboring the evolved variant of interest without any foreign DNA makes this an elegant system for directed evolution applications.
CRISPR/Cas directed evolution of SF3B1
To provide proof of concept for CDE, we evolved the spliceosome component SF3B1 for resistance to splicing inhibitors. In plants, naturally occurring splicing inhibitors, including pladienolide B (PB) and herboxidiene (GEX1A), have massive effects on the splicing machinery, resulting in transcriptome-wide splicing repression [17, 18]. These polyketide natural products, produced by Streptomyces sp., have been applied to plants as herbicides and to mammalian cells as anti-cancer therapies [19,20,21]. Work in mammalian cells showed that these splicing inhibitors target the core splicing factor SF3B1, of the SF3B complex of the U2 snRNP of the spliceosome [22]. We have tested the effects of PB and GEX1A on Oryza sativa (rice) seed germination and primary root (PR) length, and our data showed that these splicing modulators significantly inhibit seed germination and PR growth in a dose-dependent manner (Additional file 1: Figure S1).
Due to the strong conservation with the mammalian SF3B complex, we hypothesized that the rice SF3B complex is targeted by these splicing inhibitors. We searched the rice genome for SF3B genes and found that OsSF3B1 is highly conserved compared with its mammalian homolog (Additional file 1: Figure S2). To test our CDE platform on the evolution of SF3B1 resistance to splicing inhibitors, we used the CRISPR/Cas9 system to generate SF3B1-resistant variants. Because splicing inhibition leads to inhibition of plant growth and development, we selected plants during regeneration of callus using a splicing inhibitor to identify resistant variants.
To generate SF3B1 variants, we designed 119 sgRNAs targeting all possible PAM-adjacent sites in the whole coding sequence (CDS) of SF3B1 (Additional file 2: Table S1). The sgRNA library was constructed via oligonucleotide synthesis and annealing of all possible targets in SF3B1 into the sgRNA scaffold, under the control of U3 promoter in the binary vector pRGEB32 (Additional file 2: Table S2). Cas9 was produced under the control of the OsUbiquitin promoter on the same plasmid. The plasmids of this library were pooled and transformed into the Agrobacterium tumefaciens EHA105 strain for stable transformation of rice (cv. Nipponbare) embryonic callus. Subsequently, we performed callus transformation and selection for stable transformation. From the transformed callus, we regenerated whole plants on different concentrations of GEX1A. We used different concentrations to provide variable levels of selective pressure to trigger NHEJ repair and generation of SF3B1 variants resistant to GEX1A. We sub-cultured 15,000 transformed calli onto medium supplemented with 0.4 μM and 0.6 μM GEX1A, concentrations that are sufficient to inhibit wild-type callus growth. We recovered 21 rice shoots on 0.4 μM GEX1A (Additional file 2: Table S4). Our data thus showed that our directed evolution platform was successful in regenerating GEX1-resistant seedlings that were likely to possess an SF3B1 mutation (Fig. 2a).
To confirm that these seedlings carried mutations in the SF3B1 gene, we identified the sgRNA sequence in each resistant seedling through PCR amplicon sequencing. This allowed us to determine the target sequence in the SF3B1 gene for genotyping and to test the presence, nature, and identity of the causal mutation (Fig. 2b). We selected several mutants, including SGR1 (SF3B1-GEX1A-Resistant), SGR2, and SGR3 for further study and analysis (Fig. 2b). Our sgRNA sequencing data revealed that these seedlings carried a single sgRNA. To determine the mutations triggered by these sgRNAs, we PCR-amplified the region encompassing the target sequence from T0 plants and subjected these PCR amplicons to Sanger sequencing.
Our data showed that the GEX1A-resistant seedlings possessed in-frame SF3B1 mutations that are likely to be functional to support RNA splicing, but could affect drug binding (Fig. 2b). For example, the SGR1 mutant had a 3-bp mutation causing a predicted single amino acid deletion of Q157. The SGR2 mutant had a deletion of 10 amino acids, DAPDATPGIG (amino acids 223 to 232). SGR3 had a single amino acid deletion of K1050; mutation in the same K1071 of the human homolog HsSF3B1 confers resistance to splicing inhibitors [21]. In an experiment without GEX1A selection, we tried to recover knockout mutants of SF3B1. In the T0, we recovered seedlings harboring a heterozygous single-nucleotide deletion mutation with sgRNA-18 and an in-frame mutant with sgRNA-50 causing a 15 amino acid deletion LPLMKPEDYQYFGTL (442–456). The heterozygous knockout mutant was not heritable in the seed progeny. However, the in-frame mutant variant with the 15 amino acid deletion exhibited no resistance to GEX1A (Fig. 2 and Additional file 1: Figure S4). Because we were unable to recover any functional knockout (out-of-frame) mutants of SF3B1, we concluded that the loss of functional SF3B1 is embryonic lethal and that this is an essential gene for seed viability.
Domain-focused directed evolution
We then tested whether domain-focused directed evolution, in which targeted mutagenesis is conducted on key protein domains known to mediate key interactions for function, stability, or activity, was achievable in our CDE platform. We examined whether targeting SF3B1 protein domains, known to mediate drug–protein interactions, in our CDE platform could generate SF3B1 variants resistant to GEX1A. We capitalized on our prior knowledge of the key SF3B1 domains that mediate drug interactions and spliceosome inhibition and designed sgRNAs targeting these domains (Additional file 1: Figure S3). HEAT repeats (HR) 15–17 are highly conserved (Additional file 1: Figure S2 and S3) and splicing inhibitor-resistant mutations identified in mammalian cell cultures clustered around this region [21]. We targeted this region by using a single sgRNA (HR-target) or a PTG (polycistronic tRNA-gRNA) fragment that contained two sgRNAs (Additional file 1: Figure S3). After application of our CDE platform and rice callus transformation, we recovered SF3B1 variants highly resistant to GEX1A, including SGR3 to SGR6 (Fig. 2c, Additional file 1: Figure S5 and Additional file 2: Table S3). The SGR3 mutant was also recovered from GEX1A-resistant seedlings of a single sgRNA (Fig. 2a). Our data show that, compared with sgRNAs targeting random regions, the sgRNAs targeting HR15–17 are more effective in producing rice seedlings capable of survival on GEX1A-supplemented media. Genotyping analysis revealed that these mutations included SGR4 with three amino acid substitutions (K1049R, K1050E, and G1051H). SGR5 had a substitution (H1048Q) and a deletion (K1049). SGR6 was recovered from PTG transformation and had two substitutions (H1048Q and A1064S), and a deletion (K1049, Fig. 2c).
To test the heritability of these mutations and the survival of the homozygous SF3B1 mutant variants, we conducted a phenotypic and genotypic analysis of the seed progeny of the SGR1, SGR2, SGR3, SGR4, SGR5, and SGR6 mutants. Genotyping analysis revealed that these mutants were heritable in a homozygous fashion, indicating that these SF3B1 variants support efficient splicing, because these mutants are phenotypically indistinguishable from wild-type plants (Additional file 1: Figure S6).
SF3B1 mutant variants confer variable resistance to GEX1A
Next, we analyzed the responses of SGRs to GEX1A treatments. Therefore, we conducted germination and root inhibition assays at several GEX1A concentrations (Fig. 3 and Additional file 1: Figures S7 and S8). As indicated earlier, GEX1A affected rice germination at 2.5 μM and severely inhibited seed germination at 5 μM (Additional file 1: Figure S1). Here, we observed that the germination of all SGRs was not affected at 2.5 μM GEX1A but the germination of SGS1 and the wild-type rice seeds was inhibited (Fig. 3 and Additional file 1: Figures S7 and S8). When we increased the GEX1A concentration to 10 μM, the germination of all SGRs was inhibited except SGR4, which was completely unaffected by GEX1A (Fig. 3 and Additional file 1: Figures S7 and S8). Similar results were observed for the root inhibition assay (Additional file 1: Figure S8), where all the SGRs showed resistance to 0.3 and 0.5 μM GEX1A. However, the root growth of SGS1 and wild-type seedlings completely ceased at 0.3 μM of GEX1A (Additional file 1: Figure S8). Intriguingly, root growth was inhibited for all SGRs at 1 μM GEX1A, except SGR4, which did not exhibit root growth inhibition at any concentration. These data indicate that the SF3B1 variants exhibit variable degrees of resistance to GEX1A, with SGR4 conferring the strongest resistance.
Several reports have indicated that GEX1A, PB, and Spliceostatin A share the same pharmacophore and bind to the same site in the SF3B1 protein [23, 24]. To investigate whether GEX1A and PB share the same binding site in SF3B1, we tested whether GEX1A-resistant mutants were also resistant to PB by conducting primary root growth assays on media supplemented with PB (Additional file 1: Figure S9). SF3B1 mutant variants were resistant to PB (Additional file 1: Figure S9). The in-frame sensitive GEX1A mutant was also sensitive to PB. These data provide compelling evidence that these splicing drugs target the same SF3B1 domain.
Structural basis of SGR resistance to GEX1A
Recent reports have provided extensive structural studies on mammalian SF3B1 and the SF3B1:PHF5 complex and their interaction with various splicing inhibitors [24, 25]. Since rice and mammalian SF3B1 are > 80% identical, we used these mammalian structures to study the structural basis of the resistance of the rice SF3B1 mutant variants to GEX1A and PB (Fig. 3b, Additional file 1: Figures S10 and S11). Based on these structural models, we assessed the molecular effect of six OsSF3B1 variants on interactions with PB and herboxidiene. The SGS1 variant results in the deletion of an N-terminal loop-helix element (Fig. 3b, Additional file 1: Figure S11). Although this deletion might mildly reduce protein stability, it is remote from the drug binding site and hence should not lead to drug resistance, in agreement with our experimental observations. The SGR3 K1050 variant is located at the rim of the drug binding surface and reinforces PHF5A interactions through an ionic bond to PHF5A D27 (Fig. 3b, Additional file 1: Figure S11). Deletion of K1050 perturbs the binding pocket’s stereochemistry and weakens contacts to PHF5A.
The SGR4 variant contains two mutations that are not predicted to have marked effects: K1049R is a substitution with similar physico-chemical properties, and although G1051H leads to inclusion of a much larger side chain, this change does not cause steric clashes, being exposed to the solvent (Fig. 3b, Additional file 1: Figure S11). Conversely, K1050E will lead to electrostatic repulsion with GEX1A, weakening the drug binding interactions, and additionally introduce a repulsion with PHF5A D27 and weaken the OsSF3B1:PHF5A interaction.
In the SGR6 variant (H1048Q, K1049 deletion, A1064S), H1048 and K1049 are located at the rim of the drug binding site, at the interface between OsSF3B1 and PHF5A (Fig. 3b, Additional file 1: Figure S11). The double mutation H1048Q/K1049del will significantly destabilize the drug binding pocket and affect the binding to PHF5A. The third mutation A1064S is located 14 Å away from the drug binding site, and the relatively homologous substitution with a serine would only lead to minor local structural rearrangements.
The SGR Q157 deletion and DAPDATPGIG (223–232) variants carry mutations in the disordered loop region in the N-terminal of the protein, not included in our molecular models (Additional file 1: Figure S10). We hypothesize that these deletions might affect additional interactions or regulatory roles that this region might have. Collectively, our structural analysis demonstrated that the drug-resistant variants contained driver and passenger mutations and that much of the effect was mediated through an OsSF3B1 region that is at the intersection between drug and PHF5A binding. Thus, the OsSF3B1 mutations appeared to have a double effect by destabilizing binding of both drug and PHF5A. We anticipate that these data will be useful to predict and identify drug resistance in various applications.
SGR4 mutant variant exhibits efficient splicing in GEX1A treatment
GEX1A and PB inhibit splicing and cause significant intron retention in Arabidopsis thaliana. Therefore, we wanted to test the effects of GEX1A on the seedlings of wild-type rice and SF3B1 mutant variants and examine their molecular effects on splicing repression. Subsequently, we treated 1-week-old wild-type rice seedlings and SF3B1 variants with 0.3 μM GEX1A for 6 h. After extracting the RNA, cDNA was synthesized and semi-quantitative PCR was conducted on exons flanking selected introns. We have tested several genes with alternatively spliced pre-mRNAs, including those encoding: MYB family transcription factor (LOC_Os03g55590), SKIP interacting protein 3 (LOC_Os04g37790), Oryzain beta chain precursor (LOC_Os04g57440), Glucan endo- 1,3-beta-glucosidase precursor (LOC_Os07g35350), EF hand family protein (LOC_Os08g44390), and Alpha-amylase precursor (LOC_Os08g36910), and RING finger and CHY zinc finger domain-containing protein 1 (LOC_Os10g31850) to test whether they exhibit intron retention upon GEX1A treatments. Our data show that the GEX1A treatment enhanced the intron retention events during pre-mRNA splicing of these genes in wild-type rice seedlings (Fig. 4).
Because the SGR4 variant exhibited very strong resistance to GEX1A treatment in the germination and root-length assays, we treated 1-week-old seedlings of SGR4 with 0.3 μM GEX1A for 6 h and conducted RT-PCR to determine intron retention patterns. Interestingly, our data show no intron retention events for plants carrying the SGR4 variant (Fig. 4) further corroborating our phenotyping and structural analysis data. These data validate our hypothesis and the CDE platform and indicate that GEX1A is not capable of binding to the SF3B1 SGR4 variant, resulting in proficient pre-mRNA splicing.