Cotton (Gossypium hirsutum L. cv. ‘TM-1’) seeds were imbibed in deionized water for 8 h before being allowed to germinate at 28°C for 24 h in the dark. Following germination, the seedlings were grown in soil at 22/25°C (night/day) under 12 h/12 h light/dark conditions. Ten-day-old seedlings were used for transient transformation experiments. G. hirsutum L. cv. ‘YZ-1’ was used for stable transformation.

We used the pYLCRIPSR/Cas9 multiplex binary vector system containing a plant codon optimized Cas9 gene (Ma et al., 2015). To confirm target specificity in the genome of cotton during target site selection, a BLAST search was conducted using the cotton genome database¹. The more detailed procedure was performed as previously described (Ma et al., 2015). Subsequently, the candidate sgRNA sequences were subjected to secondary structure analysis using an RNA folding platform². Overlapping PCR was conducted to amplify the targeted sequences for the sgRNAs, including the two promoters utilized in this study (AtU6-29 and or AtU3b) using the primers listed in Supplementary Table S1. PCR products containing the targeted sequences were ligated into the CRISPR/Cas9 expression cassette via Golden Gate cloning, and more details were described in a previous report (Ma et al., 2015).

CRISPR/Cas9 vectors were transferred into Agrobacterium tumefaciens (GV3101 and LBA4404) for transient and stable transformation experiments, respectively. For transient transformation, GV3101 was infiltrated into cotyledons of 10-day-old cotton seedlings using a needless syringe. The seedlings were incubated in a plant growth chamber at 25°C. After incubation for 48 h, the infiltrated cotyledons were harvested for genomic DNA isolation and PCR/RE analysis (Gao et al., 2011). Stable transformation was performed in G. hirsutum L. cv. ‘YZ-1’ with LBA4404 as previously described (Jin et al., 2005).

Genomic DNA was isolated from infiltrated cotyledons using a DNA extraction kit (Tiangen Biotech, Beijing). PCR was used to amplify a genomic fragment containing the target site and appropriate restriction enzymes used to digest PCR products in order to confirm mutations at the target site. The PCR amplicons were also cloned into a TA-cloning vector and sequenced (Sangon Biotech, Shanghai).

Total RNA was isolated from transgenic cotton lines using the Aidlab RNA extraction kit (Aidlab Biotechnologies, China). First-strand cDNA was performed from 1 μg of total RNA using the M-MLV Reverse Transcription System (Promega, United States). PCR was performed as follows: 95°C for 5 min, followed by 28 cycles of amplification (95°C for 10 s, 57°C for 30 s, and 72°C for 30 s), and a final incubation at 72°C for 7 min. The cotton gene ubiquitin7 (GhUB7, Accession: DQ116441) was amplified as internal control.

The available cotton transformation methods are technically demanding and time-consuming with an average of 10 months to produce T₀ transgenic lines (Jin et al., 2005). Given the length and difficulty involved in obtaining stable cotton transgenic lines, it was critical to develop a fast method to test the efficiency of the sgRNAs. We therefore designed a fast sgRNA validation method for our research (Figure 1). Transient expression was achieved by infiltrating 10-day-old cotton cotyledons with A. tumefaciens harboring the appropriate CRISPR/Cas9 vectors (Figure 1A). The infiltrated cotyledons were harvested 48 h later, genomic DNA isolated and a fragment containing the target site amplified by PCR. Targeted sequences had been carefully selected to contain a restriction site in the vicinity of the PAM, therefore CRISPR-induced mutagenesis would likely destroy the enzyme recognition site. PCR amplification products were digested with the appropriate restriction enzyme and analyzed by gel electrophoresis to detect the loss of the restriction site. Finally, the gene-edited products, appearing in the electrophoresis as a high molecular size band were isolated and sequenced to re-confirm the modification and determine the nature of the CRISPR/Cas9-induced mutations (Figure 1B). This method proved to be very convenient to quickly validate the efficiency of the different sgRNAs, being capable to simultaneously test 4–6 different targets in less than 1 week using transient expression in cotton cotyledons.

In order to validate the transient transformation-based sgRNA validation method we choose the elongation factor-1 protein (GhEF1) as a target (D07G1160). GhEF1 catalyzes the binding of aminoacyl-tRNAs to the A-site of the ribosome during protein synthesis. The chosen 20-bp target sequence (sgRNA1-GhEF1) was located in exon #2 of the GhEF1 gene and contained a StuI restriction site 3 bases away from the PAM for detection of mutations (Figure 2A). Transient expression with binary vectors containing expression cassettes for Cas9 alone or together with sgRNA1-GhEF1 produced PCR amplicons of approximately 0.87 kb in size (Figure 2B, lanes 1 and 2). As expected, digestion of the amplification product with StuI produced two fragments of 554 and 314 bp in the absence of target sequence sgRNA1-GhEF1 (Figure 2B, lane 3), while an extra, undigested fragment, was observed in the presence of such a sequence in the CRISPR/Cas9 cassette (Figure 2B, lanes 4–6). The undigested fragment was purified from the gel and cloned for sequence analysis. A total of 43 clones were analyzed with 63% of them harboring deletions (from 1 to 8 nucleotides in length), while 37% contained insertions (1 nucleotide) (Figure 2C). In total, 10 different mutations were detected, of which single base deletions (17/43) and single base insertions (15/43) were the most common types. Further validation of the transient transformation system was achieved using the phytoene desaturase (PDS) gene (D07G1160) (Figures 3A,B). In this case, 35 clones were sequenced revealing the presence of eight different mutation types, 57% of them deletions (from 1 to 6 nucleotides in length), and 43% insertions (1 nucleotide) (Figure 3C).

The high number of homoeologous segments present in the genome of the allotetraploid cotton results in the existence of two or more copies for most genes (Wang et al., 2014; Gao et al., 2016a). In order to obtain a functional mutant in cotton, it is imperative to edit homoeologous sequences simultaneously, and in most cases, it is possible to find highly conserved CRISPR target sequences in the homeolog genes (Gao et al., 2016a). To test whether simultaneous mutation of homoeologous genes is feasible in cotton we chose Chloroplastos alterados 1 (GhCLA1) which is involved in chloroplast development and have two homoeologous sequences (D10G1640 from the sub-genome D and A10G2292 from the sub-genome A, Figure 4A). We designed a target sequence for D10G1640 that contained a single nucleotide difference at the 12th position upstream of the PAM site in A10G2292. The results of sequencing revealed successful editing events in D10G1640 as well as A10G2292, even with the imperfect target match, once more confirming the ability of CRISPR/Cas9 to edit homeologous genes in polyploids (Wang et al., 2014). As in the previous two experiments (Figures 4B,C), the frequency of deletion mutations was higher (70.45%) than insertions (27.17%).

Combined statistical analysis of our transient expression results indicates that the most common CRISPR/Cas9-induced mutations are deletions (69.93%) ranging from 1 to 6 nucleotides, while insertions accounted for (35.25%), all of them being single-nucleotide insertions (Table 1). Only one instance of nucleotide replacement was detected among the 122 mutations analyzed.

CRISPR/Cas9 has been successfully used to create multiple mutations simultaneously (Zhang Y. et al., 2016), a feature that could prove extremely valuable for crops with lengthy and complicated transformation methods such as cotton. To validate the feasibility of creating double mutants, we used the previously described target sites for the GhEF1 and GhPDS genes (Figures 2, 3) to build two sgRNA expression cassettes (sgRNA1-GhEF1 and sgRNA1-GhPDS) and cloned both of them into a single CRISPR/Cas9 expression vector (Figure 5A). To minimize the probability of silencing due to the presence of repeated sequences in the transient expression cassette, expression of the GhPDS and GhEF1 sgRNAs was driven by the AtU6-29 and AtU3b promoters respectively. Transient expression of the dual sgRNA vector in cotton cotyledons induced mutations in both targeted genes as indicated by the presence of a ‘non-digested’ amplification product for both genes (BfaI for GhPDS and StuI for GhEF1) (Figure 5B). The undigested band was purified, cloned, and subjected to sequencing analysis (Figure 5C). The sequencing results showed the presence of multiple mutation types in both genes illustrating the suitability of the CRISPR/Cas9 system for the production of double/multiple mutants for functional genomics in cotton.

Simultaneous targeting of two sites within the same gene can improve the mutagenesis efficiency of CRISPR/Cas9 and lead to the deletion of the gene fragment contained between the two targeted sites, thus increasing efficiency and providing an easier detection method (Zhang B. et al., 2016). To study whether CRISPR/Cas9 can be used for gene fragment deletion in cotton, two sgRNAs (sgRNA2-GhPDS and sgRNA3-GhPDS) were designed targeting sequences sites within the 8th exon of the GhPDS gene (48 bp apart) (Figure 6A). The target site of sgRNA2-GhPDS contained an AvaI restriction site for convenient detection of the mutation. Transient expression experiments were conducted with a binary vector containing both sgRNA as well as the Cas9 expression cassettes in cotton cotyledons. Genomic DNA was purified, and PCR performed using a pair of primers external to the targets to amplify a genomic fragment containing both target sites. Analysis of the PCR amplicons showed that co-expression of Cas9 and the two sgRNAs resulted in the appearance of a product non-digestible with AvaI (Figure 6B). The high molecular weight bands present in lanes 4, 5, and 6 of Figure 6B were purified and re-amplified by PCR revealing the existence of two molecular species with slightly different sizes (Figure 6C; lanes 3, 4, and 5). We hypothesized that the lower molecular size amplicons were produced by deletion of the genomic fragment between the two targeted sites and were therefore purified, cloned, and sequenced. Sequence analysis of 32 clones confirmed our hypothesis showing the existence of deletion ranging from 50 to 61 nucleotides between both target sites (Figure 6D).

Our developed transient transformation system was quite efficient in elucidating the suitability of CRISPR/Cas9 potential target sites in cotton cotyledons. However, this method does not produce stably mutated plants, and it is therefore not suitable for phenotypic studies. We used A. tumefaciens-mediated hypocotyl transformation to produce stable transgenic cotton lines and further confirm the feasibility of the CRISPR/Cas9 system in cotton. The GhCLA1 was selected as a target given the easily observable photobleaching phenotype caused by its inactivation (Gao et al., 2011). Some of the calli regenerated on selection medium approximately 3 months after Agrobacterium-mediated transformation with the CRISPR/Cas9 binary construct showed an albino phenotype that was maintained during the entire regeneration process in stems, leaves and entire seedlings (Figures 7A,B). Ultimately, 43 putative transgenic T₀ cotton lines were obtained, 36 of which tested positive for the Cas9 transgene by PCR (Figure 7C and Supplementary Figure S1). Two PCR primers flanking the sgRNA3-GhCLA1 target site in the 7th exon of GhCLA1 (Figure 8A) were used to amplify the genomic fragment in the transgenic lines. Sequence analysis of the amplification products revealed the presence of mutations in 29 of the 36 positive transgenic lines (80.56%) (Supplementary Note S1). Unfortunately, the severe phenotypic effects caused by the inactivation of GhCLA1 (inhibition of chloroplast biogenesis), hindered the regeneration process and only some of the initial transgenic calli were successfully regenerated into plantlets, all of which showed photobleaching and severely stunted growth (Figure 8B). RT-PCR analysis revealed the accumulation of Cas9 transcripts in albino mutants (Figure 8C). Six regenerated albino plantlets were analyzed for the presence of CRISPR/Cas9-induced mutations in the GhCLA1 target site. All T₀ plants showed mutations in the gene with deletions (78.90%) being more frequent than insertions (21.1%), in agreement with our observations from the transient expression experiments (Figure 8D). Silencing of GhCLA1 using the tobacco rattle virus (TRV)-mediated virus-induced gene silencing (VIGS) also produced an albino phenotype, but this silencing method resulted in uneven and less intense photobleaching compared to the CRISPR/Cas9 mutated plants (Supplementary Figure S2), highlighting the superiority of the CRISPR/Cas9 system over RNAi-based silencing approaches.