The highly regenerable N4.4.2 alfalfa genotype (Badhan et al., 2014) was utilized for transformation experiments throughout this study, while the cultivar AC Blue J (provided by Dr. Surya Acharya, AAFC Lethbridge) was used for crossing. Plants were grown in square pots (10.5 cm across and 12.5 cm in height) filled with Cornell soilless potting mix (Boodley and Sheldrake, 1972) in a greenhouse with supplemental lighting providing a 16 h/8 h photoperiod, a temperature of approximately 20°C/15°C day/night, and a light intensity of approximately 150 μmol/m²/s. Unless otherwise specified, plants were watered daily (without fertilizer) from the bottom for 15 min, after which time any excess water was drained off. Due to alfalfa’s obligatory outcrossing nature, the N4.4.2 background genotype, as well as all transgenic plants generated in this study, were propagated vegetatively through rooted stem cuttings to preserve particular genotypes throughout the experiment and generate “pools” of individual genotypes for downstream investigations. In every case, rooted cuttings were transferred to pots at the same time point, and were re-grown after cutting back to approximately 5 cm at least twice prior to assessments. During all trials, pots were rotated randomly every other day to minimize variations due to microclimate effects.

BLAST searches were performed against the M. sativa cultivated alfalfa at the diploid level (CADL; v0.95 and v1.0) genome database^1,2, the Alfalfa Gene Index and Expression Atlas database³ and the tetraploid alfalfa genome sequence (Chen et al., 2020; NCBI database project PRJNA540215) using the Medicago truncatula Medtr8g005960 sequence, which encodes the SPL8 homolog in this species (Gou et al., 2018). Nucleotide and amino acid alignments were carried out on identified homologs using Geneious Prime software (Biomatters Inc., San Diego, CA, United States), while domain searches were conducted using SMART⁴. Expression of alfalfa MsSPL8 homologs was screened using the gene expression profiling tool in LegumeIP V3 (Dai et al., 2021) for M. sativa (CADL; v1.0) in different tissues.

Three separate guide RNAs (gRNAs; 20-nt) were designed immediately upstream of 5′—NGG—3′ PAM sequences (Doench et al., 2014) within a conserved region of the first exon of the four highly similar MsSPL8 alleles (chr 8.2 83866861:83868757, chr 8.3 81917086:81918958; chr 8.4 80839466:80841286 and 80871132:80872948; Figure 1A). All three gRNAs exhibited 100% complementarity with all four MsSPL8 alleles (as did all primers and probes utilized in this study). The specificity and/or potential for off-target effects of each gRNA was evaluated using the Cas-OFFinder algorithm (Bae et al., 2014). Each gRNA was inserted into the binary pKSE401 vector (Xing et al., 2014; Addgene plasmid #62202). This background vector includes a cassette in which the expression of the sgRNA is driven by the Arabidopsis U626 polymerase III promoter and another in which the expression of a Zea mays codon-optimized Cas9 coding region (derived from Streptococcus pyogenes and flanked on either side by nuclear localization signals) is driven by a constitutive partially duplicated CaMV 35S promoter. This vector is designed to elicit NHEJ-mediated edits within the target gene.

In each case, 20-nt oligonucleotides corresponding to each gRNA in both orientations were designed and then synthesized by a third-party [Integrated DNA Technologies (IDT) Inc., Coralville, IA, United States]. Each oligonucleotide possessed 4-nt additions engineered at their 5′ termini to facilitate cloning into a BsaI site within pKSE401 (5′—ATTG—3′ for forward oligonucleotides and 5′—AAAC—3′ for reverse oligonucleotides; Xing et al., 2014). Equimolar ratios of each oligonucleotide pair were then annealed by incubating at 65°C for 5 min, followed by cooling at room temperature. Each double-stranded gRNA was then ligated into pKSE401 that had been linearized with BsaI (Figure 1B). All three of the resulting vectors were subjected to sequencing to confirm their identities.

The three CRISPR/Cas9 binary vectors targeting three distinct regions of MsSPL8 alleles, as well as the empty vector (pKSE401), were introduced into Agrobacterium tumefaciens LBA4404 using electroporation and transferred separately into the alfalfa N4.4.2 genotype through Agrobacterium-mediated transformation of leaf explants as described previously (Badhan et al., 2014). Selection was carried out using 50 mg/L kanamycin for 10 days followed by 75 mg/L kanamycin for the remainder of tissue culture. Kanamycin-resistant genotypes were transferred to water in 15 mL polypropylene tubes for 3–4 days for acclimation, and subsequently transplanted to pots under greenhouse conditions with supplemental light. To confirm the presence of the transgenic cassette in kanamycin-resistant genotypes, genomic DNA was extracted from leaf tissues using the BioSprint 96 DNA Plant Kit (Qiagen Inc., Toronto, ON, United States) and the presence of each transgene was verified by PCR using Invitrogen Platinum SuperFi Green PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, United States) and primers Cas9F1 and Cas9R1 (see Supplementary Table 1 for primer sequences; 621-bp amplicon). These primers anneal to the Cas9 coding region within the pKSE401 transgenic cassette, and allow the differentiation of genotypes based on the presence/absence of the transgene. Thermal parameters included an initial denaturation step of 98°C for 30 s, 35 cycles of 98°C for 10 s, 57°C for 10 s, and 72°C for 30 s, and a final extension step of 72°C for 1 min. Positive control PCRs were conducted using GoTaq Green Master Mix (Promega Corp., Madison, WI, United States) and primers WD40F2 and WD40R2 (see Supplementary Table 1 for primer sequences; 865-bp amplicon) to ensure DNA quality. Thermal parameters included an initial denaturation step of 95°C for 2 min, 35 cycles of 95°C for 30 s, 51°C for 30 s and 72°C for 1 min, and a final extension of 72°C for 5 min. Primary transformants were labeled SPL8-gRNA1, SPL8-gRNA2 or SPL8-gRNA3 according to the gRNA used, while genotypes bearing the pKSE401 empty vector were termed EV.

Gene editing frequencies (GEF) were assessed using GEF-droplet digital PCR (GEF-ddPCR; Mock et al., 2016), which used two distinct probes that anneal within a single amplicon to quantify wild-type and NHEJ-affected alleles. Primers were designed using the PrimerQuest Tool (IDT). Primers MsSPL8Fwd2 and MsSPL8Rev2 were used to amplify a 375-bp region surrounding the potential gRNA1 and gRNA2 editing sites, while primers MsSPL8Fwd3 and MsSPL8Rev3 were used to amplify a 122-bp region surrounding the potential gRNA3 editing site (see Supplementary Table 1 for primer sequences). Locked nucleic acid (LNA) probes (Tm = 66°C, 5′ hexachloro-fluorescein [HEX] labeled) were designed to bind to predicted Cas9 cut sites in each of the three gRNA target regions as a means of detecting indels. Control probes (Tm = 68°C, 5′ fluorescein amidite [FAM] labeled) were also designed to bind to a region within each amplicon that would not be impacted by gene editing (see Supplementary Table 2 for probe sequences). All probes bore a 3′ Iowa Black fluorescent quencher. Mini-genes representing the gRNA3 target region with a single nucleotide insertion (123-bp in length) or deletion (121-bp in length) were synthesized (IDT) and utilized as positive controls to validate the sensitivity and specificity of the method in alfalfa.

Genomic DNA from leaves derived from a single shoot of all confirmed independent T₀ transgenic genotypes, as well as five independent transgenic EV genotypes, was used as template for the initial screening of all three gRNA target sites. Subsequently, genomic DNA from leaves derived from eight separate and randomly chosen primary branches of three independent T₀ SPL8-gRNA1 genotypes (SPL8-gRNA1-1, -2 and -3), as well as an EV control, respectively, was assayed as a means of examining the possibility of mosaicism in genotypes bearing a mutation at the gRNA1 cleavage site.

Assays were performed using ddPCR Supermix for Probes (no dUTP; Bio-Rad Laboratories Ltd., Mississauga, ON, Canada) following the manufacturer’s instructions. Briefly, 20 μL reactions containing 900 nM of each primer, 250 nM of the appropriate probe and 40 ng of genomic DNA were utilized for assays, and droplets were generated using the Automated Droplet Generator (Bio-Rad). Thermal cycling was performed using a C1000 Touch Thermal Cycler with 96-deep well reaction module (Bio-Rad) with parameters of 95°C for 10 min, 40 cycles of 94°C for 30 s and 60°C for 1 min, followed by enzyme deactivation at 98°C for 10 min. Droplets were then analyzed with the QX200 Droplet Reader (Bio-Rad). Analysis thresholds were established for each assay and the ratio of NHEJ-sensitive vs. NHEJ-insensitive droplets, which is based on the concentrations of events per μL, was calculated using the QuantaSoft Analysis Pro software V 1.0.596 (Bio-Rad). To quantify the relative number of edited MsSPL8 alleles compared to total alleles within each sample, GEF was calculated as the percentage of edited droplets. In the case of T₀ SPL8-gRNA1 genotypes, all subsequent experiments were carried out using plants derived from vegetative stem cuttings of the shoot exhibiting the highest level of GEF in the original T₀ genotype.

To provide further confirmation of editing in SPL8-gRNA1 genotypes, T7E1 mismatch cleavage assays were also carried out using genomic DNA isolated from SPL8-gRNA1-1, -2 and -3 genotypes (with GEF of approximately 75, 50, and 25%, respectively), as well as three independent EV controls, as template. Primers MsSPL8F2 and MsSPL8Rev2 were designed to amplify a 255-bp region of MsSPL8 encompassing the predicted editing location for gRNA1 (see Supplementary Table 1 for primer sequences). In the presence of an indel at the predicted cut site, and hence cleavage of the heteroduplex by the T7E1 enzyme, fragments of 156-bp and 99-bp in length would be generated. PCR amplification was carried out using Invitrogen Platinum SuperFi Green PCR Master Mix (Thermo Fisher Scientific) and a thermal program of 98°C for 30 s, 35 cycles of 98°C for 10 s, 55°C for 10 s and 72°C for 30 s, and a final extension at 72° for 5 min. Heteroduplexes of each PCR amplicon were formed by denaturing for 10 min at 95°C, cooling to 85°C at a ramp rate of 2°C/s, and then cooling to 25°C at a ramp rate of 0.3°C/s before digestion with 10 U of T7EI (New England Biolabs, Whitby, ON, Canada) at 37°C for 15 min. Digested products were separated and visualized on a 2% agarose gel.

T₀ transgenic SPL8-gRNA1-1, -2 and -3 genotypes (exhibiting GEF of approximately 75, 50, and 25%, respectively) were examined to confirm on-target mutations, and assessed for possible off-target mutations, using Sanger sequencing. To determine precise editing events at the gRNA1 target site, a 1.4-kb region of MsSPL8 was amplified from genomic DNA derived from each genotype (including EV controls) using Invitrogen Platinum SuperFi Green PCR Master Mix (Thermo Fisher Scientific) and primers MsSPL8F1 and MsSPL8R2 (see Supplementary Table 1 for primer sequences). The amplicons were subsequently cloned into the pGEM-T-Easy vector (Promega), and clones from 21–33 colonies were subjected to Sanger sequencing (Génome Québec Centre D’Expertise et de Service, Montreal, QC, Canada).

In addition, the three most likely potential off-target sites were also chosen for evaluation, possessing two single nucleotide mismatches and a 2-nt RNA bulge (two sites), or three single nucleotide mismatches and a 1-nt RNA bulge (one site; Supplementary Figure 1) based on findings obtained using the Cas-Offinder program. In each case, an approximately 700-bp region spanning the potential off-target site was amplified from the same three transgenic SPL8-gRNA1 genotypes, as well as untransformed N4.4.2 and an EV genotype using primers MsOffTarget1FWD1 and MsOffTtarget1REV1, MsOffTtarget2FWD1 and MsOfTftarget2REV1, and MsOffTarget3FWD1 and MsOffTarget3REV1, respectively (see Supplementary Table 1 for primer sequences). The resulting amplicons were then cloned as described above and clones from 8–12 colonies from each genotype and for each possible off-target site were sequenced to assess for mutations.

For evaluation of plant height, branch number and internode length, two separate trials were carried out 30 days (pre-flowering) and 60 days (post-flowering) following the second cut. For the trial carried out 30 days following cutting, five biological replicates of well-watered SPL8-gRNA1-1 (∼75% GEF) and SPL8-gRNA1-2 (∼50% GEF), respectively, as well as five biological replicates for each of three independent EV control genotypes (15 biological replicates total), were analyzed in each case. For the trial carried out 60 days following cutting, ten biological replicates of well-watered SPL8-gRNA1-1 and -2 genotypes, respectively, as well as ten biological replicates for each of three independent EV control genotypes (30 biological replicates total), were assessed. Branch number was determined by counting the total number of branches per plant, while the longest shoot of each plant was utilized to measure plant height and the length of the longest internode.

The onset of flowering was recorded when petals first began to emerge from flower buds following the second cut. Subsequent to flowering (60 days following cutting), all aboveground biomass was removed and weighed to determine fresh weight (FW), and was then dried at 65°C for 1 week and weighed to determine dry weight (DW). Root length was evaluated by measuring the longest primary root, and root DW was resolved following drying at 65°C for 1 week.

Leaf length, width and area were measured using the middle leaflet of the third fully expanded trifoliate leaf from the shoot tip. Measurements were made 30 days following the second cut using five biological replicates (derived from vegetative stem cuttings) of each SPL8-gRNA1 genotype, and five biological replicates from each of three independent EV transformants, with three leaves assessed per plant. Leaf area was determined using the Petiole plant leaf area meter app (version 2.0.1⁵) and leaf dry weight was established after drying at 80°C for 24 h. Specific leaf area (SLA) was calculated by dividing the leaf area by dry weight in each case.

Prior to the commencement of drought treatment, a volumetric soil moisture content of 50% was established in each pot using a ML3 ThetaKit soil moisture meter with 6 cm probes (Hoskin Scientific Ltd., Burnaby, BC, Canada), after which time water was withheld and volumetric soil water contents were measured daily until they reached approximately 1% (approximately 2–3 weeks). Assessments of water deficit symptoms were carried out on the basis of soil moisture contents rather than days following the withholding of water to minimize possible effects due to possible variation in plant water uptake. Subsequently, plants were re-watered normally for approximately 2 weeks and survival was assessed. Furthermore, volumetric soil moisture content was noted at the first instance of stress symptoms, which initially included the appearance of wilted or drooping leaves and shoots, in both edited and EV genotypes. Well-watered plants were utilized as controls. Six biological replicates of SPL8-gRNA1-1 (∼75% GEF) and SPL8-gRNA1-2 (∼50% GEF) genotypes, as well as three independent EV control genotypes (18 biological replicates total), were utilized in each experimental set. The withholding of water was initiated 14 days following the second cut.

Plant water status was estimated by measuring the relative water contents (RWC) of third fully expanded trifoliate leaves from six biological replicates, with two leaves assessed per replicate plant, of SPL8-gRNA1-1 and -2 genotypes, respectively, as well as 18 biological replicates of EV controls (six biological replicates from three independent transformants). Measurements were carried out as described previously (Barrs and Weatherley, 1962) when soil moisture contents reached approximately 50% (well-watered), 12–16% (moderate drought) and 7% (severe drought). In brief, the fresh weight (FW) of trifoliate leaves was documented immediately following detachment, turgid weights (TW) were determined after submersing petioles in water in an enclosed 1.5 mL microcentrifuge tube for 4 h, and dry weights (DW) were resolved by drying turgid leaves at 80°C overnight. Relative water content (%) was then calculated as [(FW-DW)/(TW-DW)] × 100.

Detached leaf water loss assays were carried out essentially as described previously (Singer et al., 2021b) using fully expanded trifoliate leaves (third from the shoot tip) from eight biological replicate plants of each SPL8-gRNA1 genotype (two leaves per plant; sixteen replicates total for each) and seventeen biological replicate control plants (derived from three independent transformants with two leaves sampled per plant; thirty-four replicates total) under well-watered conditions. Assays were carried out by weighing leaves immediately upon harvest (W_initial), placing them on a benchtop, and then weighing every 30 min for 240 min. The rate of water loss was calculated as (W_initial – W_{atparticulartime}/W_initial) × 100.

Light-saturated photosynthetic rate (Asat), stomatal conductance (g_s), and transpiration rate (E) were evaluated using a LI-6800 (LI-COR Inc., Lincoln, NE, United States). The middle leaflet of a third fully expanded trifoliate leaf was used for measurements on four biological replicate plants of each SPL8-gRNA1 genotype, and four biological replicate plants of each of three EV control genotypes (twelve replicates total). All assessments were conducted between 10:30 am and 12:00 pm under greenhouse conditions. Measurements were made on one set of plants grown under well-watered conditions (soil moisture content of approximately 50%) and another subjected to drought treatment (average soil moisture content of approximately 11%). Within the chamber, light intensity was held at 1,500 μmol m^–2 s^–1, relative humidity at 50%, CO₂ level at 410 μmol CO₂/mol air, and heat exchanger temperature was set to 25°C. Leaflets were allowed to stabilize for 1.5 min within the chamber prior to measurement. All measurements were adjusted for leaf area, which was determined using the Petiole Pro plant leaf area meter app (version 1.4.15⁶).

Due to the outcrossing nature of alfalfa, we carried out manual reciprocal crosses between SPL8-gRNA1-1 (∼75% GEF) and the cultivar AC Blue J, as well as between SPL8-gRNA1-2 (∼50% GEF) and AC Blue J, using non-emasculated flowers. In each case, four distinct genotypes of AC Blue J were used along with eight biological replicates of each mutant genotype (derived from vegetative stem cuttings). A total of 32–64 seeds from each cross were sown and grown in the greenhouse as described in a previous section. Genomic DNA was harvested from a single leaf of each resulting F₁ plant and PCR was conducted as described for T₀ plants to screen for the presence of the transgene. T7E1 and ddPCR assays were carried out to assess editing frequencies at the gRNA1 target site using genomic DNA derived from F₁ genotypes lacking a transgene as described previously for T₀ plants.

Statistical differences between the means of SPL8-gRNA1 genotypes and EV controls were determined using student’s t-tests assuming unequal variance. Means were considered significantly different at P ≤ 0.05.

As a means of eliciting mutations in alfalfa MsSPL8, three gRNAs were designed to simultaneously target four MsSPL8 alleles using CRISPR/Cas9 technology. These four alleles corresponded to MSAD_231217 and MSAD_231229 in the M. sativa genome sequence at the diploid level (CADL), or chr 8.2 83866861:83868757, chr 8.3 81917086:81918958; chr 8.4 80839466:80841286 and 80871132:80872948 from a tetraploid reference genome sequence (Chen et al., 2020). These four alleles exhibited 98.0% identity within their coding regions at the nucleotide level and 98.2% identity at the amino acid level. Using the freely available LegumeIP database and the M. sativa CADL genome (v1.0), the expression profiles of MSAD_231217 and MSAD_231229 in a selection of tissue types indicated that both transcripts were highly expressed in flowers, and to a lesser extent in elongated stems, while only low levels of expression were observed in roots, nodules and leaves (Supplementary Figure 2). The coding regions of the MsSPL8 alleles possess three exons and two introns, and the predicted proteins incorporate a single SBP domain that is identical among MsSPL8 alleles and also to that of M. truncatula Medtr8g005960. All three gRNAs were designed to target regions of the coding sequence within exon 1, and exhibited 100% complementarity to all four MsSPL8 alleles. While gRNA1 and gRNA2 targeted regions upstream of the SBP domain, gRNA3 targeted a region within this domain (Figure 1A).

In order to target MsSPL8 alleles using CRISPR/Cas9 in alfalfa, three pKSE401-based binary vectors bearing distinct gRNAs targeting all four MsSPL8 alleles (SPL8-gRNA1, SPL8-gRNA2 and SPL8-gRNA3; Figure 1B) were generated. Each vector, along with an EV control consisting of pKSE401 without any gRNA sequence, was introduced into alfalfa, and PCR assessment of kanamycin-resistant genotypes provided confirmation for 26, 12, 12 and 17 independent SPL8-gRNA1, SPL8-gRNA2, SPL8-gRNA3 and EV transgenic genotypes, respectively.

In order to quantitatively assess the frequency of editing in our transgenic genotypes, GEF-ddPCR was carried out using genomic DNA from each of the confirmed independent transgenic SPL8-gRNA genotypes, as well as five independent EV negative control genotypes (Figures 2A,B). Of the transgenic genotypes assessed, 10 of 26 (38.5%) SPL8-gRNA1 genotypes, 5 of 12 (41.7%) SPL8-gRNA2 genotypes, and 2 of 12 (16.7%) SPL8-gRNA3 genotypes exhibited GEF ≥ 0.5%, which was chosen as the threshold indicating the presence of at least a small number of edits in a sample. None of the five EV controls exhibited a GEF above 0.2%. In the case of SPL8-gRNA1, SPL8-gRNA2 and SPL8-gRNA3 genotypes, GEF of up to 76.5, 9.5, and 1.3%, respectively, were observed (Figure 2B). Since gRNA1 appeared to exhibit superior editing efficiency compared to gRNA2 and gRNA3, we selected three independent T₀ transgenic SPL8-gRNA1 genotypes with relatively high GEF [labeled SPL8-gRNA1-1 (76.5% GEF), SPL8-gRNA1-2 (54.7% GEF), and SPL8-gRNA1-3 (26.1% GEF)] for subsequent experiments.

Since there are four MsSPL8 alleles in alfalfa, a GEF of less than approximately 25% indicates probable mosaicism for editing events within a single plant. To evaluate whether edited genotypes exhibiting a GEF above 25% were also mosaics, genomic DNA was isolated from leaves derived from eight distinct branches of SPL8-gRNA1-1, SPL8-gRNA1-2 and SPL8-gRNA1-3 genotypes, respectively, and subjected to GEF-ddPCR. While SPL8-gRNA1-1 exhibited a rather wide range of GEF among branches (between 24.6% and 76.2%), SPL8-gRNA1-2 and -3 displayed much narrower ranges (between 50.4–57.5% and 25.7–29.2%, respectively) (Figure 2C). In an attempt to minimize variation in GEF in the SPL8-gRNA1-1 genotype, all subsequent experiments were carried out using plants derived from vegetative stem cuttings of the branch exhibiting a 76.2% GEF in the original T₀ transformant.

To provide further confirmation of NHEJ-mediated gene editing in MsSPL8 alleles within SPL8-gRNA1 genotypes, we carried out T7E1 mismatch cleavage assays on genomic DNA isolated from SPL8-gRNA1-1, -2 and -3 plants, as well as three independent EV controls. Single bands of approximately 250-bp were observed in all EV controls, which indicates that no indels were present in any MsSPL8 allele within the amplified region. Conversely, all three SPL8-gRNA1 genotypes assessed exhibited multiple bands, including uncut amplicon of approximately 250-bp, as well as fragments consistent with the presence of an indel at the gRNA1 target site (approximately 150-bp and 100-bp; Figure 3A), which confirms the presence of edits at the target site in these genotypes.

To identify the precise genetic changes that occurred at the gRNA1 target site in edited alfalfa genotypes, we subsequently cloned and sequenced a region encompassing the gRNA1 target site from SPL8-gRNA1-1 (76.4% GEF), SPL8-gRNA1-2 (57.5% GEF) and SPL8-gRNA1-3 (29.2% GEF) genotypes, as well as untransformed N4.4.2. Clones from between 21 and 33 colonies were sequenced in every case. No clone derived from N4.4.2 exhibited any genetic variation within the MsSPL8 gRNA1 target region. In contrast, in the case of SPL8-gRNA1-1, we detected three distinct mutations 2- to 4-bp upstream of the gRNA1-associated PAM (a 6-bp deletion, a 9-bp deletion and a 1-bp deletion) along with wild-type sequence. In SPL8-gRNA1-2, we observed two distinct mutations at this same site (a 1-bp insertion and a 5-bp deletion) along with wild-type sequence. In SPL8-gRNA1-3, a single mutation at the predicted gRNA1 cut site (6-bp deletion) and wild-type sequence were observed (Figure 3B). While the 6-bp and 9-bp deletions would result in missing SR or RSP amino acid residues just upstream of the SBP domain, the 1-bp deletion, 1-bp insertion and 5-bp deletion would lead to a frameshift and the presence of premature stop codons. All of these stop codons are located either upstream of, or within, the SBP domain (Figure 3C).

To assess for possible off-target effects in the SPL8-gRNA1 alfalfa genotypes, we initially screened the M. sativa CADL v0.95 genome sequence with that of gRNA1 using the Cas-Offinder program. No matches were identified with a 1-nt mismatch and a 0-, 1- or 2-nt RNA bulge, or with two mismatches and 0- or 1-nt RNA bulge. However, possible off-target sites were identified with two single nucleotide mismatches and a 2-nt RNA bulge, and three single nucleotide mismatches and a 1-nt RNA bulge. To confirm that off-target mutations had not occurred in our three selected SPL8-gRNA1 genotypes, we cloned and sequenced three regions that spanned possible off-target sites for this gRNA (Supplementary Figure 1) from SPL8-gRNA1-1, -2 and -3, as well as untransformed N4.4.2 controls. No genetic variation was noted within the three possible off-target regions in any of the 8–12 clones sequenced for each genotype.

Due to a lack of a frameshift mutation in the SPL8-gRNA1–3 genotype, all further experiments focused on SPL8-gRNA1-1 and SPL8-gRNA1-2 genotypes.

To determine whether SPL8-gRNA1-1 and SPL8-gRNA1-2 genotypes differed from EV plants in terms of their morphology, we assessed various aboveground and belowground characteristics under well-watered conditions. Although no changes in the number of branches were observed prior to flowering, both SPL8-gRNA1-1 and SPL8-gRNA1-2 genotypes displayed decreases in internode length (35.7% and 27.9% relative decreases, respectively), and SPL8-gRNA1-1 exhibited a significant reduction in plant height (26.1% relative decrease), compared to EV controls (Figures 4A,B). Similarly, post-flowering, significant reductions in plant height and internode length (Figures 4A,C), shoot fresh weight (FW) and dry weight (DW) (Figure 4D), and root length and DW (Supplementary Figure 3) were noted in SPL8-gRNA1-1, but not SPL8-gRNA1-2, compared to EV controls. As was the case pre-flowering, no significant differences were observed between either SPL8-gRNA1 genotype and EV controls with respect to branch number post-flowering (Figure 4C). Taken together, these results suggest that shoot and root growth phenotypes in SPL8-gRNA1 genotypes may be gene dosage-dependent, with the genotype exhibiting an approximately 75% GEF in MsSPL8 alleles exhibiting more severe alterations than that with an approximately 50% GEF.

In addition to growth phenotypes, both edited SPL8-gRNA1 genotypes exhibited significant reductions in the time required for flowering onset compared to EV controls, with SPL8-gRNA1-1 and -2 flowering on average 42.6 and 41.6 days after cutting and EV genotypes flowering on average 48.5 days after cutting (Figures 4A,E). Furthermore, leaf characteristics were consistently altered in both SPL8-gRNA1-1 and SPL8-gRNA1-2 genotypes compared to EV controls (Figure 5A). Specifically, SPL8-gRNA1-1 and SPL8-gRNA1-2 exhibited significant reductions in leaf width (27.1% and 25.9% relative decreases, respectively; Figure 5B), leaf length (6.5% and 8.8% relative decreases, respectively; Figure 5C) and leaf area (29.2% and 29.4% relative decreases, respectively; Figure 5D) compared to EV genotypes.

In order to evaluate whether genotypes with mutations in MsSPL8 alleles were able to better withstand drought stress than EV genotypes, we withheld water from plants with an initial soil moisture content of approximately 50% and assessed the soil moisture content at which they began to exhibit signs of stress, including wilted or drooping leaves and shoots (Figure 6A). Both SPL8-gRNA1-1 and SPL8-gRNA1-2 genotypes wilted at a significantly lower soil moisture content (an average of 6.2% and 5.6% volumetric soil moisture content, respectively) than EV controls (an average of 9.1% volumetric soil moisture) (Figures 6A,B). In line with this, while leaf RWC did not differ between genotypes under well-watered conditions, both SPL8-gRNA1 genotypes exhibited significant increases in leaf RWC under moderate (approximately 14% volumetric soil moisture content) and severe (approximately 7% volumetric soil moisture content) drought conditions (Figure 6C), which indicates that the edited genotypes were better able to retain water in their leaves in the face of water deficit. This was also apparent in the physical appearance of plants at soil moisture levels of approximately 8%, at which point EV control plants consistently displayed obvious water-deficit symptoms, while SPL8-gRNA1 plants remained visibly unaffected (Figure 6A). Furthermore, while no significant difference was noted in light-saturated photosynthetic rate between EV and SPL8-gRNA1 genotypes under well-watered conditions, only EV genotypes displayed a significant reduction in light-saturated photosynthetic rate under drought compared to control conditions (Figure 6D). This suggests that drought treatment negatively impacted photosynthesis in EV genotypes, but not SPL8-gRNA1 genotypes. Finally, when plants were allowed to reach a soil moisture content of 1% and were then re-watered for approximately 2 weeks, only 50% of EV genotypes survived, while all SPL8-gRNA1 genotypes flourished, exhibiting necrosis in only a small number of leaves (Figures 6E,F), which further illustrates the enhanced drought tolerance of the edited genotypes in this study.

To determine whether differences in water use/loss were involved in the superior drought tolerance of SPL8-gRNA1 genotypes, we first assessed the rate at which soil moisture was lost from pots bearing different genotypes, and found no significant differences among genotypes (Figure 7A). Similarly, while SPL8-gRNA1-1 plants exhibited a significant decrease in detached leaf water loss rates compared to EV plants at early time points, no significant differences were observed at subsequent time points, or between SPL8-gRNA1-2 and EV plants (Figure 7B). Furthermore, while no significant differences were apparent in stomatal conductance or transpiration rate between SPL8-gRNA1 and EV genotypes either under well-watered or drought conditions, EV genotypes, but not SPL8-gRNA1 genotypes, exhibited significant reductions in both stomatal conductance and transpiration rate under drought compared to control conditions (Figures 7C,D). Taken together, these results suggest that unlike EV genotypes, SPL8-gRNA1 genotypes did not respond to the same level of drought stress by reducing stomatal conductance and transpiration rate, and that traits related to water use and/or water loss were not solely responsible for the improvement of SPL8-gRNA1 genotypes in terms of their ability to withstand water deficit.

Heritability of the MsSPL8 edits was evaluated by crossing SPL8-gRNA1-1 and SPL8-gRNA1-2 genotypes, respectively, with several genotypes of the cultivar AC Blue J. A total of 202 seedlings were established from the crosses, and PCR analysis demonstrated that 23.8% of them lacked a transgene. Furthermore, T7E1 and ddPCR assays indicated that of the transgene-negative plants, 77.1% bore an edit in at least one MsSPL8 allele (Table 1). None of the AC Blue J plants assessed exhibited a transgene or editing of this gene in any case, and GEF of 74.5% and 55.7% were confirmed for SPL8-gRNA1-1 and SPL8-gRNA1-2 genotypes, respectively (Table 1 and Supplementary Figure 4).

Intriguingly, with the exception of one genotype that exhibited a GEF of 34%, all transgene-negative seedlings fell into one of five GEF classes: 0% (0–0.6% GEF), 25% (24.0–28.7% GEF), 50% (48.7–53.2% GEF), 75% (74.5%) and 100% (99.9%). In both instances where SPL8-gRNA1 genotypes acted as the recipient for AC Blue J pollen, all transgene-negative plants possessed an edit in MsSPL8. Conversely, where AC Blue J was utilized as the recipient for SPL8-gRNA1-1 and SPL8-gRNA1-2 pollen, respectively, 40.0% and 30.0% of transgene-negative plants bore no edits (Table 1 and Supplementary Figure 4). Of the transgene-negative edited genotypes overall, 29.7% bore no edit, 59.5% possessed a GEF of approximately 25%, 35.1% possessed a GEF of approximately 50%, 2.7% possessed a GEF of approximately 75% and 2.7% possessed a GEF of 100% (Table 1 and Supplementary Figure 4). The genotypes with GEF of approximately 75% and 100% were derived from crosses where AC Blue J was utilized to pollinate SPL8-gRNA1-1 and SPL8-gRNA1-2, respectively, which suggests that a low level of selfing likely took place in the SPL8-gRNA1 genotypes. Taken together, these results confirm that edited genotypes lacking a transgene can be achieved in the next generation, and that editing was present in a range of allelic combinations.