The experiments were performed using a near-isogenic line of the semi-winter common wheat variety ‘Chinese Spring’ that is homozygous for recessive vrn-A1 and vrn-B1 alleles and homozygous for the dominant Vrn-D1 allele. Donor wheat plants were grown in a temperature-controlled greenhouse (18–25°C) under a 16 h photoperiod with additional lighting (up to 150 lmol/m⁻² s⁻¹).

The fragment of the promoter region of the vrn-A1 gene was amplified from the genomic DNA of ‘Chinese Spring’ using the primers listed in Supplementary Table S1 . The PCR products were digested using EcoRI and KpnI (Thermo Fisher Scientific, USA), cloned at the corresponding sites in the pUC18 vector, and sequenced. The sequence of the vrn-A1 promoter region is shown in Supplementary Figure S1 . The potential targets were selected for the design of candidate sgRNAs in the two regions of the vrn-A1 promoter. They include the VRN box region (yellow letters) and 59 bp sequence located between −62 and −121 bp (green letters) ( Supplementary Figure S1 ). Potential targets for sgRNAs were selected using available algorithms and programs (Doench et al., 2016).

The cleavage templates were made by adding nucleotides sequences of TTC TAA TAC GAC TCA CTA TAG and GTT TTA GAG CTA GAA ATA GC to the 5’-end the 3’-end of each predicted sgRNA sequence (without PAM), mixed with primer listed in Supplementary Table S1 , and amplified. The RNA of 14 sgRNAs was then synthesized from amplified template DNA using HiScribe T7 Quick High Yield RNA Synthesis Kit (New England Biolabs) according to the manufacturer’s instructions. Synthesized sgRNA (30 ng) and 30 ng of Cas9 nuclease (NEB) were mixed in a volume of 30 µL and incubated for 10 min at room temperature to allow the assembly of RNP complexes. The mix was then incubated for 2 h at 37 °C in NEB reaction buffer with 3 nM of the plasmid (linearized with SspI) carrying the target sequence. The reaction was stopped by adding 1 µL of proteinase K (30 mg/mL), and the digestion products were separated on a 1.2% agarose gel.

The Cas9 expression vector pGCB ( Figure 1B ) is based on the psGFP-BAR plasmid harboring the green fluorescent protein gene (GFP) driven by the rice Act1 promoter and the BAR gene (phosphinothricin acetyltransferase) under the maize Ubi1 promoter (Richards et al., 2001). A codon-optimized sequence of Cas9 driven by the maize Ubi1 promoter was cloned from the pTaCas9 plasmid (the plasmid was a gift from Daniel Voytas) (Čermák et al., 2017; Addgene plasmid # 91169) and assembled into the psGFP-BAR between the GFP and BAR genes using standard cloning methods.

To construct the target-specific sgRNA vectors, the 20 bp oligos (vr-30, vr-31, and vr-34) were synthesized and cloned into the BpiI site of the plasmid pU6-gRNA (Addgene plasmid #53062). The successful construction of the three vectors (pgRNA-VRNA1#30, pgRNA-VRNA1#31, and pgRNA-VRNA1#34) harboring the corresponding candidate sgRNA sequence under the U6 promoter was confirmed by Sanger sequencing, and they were separately used for co-transformation with pGCB (1:1, v/v).

The biolistic-mediated transformation was performed with the help of a particle inflow gun using 7–15-day-old embryogenic calli induced from isolated immature embryos as earlier described (Miroshnichenko et al., 2020). To avoid the formation of escapes, a dual selection strategy was used that is based on the continuous visual selection of transgenic tissues/plants displaying GFP expression on a phosphinotricin-enriched medium.

Genomic DNA was isolated from the leaf tissue of T0 transgenic plants using the cetyltrimethylammonium bromide (CTAB) method. The integration of Cas9 and sgRNA cassettes was examined by PCR using specific primers listed in Supplementary Table S1 . For assessing mutagenesis, the genomic regions surrounding the target sites of vrn-A1, vrn-B1, and vrn-D1 promoters were PCR-amplified using the corresponding primers. The PCR amplicons were sequenced with an Illumina platform through a commercial service provided by Evrogen JSC (Russia). To identify cis-regulatory elements that could be potentially disrupted by induced mutations, the relevant part of the vrn-A1 gene promoter was analyzed using the PLACE (Higo et al., 1999; http://www.dna.affrc.go.jp/PLACE/) and PlantCARE databases (Lescot et al., 2002; http://bioinformatics.psb.ugent.be/webtools/plantcare/html/).

To determine the days to heading, the experimental plants (T3/M3 homozygous populations) were grown in a photoperiod-controlled glasshouse without vernalization treatment at 16-18°C/19-22°C (night/day) under a long photoperiod (16 h light). Plants were grown in 15 cm pots, with three plants per pot, using a randomized design and 6 to 24 replicates (pots) depending on seed availability. Heading time for each plant was recorded when the primary spike was fully emerged from the flag leaf sheath. Flowering time was recorded when the first anthers were visible on the primary spike.

Three GE lines carrying various deletions on the promoter were tested in climate chambers conditioned with long days during the whole life cycle. Two photoperiod regimes were used: 18 h for light at 22 ± 0.5°C, and 22 h for light at 17 ± 0.5°C). Five replicates (1.5 L pots) with five plants per pot were grown using a randomized design in a ‘Fitotron’ climate chamber (Weiss Technik, Germany).

For the vernalization experiments, M4 plants were grown in a greenhouse at 18-20°C/22-24°C (night/day) until the appearance of the third leaf (15 days) and were then transferred into a growth chamber for 15, 30, and 45 days and vernalized at 4-6 °C under a 12 h day length regime. A proportion of plants was left in the greenhouse as the unvernalized control. After this period, both vernalized and unvernalized plants were grown in a greenhouse at 18-20°C/22-24°C (night/day) under a long photoperiod (16 h light).

Digital phenotyping was carried out using the Trait finder system equipped with two PlantEye 3D scanners (Phenospex, Netherlands) by scanning the surface area of experimental plants with a resolution of less than 1 mm³ and measuring the volume of digital biomass and the reflection of light in several spectra bands including blue (460-485 nm), green (530-540 nm), red (620-645 nm), and near-infrared (720-750 nm) spectral regions. To detect potential variations in the spectral properties of wheat plants, the normalized difference vegetation index (NDVI) and plant senescence reflectance index (PSRI) were measured from the seedling to tillering stage of plants based on four repeated measurements of individual pots at each time point.

Total RNA was extracted from frozen plant tissues using TRIzol RNA Isolation Reagent (Thermo Fisher Scientific). The middle part of the fifth leaf was used for analysis. The cDNA was reverse-transcribed from 2 μg of total RNA using RevertAid Reverse Transcriptase (Thermo Scientific). qPCR was performed using SYBR Green qPCR SuperMix (Thermo Fisher Scientific) on a QuantStudio™ 5 Real-Time PCR Cycler (Thermo Fisher Scientific) following the manufacturer’s instructions. The wheat TaWIN1 gene (Tenea et al., 2011) was selected as the internal control. The qPCR analyses were performed on three technical replicates and analyzed using QuantStudio TM Design and Analysis (Applied Biosystems, Thermo Fisher Scientific). The primers used for qPCR are listed in Supplementary Table S1 .

Grain protein and starch content were measured by infrared reflectance spectroscopy using InfraLUM FT-12 analyzer (Lumex, Russia).

To determine statistical significance levels, the ANOVA procedure followed by Turkey’s multiple range test was performed for comparison of heading time and relative expression. Significant differences in the NDVI, PSRI, and grain content between lines were determined via t-test with p<0.05 as a threshold.

Two promoter regions of the recessive vrn-A1 gene from the semi-winter hexaploid wheat ‘Chinese Spring’ were chosen as targets for CRISPR/Cas9-mediated modification. The first region of 18 bp included the VRN box sequence ( Supplementary Figure S1 ). The second region is a 59 bp conserved fragment whose modification in common wheat varieties has not been previously described ( Supplementary Figure S1 ) but has been reported for some spring accessions of T. timopheevii, T. dicoccum, and T. compactum. According to the algorithm of CRISPR/Cas9 target sequence design (20 bp upstream of the NGG PAM motif), we found five potential targets in the fragment containing the VRN box; nine sites were discovered between nucleotide positions −62 and −121 bp of the promoter ( Supplementary Figure S2 and Supplementary Table S2 ).

To eliminate low-potential targets, we tested the ability of predicted sgRNAs to guide the Cas9 protein to cleave the target sequences via an in vitro cleavage assay. It turned out that the designed ribonucleoprotein complexes corresponding to the target sequences of the VRN box have a very low ability to cleave the DNA template (0-10%). The rather high cleavage efficiency (equal to or higher than 50%) was demonstrated by three synthesized sgRNA-Cas9 complexes corresponding to sequences vr-30, vr-31, and vr-34 located in the second target region of the promoter ( Figure 1A and Supplementary Figure S3 ). The sgRNAs mentioned above as being the most active were used to create genetic vectors to produce wheat plants with novel mutations. Each target sequence of vr-30, vr-31, or vr-34 was synthesized and fused to the default sgRNA scaffold under the control of the U6 promoter ( Figure 1B ) in the pU6-gRNA vector (Shan et al., 2013).

Three uniform plasmids carrying one of the sgRNAs were independently co-transformed with Cas9-expression cassette pGCB ( Figure 1B ) into ‘Chinese Spring’ embryogenic calli using a particle inflow gun. Following the dual selection protocol (Miroshnichenko et al., 2020) we obtained 23 independent T0 transgenic plants carrying both the Cas9 gene and the sgRNA sequence ( Supplementary Table S3 and Supplementary Figure S4 ). All Cas9+sgRNA positive plants, including nine T0 plants carrying vr-30, eight T0 plants carrying vr-31, and six T0 plants carrying vr-34, were screened for indels. The target region of the vrn-A1 promoter was amplified from the genomic DNA of T0 transgenic plants, and the PCR products were Sanger-sequenced to detect possible alterations. According to the analysis, independent events transformed with vr-30 or vr-34 sgRNA did not carry any modification of the target sequence. In contrast, the DNA sequencing charts of independent transgenic events transformed with pgRNA-VRNA#31 exhibited new peaks of bases, specifically 3–4 bp upstream of the PAM site ( Supplementary Figure S5 ). In total, we obtained 4 primary mutant plants from 603 explants, and the efficiency of CRISPR/Cas9 editing was 0.7% ( Supplementary Table S3 ).

Alignment of the PCR amplicons against the reference DNA sequence showed that two plants were monoallelic mutants, with each carrying a 1 bp insertion, and one plant was a biallelic mutant with a 1 bp deletion and a 1 bp insertion ( Figure 2 ). Another mutated plant demonstrated monoallelic disruption of the NGG-PAM sequence due to the deletion of 4 nucleotides. The algorithm that we used to find targets for introducing mutations allows the selection of the most appropriate sequences, minimizing the off-target effects of CRISPR/Cas9 manipulations (Doench et al., 2016). A BLAST search of the reference genome sequence (genome assembly release 6.05.21) of the ‘Chinese Spring’ wheat cultivar (IWGSC CS RefSeq v2.1, GCF_018294505.1) showed that there were no perfectly matched targets with complete similarity to the vr-31 sgRNA sequence in wheat. Only one potential non-target site with a single nucleotide mismatch was found. It was located in the promoter sequence of the homeologous vrn-B1 gene. Sanger sequencing of the corresponding amplicons produced from the B subgenome of four mutated plants #1, #2, #7, and #10 showed no changes in the promoter region of homeologous vrn-B1 gene ( Supplementary Figure S6 ). To confirm the specificity of vrn-A1 editing, the promoter region of the Vrn-D1 gene, which bears some similarity to the vr-31 sgRNA sequence, was also sequenced, but no off-target events were found ( Supplementary Figure S6 ).

Inheritance of mutant alleles was further investigated in M1 plants through the production of homozygous transgene-free populations for heading time evaluation. The segregation of transgenes was tracked based on a PCR assay for the presence of GFP (element of pGCB) and sgRNA scaffold region (element of pgRNA-VRN#31) sequences, and only plants lacking both transgenic sequences were accepted as being transgene-free. The proportion of transgene-free plants ranged from 16% to 78% in the progeny of three primary mutated events ( Supplementary Figures S8–S10 ). Due to the limited number of seeds produced by the 31-1 primary plant, it was not possible to identify transgene-free progeny, as all germinated T1 plants inherited the Cas9 sequence ( Supplementary Figure S7 and Supplementary Table S4 ).

The mutations detected in 31-1, 31-2, and 31-7 primary plants were transmitted to the T1 generation without the occurrence of additional editing. The expected segregation pattern for monoallelic and biallelic mutations detected in 31-2 and 31-7 primary plants easily conforms to Mendelian distribution (1 (homozygous for mutation 1):2 (biallelic):1 (WT or homozygous for mutation 2)) ( Supplementary Figures S8, S9 and Supplementary Table S4 ). The segregation ratio of mutations in T1 plants of 31-1 did not follow the pattern for Mendelian inheritance ( Supplementary Figure S7 ; Supplementary Table S4 ), and this could be explained by the limiting number of available T1 seeds produced by the primary plant. At the same time, the occurrence of additional editing was found in the progeny of 31-10 primary plant. Overall, five additional mono- and biallelic indel variants were discovered ( Supplementary Figure S10 ). Aside from the 1 bp insertion that arose in 31-10 primary plant, two new types of 1 bp insertions, a 1 bp deletion and an 8 bp deletion, were detected in the vrn-A1 promoter of T1 plants ( Figure 2 ). All new modifications are proximal to the PAM site of vr-31 sgRNA. We performed Sanger sequencing analysis of some progeny of the self-pollinated T1 plants (including transgenic individuals carrying WT allele) and found no additional edits in the T2 generation ( Supplementary Table S5 ). Indel mutations discovered in the T1 plants of 31-10 were easily heritable, and T2 plants carrying homozygous alleles of the additional edits were successfully detected ( Supplementary Table S5 ).

In total, after sequencing the T1 and T2 progenies of four primary events, we detected seven CRISPR/Cas9-edited homozygous alleles of vrn-A1 promoter ( Supplementary Table S6 ). Some of the new edits, such as the 1 bp deletion, were generated as a result of independent editing. Tracking of the segregation of transgenes encoding BAR, GFP, Cas9, and sgRNA resulted in the acquirement of six GE populations (M3-M4 progenies) carrying CRISPR/Cas9-edited homozygous alleles without transgenic inserts ( Supplementary Figure S11 ). In parallel, the selection of several transgene-free null-segregants was performed through genetic segregation of various GE monoallelic plants for use in comparative analysis.

To test the influence of the introduced indels on the heading time, we planted nine GE populations ( Supplementary Table S5 ) in a climate-controlled greenhouse (day/night temperature of 22-25 °C/18-20 °C with 16/8 h photoperiod) together with corresponding null-segregants derived from the same primary event (if available) for comparison with nontransformed ‘Chinese Spring’ plants. Plants were cultivated without vernalization treatment. Starting from the 58th day after germination, the first emergence of heads was observed in individual plants, whereas the peak of heading occurred at 63-66 days. In this experiment, the average heading time of ‘Chinese Spring’ plants was 63.8 days (n=61). A similar timing of first head appearance (63.6 days, n=45) was also observed in nontransgenic WT31-10 plants that had not inherited induced mutations and represent the seed generation of the primary 31-10 plant. The null-segregants WT31-2 (the nontransgenic siblings of the 31-2 primary plant) showed somewhat later heading at 65.3 days (n=27); however, the observed difference (CS vs. WT31-2) was not significant according to the ANOVA test (p=0.2214) ( Supplementary Table S7 ).

No clear differences in heading dates were observed between ‘Chinese Spring’ and the group of GE lines carrying different single nucleotide insertions ( Figure 3A ). The average time for the first spike appearance was 65.0 days for GE31-7 i1 (n=45), 64.0 days for the GE31-10 i1 (n=45), 63.9 days for the GE31-10-9 i1 (n=45), and 63.3 days for the GE31-10-15 i1 (n=30). Since the three last lines resulted from the genome editing of the chimeric T0 event 31-10, it was possible to compare the resulting heading dates with the null-segregant WT31-10. Statistical analysis showed no significant difference when compared with either ‘Chinese Spring’ (p=0.9996 for CS vs. GE31-10-9 i1; p>0.9999 for CS vs. GE31-10-7 i1; p> 0.9999 for CS vs. GE31-10-15 i1) or WT31-10 plants (p=0.9972 for WT31-10 vs. GE31-10-9 i1; p>0.9999 for WT31-10 vs. GE31-10-7 i1, p=0.9986 for WT31-10 vs. GE31-10-15 i1). Since the primary T0 plant 31-7 is biallelic, it was not possible to compare the heading time of GE 31-7 i1 with an appropriate null sibling, whereas the deviation from ‘Chinese Spring’ was not confirmed (p=0.3494, CS vs. GE31-7 i1).

The average heading time for the group of GE lines carrying the same deletion of a single nucleotide in the target region of the vrn-A1 promoter varied from 64.7 days in GE 31-7 d1 (homozygous progeny of the primary plant 31-7) to 64.3/64.2 days in GE31-10-19 d1/GE31-10-2 d1 (progeny of chimeric event 31-10). Since the average number of days to head emergence in these GE lines was similar to that of ‘Chinese Spring’, no statistical significance was found ( Supplementary Table S6 ). Similarly, the comparison between heading dates of GE31-10-19 d1/GE31-10-2 d1 lines and WT31-10 (null siblings of the same primary event) did not reveal any significant differences ( Figure 3 ).

The absence of four nucleotides in the promoter region of the vrn-A1 gene in the GE31-2 d4 line also had no effect on the heading time ( Figure 3 ). Although the GE31-2 d4 plants headed relatively later (64.6 days) than the wild-type ‘Chinese Spring’ plants (63.8 days) but slightly earlier than the null sibling plants of WT31-2 (65.3 days) originated from the same primary event, the observed deviations were within the experimental error (p=0.9895, WT31-2 vs. GE31-2 d4; p=0.7648, CS vs. GE31-2 d4).

Contrary to the six CRISPR/Cas9-induced mutations listed above, the loss of 8 nucleotides between −125 and −117 bp of the recessive vrn-A1 promoter affected the heading/flowering time of wheat ( Figures 3A, B ). Plants of the GE31-10-19 d8 line headed, on average, 2.3-2.5 days earlier than ‘Chinese Spring’ and the corresponding null sibling WT31-10. Moreover, significant differences with a high level of confidence (p-value < 0.0001) were also found between the average head emergence date of the GE31-10-19 d8 line and the heading dates of most other mutated lines ( Supplementary Table S7 ).

When the wheat lines carrying various nucleotide deletions were cultivated in climate chambers under increased day length of 18 h (at 22 °C) or 22 h (at 17 °C), the GE31-10-19 d8 line also demonstrated earlier heading than plants of ‘Chinese Spring’ and GE lines GE31-7 d1 and GE31-2 d4 carrying 1 or 4 bp deletions ( Figure 4 ). Despite the fact that digital phenotyping revealed more active increases in plant biomass under a 22 h photoperiod growth regime, no significant differences in the morphological stages of plant development between the GE lines and ‘Chinese Spring’ were found between the different day/night/temperature regimes ( Figures 4A, B ). The spectral NDVI index of GE lines and ‘Chinese Spring’ was similar throughout the period of automatic digital phenotyping until the tilling stage, when the NDVI index value of GE31-10-19 d8 started to decrease, especially at 22 h photoperiod ( Figures 4C, D ), whereas its digital plant biomass still showed a tendency to increase ( Figures 4B, C ). Plants of GE31-10-19 d8 reached the tillering stage 2-5 days earlier than the GE lines with minor nucleotide deletions and wild-type ‘Chinese Spring’. As a result, at 18 h photoperiod, the average heading time in plants of GE31-10-19 d8 was 83.5 days in contrast to 86.0-87.6 days for GE31-7 d1, GE31-2 d4, and ‘Chinese Spring’ plants ( Figure 4E ). Under the 22 h photoperiod regime ( Figure 4F ), the difference also was evident, as the average heading date of 67.4 days in GE31-10-19 d8 was shorter than 71.3 days in ‘Chinese Spring’ (p < 0.0001), while the differences between ‘Chinese Spring’ and two other lines were not statistically significant (p=0.1454, CS vs. GE31-7 d1; p=0.8251, CS vs. GE31-2 d4).

To further assess the relationship between the changes in growth habit and the CRISPR/Cas9-induced 8 bp deletion in the promoter region of vrn-A1 gene, we compared the heading dates of unvernalized GE31-10-19 d8 plants with that of plants vernalized for 15, 30, and 45 days. Even the short vernalization of 15 days advanced the head emergence in such a way that all accessions, including GE31-10-19 d8, nontransgenic null sibling WT31-10, and ‘Chinese Spring’ headed earlier than the corresponding unvernalized plants ( Figure 5A ). Regardless of the duration of cold treatment (15, 30, or 45 days), the time to plant heading in GE31-10-19 d8 was similar to that of both ‘Chinese Spring’ and the null sibling WT control. In parallel, the unvernalized GE31-10-19 d8 plants headed, on average, 3-4 days earlier than the unvernalized ‘Chinese Spring’ or nontransgenic null sibling ( Figure 5A ).

To investigate the effect of mutations on the relative expression of the vrn-1 gene, we assessed the expression levels in unvernalized plants at the early tilling stage in the fifth fully expanded leaf. The quantitative gene expression results showed no significant changes in the vrn-A1 gene expression levels in CRISPR/Cas9-mutated plants carrying the 8 nt deletions compared with WT plants ( Figure 5B ).

Analysis of the harvested seeds showed that the presence of a 8 bp deletion in the vrn-A1 promoter does not affect significantly the seed quality ( Supplementary Figure S12 ). The thousand grain weight of GE31-10-19 d8 line (34.5 g) did not significantly differ from that of the nontransgenic null sibling (33.1) and wild-type ‘Chinese Spring’ (35.9 g) plants ( Supplementary Figure S12 ). Although some increase in the total protein content of the grains was observed between the GE31-10-19 d8 line (18.7%) and wild-type ‘Chinese Spring’ (18.1%), the nontransgenic null sibling showed nearly the same average value (18.6%) as the plants carrying the new mutated VRN-A1 allele. Similarly, no significant differences were observed across the GE line and two controls (48.4-49.1%) when the total starch content was analyzed ( Supplementary Figure S12 ).