To identify PEBP family members in Chinese cabbage (BrPEBP), we analyzed our previous RNA-seq dataset (Jung et al., 2021) with the amino acid sequence of Arabidopsis FT (AtFT, encoded by At1g65480) and identified 12 candidate genes ( Table S1 ). In agreement with the previous study in the B. rapa variety ‘yellow sarson R-o-18’ (Del Olmo et al., 2019), the protein encoded by Bra022475 (referred to as BraA.FT.a in this study) showed the highest identity to AtFT with 86%, followed by the proteins encoded by Bra004117 (BraA.FT.b) and Bra015710 (BraA.TSF) with 82%, and that encoded by Bra010052 (BraA06g025510) with 60%. Other PEBP-like proteins shared less than 60% in identity with AtFT and were not considered further.

To characterize the structural and functional divergence of the BrPEBP candidates, we performed a phylogenetic tree analysis using the Bayesian evolutional analysis software called BEAST 2.5 (Bouckaert et al., 2019). The phylogenetic analysis also included five Arabidopsis PEBP family members, TSF, BFT, CENTRORADIALIS (CEN), and MFT for putative functional assignment (Jin et al., 2021). The BrPEBP candidates were divided into three groups named FT-, TFL-, and MFT- clade with three Bra022475, Bra004117 and Bra015710 belonging to FT clade. In detail, the proteins encoded by Bra022475 and Bra004117 were the closest to AtFT, prompting us to rename their encoding genes BrFT1 and BrFT2, respectively. The protein encoded by Bra015710 was close to AtTSF than AtFT, while the protein encoded by Bra010052 was close to AtBFT in TFL clade; the corresponding genes were thus renamed BrTSF and BrBFT, respectively. The other Chinese cabbage proteins in the tree were closer to AtCEN, AtTFL, or AtMFT ( Figure 1A ). We thus focused on BrFT1, BrFT2, BrTSF, and BrBFT for further analysis as potential AtFT-like genes.

To characterize the function of the BrPEBP genes above, we examined their transcript levels using our previous RNA-seq dataset derived from the early-bolting inbred line ‘4004’ and the late-bolting inbred line ‘50’ grown under normal conditions (continuous condition at 23°C, 16 h light/8 h dark) or exposed to vernalization (at 4°C for 35 days at the same light conditions) (Jung et al., 2021). We determined that BrFT1 and BrFT2 are relatively highly expressed in inbred line 4004, but not in inbred line 50 in response to vernalization in accordance with the flowering phenotypes between the two inbred lines. From this result, we suggest a promoting role in flowering for these two genes ( Figure 1B , lane 3). BrTSF, BrBFT, and other PEBP-like genes were rarely expressed under either normal or vernalization conditions ( Figure 1B ). To confirm these expression levels, we performed reverse transcription quantitative PCR (RT-qPCR) analysis of the four BrFT-like genes in 4004 and 50 plants exposed to vernalization or maintained at 22°C. The expression level of BrFT1 rose about 9-fold in the 4004, but not in 50 upon vernalization. Although BrFT2 was expressed 13-fold more highly in 4004 than in 50, BrFT2 expression levels increased in both lines in response to vernalization, with a 19-fold and 29-fold increase in 4004 and 50, respectively. Notably, BrTSF expression did not appear to respond to vernalization in the 4004 line and decreased 5-fold in line 50. BrBFT showed opposite responses to vernalization in its transcript levels in line 4004, experiencing a 14-fold drop following vernalization compared to normal conditions, whereas its expression level increased about 3.5-fold after vernalization treatment in line 50 ( Figure 1C ). We conclude that BrFT1 and BrFT2 may differ from BrTSF and BrBFT in terms of their response to vernalization, raising the possibility that these two groups of genes may differently regulate flowering time.

To assess the roles of BrFT-like genes in the regulation of flowering time, we individually overexpressed the full-length genomic sequence from BrFT-like genes (gBrFT1, gBrFT2, gBrTSF, and gBrBFT) in the Arabidopsis ft-10 mutant ( Figure S1A ) under the control of the cauliflower mosaic virus (CaMV) 35S promoter ( Figure S1B ). Overexpressing BrFT1 or BrFT2 in the ft-10 mutant background accelerated flowering compared to the vector control, or the overexpression of BrTSF or BrBFT, in both the T₁ and T₂ generations ( Figures S2A, B ). We confirmed the overexpression of individual BrFT-like genes in T₂ transgenic lines by RT-qPCR analysis ( Figure S2C ). We then selected homozygous T₃ transgenic lines by selecting seedlings on half-strength Murashige and Skoog (MS) medium containing hygromycin and confirmed the presence of the transgene by PCR analysis ( Figure S2D ).

We selected one homozygous transgenic line per BrFT-like gene: BrFT1-OE #5-6/ft-10 for BrFT1, BrFT2-OE #2-1/ft-10 for BrFT2, BrTSF-OE #7-5/ft-10 for BrTSF, and BrBFT-OE #4-3/ft-10 for BrBFT, with Vec #10-1/ft-10 as empty vector control. Homozygous BrFT1-OE #5-6/ft-10 or BrFT2-OE #2-1/ft-10 transgenic plants flowered much earlier than the Vec #10-1/ft-10 control when grown under long-day conditions ( Figure 2A ), flowering after 22 days or 15 ± 7 days, respectively, and produced far fewer leaves than the Vec #10-1/ft-10 control with a number coming close to that of the wild-type Col-0 accession (background of ft-10) ( Figure 2B ). In sharp contrast, homozygous lines overexpressing BrTSF or BrBFT (lines BrTSF-OE #7-5/ft-10 and BrBFT-OE #4-3/ft-10) flowered at the same time as Vec #10-1/ft-10, at 35 ± 3 days after sowing, and produced the same number of leaves as the empty vector control.

To corroborate the relative expression levels of BrFT-like genes in flowering time, we performed RT-qPCR analysis and confirmed that each BrFT-like gene is overexpressed in its corresponding transgenic line in the ft-10 background ( Figure S2E ). These results suggest that BrFT1 and BrFT2 are important in flowering time regulation and function as floral activators. We obtained the same results with the independent T₃ transgenic lines BrFT1-OE #10-4/ft-10 and BrFT2-OE #8-3/ft-10, which both accelerated flowering to a similar extent as BrFT1-OE #5-6/ft-10 and BrFT2-OE #2-1/ft-10 ( Figure S3 ). By contrast, BrTSF and BrBFT did not appear to contribute to flowering time regulation, at least when overexpressed in Arabidopsis.

To determine whether BrFT-like genes induce a subset of genes whose expression is known to be regulated by the FT–FD complex, we analyzed the expression levels of floral meristem identity genes in Arabidopsis transgenic lines overexpressing BrFT1 or BrFT2 by RT-qPCR analysis. Indeed, the transcript level of AtSOC1 (SUPPRESSOR OF OVEREXPRESSION OF CO 1), whose expression can be directly induced by the FT–FD complex (Lee and Lee, 2010), significantly increased in BrFT1 and BrFT2 transgenic plants compared to the vector control line and even higher than the Col-0 control. The expression of LEAFY (LFY), the master regulator of floral fate and another target gene of the FT–FD complex (Zhu et al., 2020), was also highly induced by the overexpression of BrFT1 or BrFT2 in ft-10 and reached levels close to that of Col-0. In addition, the floral homeotic genes AP1, FRUITFUL (FUL), and SEPALLATA 3 (SEP3), whose expression is activated in response to the FT–FD complex (Teper-Bamnolker and Samach, 2005; Wigge et al., 2005), increased relative to the vector control line and reached levels comparable to Col-0 when BrFT1 or BrFT2 was overexpressed ( Figure 3 ; green and yellow color series graphs).

The splice FLOWERING LOCUS M (FLM) isoform AtFLM-δ encodes a floral promoter in Arabidopsis (Capovilla et al., 2017); overexpressing BrFT1 or BrFT2 in ft-10 increased AtFLM-δ levels 2- to 3-fold relative to Col-0 and the empty vector control line ( Figure 3 ; green and yellow color series graphs). By contrast, overexpression of BrTSF or BrBFT did not affect the expression levels of floral homeotic identity genes or FLM-δ in the ft-10 mutant background, with the exception of AtAP1 ( Figure 3 ; sky blue and pink color graphs). Interestingly, the BrBFT transgenic line showed an extremely low expression of AtAP1 compared to all other transgenic lines and Col-0 WT. We conclude that the overexpression of BrFT1 or BrFT2 can induce the expression of a subset of floral meristem identity genes, as does AtFT.

The similar phenotypes upon overexpression of BrFT1 or BrFT2 in the ft-10 mutant raised the possibility that they might regulate flowering time redundantly in Chinese cabbage. To test this hypothesis, we used simultaneous CRISPR/Cas9-mediated mutagenesis of BrFT1 (Bra022475) and BrFT2 (Bra004117) genes using the Chinese cabbage inbred line ‘20’ as in our previous study (Jung et al., 2021). We designed one single-guide RNA (sgRNA) targeting the first exon of BrFT1 and BrFT2 to edit both genes simultaneously ( Figure S4A ). We then transformed Chinese cabbage hypocotyls and regenerated whole plants, as described in Materials and Methods. We obtained several T₀ plants that we genotyped for the presence of mutations at the target sites by genotyping PCR and whole-genome sequencing ( Figure 4A ). We selected line Brad39, which mutation was of > 99% for BrFT1 and > 50% for BrFT2 with insertion/deletion accounting for the largest proportion of mutations using next-generation sequencing (NGS) ( Figure 4B , Figure S4B ). We then grew the genome-edited ‘Brad39’ T₀ plant under the same growth conditions as its isogenic wild-type parent 20 to obtain T₁ seeds, but Brad39 did not bolt even after 3 months of growth ( Figure 4C ). Even after 6 months in LD conditions, Brad39 failed floral organogenesis thus never produced T₁ seeds, as the switch to the reproductive stage never took place ( Figure S5 ). We thus concluded that of the four BrFT-like homologs in Chinese cabbage, BrFT1 and BrFT2 may be redundant positive regulators of flowering whose simultaneous loss of function impairs floral organogenesis.

Since FT regulates the transcription of a subset of downstream genes by interacting with FD (Abe et al., 2005), we further investigated whether BrFT-like proteins interact with BrFD protein. Previous reports have indicated that Arabidopsis FT harbors four segments (segments A to D) in its C terminus and that FT interacts with FD through the formation of an external loop formation of 14 amino acids from segment B, while the L/IYN motif in segment C is crucial for FT activity (Ahn et al., 2006; Wang et al., 2017). We first aligned BrFT-like proteins, AtFT, AtTSF, AtBFT, and the rice floral integrator Heading date 3a (OsHd3a) to explore the extent of sequence conservation ( Figure S6 ). In segment B, the 14–amino acid sequence from BrBFT was more consistent with that of AtBFT than AtFT; moreover, in segment C, BrBFT did not have the same motif as other BrFT-like proteins and BrTSF showed the motif NYN, thus harboring a mismatched sequence. We hypothesized that the differences in sequence between BrTSF/BrBFT and BrFT1/BrFT2 might determine their interaction potential with FD.

To determine whether BrFT-like proteins interact with BrFD, we performed a yeast two-hybrid (Y2H) assay. To this end, we individually cloned the full-length coding sequences of BrFT-like genes into pGBKT7 (BD vector), while the full-length coding sequence of BrFD was cloned into pGADT7 (AD vector). We introduced the appropriate pairs of constructs into yeast cells and tested protein–protein interaction, which revealed that BrFT1 and BrFT2 can interact with BrFD, whereas BrTSF and BrBFT did not ( Figure 5A ). Next, to delineate the exact differences between BrFT-like proteins that dictate their interaction with BrFD, we took a closer look at the protein alignment of BrFT-likes with AtFT and noticed that three amino acids (aa), Val-121, Gly-137, and Leu-150, are distinct between positive regulators of flowering (BrFT1 and BrFT2) and undefined function proteins (BrTSF and BrBFT) ( Figure 5B ). To examine whether the substitution of these three aa in BrFT-like proteins might change their interaction with FD, we performed bimolecular fluorescence complementation (BiFC) assays by transiently infiltrating Nicotiana benthamiana leaves with constructs encoding BrFT-like proteins fused to the N-terminal half of enhanced yellow fluorescence protein (BrFT-like-nEYFP) and BrFD fused to the C-terminal half of EYFP (cEYFP-BrFD). We observed green fluorescence in the nucleus when BrFT1 and BrFT2 constructs were co-expressed with the BrFD and AtFD constructs, but not when the FD constructs were co-expressed with either BrTSF or BrBFT, in line with the Y2H result ( Figure 5C and S7A ). Next, we reciprocally changed the three aa of BrFT1, BrFT2, BrTSF, and BrBFT and repeated the BiFC assay ( Figure 5B : each 3m). Surprisingly, BrFT1^3m and BrFT2^3m lost their ability to bind to the two FDs, while BrBFT^3m gained binding activity toward BrFD and AtFD ( Figure 5D , S7B ). Changing these three aa in BrTSF failed to confer the ability to interact with BrFD or AtFD for an unknown reason.

Although BrTSF overexpression did not rescue the late flowering of ft-10 and its encoded protein failed to interact with FD, we grew transgenic lines overexpressing BrTSF under short-day (SD) conditions, since Arabidopsis TSF promotes flowering under this condition (Yamaguchi et al., 2005). When grown in SDs, most BrFT1-OE #5-6/ft-10 and BrFT2-OE #2-1/ft-10 plants reached the flowering stage 5 weeks after germination, thus 2 to 3 weeks faster than Col-0. However, the overexpression line BrTSF-OE #7-5/ft-10 showed no flowering after 8 weeks, like Vec #10-1/ft-10 plants. Similarly, the other overexpression line BrBFT-OE #4-3/ft-10 failed to flower after 9 weeks in SDs ( Figures S8A, B ). In fact, none of the plants from the Vec #10-1/ft-10, BrTSF-OE #7-5/ft-10, or BrBFT-OE #4-3/ft-10 transgenic lines reached the flowering stage in SDs after close to 13 weeks (90 days), in contrast to Col-0 and lines overexpressing BrFT1 and BrFT2 ( Figures S8C, D ). We conclude that BrTSF is incompetent to induce flowering in Arabidopsis ft-10 mutant under both LD and SD conditions.

Arabidopsis (Arabidopsis thaliana) Col-0 and ft-10 seeds were sown on soil after being stratified for 2 days at 4°C in the dark, placed in a growth room (23°C, long-day conditions; 16 h light/8 h dark or short-day conditions; 8 h light/16 h dark), and grown for 8–9 weeks. The flowering phenotype was assessed based on the number of rosette leaves and days until bolting, which were recorded when the length of the main stem was ≥ 0.5 cm. Phenotyping was performed in three independent biological replicates (with at least 10 plants per replicate).

The early-bolting Chinese cabbage (Brassica rapa ssp. pekinensis) inbred line ‘20’ was used in this study. Seeds were obtained from NongHyup Seed (Anseong, Gyeonggi-do, Korea). Seeds were sown on sterilized soil and placed in a growth room maintained at 23°C and in long-day conditions (16 h light/8 h dark). After 2 weeks, vernalization was initiated by placing the trays in a cold room at 4 ± 1°C and in a 12-h-light/12-h-dark photoperiod for 35 days. After vernalization, the trays were transferred to a vinyl house and grown for 3 months.

The full-length regions of BrFT-like genes were amplified from ‘20’ Chinese cabbage genomic DNA by PCR with Lamp Pfu DNA polymerase (BioFACT, Daejeon, Korea). The PCR products were individually cloned into a modified pCAMBIA1300 vector in which the cauliflower mosaic virus (CaMV) 35S promoter and NOS terminator had been cloned into the multiple cloning site. All constructs were verified by sequencing and transformed into Agrobacterium (Agrobacterium tumefaciens) strain GV3101. The constructs were transformed into ft-10 (CS9869; ABRC, Columbus, OH, USA) by the floral dip method (Clough and Bent, 1998). T₁ seeds were sown onto a half-strength Murashige and Skoog (MS) containing 0.5% agar plate containing 25 mg/L hygromycin for the selection of transgenic seedlings. T₂ plants showing a 3:1 segregation ratio of hygromycin resistance to sensitivity were selected and allowed to self to collect homozygous T₃ seeds. The expression of BrFTs was confirmed by PCR.

Amino acid sequences of Arabidopsis and Brassica rapa PEBPs were obtained from TAIR10 (https://www.arabidopsis.org) and the Brassica database (http://brassicadb.cn), respectively. All proteins were used for constructing a phylogenetic tree and sequence alignment. The phylogenetic tree of PEBP proteins was constructed using Bayesian evolutionary analysis with divergence time analysis in BEAST 2.5 software (version 2.7.1) (Bouckaert et al., 2019). The sequence alignment was analyzed using BioEdit (version 7.2). Rice Hd1a (LOC_Os06g06320) sequence was obtained for sequence alignment using the Rice Genome Annotation Project (http://rice.uga.edu).

Transcriptome deep sequencing (RNA-seq) data were analyzed as previously reported (Jung et al., 2021). Gene expression data for the inbred lines ‘4004’ and ‘50’ were used for the analysis of BrFT genes and were represented as a heatmap of the normalized read counts from three biological replicates.

Total RNA was extracted from Arabidopsis rosette leaves of 2-week-old seedlings using the Wizprep™ Plant RNA mini Kit (wizbiosolutions, Gyeonggi-do, Korea). Total RNA was treated with RNase-free DNase I (Thermo Fisher scientific, Waltham, MA, USA) to remove traces of genomic DNA. PrimeScript™ RT Master Mix (TaKaRa, Shiga, Japan) was used for first-strand cDNA synthesis. Subsequently, qPCR was performed on a Bio-Rad CFX real-time PCR system (Bio-Rad, Hercules, CA, USA) using SYBR Prime Q-Mastermix (Genetbio, Daejeon, Korea), according to the manufacturer’s instructions. Relative expression levels were determined by normalizing the expression of each gene of interest against AtACT2 transcript levels. All PCR determinations were performed from at least three different biological replicates, each with three technical replicates, under the same conditions per experiment.

The relative expression levels of BrFT-like genes were confirmed in Figure 1 using the same conditions as a previous study (Jung et al., 2021). The leaves of the two Chinese cabbage inbred lines ‘4004’ and ‘50’ with different flowering times were collected. Total RNA extraction and first-strand cDNA synthesis were performed as above. All primers used in this study are listed in Table S3 .

The BiFC assay was conducted as previously described (Walter et al., 2004). The BrFD and AtFD full-length coding sequences were cloned into pSPYCE-35S, and the full-length coding sequences of BrFT-like genes (BrFT1, BrFT2, BrTSF, and BrBFT) were cloned from cDNA into pSPYNE-35S. The resulting constructs were introduced in Agrobacterium strain GV3101. Positive Agrobacterium colonies were cultured in YEP medium, pelleted by brief centrifugation, and resuspended to a final optical density of 0.8 in infiltration buffer (10 mM MgCl₂, 10 mM MES-KOH pH 5.7, and 200 µM acetosyringone). The appropriate pairs of cultures were then co-infiltrated with the P19 silencing suppressor into N. benthamiana leaves. After 48 h, YFP fluorescence was observed from the infiltrated leaves using a confocal laser scanning microscope (LSM800; Zeiss, Oberkochen, Germany). The settings for the confocal microscope were as follows: GFP, excitation of 488 nm and emission of 509 nm; RFP, excitation of 553 nm and emission of 573 nm.

The full-length coding sequences of BrFT-like genes (BrFT1, BrFT2, BrTSF, and BrBFT) were individually cloned into vector pGBKT7. The full-length coding sequence of BrFD (Bra010504) was cloned into pGADT7. The primers used for cloning are listed in Table S3 . Each construct harboring one BrFT-like gene was co-transformed with the BrFD or AtFD plasmid into yeast strain AH109, and positive colonies were selected on synthetic defined (SD) medium lacking leucine and tryptophan with dextrose (SD –LT) at 28°C for 7 days. Selected colonies were spotted onto agar plates containing either SD –LT, SD –LTH (SD medium lacking leucine, tryptophan, and histidine), or SD –LTH containing 0.5 mM 3-amino-1,2,4-triazole (3-AT). After plating, the cells were allowed to grow for 7 days. SD –LTH or SD –LTH +3-AT plates were used to test for protein−protein interactions. BrFD cloned into the pGBKT7 empty vector was used as negative control, and the OsCYP18-2-BD and OsSKIP-AD constructs were used as a positive control as previously described (Lee et al., 2015).

Cas-Designer (http://www.rgenome.net/cas-designer/) was used to design a specific single-guide RNA (sgRNA) against the Chinese cabbage FT genes (Bra022475 and Bra004117). Thereafter, the selected sgRNA (sgRNA: 5′-AAGCCAAGAGTTGAGAT-3′) targeting both BrFT genes was synthesized with a restriction enzyme sequence for cloning into the pHAtC vector (Kim et al., 2016) linearized with the restriction enzyme AarI (CACCTGC (4/8) ^). The resulting vector was transformed into Agrobacterium strain LBA4404 strain and then used for transformation of Chinese cabbage plants.

The Chinese cabbage inbred line ‘20’ from NongHyup Seed in Korea (Anseong) was used for transformation according to a previously published method (Lee et al., 2004). Hypocotyls were incubated for ~1–2 days in darkness before being cut into 0.5- to 1-cm-long segments. Co-culture was performed with the transformed Agrobacterium cultures in the dark for 2 days. After washing, the explants were cultivated on callus induction medium (MS salts with 3% [w/v] sucrose, 5 mg/L benzyl adenine [BA], 1 mg/L naphthaleneacetic acid [NAA], and 300 mg/L cefotaxim) in the dark for 3 days. The induced calli were transferred to shoot induction medium (MS salt with 3% [w/v] sucrose, 10 mg/L BA, 1 mg/L cefotaxime, and 10 mg/L hygromycin). Once shoots developed, the plantlets were cultured on root inducing medium (MS salt with 3% [w/v] sucrose, 0.1 mg/L NAA, and 0.1 mg/L gibberellin).

Statistical analyses were performed in GraphPad Prism 8.0.2 (GraphPad Software, San Diego, CA, USA). Phenotypic analysis was performed by analysis of variance (ANOVA), and different lowercase letters indicate significant differences between samples (P < 0.05, one-way ANOVA followed by Tukey’s test). For the analysis of RT-qPCR results, statistical significance was based on two-tailed Student’s t tests, with differences considered significant at a P-value of <0.05 (*P < 0.05, **P < 0.01, and ***P < 0.005), and on ANOVA, with different lowercase letters indicating significant difference between samples (P < 0.05, one-way ANOVA followed by Tukey’s test). At least three replicates were performed, and the data are shown as means ± standard error of the mean.

Sequence data from this article can be found in the Brassicaceae Database (BRAD) and The Arabidopsis Information Resource (TAIR10) under the following accession numbers: BrFT1 (Bra022475), BrFT2 (Bra004117), BrTSF (Bra015710), BrBFT (Bra010052), BrFD (Bra010504), AtFT (At1g65480), and AtFD (At4g35900). RNA-seq data were previously submitted to the Gene Expression Omnibus (GEO) database (Jung et al., 2021) under GEO accession numbers GSE106444 and GSE139375 and were reanalyzed here.