The annotated genome data of P. frutescens was obtained from the GenBank (https://www.ncbi.nlm.nih.gov/genbank/). Amino acid sequences of Arabidopsis and rice PEBPs were retrieved from the TAIR (http://www.arabidopsis.org/) and RGAP (http://rice.uga.edu/) databases, respectively. To identify PfPEBP, the Hidden Markov Model (HMM) profile of the PBP domain (PF01161) was obtained from the Pfam database and used as the query. In addition, amino acid sequences encoded by the genome of P. frutescens were also searched using HMMER v3.3.2 (E-value ≤ 1.0 × e⁻⁵). BLASP was performed against P. frutescens genome using Arabidopsis PEBPs as queries (E-value ≤ 1.0 × e⁻⁵). Finally, PfPEBP candidates were submitted to CD-Search (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi), SMART (http://smart.embl-heidelberg.de/) and Pfam (http://pfam.xfam.org/search) for conserved domain analysis.

The basic physicochemical properties of PfPEBP were analyzed using the ProtParam tool in Expasy (https://www.expasy.org/), including putative molecular weight (MW), isoelectric point (pI), and grand average of hydropathicity (GRAVY). Subcellular localizations were predicted using the Plant-mPLoc prediction tool (http://www.csbio.sjtu.edu.cn/bioinf/plant-multi/). The chromosomal mapping was performed based on genome annotation information and visualized with the TBtools (Chen et al., 2020). Gene structure was analyzed using the GSDS 2.0 server (http://gsds.gao-lab.org/). Finally, the conserved motifs were predicted using the MEME Suite (https://meme-suite.org/meme/tools/meme) with the following parameters: optimum width 10–50 amino acids, any number of repetitions of a motif, and maximum number of motifs set at 5.

Amino acid sequences of PfPEBP were aligned using the Clustal Omega and visualized using the Jalview. Phylogenetic tree of 25 PEBP proteins from Perilla, Arabidopsis and rice were constructed using the Neighbor-Joining (N-J) method (1000 bootstrap replicates) in MEGA X based on JTT+G substitution matrix. The phylogenetic tree was visualized with the iTOL tool (https://itol.embl.de).

To understand the cis-acting elements in the promoter region of PfPEBP, a 2000 bp upstream fragment of the initiation codon (ATG) was retrieved and predicted by the PlantCARE (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/). In addition, the cis-acting elements of rice OsFTL1, OsFTL2, OsFTL3, and Arabidopsis AtFT were also predicted and visualized using the TBtools.

The homologous proteins of PfFTs and PfTFL1s in Arabidopsis were determined by BLASTP. Subsequently, prediction of PPI networks with PfFTs and PfTFL1s as hub proteins were performed using STRING v11.5 (https://string-db.org/) (Szklarczyk et al., 2019).

Seeds of Perilla and Arabidopsis [Col-0 (Columbia ecotype), Ler (Landsberg erecta ecotype) and ft-1 mutant (Ler ecotype)] were kept in the Key Laboratory of Oil Peony Germplasm Innovation and Utilization of Chongqing Normal University. Perilla seeds were sown in plastic pots with nutrient soil and coarse soil (1:2, v:v) and grown in a incubator: 24°C/20°C (Light/Dark), 16 h/8 h (LD: Light/Dark) or 8 h/16 h (SD: Light/Dark). The cultivation and growth conditions of Arabidopsis (Col-0, Ler and ft-1) were performed as described by Xu et al. (Xu et al., 2022).

When grown to 6 true leaves, leaves were collected every 4 h for 48 h and used for diurnal rhythmicity analysis of PfFT1 expression. To analyze the tissue specific expression patterns of PfFT1 under different photoperiods, the roots, stems, leaves, flowers at flowering stage were collected at 56/141 days after sown (DAS), respectively. Mature seeds were collected at 86 DAS (SD condition), 171 DAS (LD condition). All samples were snap-frozen in liquid nitrogen and stored at −80°C until use.

To investigate the expression profiles of PfFT1 in different tissues and photoperiodic conditions, quantitative real-time polymerase chain reaction (RT-qPCR) was performed. Total RNA was extracted using the RNAprep pure Plant Kit (Tiangen, Beijing) and reverse transcribed into cDNA using the PrimeScript RT Reagent Kit with gDNA Eraser (Takara, Japan). The RT-qPCR reaction system (20µL) contains 10 ng cDNA, 10 µL SYBR Green Supermix (BioRad, USA), 1 µL primer pairs (10 µM/each), and ddH₂O. The reaction conditions were 95 °C for 2 min, followed by 40 cycles at 95 °C for 10 s, 60 °C for 20 s and 72 °C for 20 s. Perilla PfActin7 was used as the internal reference gene. Three biological and three technical replicates (n=3×3) were performed for each sample. Primers used in this study were listed in Supplementary Table S1 . The relative expression level of PfFT1 was calculated using the 2^−ΔΔCT method (Livak and Schmittgen, 2001).

The full-length coding sequence of PfFT1 was amplified using the leaf cDNA as the template. The PCR reaction system contains 1 µL cDNA, 10 µL Premix Taq enzyme (TAKARA, Japan), 1 µL primer pairs (10 µM/each), and 8 µL ddH₂O. The reaction conditions were 94 °C for 5 min, followed by 30 cycles at 94°C for 30 s, 57°C for 30 s and 72°C for 60 s, and a final extension at 72 °C for 10  min. All PCR products were detected by 1% agarose gel electrophoresis, and purified using the SanPrep Column DNA Gel Extraction Kit (Sangon, Shanghai). The recombinant vector pCAMBIA1303-PfFT1 was constructed using the In-Fusion HD Cloning Kit (Takara, Japan) following the manufacturer’s instructions, and then transformed into Agrobacterium tumefdciens GV3101 (WEIDI, China) by the freeze-thaw method. Arabidopsis Col-0, Ler and ft-1 were transformed using the floral dip method (Clough and Bent, 1998). The harvested plant seeds were planted in 1/2 MS medium containing 20 mg/L hygromycin B. Plants that can grow 4 or more true leaves were considered as positive seedlings and transplanted to soil. Leaf genomic DNA was extracted and amplified against the T-DNA region of the pCAMBIA1303 vector to detect the target gene PfFT1. The transgenic Arabidopsis was then stained using the GUS stain Kit (Coolaber, China). Plants validated by the above two steps were regarded as the positive T₀ generation. Following the same protocol, the T₃ generation without trait segregation was considered as a homozygous transgenic line and used for further studies.

To analyze the effect of PfFT1 on flowering, wild-type Col-0, transgenic Col-0 (TC); ft-1 mutant, complemented mutant (CM); Ler, and transgenic Ler (TL) plants were grown in an artificial climate chamber under the LD condition (22°C/20°C, Light/Dark). At least 10 plants from each line were selected to record the bolting and flowering time, number of rosette leaves at flowering.

Total RNA was extracted from leaves of 20-day-old Arabidopsis (transgenic or non-transgenic line) grown under the LD conditions, and reverse transcribed into cDNA and used for gene expression analysis of PfFT1, AtFT, AtAP1, AtFUL, AtSOC1 and AtLFY. The RT-qPCR reaction system (20µL) consisted 10 ng cDNA, 10 µL SYBR Green Supermix, 1 µL primer pairs (10 µM/each), and ddH₂O. The reaction condition was 95 °C for 1 min, followed by 40 cycles at 95 °C for 10 s, 60 °C for 30 s and 72 °C for 20 s. Arabidopsis AtActin8 was used as the internal reference gene. Three biological and three technical replicates (n=3×3) were performed for each sample. Primers were listed in Supplementary Table S1 . The relative gene expression levels were calculated using the 2^−ΔΔCT method.

All data were analyzed by Student’s t-test using IBM SPSS v26.0. Significant difference was considered when P < 0.05. Graphs were plotted using GraphPad Prism v8.0.

A total of 10 PfPEBP genes were identified and named according to the similarity with the homologous genes in Arabidopsis, including 2 PfFT, 4 PfTFL1, 2 PfCEN and 2 PfMFT ( Supplementary Table S2 ). The corresponding protein length ranges from 149 (PfCEN2) to 176 (PfMFT1) amino acids, the molecular weight of PfPEBP ranges from 17.0 to 19.9 kDa, and the pI ranges from 7.74 to 9.51. All GRAVY values were less than 0, indicating that all PfPEBP are hydrophobic. Subcellular localization prediction showed that PfFTs, PfTFL1s and PfCENs, and PfMFTs were located in the nucleus, cytoplasm, cytoplasm-nucleus, respectively ( Supplementary Table S2 ).

To understand gene structural diversities of PfPEBP genes, chromosomal location, number of exons and introns, and distribution of conserved domains were investigated. Chromosome mapping showed that 10 PfPEBP genes were distributed on Chr02, Chr05, Chr11, Chr14, Chr15, Chr17 and Chr18 ( Figure S1 ). The phylogenetic relationships of the PfPEBP family was shown in Figure 1A . Each PfPEBP gene was composed of 4 exons and 3 introns ( Figure 1B ). Motifs 2-5 were conserved in all PfPEBP ( Figure 1C ). Only PfCEN2 lacks motif 1. Each motif contained 11-50 amino acids ( Figure 1D ). Members in the same subfamily shared a similar motif pattern, suggesting that they might have similar functions. Comparison analysis revealed that the conserved PEBP domain contained complete motifs 2-5, indicating that they are crucial for their functions, and also explaining the conserved domain patterns of PfPEBP members.

Amino acid sequence alignment showed that the identities between PfFTs and PfTFL1s were more than 55% ( Figure 2 ). All PfPEBP have conserved motif DPDxP and GxHR, which are critical for anion-binding. Particularly, motif GxHR has a preference for Ile residues. The conserved motif LYN in PfFTs was localized in segment C, which was essential for the enzymatic activity. Fragment D of PfFTs contains the motif SGTGGR, while that of PfTFL1s were replaced by the motif TAARRR. Further analysis showed that all PfFTs had the key amino acid residues Tyr⁸⁴, Gln¹³⁹, but they were replaced by His⁸⁵, Asp¹⁴² or Asn¹⁴² in PfTFL1s.

To explore the evolutionary relationship of the PfPEBP family, a phylogenetic tree of PEBP from P. frutescens, A. thaliana, and O. sativa was constructed. The results showed that 35 PEBP proteins were clustered into three subfamilies: FT-like, TFL1-like and MFT-like ( Figure 3 ). The FT-like subfamily has 18 members, including AtFT, OsFTLs, and PfFTs. The TFL1-like subfamily has 12 members, including AtCEN, AtTFL1, AtBFT, OsCENs, PfCENs and PfTFL1s. The MFT-like subfamily has five members, including AtMFT, OsMFTs, and PfMFTs.

To further understand the regulatory patterns of PfPEBP genes, the cis-regulatory elements within a fragment 2000 bp upstream of the start codon (ATG) of each gene were analyzed. Based on their functions, they were divided into three categories: growth and development, hormone responses, and biotic/abiotic stress responses ( Figure 4A and Supplementary Table S4). Light-responsive elements, such as Box4 and G-Box were present in the promoter of all PfPEBP genes, suggesting that they are light signal reponse regulators. Circadian-responsive element (circadian) only occurred the promoter of PfMFT genes, whereas the anaerobic-inducible element (GC-motif) was only present in the promoter of PfTFL1 and PfMFT genes. Further analysis found that the distribution of hormone response elements was quite different. The auxin responsive element (AuxRR-core) was only found in PfFT and PfMFT genes. These findings revealed that the PfPEBP genes could respond light, hormones and stresses.

Comparion analysis revealed that Arabidopsis (LD plant) has specific light-response elements, such as AE-Box and Gap-motif, while Perilla and rice (SD plant) have specific light-response elements, such as ACE, AT1-motif, Box-4, Chs-CMA1a, GATA-motif, TCCC-motif, and sp1 ( Figure 4B and Supplementary Table S5 ). Differences in cis-acting elements in promoter regions of PEBP genes in LD and SD plants might lead to their different responses to photoperiod.

Comparison of amino acid sequences is helpful to understand functional similarities, amino acid sequence identities between PfFTs and AtFT was 78.16%, and that of PfTFL1s and AtTFL1 was 68.60-72.83% ( Figure S2 and Supplementary Table S6 ). High sequence identity and conservation of key motifs provided an important basis for PPI network prediction, which provides a more intuitive understanding of their functions. STRING analysis showed that both PfFTs and PfTFL1s potentially interact with FD, AP1, LFY, CO, AGL20, DECOY, RPL23/L15e, and RPL29. In addition, PfFTs also interacted with FLC, GI, and SVP ( Figure 5 ). This result suggested that PfFTs and PfTFL1s may have different functions in regulating flowering through similar pathways.

To understand effects of photoperiod on PfFT1 expression, its expression within 48 h under LD or SD conditions was analyzed by RT-qPCR ( Figure 6A ). The result revealed that PfFT1 have specific rhythmic expression patterns under different photoperiod conditions. Under the SD conditions, its expression level was increased with the increase of light duration within 0-8 h, then gradually decreased under the dark condition (8-24 h). Within 24-48h, its expression level peaked after 8 h of light (32 h), but it was higher than that within 0-24 h. Under the LD condition, the expression level reached a peak after 4 h light exposure (significantly lower than the peak value under the SD condition), and then gradually decreased to a very low level. In conclusion, the expression level of PfFT1 under the LD condition was extremely low, which strongly supported that the Perilla is a SD plant.

To further explore the relationship between the PfFT1 expression and flowering, its tissue expression pattern under different photoperiods was analyzed ( Figures 6B, C ). Under the LD condition, the highest expression was detected in seeds, then followed by flowers, while it was extremely low in roots, stems and leaves. Under the SD condition, highest expression was found in leaves, followed by flowers and seeds. But the expression was very low in roots and stems. These results suggested that PfFT1 could promote flowering under SD conditions, whereas it might also regulate seed development under LD conditions.

To study the regulatory role of PfFT1 on flowering, PfFT1 gene driven by CaMV 35S promoter was transferred into Arabidopsis Col-0, Ler, and ft-1, respectively ( Figures S3A, B ). After hygromycin screening, GUS staining and PCR identification, a total of 18 TC, 10 CM and 15 TL transgenic lines were obtained ( Figures S3C–F ). Subsequently, three T₃ plants of each transgenic line were selected based on the expression level of PfFT1 for further analysis. Compared with the controls, all transgenic plants exhibited varying degrees of early flowering ( Figures 7A–C ). Compared with the wild-type Col-0, the bolting and flowering time of TC lines were 5-8 day earlier with 2-3 less rosette leaves at flowering ( Table 1 ). Overexpression of PfFT1 not only promoted early flowering of Col-0, but also rescued the flowering time defect of ft-1 mutant. All CM lines started bolting at day 15-16, the flowering time was 30 d earlier than that of ft-1 line, and the number of rosette leaves at flowering was only 4-5. However, the ft-1 mutant bloomed around day 50 with 18 rosette leaves ( Table 2 ). These results indicated that the overexpression of PfFT1 could complemented the delayed flowering in ft-1 mutant. In addition, it was found that the bolting and flowering time of TL lines were 3-5 d earlier than that of Ler type ( Table 2 ).

RT-qPCR analysis showed that PfTF1 was expressed in all TC, CM and TL lines ( Figures S4A–C ), but was not detected in wild-type Col-0, and Ler type. Expression of AtFT and PfFT1 were not detected in ft-1 mutant ( Figures S4B, E ).

To verify whether the introduction of PfFT1 could regulate the expression of endogenous flowering-related genes in Arabidopsis, the expression level of AtAP1, AtFUL, AtSOC1 and AtLFY were analyzed ( Figures 8A–L ). In TC lines, high expression of AtSOC1 and AtAP1 was observed ( Figures 8A, G ). Compared with the ft-1 mutant, AtAP1, AtFUL, AtSOC1 and AtLFY were significantly up-regulated in CM lines ( Figures 8B, E, H, K ). Compared with the non-transgenic ft-1 line, the flowering time of transgenic lines was significantly earlier. It was also earlier than the average flowering time of Ler (25.27 ± 0.91) ( Table 2 ). Compared with the non-transgenic Ler, AtAP1 was highly expressed in OL lines, but AtSOC1 was less expressed ( Figures 8C, I ).