The experimental materials were collected from the experimental field (34.60404° N, 112.42325° E) of Jingyuan Farm, Henan University of Science and Technology. Four tree peony cultivars were selected: the single-petal cultivar ‘Ziban’, the semi-double-petal cultivar ‘Baqiandaichun’, the double-petal cultivar ‘Doulv’, and the multi-flowered cultivar ‘Luoyang Hong’. The sepals, outer petals, inner petals, and stamens (Se, OP, IP, St) of the single-petal cultivar ‘Ziban’ and the sepals, petals, petaloid petals, and stamens (Se, Pe, Pet, St) of the semi-double-petal, double-petal, and multi-flowered cultivars ‘Baqiandaichun’, ‘Doulv’, and ‘Luoyang Hong’ were quickly frozen in liquid nitrogen and stored at −80°C.

Total RNA was extracted from each tissue of tree peony cultivars ‘Luoyang Hong’, ‘Ziban’, ‘Baqiandaichun’ and ‘Doulv’ using an RNA extraction kit (TianGen, Beijing). The resulting RNA was reverse transcribed into cDNA using a reverse transcription kit (Accurate, Hunan). The tree peony cultivar ‘Luoyang Hong’ cDNA was used as a template to amplify PsSEP3L1 using a high-fidelity enzyme (TransGen Biotech, Beijing). The target DNA fragment was excised from the gel at the expected size and recovered using a gel recovery kit equipped with a microcolumn concentration system (Zoman, Beijing). The primers used in this study are listed in Table S1.

The amino acid sequence, physicochemical properties, transmembrane domain, phosphorylation sites, secondary structure, tertiary structure, conserved domain, signal peptide, hydrophilicity, hydrophobicity, subcellular localization and phylogenetic tree of PsSEP3L1 were predicted and analyzed using software and online platforms (Table S2).

Four varieties of tree peonies were selected for qRT-PCR expression analysis of PsSEP3L1: the single-petal cultivar ‘Ziban’, the semi-double-petal cultivar ‘Baqiandaichun’, the double-petal cultivar ‘Doulv’, and the multi-flowered cultivar ‘Luoyang Hong’. Sepals, outer petals, inner petals, and stamens (Se, OP, IP, and St) from the single-petal cultivar, and sepals, petals, petaloid petals, and stamens (Se, Pe, Pet, and St) from the semi-double-petal and double-petal cultivars at full bloom were selected. Relative expression levels were calculated using the 2^−ΔΔCt method and statistically analyzed with SPSS software. The primers used in this study are listed in Table S1.

The CDS of PsSEP3L1, excluding the stop codon, was amplified to generate the 35S::PsSEP3L1-GFP fusion construct using gene-specific primers listed in Table S1. Agrobacterium cultures harboring 35S::GFP and 35S::PsSEP3L1-GFP were infiltrated into leaves of Nicotiana benthamiana respectively. To induce protein expression, the infiltrated Nicotiana benthamiana leaves were incubated at 25°C in the dark for 48 hours. The subcellular localization of PsSEP3L1 in tobacco leaf cells was then examined using a laser scanning confocal microscope.

According to the PsSEP3L1 sequence, the PsSEP3L1-1300 overexpression vector and the PsSEP3L1-TRV2 silencing vector were successfully constructed (Table S1). The tree peony cultivar ‘Luoyang Hong’ was used as the experimental material. The study found that during the flower organ development process of tree peony, the big-bell stage represents a critical developmental period of flower type formation [28,29]. Therefore, we selected the big-bell stage for experimental investigation. TRV1 was mixed in equal volumes with the bacterial solutions containing PsSEP3L1-TRV2 or TRV2. The bacterial solutions containing the empty 1300 vector and PsSEP3L1-1300 were resuspended in infection solution and incubated in the dark for 3–4 h. The bacterial suspension was then injected into the pedicels using a 1 mL sterile syringe every 5 to 7 days until the color emergence stage of flower buds and pedicels. When the flowers reached the full-bloom stage, phenotypic observations were recorded, the number of petals was counted for analysis, and samples were immediately taken for qRT-PCR to detect the changes in gene expression. Six independent biological replicates were conducted for the transient transformation experiment. The t-test was performed using SPSS 16 to verify statistical differences and the results were visualized using Prism 5.0.

Based on the database previously established by our group, the genes PsSEP1L, PsAP1, PsAP3, and PsAG, which may interact with PsSEP3L1, were screened. Restriction sites EcoRI and BamHI were selected according to the pGBKT7 and pGADT7 vector sequences, and cloning primers were designed according to the gene sequence and the vector restriction site (Table S1). The pGBKT7-PsSEP3L1, pGADT7-PsSEP3L1, pGADT7-PsSEP1L, pGADT7-PsAP1, pGADT7-PsAP3, and pGADT7-PsAG fusion vectors were constructed. Yeastmaker™ Yeast Transformation System 2 was used to prepare Y2H Gold yeast competent cells and perform yeast transformation. The recombinant plasmids pGADT7 and pGBKT7 were co-transfected into yeast cells. The transformed yeast strains were coated with SD/-Trp solid medium and SD/-Leu/-Trp/x-α-gal solid medium to verify whether gene autoactivation existed. After the single colony grew on SD/-Trp solid medium, it was picked and transferred into SD/-Trp liquid medium. After 10 h, the OD value of the bacterial solution was detected to verify whether the bait gene was toxic. After co-transformation, the bacterial solution was coated on SD/-Leu/-Trp solid medium, serially diluted in four gradients, and then dot-coated on SD/-Leu/-Trp and SD/-Ade/-His/-Leu/-Trp/X-α-Gal solid media, respectively. After culturing at 30°C for 3 to 5 days, blue and white spots were observed.

The CDS of PsSEP3L1 was inserted into the cYFP vector, while the CDSs of PsSEP1L, PsAP1, PsAP3, and PsAG were individually inserted into the nYFP vector, generating the fusion constructs cYFP-PsSEP3L1, nYFP-PsSEP1L, nYFP-PsAP1, nYFP-PsAP3, and nYFP-PsAG. cYFP-PsSEP3L1 was co-infiltrated with nYFP-PsSEP1L, nYFP-PsAP1, nYFP-PsAP3, or nYFP-PsAG into tobacco leaves respectively. Conduct both positive and negative controls simultaneously. After 48–72 h of cultivation, yellow fluorescent protein (YFP) signals were observed using a confocal laser scanning microscope. The primers used in this study are listed in Table S1.

Using cDNA from the tree peony cultivar ‘Luoyang Hong’ as a template, PsSEP3L1 was obtained by homologous cloning, and its full-length sequence was 752 bp (Fig. 1A). Bioinformatic analysis showed that the open reading frame (ORF) of PsSEP3L1 was 738 bp, encoding 245 amino acids. Prediction of the physicochemical properties of PsSEP3L1 protein indicated that its molecular formula was C₁₂₄₇H₁₉₇₂N₃₅₄O₃₇₇S₈, with a total of 3958 atoms. The theoretical isoelectric point (pI) of PsSEP3L1 is 8.64, and its relative molecular mass is 28,212.0 Da. The protein contains 34 positively charged residues (Arg + Lys) and 31 negatively charged residues (Asp + Glu). With an instability index of 41.53, the protein is predicted to be unstable. Hydrophilicity and hydrophobicity analyses indicate that PsSEP3L1 is a hydrophilic protein (Fig. 1D). Transmembrane structure prediction using TMHMM showed that the protein does not contain a transmembrane domain (Fig. 1E). Phosphorylation site prediction showed that PsSEP3L1 contained 45 putative phosphorylation sites on threonine residues (Fig. 1F). Protein structure prediction showed that the secondary structure of PsSEP3L1 consisted mainly of α-helix (56.33%), random coil (30.61%), extended strand (8.98%), and β-corner (4.08%) (Fig. 1B). Swiss-model analysis revealed a predicted tertiary protein structure (Fig. 1C). Conserved domain analysis revealed that PsSEP3L1 contained a MADS-box and a K-box conserved domain (Fig. 1H). Subcellular localization studies revealed that the PsSEP3L1 protein is localized in the nucleus (Fig. 1G).

The predicted amino acid sequence of PsSEP3L1 was analyzed using BLAST (National Center for Biotechnology Information, NCBI). The results showed that PsSEP3L1 shared 62.10% sequence identity with the SEP3 protein of Malania oleifera and 60.47% similarity with that of Ipomoea batatas. Amino acid sequences of SEP3 proteins from the nine plant species most closely related to PsSEP3L1 were retrieved from the NCBI database and aligned using DNAMAN (Fig. 2A). The resulting phylogenetic analysis showed that the protein encoded by PsSEP3L1 is closely related to that of Manihot esculenta, but more distantly related to those of Vitis riparia and Herrania umbratica (Fig. 2B).

To explore the expression pattern of PsSEP3L1 in tree peony, its expression levels in different cultivars and floral organs were analyzed using qRT-PCR. The results showed that the relative expression level of PsSEP3L1 differed significantly among different cultivars and tissues (Fig. 3). In the single-petal cultivar ‘Ziban’, the relative expression of PsSEP3L1 in the outer and inner petals was 2.41- and 3.00-fold higher than that in the sepals, respectively. In the semi-double-petal cultivar ‘Baqiandaichun’, its expression in petals, petaloid petals, and stamens was 5.04, 6.88, and 5.38 times higher than that in sepals. In the double-petal cultivar ‘Doulv’, the relative expression of PsSEP3L1 in petals, petaloid petals, and stamens was 6.41, 5.52, and 3.98 times higher than that in sepals. In the multi-flowered cultivar ‘Luoyang Hong’, the relative expression level of PsSEP3L1 was 13.76-fold higher in petals, 23.73-fold higher in petaloid petals, and 10.61-fold higher in stamens compared with that in sepals (Fig. 3). Collectively, these analyses showed that PsSEP3L1 exhibited the highest expression levels in petals and petaloid petals.

The above analysis revealed that PsSEP3L1 was expressed in all floral organs with significant differences in the single-petal varieties, with higher relative expression in the outer and inner petals. In cultivars with semi-double, double, and multi-petal flowers, the relative expression levels of PsSEP3L1 in petals, petaloid stamens, and stamens were significantly higher than those in sepals. Therefore, these findings indicated that PsSEP3L1 is mainly expressed in petals, petaloid petals, and stamens, and may play a role in the formation of tree peony floral organs.

PsSEP3L1 was transiently expressed in N. benthamiana leaves, and its subcellular localization was examined using laser scanning confocal microscopy. Fluorescence signals observed in leaves infiltrated with 35S::PsSEP3L1-GFP indicated that PsSEP3L1 is localized in the nucleus (Fig. 4). This finding is consistent with the prediction.

To further investigate the function of PsSEP3L1, transient overexpression and virus-induced gene silencing (VIGS) assays were used on the flowers of the tree peony cultivar ‘Luoyang Hong’ at the big-bell stage. Observation and statistical analysis of flower organs in the flowering stage of transient plants (Fig. 5A) showed that overexpression of PsSEP3L1 significantly increased the number of tree peony petals (Fig. 5B), while silencing this gene resulted in a significant reduction in the number of petals (Fig. 5C), indicating that PsSEP3L1 positively regulates the number of tree peony petals.

RNA was extracted from petals, petaloid petals, and stamens (Pe, Pet, and St) of different transient plants, and cDNA was synthesized for quantitative real-time PCR analysis. The relative expression of PsSEP3L1 in overexpressed plants was significantly higher than that in the empty vector. The relative expression levels in petals, petaloid petals, and stamens of overexpressed plants were 2.19-fold, 3.60-fold, and 1.83-fold higher than those in the empty vector, respectively, and the highest relative expression level was found in petaloid petals (Fig. 5D). Conversely, in PsSEP3L1-silenced plants, the relative expression levels of PsSEP3L1 in petals, petaloid petals, and stamens were decreased by 33.23%, 42.32%, and 40.48%, respectively, compared with the TRV2 empty vector, and the decrease was the largest in petaloid petals (Fig. 5E). Collectively, these results demonstrate that PsSEP3L1 plays a role in regulating petal number in tree peony.

To explore the regulatory mechanism of PsSEP3L1 in flower organ number in tree peony, yeast two-hybrid experiments were performed. Using cDNA from the tree peony cultivar ‘Luoyang Hong’ petals as a template, the cDNA was cloned into pGBKT7 and pGADT7 vectors via homologous recombination. The pGBKT7-PsSEP3L1, pGADT7-PsSEP1L, pGADT7-PsAP1, pGADT7-PsAP3, and pGADT7-PsAG fusion vectors were successfully constructed. The following combinations were co-transformed into yeast cells: pGADT7-T with pGBKT7-LAM (negative control), pGADT7-T with pGBKT7-p53 (positive control), pGADT7-PsSEP1L with pGBKT7-PsSEP3L1, pGADT7-PsAP1 with pGBKT7-PsSEP3L1, pGADT7-PsAP3 with pGBKT7-PsSEP3L1, and pGADT7-PsAG with pGBKT7-PsSEP3L1. The co-transformed Y2HGold yeast strain grew normally on SD/-Leu/-Trp medium. Among them, the combinations pGADT7-T with pGBKT7-p53, pGADT7-PsSEP1L with pGBKT7-PsSEP3L1, pGADT7-PsAP3 with pGBKT7-PsSEP3L1, and pGADT7-PsAG with pGBKT7-PsSEP3L1 were able to grow on SD/-Leu/-Trp/-His/-Ade medium and produced blue colonies (Fig. 6A), indicating that PsSEP1L, PsAP3, and PsAG interact with PsSEP3L1. BiFC further assay indicated that PsSEP3L1L with PsSEP1L, PsAP3 and PsAG inter-acted in the nucleus of N. benthamiana (Fig. 6B). Given that PsSEP3L1 is a member of the MADS-box family, we speculate that it may function in combination with other MADS-box proteins to form multi-protein complexes, thereby co-regulating floral organ development in tree peony.