Tomato AC (S. lycopersicum cultivar “Ailsa Craig”) with two locules was used as material for constructing lc ^CR mutants, lc ^CR T₁ generation was grown in a greenhouse (16 h of light at 25°C, 8 h of dark at 15°C), and conventional cultivation management was carried out.

Young leaves from the lc ^CR T₁ generation were sampled for target detection. The shoot apicals from both lc ^CR T₁ generation and AC were taken during the carpellary formation stage for qRT-PCR analysis; three biological replicates were set up for each sample. The shoot apical from AC were taken at different stages of flower bud differentiation, including early flower bud differentiation (S₁), middle flower bud differentiation (sepal and petal formation stage (S₂)), and late flower bud differentiation stage (carpel formation stage (S₃)). Every 15–20 shoot apical samples were collected for library construction and qRT-PCR analysis. Both samples were stored in an ultra-low-temperature refrigerator to be determined.

One single guide RNA (sgRNA) was designed to target lc for obtaining genome-edited mutants. The sgRNA sequence is listed in the Supplementary Table . The U6-lc-sgRNA was cloned into CRISPR/Cas9 binary vector pCBC-DT1T2-tomatoU6 to generate pCBC-DT1T2-tomatoU6-lc. This construct was then transformed into competent cells of Agrobacterium (Agrobacterium tumefaciens) strain LBA4404 using electroporation. Positive colonies obtained were used to transform the wild-type tomato AC through leaf disc cocultivation. Each primary transformant (T ₀) was genotyped by sequencing PCR products amplified from the relevant target region.

The plant genomic DNA extraction kit (Tiangen, Beijing, China) was used to extract genomic DNA, following the manufacturer’s instructions (Code No. DP305, Tiangen). The extracted products were sequenced and compared with Cas9/gRNA using DNAMAN9.0 to determine whether the target was mutated and the specific mutation mode. Primers used are listed in the Supplementary Table .

lc ^CR T ₁ generation mutants and wild-type plants were selected for 10 strains, respectively, and the number of sepals, stamens, petals, and locules in the first inflorescence flowers and fruit were investigated. Additionally, the number of sepals, petals, and fruit locules of flowers in the first and second inflorescence for SlSEP3 silenced strains and wild-type plants were investigated. Data statistics and significant differences were calculated and analyzed by Excel software.

Total RNA from shoot apical of lc ^CR mutants and AC was extracted using TRIzol reagent (Invitrogen, Carlsbad, California, USA). The cDNA was synthesized using PrimeScript RT Master Mix (TaKaRa, Kusatsu, Shiga, Japan) and diluted fourfold in ddH₂O. Gene-specific primers were designed using the National Center for Biotechnology Information (NCBI) website (https://www.ncbi.nlm.nih.gov/tools/primer-blast/index.cgi). Real-time quantitative analysis was performed using TB Green^® Premix Ex Taq II (TaKaRa, Japan). The 20-μL reaction system contained 2 μL of diluted cDNA, 10 µL of 2 × TB Green Premix Ex Taq, and 0.3 μM of each primer with a qTOWER3/G real-time system (Analytik Jena, Jena, Thuringia, Germany). The reaction program consisted of 95°C for 30 s, 95°C for 5 s (45 cycles), 60°C for 34 s (45 cycles), and Melt for 15 s. The tomato actin (NCBI: NM_001330119.1) was used as an internal control. Primers are listed in the Supplementary Table .

Total RNA was extracted from shoot apicals at different stages of flower bud differentiation using TRIzol reagent (Invitrogen, USA). The extracted RNA was then evaluated for integrity, purity, and concentration. An equal amount of RNA from each stage was mixed, and 1 μg was used as a template. The first strand of cDNA was synthesized using SMART cDNA Library Construction Kit (Taraka, Japan) according to the manufacturer’s instructions (Code No. 634901, Taraka). The second strand cDNA was amplified using Advantage 2 PCR Kit (Taraka, Japan) following the manufacturer’s instructions (Code No. 639206, Taraka). Double-stranded cDNA was purified using the DNA Fragment Kit (Taraka, Japan) according to the manufacturer’s instructions (Code No. 9761, Taraka), and SfiI was used for enzyme digestion. Short fragments were purified using CHROMA SPIN-1000-TE (Taraka, Japan) according to the manufacturer’s instructions (Code No. 636079, Taraka). The cDNA sample with the removed short fragment was attached to three pGADT7-SfiI containing different reading frames using a DNA ligation Kit (Taraka, Japan) according to the manufacturer’s instructions (Code No. 6022, Taraka). Three primary cDNA libraries were transformed into Escherichia coli HST08 premium electro-cells by electric shock according to the manufacturer’s instructions (Code No. 9028, Taraka). The capacity of the library was calculated by diluting the transformed products 10³-fold and culturing them on an LB agar medium. The number of single colonies obtained was multiplied by dilution times and the total volume of solution (mL) and then divided by the volume of culture medium (mL) to calculate the capacity of the library. To use 2×Taq PCR Mastermix (Tiangen, Beijing) for PCR amplification, 48 single colonies were randomly selected. The primary cDNA library was transformed into HST08 again, and all single colonies were collected for germiculture. The Plasmid Purification Kit (Taraka, Japan) was used to extract the plasmid according to the manufacturer’s instructions (Code No. 9760, Taraka). Finally, the concentration of the final cDNA library was detected.

The bait fragment used for yeast one-hybrid library screening was designed based on the CArG element found in the wild-type lc. It consisted of three tandem copies of a specific nucleotide sequence, which is listed in the Supplementary Table . Two oligonucleotide chains of the bait fragment were synthesized using the gene-synthesis technique and mixed in a 1:1 ratio. The reaction program consisted of 95°C for 30 s, 72°C for 2 min, 37°C for 2 min, 25°C for 2 min, and storage at 4°C. To generate the bait vector, the bait sequence was cloned into the pAbAi vector, which was digested with XhoI and SmaI restriction enzymes. The resulting construct was named pAbAi-lc. Next, the bait vector pAbAi-lc was digested with BbsI and BstbI enzymes. The p53 vector was used as a positive control. The linearized pAbAi-lc and p53 vectors were then transformed into Y1HGold yeast-competent cells using Yeast Transformation System 2 (Taraka, Japan) according to the manufacturer’s instructions (Code No. 630439, Taraka). The transformed products were detected using Matchmaker Insert Check PCR Mix 1 (Taraka, Japan). The concentration of aureobasidin A (AbA) in the bait and positive control yeast strains was determined using Matchmaker Gold Yeast One-Hybrid Library Screening System Kit (Clontech, Japan, Mountain View, CA, USA) according to the manufacturer’s instructions (Code No. 630491, Clontech). For library screening, 15 μg of the cDNA library was transformed into Y1HGold[pAbAi-lc] yeast-competent cells. The resulting product was diluted 10-, 100-, and 10³-fold, respectively, and then cultured on SD/-Leu and SD/-Leu/AbA medium (90 mm). The number of colonies was counted on the 100-fold dilution medium, and the rest of the reaction liquid was cultured on SD/-Leu/AbA medium (150 mm). The number of colonies obtained was calculated using the formula: number of single colonies × dilution times × suspension volume (mL) ÷ volume on medium (mL). Finally, the colonies on the 150-mm medium were detected using Matchmaker Insert Check PCR Mix 2 (Taraka, Japan).

The coding sequence of SlSEP3 was inserted into the pGADT7 vector (Clontech), while the CArG element in the wild-type lc was cloned into the pAbAi vector (Clontech) individually. The assay was performed with the Matchmaker Gold Yeast One-Hybrid Library Screening System Kit (Clontech, Japan) according to the manufacturer’s instructions (Code No.630491, Clontech). Primers used are listed in the Supplementary Table .

The coding sequence of SlSEP3 was cloned into the pGADT7 vector and transformed into Y1HGold yeast-competent cells. Positive colonies were cultured in a YPDA liquid medium containing 50 mg mL⁻¹ ampicillin at 30°C until the OD₆₀₀ value reached 1.0. Yeast protein was extracted using a yeast protein extraction kit (Solarbio, Beijing, China) following the manufacturer’s instructions (Code No. BC3780, Solarbio). The extracted protein was purified using an Amicon^® Ultra-4 centrifugal filter (Merck, Burlington, MA, USA) to remove small molecule salts and heteroproteins according to the manufacturer’s instructions (Code No. UFC8010, Merck). The purified protein was dissolved in 1× PBS buffer and stored at −20°C in a refrigerator.

Electrophoretic mobility shift assay (EMSA) assay was performed using a Chemiluminescent EMSA Kit (Beyotime, Shanghai, China) according to the manufacturer’s instructions (Code No. GS009, Beyotime). The biotin-labeled probe was prepared by SBS Genetech Co. Ltd. (Beijing, China). All oligonucleotide probes used are listed in the Supplementary Table .

Protein sequences of E-class genes involved in flower development in Arabidopsis and tomatoes were retrieved from the NCBI website. Multiple sequence alignment was conducted using Clustalw, and the neighbor-joining model was employed for constructing the phylogenetic tree.

SlSEP3 silencing fragments were designed using the virus induced gene silencing (VIGS) tool on the Sol Genomics Network (https://solgenomics.net/) and amplified by PCR, then ligated into the pTRV2 vector to generate the pTRV2-SlSEP3 constructs. The pTRV2-SlSEP3, pTRV2 empty vector, and pTRV1 empty vectors were transformed into Agrobacterium (Agrobacterium tumefaciens) strain GV3101 individually. A mixture of cultures containing pTRV1 and pTRV2-SlSEP3 was used for infection as previously described (Liu et al., 2002). Two infections were performed, one on germinated seeds and the other on stems. Silencing efficiency was determined using qRT-PCR on both young buds and leaves below the infected site. The primers used are listed in the Supplementary Table .

The coding sequence of the histone deacetylase SlHDAs was cloned into the pGADT7 vector (Clontech, Japan), while the coding sequence of SlSEP3 was cloned into the pGBKT7 vector (Clontech, Japan). The assay was performed with the Matchmaker Gold Yeast Two-Hybrid System Kit (Clontech, Japan) according to the manufacturer’s instructions (Code No. 630489, Clontech, Japan). Primers used are listed in the Supplementary Table .

To investigate the molecular function of lc in affecting tomato locule number, four lc mutants (#11, #12, #9, and #22) were generated using the CRISPR/Cas9 system under AC background. The lc target region in T ₁ generation plants was amplified and sequenced, revealing a 5-bp deletion, 1-bp insertion, 1-bp deletion, and 2-bp deletion in the CArG motif of the target sequences of T ₁ generation lines ( Figure 1A ). Compared with WT, the homozygous CR-lc#11 mutant showed an increase in floral organ number ( Figures 1B, C ; Supplemental Dataset ), with some partial flowers composed of five to seven sepals, five to six stamens, and five to seven petals ( Figure 1C ). Additionally, the locule number increased from two to three ( Figure 1D ), with 14.2% of the fruits from the CR-lc#11 mutant exhibiting three locules ( Figure 1E ; Supplemental Dataset ). However, most flowers in the CR-lc#11 mutant remained unchanged from the WT, with flowers composed of four to five sepals, four to five stamens, and four to five petals. In addition, we also analyzed the phenotype of flower organ number and locule number in homozygous CR-lc#12 mutant ( Supplementary Figure S1 , Supplemental Dataset ). Similar to the CR-lc#11 mutant, the number of petals, sepals, and stamens increased compared with WT in the CR-lc#12 mutant. Furthermore, 10.7% of the fruits exhibited three locules. These findings suggest that the lc mutation affects the differentiation of stem cells in partial flowers.

In the CR-lc#11 mutant, we found that SlWUS expression was significantly increased during the differentiation of stem cells, while the expression of SlCLV3, SlTAG1, and SlCLE9 was not changed ( Figure 1F ; Supplemental Dataset ). It suggested that lc might change locule number by affecting SlWUS transcription.

To investigate the upstream regulation of tomato lc, we first constructed a shoot apical cDNA library for screening proteins that bind to lc. We extracted shoot apical RNA from different stages of flower bud differentiation in tomato AC ( Supplementary Figure S2A ) and synthesized double-stranded cDNA from equal amounts of total RNA. The resulting electrophoretic bands of double-stranded cDNA fragments were dispersed, with a majority concentrated in the 500- to 3,000-bp range ( Supplementary Figure S2B ). We purified the cDNA by removing highly expressed components and retaining the effective fragments within the enriched interval ( Supplementary Figure S2C ). The normalized full-length cDNA was then digested with SfiI ( Supplementary Figure S2D ) and ligated to a three-frame pGADT7-SfiI vector to generate the primary cDNA library. We ultimately transformed the primary cDNA library into competent cells from Escherichia coli and measured its capacity, which was calculated based on the number of colonies. The primary cDNA library capacity of the three reading frame vectors was 2.6 × 10⁶ cfu, 1.6 × 10⁶ cfu, and 2.4 × 10⁶ cfu, respectively. Finally, we detected the length of inserted fragments, which was mainly distributed in the 500- to 3,000-bp range ( Supplementary Figure S2E ). We mixed the colonies for germiculture in an LB liquid medium, the concentration of the final cDNA library plasmid was 1 mg mL⁻¹.

Secondly, we performed a yeast one-hybrid screen. We selected an OD₆₀₀ value of 0.002 to screen for the appropriate concentration of AbA ( Supplementary Figure S3A ). We found that 150 ng mL⁻¹ of AbA completely inhibited yeast growth. To ensure the reliability of subsequent library screens, we selected a concentration of 200 ng mL⁻¹ of AbA. We transformed the final cDNA library into a bait yeast strain and used the P53 positive yeast strain as a control. The number of positive control colonies on SD/-Leu and SD/-Leu/AbA²⁰⁰ agar mediums was roughly the same, but the number of bait yeast strain colonies on SD/-Leu/AbA²⁰⁰ agar medium was significantly lower than that on SD/-Leu agar medium. We counted 137 single colonies on SD/-Leu/AbA²⁰⁰ agar medium ( Supplementary Figure S3B ). The number of obtained clones was calculated to be 137 cfu/0.1 mL × 100 mL × 15 mL = 2.055 × 10⁶, which was greater than 10⁶, indicating that the library screen test was successful.

Finally, we selected every single colony on the 150-mm agar medium for PCR detection and sequencing. The obtained results were analyzed on the Tomato Genomics Network, and we obtained statistical information about the function of the screened factors ( Table 1 ). Eventually, two candidate transcription factors were identified: Solyc05g015750, which encodes SEPALLATA3 (SEP3) of the tomato MADS-box transcription factor, and Solyc02g092370, which encodes GARS11 of the tomato GRAS family transcription factor. Based on the function of transcription factors, we considered SlSEP3 to be a potential candidate for binding to lc.

To ensure the reliability of yeast one-hybrid screening results, we implemented a yeast one-hybrid assay and an electrophoretic mobility shift assay. We recombined the coding region of SlSEP3 with the pGADT7 vector and performed a yeast one-hybrid assay with the CArG element in wild-type lc. We used an empty pGADT7 vector as the negative control and mutant lc as a control ( Figure 2A ). The results demonstrated that the SlSEP3 transcription factor interacted with the wild-type lc but not with the mutant lc. Furthermore, we transferred the recombinant expression vector pGADT7-SlSEP3 into yeast Y1HGold-competent cells for cultivation and obtained a purified recombinant HA-tagged SlSEP3 protein from yeast for the EMSA. To exclude any influence from HA-tagged yeast protein, it was set as the experimental control. We confirmed that HA-SlSEP3 yeast protein bound to the wild-type lc, and the binding ability decreased when the concentration of the cold competition probe was increased to 200 times ( Figure 2B ). Taken together, the results from the yeast one-hybrid assay and EMSA provided strong evidence supporting the reliability of the yeast one-hybrid screening.

In order to elucidate the role of SlSEP3 in regulating tomato locule number, we performed qRT-PCR analysis to examine the expression level of SlSEP3 and SlWUS in the shoot apical ( Figure 3 ). Our findings revealed that SlWUS expression was downregulated during flower bud differentiation, particularly after the early stage, whereas the expression of SlSEP3 was upregulated, especially after the sepal and petal formation. Notably, the expression patterns of SlSEP3 and SlWUS were found to be opposite to each other. Given that SlSEP3 functions as a transcription factor that interacts with lc, it is plausible to speculate that SlSEP3 may serve as an inhibitor of SlWUS. In addition, we performed a phylogenetic analysis of the E-class genes that influence tomato flower development. There are five E-class genes in tomato, and upon comparing them with the AtSEP in Arabidopsis, it was observed that tomato SlSEP3 clustered preferentially with AtSEP3. This suggests that tomato SlSEP3 shares a similar function with AtSEP3 in Arabidopsis. By analyzing the expression patterns of other E-class genes at different stages of floral bud differentiation in wild-type tomatoes, we found that the SlMADS1 and SlSEP3 might have similar expression patterns. However, further research is needed to determine whether there is functional redundancy between these two genes in regulating fruit locule number.

To further investigate whether SlSEP3 is a negative regulator of fruit locule number, we generated a set of tomato lines with individually silenced SlSEP3 using virus-induced gene silencing and analyzed SlSEP3 expression in young buds and leaves by qRT-PCR ( Figures 4A, B ; Supplemental Dataset ). Compared with wild-type plants, pTRV-SlSEP3 plants displayed an increase in the number of floral organs in both the first and second inflorescence ( Figures 4C , 5A, B ; Supplemental Dataset ). In the first inflorescence, only 20% of the flowers in wild-type plants had six petals and six sepals, whereas 60% of flowers in pTRV-SlSEP3 plants had six petals and 30% of the flowers had six sepals ( Figure 5A ). Moreover, the number of petals in the second inflorescence of pTRV-SlSEP3 plants was significantly increased ( Figure 5B ), particularly in pTRV-SlSEP3-3 plants where seven and eight petals were observed ( Figure 4C ). Furthermore, there was an increase in the number of locules in both the first and second inflorescence of pTRV-SlSEP3 plants, with 20% of the fruits in the first inflorescence and 30% of the fruits in the second inflorescence of pTRV-SlSEP3 plants having three locules ( Figures 4D , 5C, D ). These results suggested that SlSEP3 acts as a negative regulator of tomato locule number.

To further investigate the regulation of SlWUS expression by the tomato SlSEP3 transcription factor after binding to lc, we conducted an analysis of the interaction between SlSEP3 and tomato histone deacetylase, drawing upon the functions of homologous genes in Arabidopsis ( Figure 6 ). We cloned the coding sequences of SlHDA1–9 and SlHDT1–3 from the tomato histone deacetylase family and performed yeast two-hybrid assays. The results provide preliminary evidence that SlSEP3 could interact with several tomato histone deacetylases, including SlHDA1, SlHDA3, SlHDA4, SlHDA5, SlHDA6, SlHDA7, and SlHDA8 in yeast cells.