Genome-Wide Characterization of PEBP Gene Family and Functional Analysis of TERMINAL FLOWER 1 Homologs in Macadamia integrifolia

Edible Macadamia is one of the most important commercial nut trees cultivated in many countries, but its large tree size and long juvenile period pose barriers to commercial cultivation. The short domestication period and well-annotated genome of Macadamia integrifolia create great opportunities to breed commercial varieties with superior traits. Recent studies have shown that members of the phosphatidylethanolamine binding protein (PEBP) family play pivotal roles in regulating plant architecture and flowering time in various plants. In this study, thirteen members of MiPEBP were identified in the genome of M. integrifolia, and they are highly similarity in both motif and gene structure. A phylogenetic analysis divided the MiPEBP genes into three subfamilies: MFT-like, FT-like and TFL1-like. We subsequently identified two TERMINAL FLOWER 1 homologues from the TFL1-like subfamily, MiTFL1 and MiTFL1-like, both of which were highly expressed in stems and vegetative shoots, while MiTFL1-like was highly expressed in young leaves and early flowers. A subcellular location analysis revealed that both MiTFL1 and MiTFL1-like are localized in the cytoplasm and nucleus. The ectopic expression of MiTFL1 can rescue the early-flowering and terminal-flower phenotypes in the tfl1–14 mutant of Arabidopsis thaliana, and it indicates the conserved functions in controlling the inflorescence architecture and flowering time. This study will provide insight into the isolation of PEBP family members and the key targets for breeding M. integrifolia with improved traits in plant architecture and flowering time.


Introduction
The phosphatidylethanolamine binding protein (PEBP) family is ancient, and the encoded protein sequences are highly conserved, being present in plants, animals and microorganisms [1][2][3]. In plants, the PEBP family can be divided into three subfamilies: MOTHER OF FT AND TFL1-like (MFT-like), FLOWERING LOCUS T-like (FT-like) and TERMI-NAL FLOWERING 1-like (TFL1-like), which are involved in both the regulation of flowering time and the control of plant architecture [4,5]. MFT-like genes are the ancestors of FT-like branch proportion. Thus, TFL1 controls several aspects of plant development, including the juvenile phase, flowering time, inflorescence architecture, shoot growth pattern and life history strategies [27]. Given that woody crops generally have large architectures and long juvenile periods, they are not conducive to commercial production. PEBP family genes have potential value for breeding optimal flowering times and ideal plant types.
The genus Macadamia F. Muell. belongs to the Proteaceae family [34]. Among four common nut species of this genus, M. ternifolia and M. jansenii are small trees that produce inedible nuts, whereas M. integrifolia, M. tetraphylla and their hybrids are cultivated worldwide for their edible premium kernels [35]. Macadamia nuts have a unique flavor and are rich in lipids, proteins and important micronutrients and are favored by consumers [36,37]. M. integrifolia is a dicotyledon plant with a genome (2n = 28) size of about 652-896 Mb [38]. At present, the genome sizes of two cultivars are known: 745 Mb for HEAS 741 [39] and 794 Mb for HEAS 344 (Kau) [40]. The domestication history of Macadamia is short, the record is clear, and the release of genome data provides the possibility of improving the efficiency of Macadamia breeding [40]. In this study, we first identified all potential PEBP family genes of M. integrifolia and systematically analyzed their phylogenetic relationships, chromosome locations, gene structures, motifs and promoter cis-acting elements. In other species, TFL1 was reported to influence the first flowering time and plant architecture. To understand the function of the MiTFL1 gene in M. integrifolia, we analyzed the expression pattern of MiTFL1 and its ectopic expression in Arabidopsis to verify its function. This study will provide a reference for understanding the different roles of PEBP family members in flowering time and plant architecture regulation in M. integrifolia.

Identification of the PEBP Family in M. integrifolia
A total of 13 PEBP genes were identified in the whole genome of M. integrifolia by combining an HMM and a BLASTP search. These 13 PEBP genes encode 16 transcripts (Table 1) because MiMFT3, MiFT1 and MiFT3 have two transcripts each due to variable splicing. The M. integrifolia PEBP family proteins range from 172 to 207 amino acids in length and from 18.98 to 23.12 kDa in molecular weight (MW), with a minimum protein isoelectric point (PI) of 5.87 and a maximum of 9.47 ( Table 1). The Grand average of hydropathicity (GRAVY) values of less than 0 indicate that all MiPEBPs are hydrophilic (Table 1). Most of the 13 PEBP family genes in M. integrifolia contain four exons and three introns except for MiMFT3 and MiFT3, both of which contain five exons and four introns. Among them, the MiFT3 gene is longer than the other genes, with a length of more than 65,000 bp ( Figure 1A). A total of six conserved motifs were identified in the family of MiPEBP proteins. The motifs 1 to 5 were observed in almost all MiPEBP, and the sequences were consistent, indicating that MiPEBP genes are relatively conserved. Motif 6 was only present in some genes of the MFT-like and TFL1-like subfamilies but was not present in the FT-like subfamily ( Figure 1B). In Figure 1C, the MiPEBP genes were mainly distributed on seven chromosomes, which were chromosomes 01, 03, 05, 06, 08, 10 and 11, respectively, and the distribution was uneven. Chromosomes 03, 06 and 08 contained two MiPEBP genes, while the remaining chromosomes contained only one gene for each. In addition, MiFTL1-like and MiFT1 genes were located on two unanchored scaffolds.

Phylogenetic Analysis and Classification of the PEBP Gene Family in M. integrifolia
To investigate the relationships among the MiPEBP proteins, a phylogenetic tree was constructed via the neighbor joining method based on a multiple-sequence alignment of 126 sequences from Oryza sativa, Zea mays, Brachypodium distachyon, Sorghum bicolor, Arabidopsis thaliana, Vitis vinifera, Solanum lycopersicum, Malus domestica and Macadamia integrifolia. These 126 PEBP protein sequences were found to be classified into MFT-like, FT-like and TFL-like subfamilies (Figure 2A). Each species contains a different number of PEBP genes (Figure 2A). Compared to dicotyledonous plants, monocotyledonous plants exhibit

Phylogenetic Analysis and Classification of the PEBP Gene Family in M. integrifolia
To investigate the relationships among the MiPEBP proteins, a phylogenetic tree was constructed via the neighbor joining method based on a multiple-sequence alignment of 126 sequences from Oryza sativa, Zea mays, Brachypodium distachyon, Sorghum bicolor, Arabidopsis thaliana, Vitis vinifera, Solanum lycopersicum, Malus domestica and Macadamia integrifolia. These 126 PEBP protein sequences were found to be classified into MFT-like, FT-like and TFL-like subfamilies (Figure 2A). Each species contains a different number of PEBP genes (Figure 2A). Compared to dicotyledonous plants, monocotyledonous plants exhibit

Phylogenetic Analysis and Classification of the PEBP Gene Family in M. integrifolia
To investigate the relationships among the MiPEBP proteins, a phylogenetic tree was constructed via the neighbor joining method based on a multiple-sequence alignment of 126 sequences from Oryza sativa, Zea mays, Brachypodium distachyon, Sorghum bicolor, Arabidopsis thaliana, Vitis vinifera, Solanum lycopersicum, Malus domestica and Macadamia integrifolia. These 126 PEBP protein sequences were found to be classified into MFT-like, FT-like and TFL-like subfamilies (Figure 2A). Each species contains a different number of PEBP genes (Figure 2A). Compared to dicotyledonous plants, monocotyledonous plants exhibit higher numbers of PEBE genes ( Figure 2B). In M. integrifolia, the MFT-like subfamily harbors three genes, including MiMFT1, MiMFT2 and MiMFT3. Five genes were clustered into the FT-like subfamily, which were MiFT1, MiFT2, MiFT3, MiFT4 and MiFT5. There were five genes including MiTFL1-like, MiTFL1, MiBFT1, MiBFT2s and MiBFT3 identified in the TFL-like subfamily (Figure 2A).

Analysis of Cis-Acting Regulatory Elements in Promoter Regions of MiPEBP Genes
An analysis of 3000 bp promoter sequences on the upstream of each of the 13 MiPEBP genes revealed that there are multiple types of cis-acting elements which are involved in light-responsive, stress-responsive and hormone-responsive related elements. Among these regulatory elements, light-responsive elements are the most abundant ( Figure 3). The stress-responsive elements include drought stress response, low temperature response, wounding response, anaerobic induction, defense and stress response elements

Analysis of Cis-Acting Regulatory Elements in Promoter Regions of MiPEBP Genes
An analysis of 3000 bp promoter sequences on the upstream of each of the 13 MiPEBP genes revealed that there are multiple types of cis-acting elements which are involved in light-responsive, stress-responsive and hormone-responsive related elements. Among these regulatory elements, light-responsive elements are the most abundant ( Figure 3). The stress-responsive elements include drought stress response, low temperature response, wounding response, anaerobic induction, defense and stress response elements ( Figure 3). Hormone-responsive elements are also abundant, and the number of responsive elements ranged from the highest to lowest ranked as the MeJA-responsive element, abscisic acid responsive element, gibberellin responsive element, auxin responsive element and salicylic acid responsive element ( Figure 3). Results of the promoter cis-acting element analysis indicated that the PEBP gene family in M. integrifolia plays important roles in hormonal regulation, responses to light signals and resisting abiotic stress.

Identification and Multiple Alignment of MiTFL1 Homologue
To further analyze members of the TFL1-like subfamily of M. integrifolia, proteins which are homologous to the Arabidopsis TFL1-like subfamily and the Macadamia TFL1-like subfamily were selected to construct the phylogenetic tree. The results showed that the Macadamia TFL1-like members are mainly classified into two clades. Two are homologous to the TFL1, while three are homologous to the BFT ( Figure 4A).

Identification and Multiple Alignment of MiTFL1 Homologue
To further analyze members of the TFL1-like subfamily of M. integrifolia, proteins which are homologous to the Arabidopsis TFL1-like subfamily and the Macadamia TFL1-like subfamily were selected to construct the phylogenetic tree. The results showed that the Macadamia TFL1-like members are mainly classified into two clades. Two are homologous to the TFL1, while three are homologous to the BFT ( Figure 4A).
A multiple sequence alignment of the MiTFL1s and other TFL1 homologs showed that the key amino acid residues His88 and Asp144 in TFL1, which lead to the functional divergence between TFL1 and FT in Arabidopsis, are quite conserved in MiTFL1 and MiTFL1like ( Figure 4B). Two motifs are highly conserved among proteins of the PEBP family, D-P-D-X-P and G-X-H-R, which contribute to the conformation of the PEBP family. Ligand binding site motifs are also present in exons 2 and 4 of MiTFL1 and MiTFL1-like ( Figure 4B). Exon 4 plays a key role in the function of FT/TFL1 proteins, and the B and C segments are particularly important in the determination of the functional specificity of FT and TFL1. Unlike FT, the amino acids in segment B evolved rapidly between TFL1 homologs and show similarities between MiTFL1 and MiTFL1-like ( Figure 4B).  A multiple sequence alignment of the MiTFL1s and other TFL1 homologs showed that the key amino acid residues His88 and Asp144 in TFL1, which lead to the functional divergence between TFL1 and FT in Arabidopsis, are quite conserved in MiTFL1 and MiTFL1-like ( Figure 4B). Two motifs are highly conserved among proteins of the PEBP family, D-P-D-X-P and G-X-H-R, which contribute to the conformation of the PEBP family. Ligand binding site motifs are also present in exons 2 and 4 of MiTFL1 and MiTFL1like ( Figure 4B). Exon 4 plays a key role in the function of FT/TFL1 proteins, and the B and C segments are particularly important in the determination of the functional specificity of FT and TFL1. Unlike FT, the amino acids in segment B evolved rapidly between TFL1 homologs and show similarities between MiTFL1 and MiTFL1-like ( Figure 4B).

Subcellular Localizations of MiTFL1 and MiTFL1-like
To detect the subcellular localizations of MiTFL1 and MiTFL1-like, 35S::MiTFL1-GFP and 35S::MiTFL1-like-GFP fusion plasmids were constructed and transiently expressed in tobacco leaves. The GFP signals of MiTFL1and MiTFL1-like were distributed in both the nucleus and cytoplasm and were detected via confocal microscopy. The green fluorescent protein of the control group was distributed in the nucleus and the cytoplasm ( Figure 5).

Subcellular Localizations of MiTFL1 and MiTFL1-like
To detect the subcellular localizations of MiTFL1 and MiTFL1-like, 35S::MiTFL1-GFP and 35S::MiTFL1-like-GFP fusion plasmids were constructed and transiently expressed in tobacco leaves. The GFP signals of MiTFL1and MiTFL1-like were distributed in both the nucleus and cytoplasm and were detected via confocal microscopy. The green fluorescent protein of the control group was distributed in the nucleus and the cytoplasm ( Figure 5). These results suggest that both MiTFL1 and MiTFL1-like proteins are localized in the cytoplasm and nucleus to function.
These results suggest that both MiTFL1 and MiTFL1-like proteins are localized in the cytoplasm and nucleus to function.

Expression Patterns of MiTFL1 and MiTFL1-like in M. integrifolia
To check the expression patterns of MiTFL1 and MiTFL1-like genes, different tissues of M. integrifolia trees which were about 10 years old and grown in Jinghong, Yunnan Province, were collected ( Figure 6A-I), and the transcripts of the two genes were detected via real-time quantitative PCR. Compared with the floret-and fruit-related tissues, MiTFL1-like is mainly expressed in the vegetative shoot, stem and young leaf ( Figure 6J). It is noteworthy that the transcript of MiTFL1 is highly and specifically detected in the vegetative shoot and stem and is hardly detected in the racemes and florets of different stages ( Figure 6K). This indicates that the functional diversity between MiTFL1 and MiTFL1-like may occur at the transition from the vegetative to the reproductive development stage in M. integrifolia.

Expression Patterns of MiTFL1 and MiTFL1-like in M. integrifolia
To check the expression patterns of MiTFL1 and MiTFL1-like genes, different tissues of M. integrifolia trees which were about 10 years old and grown in Jinghong, Yunnan Province, were collected ( Figure 6A-I), and the transcripts of the two genes were detected via realtime quantitative PCR. Compared with the floret-and fruit-related tissues, MiTFL1-like is mainly expressed in the vegetative shoot, stem and young leaf ( Figure 6J). It is noteworthy that the transcript of MiTFL1 is highly and specifically detected in the vegetative shoot and stem and is hardly detected in the racemes and florets of different stages ( Figure 6K). This indicates that the functional diversity between MiTFL1 and MiTFL1-like may occur at the transition from the vegetative to the reproductive development stage in M. integrifolia.  Figure 7A). These trans-

The Role of MiTFL1 in Regulating Flowering Time
The 35S::MiTFL1 transgenic lines of Arabidopsis, which were obtained via the flower dip transformation of a tfl1-14 mutant, were planted to analyze the plant phenotypes. In contrast to the Arabidopsis tfl1-14 mutants, several MiTFL1 transgenic lines did not produce visible flower buds until 15 days after the flowering of tfl1-14 mutants, and several transgenic lines of MiTFL1 never produced visible flower buds ( Figure 7A). These transgenic lines flowered significantly later compared to the tfl1-14 mutant ( Figure 7B). The flowering time line 187# was significantly later than the flower time of the wild-type ( Figure 7B). The transgenic lines showed more rosette leaves than the mutant tfl1-14 ( Figure 7C), indicating that the transgenic lines underwent longer vegetative periods. flowering time line 187# was significantly later than the flower time of the wild-type (Figure 7B). The transgenic lines showed more rosette leaves than the mutant tfl1-14 ( Figure  7C), indicating that the transgenic lines underwent longer vegetative periods.  (Tables S3 and S4). In (B,C), a, b, c, d, different letters indicate statistically significant differences between each other. The comparison was conducted via a one-way analysis of variance (ANOVA), and a significance level of p < 0.05 was considered as statistically significant.

The Phenotypic Analysis of MiTFL1 in Arabidopsis
Compared with the tfl1-14 mutant of Arabidopsis, 35S::MiTFL1 transgenic lines had higher main stem heights, longer lateral branches and more lateral branches on the main stem, indicating that the transformation of MiTFL1 essentially restored the mutant phenotypes of tfl1-14 ( Figure 8). The meristems of the transgenic lines exhibited a longer vegetative period similar to that of the wild-type, which was different from the terminal flower phenotype of the tfl1-14 mutant (Figure 8). These results indicate the essential function of MiTFL1 in maintaining meristem activity during the vegetative phase, which is similar to that of TFL1 in Arabidopsis. Meanwhile, some transgenic lines displayed abnormal flowers and siliques, with sepal and petal fusion, pedicel elongation and lack of silique fullness (Figure 8).  (Tables S3 and S4). In (B,C), a, b, c, d, different letters indicate statistically significant differences between each other. The comparison was conducted via a one-way analysis of variance (ANOVA), and a significance level of p < 0.05 was considered as statistically significant.

The Phenotypic Analysis of MiTFL1 in Arabidopsis
Compared with the tfl1-14 mutant of Arabidopsis, 35S::MiTFL1 transgenic lines had higher main stem heights, longer lateral branches and more lateral branches on the main stem, indicating that the transformation of MiTFL1 essentially restored the mutant phenotypes of tfl1-14 ( Figure 8). The meristems of the transgenic lines exhibited a longer vegetative period similar to that of the wild-type, which was different from the terminal flower phenotype of the tfl1-14 mutant (Figure 8). These results indicate the essential function of MiTFL1 in maintaining meristem activity during the vegetative phase, which is similar to that of TFL1 in Arabidopsis. Meanwhile, some transgenic lines displayed abnormal flowers and siliques, with sepal and petal fusion, pedicel elongation and lack of silique fullness ( Figure 8).

Discussion
Flowering is a key process which indicates the transition from vegetative to reproductive growth for plants, and genes of the PEBP family play important roles during this process and in the final plant architecture [3]. Up to now, the genes of the PEBP family have been identified in many plant species, such as quinoa, perilla, sugarcane, rice, garlic, etc., and the number of PEBP genes varies among these species [41][42][43][44][45]. A total of 13 PEBP family members were identified in M. integrifolia, and they were classified into three subfamilies: MFT-like, FT-like and TFL1-like. Further duplications of the PEBP genes may occur as flowering plants evolve, which would give rise to a varied number of PEBP genes among species. Dicot species tend to have fewer PEBP genes than monocot species. In the phylogenetic tree, the FT-like subfamily of dicots showed less branches than that of monocots, suggesting that the function of the FT gene may be more complex in monocots. In most MiPEBP family members, the gene structure is very conserved, with four exons and three introns. Almost all MiPEBP members contain motifs 1 to 5, which are arranged in a consistent order. This illustrates that members of MiPEBP family are relatively conserved

Discussion
Flowering is a key process which indicates the transition from vegetative to reproductive growth for plants, and genes of the PEBP family play important roles during this process and in the final plant architecture [3]. Up to now, the genes of the PEBP family have been identified in many plant species, such as quinoa, perilla, sugarcane, rice, garlic, etc., and the number of PEBP genes varies among these species [41][42][43][44][45]. A total of 13 PEBP family members were identified in M. integrifolia, and they were classified into three subfamilies: MFT-like, FT-like and TFL1-like. Further duplications of the PEBP genes may occur as flowering plants evolve, which would give rise to a varied number of PEBP genes among species. Dicot species tend to have fewer PEBP genes than monocot species. In the phylogenetic tree, the FT-like subfamily of dicots showed less branches than that of monocots, suggesting that the function of the FT gene may be more complex in monocots. In most MiPEBP family members, the gene structure is very conserved, with four exons and three introns. Almost all MiPEBP members contain motifs 1 to 5, which are arranged in a consistent order. This illustrates that members of MiPEBP family are relatively conserved in sequence similarity, consistent with the high conservation of the PEBP gene among different species. The complex physiological process of flowering is regulated by photophore, hormone biosynthesis and signaling [46]. An analysis of the promoter elements of the MiPEBP family revealed the presence of a large number of different types of cis-acting elements on their promoter regions, with an abundance and high number of light-and hormone-responsive element types, implying that they may play important roles in regulating the photoperiod pathway and gibberellin pathway of plant flowering. Thus, the characterization of PEBP gene functions is a key means of fine-tuning the flowering of M. integrifolia.
Functional studies of PEBP genes have been widely reported in different species. FT and TFL1 are two important genes acting downstream of the flowering regulation network [47]. FT and TFL1 have highly similar amino acid sequences but have antagonistic functions for flowering time in plants [48]. This functional difference is only caused by the differentiation of one critical amino acid residue and a conserved amino acid segment in the PEBP domain [48]. Several other genes have been reported to regulate flowering. Within the SAM domain, TFL1 forms a complex with the bZIP transcription factor FD and 14-3-3 proteins to repress the expression of the flowering-time-related genes LFY, AP1 and CAL, thereby repressing the floral transition [22]. The photoperiod signal is transmitted to FT by CO, and then the FT protein begins to translocate from the leaf to the SAM [49]. Similar to TFL1, FT competitively binds FD and activates the downstream flowering-related genes SOC1, AP1 and LFY to promote flowering [50]. In addition to delaying flowering time, TFL1 influences the growth habit and inflorescence architecture in plants. The TFL1-like gene CENTRORADIALIS (CEN) of Antirrhinum majus was the first PEBP gene identified in plants. The flowering time in the cen mutant was unaffected, but the plants were short and compact, and the inflorescences terminated early [51]. Rice (Oryza sativa) has four TFL1 homologous: RCN1, RCN2, RCN3 and RCN4. RCN1 and RCN2 delay flowering and increase the number of tillers [52]. Similarly, knocking out RCN genes in rice will lead to smaller panicles and fewer branches [53]. In soybean (Glycine max), GmTFL1b/Dt1 is involved in the stem determinate growth [54]. The tomato TFL1 ortholog SP regulates the indeterminate growth habit of the apical meristem and delays flowering time [55]. All these comparative studies indicate the conserved function of TFL1 in the inflorescence architecture and flowering among different plant species. The expression of MiTFL1 genes in different tissues of M. integrifolia was analyzed. The expression patterns of MiTFL1 and MiTFL1-like were similar, and both were expressed in the stem and vegetative shoot. MiTFL1-like is also expressed in young leaves and early flowers, suggesting that MiTFL1-like may play other functions during flower and leaf development. All transgenic Arabidopsis lines which overexpressed MiTFL1 showed delayed anthesis and indeterminate inflorescence. Based on the above studies, we have confirmed that there are significant differences in expression patterns between MiTFL1 and MiTFL1-like, with MiTFL1 having a conserved function in flowering regulation and plant architecture. In view of the conservative function of the TFL1 gene in flowering regulation and plant architecture, other Macadamia cultivars may also show similar expression patterns and functions.
PEBP-family-related genes mainly play important roles in regulating seed dormancy, flowering time and plant architecture [3,56], and can solve many problems in the actual production process, such as improving seed germination rate, prolonging the vegetative growth period of vegetable crops, prolonging the flowering period of ornamental plants and shortening the juvenile state of woody plants in breeding. In agricultural production, the sp mutant of tomato has been widely used in tomato breeding because of its shorter vegetative growth period, earlier flowering period and earlier fruit maturity [24]. Cucumber plants with function deletions of the CsTFL1 gene showed significantly limited vegetative growth, and the plant height was greatly reduced, which was more convenient for management and had great application potential [57]. Different from herbaceous crops, Macadamia is a large subtropical rainforest tree which originates from Australia [58]. Cultivated Macadamia nuts grow at an average height of 12-15 m and require an average of 15 years to reach peak yield [58,59]. Both the large plant size and long juvenile period hamper commercial production and breeding efficiency [35,60]. Large trees require pruning to maintain productivity, while the small tree variety enables high planting density and early production [35]. Traditional methods to breed new Macadamia cultivars are timeconsuming, laborious and expensive [61]. Genetic engineering of Macadamia will accelerate the development of new cultivars with excellent agronomic traits [62]. This research indicates that the PEBP genes have potential advantages for breeding valuable Macadamia cultivars with early flowering times and commercial plant architectures.

Plant Materials and Growth Conditions
The M. integrifolia (HEAS 863) trees used in this study were planted in a Macadamia germplasm base of the Yunnan Institute of Tropical Crops. Different organs from the stem, leaf, vegetative shoot, inflorescence, flower and nutlet were collected, placed in liquid nitrogen, and stored at −80 • C for RNA extraction.
Arabidopsis thaliana wild type Col-0 and tfl1-14 mutant seeds were sown after 3 days of treatment at 4 • C. Nicotiana benthamiana seeds were sown on wet soil. Arabidopsis and tobacco grew in a greenhouse under a 16/8 h light (150 µE m −2 s −1 )/dark cycle and 24/20 • C conditions at a relative humidity of 50-60%; the plants were well-watered and received adequate nutrition.

Identification of PEBP Family Genes in M. integrifolia
In order to identify PEBP genes from M. integrifolia, the annotated genome data of M. integrifolia were downloaded from GenBank (https://www.ncbi.nlm.nih.gov/genbank/, GCF_013358625.1, accessed on 15 June 2022). The hidden Markov model (HMM) profile of the PBP domain (PF01161) was obtained from the Pfam database and used as the query; then, M. integrifolia amino acid sequences were also searched using HMMER v3.3.2, E-value ≤ 10 −5 . The protein sequences of six PEBP members of A. thaliana were downloaded as query sequences from the Arabidopsis database (https://www.arabidopsis.org/, accessed on 15 June 2022). BLASTP was performed with the protein sequences of M. integrifolia, E-value ≤ 10 −5 . Candidate sequences obtained via the two methods were uploaded to the Pfam website (http://pfam-legacy.xfam.org/, accessed on 17 June 2022) for further comparison and screening based on the PEBP domain, and it was confirmed that all members contained conserved PEBP domains. The physical and chemical parameters of the PEBP proteins, including the theoretical isoelectric point (PI), molecular weight (MW) and grand average of hydropathicity (GRAVY), were determined using ProtParam (https://web.expasy.org/protparam/, accessed on 6 August 2022).

Gene Structures, Protein Motifs and Chromosome Locations
The exon and intron locations of the PEBP genes were analyzed by comparing the coding sequences with their genome sequences. The Multiple Em for Motif Elicitation (MEME) online tool (https://meme-suite.org/meme/tools/meme, accessed on 15 August 2022) was used to predict protein motifs. The chromosome distributions of PEBP genes were obtained based on genome GFF3 files. Finally, the gene structures, protein motifs and chromosome locations were visualized with the software TBtools (v1.120) [63].

Phylogenetic Analyses and Multiple Alignments
Phylogenetic trees were generated by using MEGA 7.0 with the neighbor-joining (NJ) algorithm. Bootstrap values from 1000 replications were used to assess group support, and the substitution model used was the Poisson model. The phylogenetic tree was visualized with the iTOL tool (https://itol.embl.de, accessed on 2 May 2023). Multiple alignments of the sequences were performed using the DNAMAN software. Gene accession numbers are listed in Supplementary Table S1.

Analysis of Cis-Acting Regulatory Elements in Promoter of MiPEBPs
The upstream 3000 bp promoter sequences of 13 MiPEBP genes were uploaded to the PlantCARE Database (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/, accessed on 12 March 2023) to search for cis-acting elements. The physical distribution of the various cis-acting elements was visualized in TBtools (v1.120) [63].

RNA Extraction and qRT-PCR
Total RNA from M. integrifolia plants was extracted via a modified CTAB method, and total RNA from Arabidopsis plants was extracted using a RnaEx™ Total RNA isolation kit (GENEray Biotech, Shanghai, China). The RNA was quantified using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Shanghai, China), and the quality was checked via electrophoresis. The first-strand cDNA was reverse-transcribed with the HiScript ® II 1st Strand cDNA Synthesis Kit (+gDNA wiper) (Vazyme, Nanjing, China). A quantitative real-time PCR was performed using Magic SYBR Green qPCR Mix (Magic-bio, Hangzhou, China) on a LightCylcer 480 device (Roche, Basel, Switzerland). AtActin and MiActin were used as internal reference genes. The data were calculated using the 2 −∆∆Ct method. Primer sequences are listed in Supplementary Materials Table S2.

Vector Construction and Plant Transformation
The full-length coding sequence of MiTFL1 was amplified via PCR from the cDNA of vegetative shoots with Phanta MaxSuper-Fidelity DNA Polymerase (Vazyme, Nanjing, China). All PCR products were detected via 1.5% agarose gel electrophoresis and purified using an EasyPure ® Quick Gel Extraction Kit (TransGen, Beijing, China). The PCR-amplified fragment products were inserted into the pCAMBIA3301 vectors between the NcoI and BstEII sites via a ClonExpress II One Step Cloning Kit (Vazyme, Nanjing, China) under the CaMV35S promoter. The recombinant plasmid was transformed into the Agrobacterium tumefaciens EHA105 strain for transformation into the A. thaliana tfl1-14 mutant. Arabidopsis plants were transformed via the floral dip method [64]. T0 transgenic seeds were sown in the soil and then selected by spraying Basta. The identification of T1 transgenic plants is shown in Supplementary Materials Figure S1.

Subcellular Localization
The CDS sequences of MiTFL1 and MiTFL1-like were amplified and connected to the pPYS22-GFP vector between the XhoI and KpnI sites via homologous recombination. The recombinant plasmids were transferred into Agrobacterium tumefaciens strain EHA105. They were centrifuged after overnight incubation and resuspended as an Agrobacterium pellet in an infiltration solution (10 mM of MgCl 2 , 10 mM of MES and 100 µM of acetosyringone, PH = 5.6) to a desired optical density (OD600 =1.0). The 4-week-old tobacco leaves were infected with the mixed solution. After 2 days in the dark, the GFP signal was observed via a laser confocal microscope and the empty vector was used as a control.