PpTCP18 is upregulated by lncRNA5 and controls branch number in peach (Prunus persica) through positive feedback regulation of strigolactone biosynthesis

Abstract Branch number is an important agronomic trait in peach (Prunus persica) trees because plant architecture affects fruit yield and quality. Although breeders can select varieties with different tree architecture, the biological mechanisms underlying architecture remain largely unclear. In this study, a pillar peach (‘Zhaoshouhong’) and a standard peach (‘Okubo’) were compared. ‘Zhaoshouhong’ was found to have significantly fewer secondary branches than ‘Okubo’. Treatment with the synthetic strigolactone (SL) GR24 decreased branch number. Transcriptome analysis indicated that PpTCP18 (a homologous gene of Arabidopsis thaliana BRC1) expression was negatively correlated with strigolactone synthesis gene expression, indicating that PpTCP18 may play an important role in peach branching. Yeast one-hybrid, electrophoretic mobility shift, dual-luciferase assays and PpTCP18-knockdown in peach leaf buds indicated that PpTCP18 could increase expression of PpLBO1, PpMAX1, and PpMAX4. Furthermore, transgenic Arabidopsis plants overexpressing PpTCP18 clearly exhibited reduced primary rosette-leaf branches. Moreover, lncRNA sequencing and transient expression analysis revealed that lncRNA5 targeted PpTCP18, significantly increasing PpTCP18 expression. These results provide insights into the mRNA and lncRNA network in the peach SL signaling pathway and indicate that PpTCP18, a transcription factor downstream of SL signaling, is involved in positive feedback regulation of SL biosynthesis. This role of PpTCP18 may represent a novel mechanism in peach branching regulation. Our study improves current understanding of the mechanisms underlying peach branching and provides theoretical support for genetic improvement of peach tree architecture.


Introduction
Tree architecture is of great interest to plant breeders because it is a critical factor that affects the management of orchards, forests, and parks [1,2]. In fruit crops, a proper balance between vegetative and reproductive growth is a key agronomic trait that has a great impact on fruit yield and quality [3]. Peach [P. persica (L.) Batsch] is the most economically important deciduous fruit crop in the world (FAO, http://faostat.fao.org); it exhibits excessive vegetative growth compared with other fruit trees, such as apple and pear. Reducing peach tree branching would reduce the need for pruning, decrease orchard management expenses, and increase fruit yield and quality [4]. Compared with standard peach trees, pillar peach trees require less pruning and are promising candidates for use in high-density peach production systems due to their narrow branch angles and fewer second-order branches [5]. However, the molecular mechanisms underlying peach branching are unclear. It is therefore important to explore the biological processes that yield low vegetative growth to allow breeding of peach cultivars that are conducive to labor-saving cultivation.
Tree architecture is mainly inf luenced by the spatial patterning of branches, which is determined by branching (branch number), branch angle, and internode length [1,[6][7][8][9]. Branching is controlled by a complex and interconnected regulatory network that responds to genetic, hormonal, and environmental factors [10,11]. Decades ago, auxin was found to regulate meristem fate and was considered the major plant growth regulator (PGR) involved in shaping plant architecture by indirectly inhibiting branching [10,12]. Exogenous auxin was also found to significantly inhibit branching [13].
The role of auxin in apical dominance was thought to be indirect [10]. Characterization of high-branching mutants in numerous species, including more axillary growth (max) in Arabidopsis, dwarf (d) in rice, and ramosus (rms) in pea revealed that the longdistance signaling molecule strigolactone (SL) is the key PGR that controls shoot branching [14][15][16]. The second messengers model posits that SLs are secondary messengers of auxin signaling [10]. SLs are synthesized through the β-carotene pathway via three oxidative reactions, which are carried out by a carotene isomerase (D27), two carotenoid cleavage dioxygenases (CCD7/MAX3 and CCD8/MAX4), and a cytochrome P450 (MAX1); bioactive SL or SL-like compounds are synthesized through multiple routes and involve in a range of enzymes including LATERAL BRANCHING OXIDOREDUCTASE (LBO) [17]. Bioactive SLs interact with the receptor protein D14/AtD14/DAD2/RMS3, a α/β-hydrolase [18,19]. A complex is formed between SL and one of the three primary targets, the D53-like SUPPRESSOR OF MAX2 1 (SMAX1)-LIKE (SMXL) proteins D53/SMXL6,7,8. During perception, D14 and D53/SMXL6,7,8 are hydrolyzed by 26S proteasomes [20][21][22]. Mutations in SL biosynthesis and signaling genes, such as OsMAX4, OsD14 [23], OsD53 [24], and AtMAX1 [25], increase branching/tillering by decreasing endogenous levels of SL or impacting SL signal transduction. This indicates that SL plays an important role in controlling branching/tillering. Transcription factors in the TEOSINTE BRANCHED1/ CYCLOIDEA/PROLIFERATING CELL FACTOR (TCP) family are downstream targets of SL signaling and are key factors involved in controlling branching. TCP family members, such as TB1, BRC1, and FC1 [9][10][11]14], contain a 59-residue homologous region named the TCP domain. BRC1 sub-family genes in Arabidopsis, rice, and pea are predominantly expressed in the axillary buds and are involved in bud dormancy; mutation of these genes increases branching/tillering [14,26,27]. Recently, these proteins were reported to be involved in a different pathway that controls branching. Brassinosteroid (BR) signaling protein (BES1) regulated branching by inhibiting BRC1 expression through direct binding to the promoter and subsequent SMXL recruitment [28]. Likewise, SQUAMOSA PROMOTER BINDING PROTEIN-LIKE 14 (SPL14) inf luences branching by regulating expression of OsTB1 through a GTAC cis-element in the promoter [29].
Long noncoding RNAs (lncRNAs), which are >200 nt and usually lack protein-coding potential, are transcribed by RNA polymerase II, III, or IV/V in plants [30]. LncRNAs contribute to regulation of gene expression by inf luencing DNA methylation, histone modification, and chromatin structure, and are potent cis-and trans-regulators of gene expression [31]. Although lncRNAs have been the focus of increased research interest in recent years, the field of lncRNA research is still in its infancy. This is especially true in plants, in which only a few lncRNAs have been studied and sufficiently described [32][33][34][35].
Although the functions of TCP genes have been characterized in model plants, the molecular mechanisms controlling branching/tillering have not been well characterized in woody plant, including peach. In this study, the mechanisms of PpTCP18 controls peach branching through regulating SL biosynthesis were investigated. GR24 treatment could obviously inhibit peach branching. PpTCP18 could regulate SL biosynthesis through a positive feedback mechanism by directly activating PpLBO1, PpMAX3, and PpMAX4 expression. Furthermore, the expression of PpTCP18 could be dramatically increased by lncRNA5. These results will shed light on how PpTCP18 may regulate peach branching at the molecular level.

The pillar peach cultivar had fewer branches and higher SL content
The number of branches within 40 nodes of 'Okubo' and 'Zhaoshouhong' were counted. Results indicated that 'Okubo' had slightly more primary branches than 'Zhaoshouhong' (Fig. 1a, b, and c). 'Zhaoshouhong' had almost no secondary branches; 'Okubo' had ∼6.8 within the first 40 nodes (Fig. 1a, b, and c).
SL contents were measured in the leaf buds of 'Zhaoshouhong' and 'Okubo'. The SL content was almost 2-fold higher in 'Zhaoshouhong' than 'Okubo' (Fig. 1d), suggesting that the lower branch number of 'Zhaoshouhong' may be related to the higher SL content.

Exogenous strigolactone significantly decreased branch number in 'Zhaoshouhong'
Exogenous SL (GR24) treatment 'Zhaoshouhong' was conducted to verify the function of SL in controlling peach branching. The number of branches and nodes were observed for 15 d in control 'Zhaoshouhong' trees and those treated six times with GR24, respectively. Branching was largely inhibited in GR24treated 'Zhaoshouhong' plants compared to the control (Fig. 2a, b, and c). By 9 d after treatment, the GR24-treated group had an average of 11.4 branches, whereas the control group had an average of 18.5 (Fig. 2c). The branching rate of GR24-treated 'Zhaoshouhong' plants was dramatically lower than control plants at 3 d after GR24 treatment (Fig. 2d), although the number of nodes did not significantly differ between control and GR24treated plants ( Fig. 2b and d). These results indicated that SL decreased the branch number of 'Zhaoshouhong' annual growth, and furthermore decreased the number of sylleptic branches.

PpTCP18 and SL synthesis-related genes showed opposite expression patterns after GR24 treatment
To identify candidate genes involved in the SL pathway, two groups of transcriptome data were analysed, including 'Zhaoshouhong' and 'Okubo', the control and GR24 treated 'Zhaoshouhong'. Exogenous GR24 treatment significantly inhibited the expression of SL synthesis-related genes such as PpLBO1, PpMAX3, and PpMAX4 (Fig. 3a, bottom three rows). Transcriptome data from leaf buds sampled before bud break suggested that PpLBO1, PpMAX3, and PpMAX4 were expressed at higher levels in 'Zhaoshouhong' than in 'Okubo' (Fig. S1, see online supplementary material). The expression profiles of SL-related genes were consistent with endogenous SL levels in 'Zhaoshouhong' and 'Okubo' (Fig. 1d).
TCP transcription factors are key factors involved in branching control [9][10][11]. To identify TCPs genes related to peach branching, 20 TCPs genes were identified in peach genome and further analyzed based on transcriptome data. The results suggested that in 'Zhaoshouhong', most PpTCPs were differentially expressed after exogenous GR24 treatment compared to the control (Fig. 3a). GR24 treatment caused upregulation of four of the 20 PpTCP genes: PpTCP3, PpTCP5, PpTCP13, and PpTCP18 (Fig. 3a). Phylogenetic analysis of 20 PpTCPs, 24 AtTCPs, one Zea mays TCP (ZmTB1), and one Oryza sativa TCP (OsTB1) revealed two subclasses. Class I contained 13 AtTCPs and 11 PpTCPs; Class II contained 11 AtTCPs, 9 PpTCPs, ZmTB1, and OsTB1. Class II could be further subdivided into two groups: CIN and TB1/CYC. PpTCP18 belonged to the TB1/CYC group and was closely related to AtBRC1 (AtTCP18), which is the key factor that controls branching in Arabidopsis [14] (Fig. 3b). PpTCP18 was therefore selected for further study. Additional transcriptome data from untreated leaf buds showed that PpTCP18, PpLBO1, PpMAX3, and PpMAX4 were expressed at higher levels in 'Zhaoshouhong' than in 'Okubo' (Fig. S1, see online supplementary material). These results suggested that PpTCP18 may participate in peach tree branching by regulating SL biosynthesis.

PpTCP18 was localized to the nucleus and cell membrane
A PpTCP18:GFP fusion construct was generated and expressed in tobacco leaves. At three days after injection, the f luorescent signal

PpTCP18 directly interacted with the promoters of PpLBO1, PpMAX3, and PpMAX4 to enhance their expression
Transcriptome analysis of 'Okubo' and 'Zhaoshouhong' leaf buds (before bud break) showed that PpLBO1, PpMAX3, and PpMAX4 had similar expression patterns to that of PpTCP18 (Fig. S1, see online supplementary material), indicating that PpTCP18 may regulate the expression of these SL biosynthesis genes. To verify this hypothesis, a bioinformatics analysis was performed on the promoter sequences of PpLBO1, PpMAX3, and PpMAX4. A TCP binding motif (GGNCCC/GGGNCC) was found in each promoter (Table S1, see online supplementary material). A yeast one-hybrid (Y1H) experiment was carried out to verify the interactions of PpTCP18 with the PpLBO1, PpMAX3, and PpMAX4 promoters. When PpTCP18 was co-transformed in yeast cells with a reporter construct containing either the PpLBO1, PpMAX3, or PpMAX4 promoter, the yeast grew well in the presence of Aureobasidin A (AbA) (200 ng/ml). When they were co-transformed with the empty vector pGADT7, the yeast did not grow well on the same medium (Fig. 4a). EMSAs were performed to further confirm the interactions between PpTCP18 and the PpLBO1, PpMAX3, and PpMAX4 promoters. The coding sequence (CDS) of PpTCP18 was cloned into the pGEX-6P-1 vector, and recombinant PpTCP18:MBP protein was purified (Fig. S3, see online supplementary material). EMSA results suggested that PpTCP18 could directly bind the probes derived from the PpLBO1, PpMAX3, and PpMAX4 promoters via the GGNCCC/GGGNCC motif. The band shifts observed for PpTCP18 incubated with the PpLBO1, PpMAX3, or PpMAX4 promoters were greatly weakened by the addition of 100-fold unlabeled probes, and disappeared entirely when the GGNCCC/GGGNCC motif was mutated to a polyA sequence (Fig. 4b). These results indicated that PpTCP18 could directly bind to the SL-responsive elements located in the PpLBO1, PpMAX3, and PpMAX4 promoters via the GGNCCC/GGGNCC motif.
To investigate whether PpTCP18 could affect the activity of SL biosynthesis-related genes, the promoters of PpLBO1, PpMAX3, PpMAX4 were cloned into dual luciferase reporter vector pGreenII 0800-LUC, respectively. Then either of them and 35:PpTCP18 were transiently expressed in tobacco leaves. Results showed that transient infiltration of tobacco leaves with Agrobacterium carrying the 35S:PpTCP18 vector enhanced the activity of the PpLBO1, PpMAX3, and PpMAX4 promoters by 2.5-, 8.7-, 3.4-fold compared with the empty vector (Fig. 4c). To further verify the transcriptional activation ability of PpTCP18, a virus-induced gene silencing (VIGS) experiment was performed to knock down PpTCP18 in peach. The results suggested that transient Agrobacterium infiltration of pillar peach leaf buds with the TRV2:PpTCP18 vector reduced expression of PpLBO1, PpMAX3, and PpMAX4, by 1.9-, 94.9-, and 4-fold, respectively, compared to the control group infiltrated with empty TRV2 vector. Endogenous SL levels in plants treated with PpTCP18 and untreated controls were 0.65 and 3.15 ng/g, respectively (Fig. 4d). Collectively, these results demonstrated that PpTCP18 could directly bind to the PpLBO1, PpMAX3, and PpMAX4 promoters and significantly enhance their expression.

PpTCP18 inhibited branching in Arabidopsis
In order to verify the function of PpTCP18 in branching, PpTCP18 overexpression vector was constructed and stably transformed into Arabidopsis. Three transgenic PpTCP18-overexpression lines (TCP18-1, TCP18-3, and TCP18-5) that had higher PpTCP18 transcript levels than the WT were selected for further analysis (Fig. S4, see online supplementary material). After two weeks in a growth chamber, the rosette leaves were counted. Results showed that PpTCP18 overexpression significantly decreased the number of rosette leaves in the transgenic lines; there were an average of 16.25, 15.67, and 15.75 leaves for PpTCP18-1, PpTCP18-3, and PpTCP18-5, respectively, compared to 24 leaves in the WT ( Fig. S5a and b, see online supplementary material). There were no obvious differences in the length and width of the rosette leaves between the transgenic lines and the WT ( Fig.  S5c and d, see online supplementary material). The primary cauline leaf branches (CI) and primary rosette leaf branches (RI) (Fig. S5e, see online supplementary material) were counted two weeks after f lowering. PpTCP18 overexpression led to significantly fewer primary rosette-leaf branches, with RI values of 5.66, 5.58, 5.41, and 9.00 for PpTCP18-1, PpTCP18-3, PpTCP18-5, and WT, respectively ( Fig. S5f and g, see online supplementary material). There were no significant differences in the CI between the transgenic lines and WT (Fig. S5h, see online supplementary material). The endogenous SL content was higher in the nodes of transgenic plants compared to the WT. Consistent with the measured SL concentrations (Fig. S5i, see online supplementary material), expression levels of SL-related genes (namely AtMAX2, AtMAX4, AtBRC1, and AtBRC2) were significantly higher in transgenic lines compared with WT; furthermore, SMXL6/7/8 expression was markedly decreased in transgenic lines, whereas there were no differences in AtBES1 expression (Fig. S5j, see online supplementary material). ABA biosynthesis-related genes (AtNCED1/2/3/4) were upregulated in transgenic lines compared to WT, but AtNCED5/6 transcripts could not be detected, perhaps due to low expression levels (Fig. S5I, see online supplementary material). These results suggested that PpTCP18 may prevent primary rosette branching in Arabidopsis by inf luencing the SL and ABA pathways.

LncRNA5 (TCONS_00066816) increased PpTCP18 expression
To select lncRNA targeted PpTCP18 and verify whether the candidate lncRNA could regulate the expression of PpTCP18. LncRNA libraries were constructed from 'Okubo' and 'Zhaoshouhong' leaf buds before bud break. In total, 14 503 lncRNAs were identified in peach, 975 of which were differentially expressed between 'Okuba' and 'Zhaoshouhong'. There were 546 upregulated and 429 downregulated lncRNAs in 'Zhaoshouhong' compared to 'Okuba' (Fig. 5a), with the former including TCONS_00066816 (lncRNA5). The Kyoto Encyclopedia of Genes and Genomes (KEGG) suggested that the genes targeted by differentially expressed lncRNAs were involved in a variety of biological processes, including starch and sucrose metabolism, f lavonoid biosynthesis, and hormone responses (Fig. 5b). Target gene prediction results showed that PpTCP18 was the target of lncRNA5. LncRNA5, located 60 839 bp downstream of PpTCP18 (Fig. 5c), showed a similar expression pattern with that of PpTCP18; lncRNA5 was expressed 5-fold higher in 'Zhaoshouhong' than 'Okuba' (Fig. 5d). The PpTCP18 promoter was cloned into the dualluciferase reporter pGreenII 0800-LUC vector to determine whether the expression of PpTCP18 could be regulated by lncRNA5 (Fig. 5e). Dual luciferase assay (DLR) results showed that PpTCP18 promoter activity was significantly increased by lncRNA5 (Fig. 5e). These results suggested that lncRNA5 could increase PpTCP18 expression.

Discussion
Pillar peach trees require less pruning, which can greatly reduce costs [5]. In this study, the pillar cultivar 'Zhaoshouhong' and the standard cultivar 'Okubo' were compared. 'Zhaoshouhong' had the same number of nodes as 'Okubo' but fewer secondary branches. This finding was similar to results reported by Bassi and Dima showing fewer second-order branches of columnar (pillar) peach trees compared to standard peach trees [5]. The endogenous SL content was significantly higher in 'Zhaoshouhong' than in 'Okubo', and GR24 treatment significantly inhibited peach branching. These results indicated that SL plays an important role in apical dominance in peach [36], and that the SL level was closely negatively correlated with branch number.
LncRNAs coordinate gene expression via chromatin modification, acting either as endogenous target mimics (eTMs) or through cis-or trans-activation [30,31]. Studies have demonstrated that lncRNAs participate in regulating fruit ripening [33], drought tolerance [45], disease resistance [46], and fruit coloring [35]. However, there have been few reports of lncRNA involvement in the SL signaling pathway. Here, we found that lncRNA5 in peach (TCONS_00066816) upregulated PpTCP18. LncRNA5 may enhance PpTCP18 expression through recognition of the PpTCP18 sequence Promoters driving the empty vector were used as negative controls. b Electrophoretic mobility shift assays (EMSAs) showed that PpTCP18 bound the GGNCCC/GGGNCC motif in the PpLBO1, PpMAX3, and PpMAX4 promoters. c Dual luciferase assays indicated that PpTCP18 activated transcription of PpLBO1, PpMAX3, and PpMAX4. Plasmids containing PpTCP18 and either PpLBO1, PpMAX3, or PpMAX4 were introduced into tobacco leaves via Agrobacterium tumefaciens-mediated transformation to assess the activation of PpLBO1, PpMAX3, and PpMAX4 promoters by PpTCP18. Firef ly luciferase (LUC)/Renilla luciferase (REN) ratios were significantly higher when leaves were co-transformed with the PpTCP18 vector than with the empty control vector, suggesting that PpTCP18 enhanced PpLBO1, PpMAX3, and PpMAX4 promoter activity. d Virus-induced gene silencing of PpTCP18 in peach leaf buds reduced expression of PpLBO1, PpMAX3, and PpMAX4, and reduced the endogenous SL concentration. Transcript levels are shown for PpTCP18, PpLBO1, PpMAX3, and PpMAX4 in peach leaf buds infiltrated with the empty (control) or TRV2:PpTCP18 vector. Values shown are the mean ± standard deviation of three biological replicates. * * P < 0.01 vs. control (Student's t-test). through sequence complementarity and R-loop formation; this would be comparable to a regulatory mechanism in apple (Malus domestica), in which MdLNC610 increases MdACO1 expression [35].
Previous studies have suggested that SL may be the auxin secondary messenger that directly inhibits plant branching [10,47]; that SL biosynthesis is subject to positive feedback regulation by auxin, whereas auxin biosynthesis is subject to negative feedback regulation by SL [38]; and that ABA biosynthesis is induced by SL [48]. In contrast to these previous studies, we here identified a novel mechanism involved in peach branch number regulation; PpTCP18 controls peach branching by positive feedback regulation of SL biosynthesis. A hypothesis was proposed to explain the role of PpTCP18 in peach branching: (i) in standard peach, the endogenous SL content was very low, leading to lower PpTCP18 expression and more the branch number; and (ii) in pillar peach, the SL concentration was higher, inducing PpTCP18 expression, which in turn positively regulated SL biosynthesis gene expression and SL content. Furthermore, PpTCP18 expression was dramatically increased by lncRNA5, greatly inhibiting branch number in pillar peach trees.

Plant materials
Representatives of two types of peach tree, 'Okubo' ('O', standard peach) and 'Zhaoshouhong' ('Z', pillar peach), were grown in the research orchard of the Henan Agricultural University in Zhengzhou, China. Phenotypic analyses of 'Okubo' and 'Zhaoshouhong' annual growth were conducted in May 2019, with leaves removed to assess branch number. Leaf buds were sampled from both cultivars in April (before bud break) for SL measurements and transcriptome analyses. Annual growth of 'Zhaoshouhong' was used for GR24 treatment, and the phloem was sampled at branch junctions for transcriptome analysis. Leaf buds after bud break (3-4 cm) were sampled from threeyear-old 'Zhaoshouhong' trees and used in virus-induced gene silencing (VIGS) experiments. Arabidopsis (Col-0) was used for stable transformation. Nicotiana benthamiana was used for dualluciferase assays (DLR).

SL measurements
SL levels were measured by NJRuiYuan Co., Ltd. as described by Ruiz-Lozano [49]. Peach leaf buds (5 g) were ground into powder with liquid nitrogen. The powder was extracted with 0.5 mL 40% acetone (v/v), followed by vortexing for 2 min, then centrifugation at 8000 × g for 5 min. The pellets were collected and extracted twice with 0.5 mL 50% acetone (v/v). The supernatant was purified with membrane filters (0.22 μm pore size). SL content was then measured by HPLC-MS/MS.

TCP gene family identification and phylogenetic tree construction
To identify TCP genes in peach, sequences of Arabidopsis TCP genes were downloaded from PlantTFDB (v4.0) [50] and used as queries in BLAST searches against the peach genome. Twenty PpTCP genes were identified in peach and named PpTCP1 through PpTCP20 based on their chromosome positions. Amino acid sequences of the TCPs from P. persica (PpTCPs) and Arabidopsis thaliana (AtTCPs) [14] were used together with ZmTB1 [51] from Z. mays and OsTB1 [52] from O. sativa (Table S2, see online supplementary material) to construct a neighbor-joining phylogenetic tree using MEGA 5.0 [50]. The maximum likelihood method was employed with 1000 bootstrap replicates.

Exogenous GR24 treatment
The synthetic SL GR24 (Coolaber, Beijing, China) was dissolved in acetone, then diluted with water to a final concentration of 20 μM. This solution was used to treat the annual growth of 'Zhaoshouhong'. Water was used as the control. Plants were treated every 3 d for a total of six treatments. The node number and branch number were counted every 3 d, and the branching rate was calculated as branch number divided by node number. Each treatment was conducted with three biological replicates, and each replicate contained five annual growth plants of 'Zhaoshouhong'.
To identify genes that respond to GR24, the whole-plant annual growth of 'Zhaoshouhong' was treated with GR24. The branch junctions of treated and control plants were sampled after treatment for 0 and 2 h, then stored in a −80 • C freezer prior to transcriptome analysis.

RNA extraction, cDNA synthesis, and gene expression analysis
RNA extraction, cDNA synthesis, and quantitative reverse transcription (qRT)-PCR experiments were performed as described by Wang et al. [53]. RNA was extracted from peach buds sampled in April using a Total RNA Kit (Tiangen, Beijing, China). Purified RNA was used for transcriptome analysis and qRT-PCR. cDNA synthesis was performed using the FastKing RT Kit (Tiangen, Beijing, China). Relative expression levels of PpTCP18, PpLBO1, PpMAX3, and PpMAX4 were analysed with qRT-PCR. Actin (Prupe.1G234000) was used as the internal control for normalizing gene expression. There were three independent biological replicates. Primers are shown in Table S3 (see online supplementary  material).

Gene cloning and promoter analysis
The PpTCP18 gene and the 2 kb promoter regions upstream of the PpLBO1, PpMAX3, and PpMAX4 start codons were cloned as described by Wang et al. [53]. Primers are shown in Table S3 (see online supplementary material). The online tool plantCARE (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/) was used to identify GGNCCC/GGGNCC motifs.

Subcellular localization of PpTCP18
The CDS of PpTCP18 was cloned into the pSAK277-GFP vector to construct the PpTCP18:GFP fusion protein driven by the CaMV 35S promoter. A Leica SP8 Confocal Microscope (Wetzlar, Germany) was used to observe f luorescent signals. mCherry and Cd3-1009 [54] were the nuclear and membrane-localized marker genes used, respectively. Primers are shown in Table S4 (see online supplementary material).

Virus-induced gene silencing (VIGS)
Transient gene expression was performed as described by Liu et al. [55]. Approximately 200 bp of the non-conserved complementary sequence of PpTCP18 was amplified and used for gene silencing. Primers are shown in Table S4 (see online supplementary material). The PCR product was purified and cloned into the TRV2 vector. Plasmids were transformed into Agrobacterium tumefaciens strain GV3101. Peach buds (3-4 cm) from 'Zhaoshouhong' were soaked in 75% alcohol for 1 min and washed three times with sterile water. The sterilized peach leaf buds were pre-cultured on Murashige and Skoog (MS) medium for 1 d, then submerged in a mixture of A. tumefaciens strains carrying TRV2 and TRV1 vectors and vacuum infiltrated (−70 kPa). Peach leaf buds were washed three times in sterile water, then cultured in MS medium for 2 d. Finally, leaf buds were frozen in liquid nitrogen and stored at −80 • C prior to qRT-PCR analysis and SL measurements. There were three independent replicates of these experiments.

Yeast one-hybrid (Y1H)
The CDS of PpTCP18 was cloned into the pGADT7 vector. The promoter regions ∼2 kb upstream of the PpLBO1, PpMAX3, and PpMAX4 start codons were amplified and ligated into the pAbAi vector. A Matchmaker™ Gold Yeast One-Hybrid Library Screening System Kit (Clontech, San Francisco, CA, USA) was used to test interactions between PpTCP18 and the promoters. Primers are shown in Table S3 (see online supplementary material).

Dual-luciferase assay (DLR)
The promoter sequences of PpLBO1, PpMAX3, PpMAX4, and PpTCP18 were cloned into the pGreenII 0800-LUC vector. LncRNA5 and the CDS of PpTCP18 were cloned separately into the pSAK277 vector. The LUC and REN luciferase activity levels were measured using a DLR kit (Promega, Madison, WI, USA) following the manufacturer's instructions. For each assay, at least six measurements were obtained. Primers are listed in Table S3 (see online supplementary material).

Electrophoretic mobility shift assay (EMSA)
The CDS of PpTCP18 was cloned into the pMAL-c5x vector containing the MBP tag, then the plasmid was transformed into Escherichia coli. Recombinant PpTCP18-MBP protein was expressed and purified as described by Wang et al. [53]. An EMSA Kit (Thermo Scientific, Waltham, MA, USA) was used following the manufacturer's instructions. Sequences containing the GGNCCC/GGGNCC motifs derived from the PpLBO1, PpMAX3, and PpMAX4 promoters (Table S5, see online supplementary material) were labeled with biotin at the 5 termini and were used as probes. The same DNA fragments without labels were used as competitors (cold probes). DNA fragments with the GGNCCC/GGGNCC motif changed to a polyA sequence were used as mutated probes.

Plant transformation and phenotype analysis
Arabidopsis plants were transformed with the pSAK277-PpTCP18 construct using an Agrobacterium-mediated transformation protocol as described by Zhang et al. [56]. Arabidopsis seeds of the T 2 generation were sown on commercial soil after cold treatment at 4 • C for 3 d. Plants were transferred to a growth chamber at 20 • C with a 16/8 h light/dark cycle. Leaves were counted at 2 weeks after transfer to the growth chamber. Primary-leaf branches (branches >0.5 cm long) were counted after the main inf lorescence was visible. Each plant line was represented by three biological replicates; each replicate contained three individual T 2 plants. For SL content measurement and SL-related gene expression analysis, only the nodes were sampled. All gene expression analyses were performed with three independent biological replicates. Primers are shown in Table S3 (see online supplementary material).

LncRNA sequencing and bioinformatic analysis
Leaf buds from 'Okubo' (O-1, O-2, O-3) and 'Zhaoshouhong' (Z-1, Z-2, Z-3) were sampled in April (before bud break) and sent to Novogene (Beijing, China) for lncRNA library construction. Differentially expressed lncRNAs were identified using DESeq [2] software. Several databases were used to predict genes targeted by differentially expressed lncRNAs, including Volcano plot and the Kyoto Encyclopedia of Genes and Genomes (KEGG). LncRNA5 was selected and analysed with a DLR assay.

Statistical analysis
Data were analysed with two-tailed Student's t-tests. Microsoft Excel 2010 was used for statistical analyses. Statistical test parameters and statistical significance are noted in the figure legends.