RcSPL1–RcTAF15b regulates the flowering time of rose (Rosa chinensis)

Abstract Rose (Rosa chinensis), which is an economically valuable floral species worldwide, has three types, namely once-flowering (OF), occasional or re-blooming (OR), and recurrent or continuous flowering (CF). However, the mechanism underlying the effect of the age pathway on the duration of the CF or OF juvenile phase is largely unknown. In this study, we observed that the RcSPL1 transcript levels were substantially upregulated during the floral development period in CF and OF plants. Additionally, accumulation of RcSPL1 protein was controlled by rch-miR156. The ectopic expression of RcSPL1 in Arabidopsis thaliana accelerated the vegetative phase transition and flowering. Furthermore, the transient overexpression of RcSPL1 in rose plants accelerated flowering, whereas silencing of RcSPL1 had the opposite phenotype. Accordingly, the transcription levels of floral meristem identity genes (APETALA1, FRUITFULL, and LEAFY) were significantly affected by the changes in RcSPL1 expression. RcTAF15b protein, which is an autonomous pathway protein, was revealed to interact with RcSPL1. The silencing and overexpression of RcTAF15b in rose plants led to delayed and accelerated flowering, respectively. Collectively, the study findings imply that RcSPL1–RcTAF15b modulates the flowering time of rose plants.


Introduction
Rose (Rosa chinensis), which is known as the queen of f lowers, is an economically and culturally important plant species. Rose plants are cultivated worldwide as garden plants or for the production of cut f lowers [1]. According to their f lowering habits, rose plants have been classified into the following three types: recurrent or continuous f lowering (CF) (f lowers bloom continuously under appropriate conditions); occasional re-blooming (OR) genotype (f lowers bloom in spring and occasionally in autumn); and oncef lowering (OF) (f lowers bloom only in spring) [2]. Previous studies on the molecular basis of CF behavior mainly focused on the gibberellin (GA) pathway as well as on the f loral repressor gene TERMINAL FLOWER 1 (TFL1) [2][3][4][5][6][7][8]. However, CF seedlings have a short juvenile phase (4-5 weeks) with a determinate growth pattern, whereas OF seedlings have a long juvenile phase (1-3 years) with an indeterminate growth pattern [9,10]. Therefore, we hypothesized that the f lowering-related age pathway may contribute to the differences in f loral development among rose types.
In plants, f lowering at the appropriate time is critical for reproductive success. Additionally, the transition to the f lowering stage is regulated by the complex interplay between environmental cues and genetic pathways [11,12]. Genetic analyses revealed six f lowering response pathways (age, vernalization, photoperiod, GA, autonomous, and ambient temperature pathways). Genes encoding f lowering time integrators, including APETALA 1 (AP1), The SPL genes identified to date reportedly encode proteins that affect plant architecture as well as diverse development-related processes, including phase transitions, GA biosynthesis and signaling, and somatic embryogenesis [27][28][29][30][31][32]. However, the functional characterization of rose SPL genes has lagged behind that of SPL genes from other plant species because of an insufficient amount of genomic information and the problems associated with rose regeneration.
The autonomous pathway promotes f lowering independently of day length. Several genes, including TATA-BINDING PROTEIN-ASSOCIATED FACTOR 15b (TAF15b), FLOWERING LOCUS KH DOMAIN (FLK), FLOWERING LOCUS CA (FCA), FLOWERING LOCUS PA (FPA), and FLOWERING LOCUS VE (FVE), have been identified as crucial factors regulating f loral development in this pathway. Among the encoded proteins, TAF15b is a component of the transcription factor IID complex (TFIID) with a Zn-finger motif and an RNA-recognition motif [33]. Although 18 putative TAF genes in A. thaliana have been analyzed, with some of them encoding proteins inf luencing various developmental processes, little is known about the TAF functions related to f loral development [34][35][36][37].
Despite the ornamental and economic value of rose, genes that regulate f lowering, especially those associated with CF and OF, remain relatively unexplored. Rosa chinensis 'Old Blush' and R. chinensis var. spontanea, which respectively exhibit CF and OF behaviors, are typically selected as the research materials for investigating f lowering. In the present study, 15 SPL genes were identified in these two varieties and their expression patterns were examined during f lower developmental stages. Among the analyzed genes, RcSPL1 expression increased significantly in both varieties. Thus, RcSPL1 was functionally characterized using transgenic A. thaliana, and rose in which the target gene was silenced or overexpressed via Tobacco rattle virus (TRV)-mediated transient expression. We also determined that RcTAF15b is an SPL1-interacting protein that helps control f lowering time. These findings provide substantial insights into the mechanisms underlying rose f loral development.

RcSPL1 expression increased continuously in rose flower developmental stages
Fifteen SPL genes were identified in the 'Old Blush' (https://lipmbrowsers.toulouse.inra.fr/pub/RchiOBHm-V2/) and R. chinensis var. spontanea (https://ngdc.cncb.ac.cn/search/?dbId=gwh&q= GWHAAAF00000000&page=1) genome databases (Supplementary  Data Table S1). Their expression patterns were examined during f lower developmental stages in 'Old Blush' and R. chinensis var. spontanea. In this study, the rose f loral development period was divided into six stages as previously described, with minor modifications [38]. In addition, internal morphology was determined using paraffin sections and scanning electron microscopy (Supplementary Data Fig. S1, Supplementary Data Table S2). Reverse transcription-quantitative PCR (RT-qPCR) data indicated that, among the identified genes, the SPL1 expression levels continuously increased as f lowers developed in both 'Old Blush' and R. chinensis var. spontanea (Fig. 1a). Moreover, at each stage, the expression level was higher in 'Old Blush' than in R. chinensis var. spontanea (Fig. 1a), indicating that SPL1 is crucial for the development of rose f lowers. The expression patterns of the other SPL genes were less distinctive. Furthermore, SPL9 expression was undetectable in both varieties. Accordingly, we focused on SPL1 for the subsequent analysis. Notably, the SPL1 sequence was the same in both varieties. Therefore, it was designated as RcSPL1 in this study.
To evaluate the evolutionary relationship between RcSPL1 and its orthologs in A. thaliana, the deduced amino acid sequences of the SPL genes were aligned using MEGA X for phylogenetic analysis, which showed that the RcSPL1 sequence was highly similar to AtSPL1 in A. thaliana (Supplementary Data Fig. S2a). The sequence alignment suggested that the SBP domain was highly conserved in RcSPL1 and FvSPL1 in woodland strawberry (Fragaria vesca) (Supplementary Data Fig. S2b). The sequence analysis also revealed a conserved nuclear localization signal and two conserved Zn-binding sites in the SBP domain ( Supplementary  Data Fig. S2c). The first and second Zn-finger structures were CCCH-type and CCHC-type, respectively (Supplementary Data Fig. S2c). These SBP domain structural characteristics are consistent with those in most other plants, implying that the SBP domain is highly conserved. Subcellular localization analysis indicated that the RcSPL1-GFP signal overlapped the mCherry signal (a nuclear marker) [39], providing evidence that RcSPL1 is a nuclear protein (Fig. 1b). The nuclear localization of RcSPL1 along with its increased expression during the f lowering process suggests that RcSPL1 might be essential for rose f loral development.

RcSPL1 expression was directly regulated by rch-miR156
MicroRNA156 was identified as a master regulator of f lowering transition. Most previously reported miR156 family members negatively regulate SPL expression by directly cleaving SPL transcripts. The psRNATarget online tool predicted that RcSPL1 is targeted by rch-miR156, with a potential cleavage site in the 3 untranslated region (UTR) (Fig. 2a). Moreover, rch-MIR156 formed a local stem-loop structure and its sequence was highly similar to that of fve-MIR156e and ppe-MIR156e from F. vesca and Prunus persica (Supplementary Data Fig. S3a and b). To confirm the regulatory effect of rch-miR156 on RcSPL1 expression, a green f luorescent protein (GFP) reporter assay was performed to monitor the cleavage of RcSPL1 by rch-miR156 in a Nicotiana benthamiana transient expression system [40]. To compensate for the varying cleavage efficiency of rch-miR156, RcSPL1-sensor-GFP was co-expressed with different pri-rch-miR156 content in N. benthamiana (T1, T2, and T3). As controls, RcSPL1-sensor-GFP was co-expressed with the pCAMBIA1300 vector (C1) and pri-rch-miR156 was co-expressed with the 130-GFP vector (C2) (Fig. 2a). RT-qPCR analysis showed that rch-miR156 expression increased significantly as the pri-rch-miR156 content increased (Fig. 2b). Examination of cells in the same-size microscopic field detected a GFP signal in 100% of the cells in the control (C1 and C2) tobacco leaves (Fig. 2c). However, the percentage of cells with GFP signal decreased significantly as the expression of rch-miR156 increased, but there was no significant difference between T1 and C1 (Fig. 2c). In addition, the GFP f luorescence intensity in N. benthamiana leaves was consistent with the above-mentioned results (Supplementary Data Fig. S4). Hence, rch-miR156 downregulated RcSPL1 expression, but only when it accumulated to a certain level. Additionally, the protein produced by the His-RcSPL1-3 UTR construct accumulated to significantly lower levels when it was co-expressed with pri-rch-miR156 than when it was co-expressed with the empty vector (pCAMBIA1300) or His-RcSPL1 co-expressed with pri-rch-miR156 in the N. benthamiana transient expression system ( Fig. 2a and d, Supplementary Data Fig. S5). These results indicated that rch-miR156 effectively cleaved RcSPL1 to directly regulate RcSPL1 expression. . spontanea during f lower development. Stage 1, earliest bud break; stage 2, shoot elongation (>20 mm) without three-leaf let leaf; stage 3, the first three-leaf let leaf is visible; stage 4, the first five-leaf let leaf is visible; stage 5, f lower bud appearance; stage 6, f lower bud with a diameter of 5-7 mm. Mean values ± standard deviation are shown from three biological replicates (n = 3). RcUBI2 was used as the internal control. b Subcellular localization of RcSPL1-GFP heterologously expressed in N. benthamiana leaves. The plasmid pSuper:RcSPL1-GFP was co-infiltrated with the nuclear marker pSuper:NF-YA4-mCherry. Green and red f luorescences were visualized by confocal microscope (Leica TCS SP8) 3 days after infiltration. Scale bar, 50 μm. The experiment was performed independently three times, and representative results are shown.
RcSPL1-OE lines than in the wild-type (WT) plants (Fig. 3a). In A. thaliana, the production of trichomes on the abaxial surface of the leaf and the formation of a serrated leaf margin have been used to mark the vegetative phase transition (juvenile-toadult transition) [41,42]. To investigate whether RcSPL1 inf luences the vegetative phase transition, WT plants and RcSPL1-OE lines were compared regarding their leaf morphology. Compared with the WT plants, the RcSPL1-OE plants greatly promoted the emergence of abaxial trichomes and serration on the leaf margin ( Fig. 3b and c, Supplementary Data Table S3), indicating that RcSPL1 accelerated the vegetative phase transition. Additionally, the RcSPL1-OE lines f lowered 7-9 days earlier than the WT plants ( Fig. 3d and e, Supplementary Data Table S3). Moreover, compared with the WT plants, the RcSPL1-OE lines had significantly fewer rosette leaves (Fig. 3f, Supplementary Data Table S3). Accordingly, the overexpression of RcSPL1 in A. thaliana significantly promoted the vegetative phase transition and f lowering.

RcSPL1-silenced and -overexpressed rose plants had opposite flowering phenotypes
To clarify the effect of RcSPL1 on rose f loral development, RcSPL1silenced and -overexpressed Rosa hybrida 'Samantha' plants were generated using a TRV-mediated transient silencing and overexpressing system. The RcSPL1 cDNA fragment (343 bp) was inserted into the pTRV2 virus-induced gene silencing (VIGS) vector, whereas the RcSPL1 open reading frame (ORF) was cloned . Asterisks represent statistically significant differences determined by Student's t-test ( * * * * P < 0.0001). ND, not detected. AtActin was used as an internal control. b Morphology of rosette leaves of WT and RcSPL1-OE lines. Adult leaves were defined as those with abaxial trichomes and serrated leaf margins. Scal bar, 1 cm. c Rosette leaf number at the initation of abaxial trichomes of WT and RcSPL1-OE lines. Mean values ± standard deviation are from 13 plants. Asterisks represent statistically significant differences determined by Student's t-test ( * * * * P < 0.0001). d Days to f lowering of WT and transgenic plants. Mean values ± standard deviation are shown from 30 plants (n = 30). Asterisks represent statistically significant differences determined by Student's t-test ( * * * * P < 0.0001). e Flowering phenotype comparison between WT and transgenic Arabidopsis. f Rosette leaf number of WT and RcSPL1-OE lines at f lowering time. Mean values ± standard deviation are shown from 30 plants (n = 30). Asterisks represent statistically significant differences determined by Student's t-test ( * P < 0.05, * * P < 0.01). g-i RT-qPCR analyses of AtAP1 (g), AtFUL (h), and AtLFY (i) transcript levels in RcSPL1-OE lines. Mean values ± standard deviation are shown from three biological replicates (n = 3). Asterisks represent statistically significant differences determined by Student's t-test ( * P < 0.05, * * P < 0.01, * * * P < 0.001, * * * * P < 0.0001). AtActin was an internal control.
into the psTRV2 vector [43]. Changes in RcSPL1 expression were preliminarily analyzed by RT-qPCR. Approximately 20 plants with downregulated or upregulated RcSPL1 expression levels were used for the phenotypic analysis. Three plants with similar RcSPL1 expression levels were selected for RT-qPCR assay (Supplementary Data Fig. S6). Compared with the corresponding control plants (TRV and sTRV), RcSPL1 transcript levels were significantly lower in the RcSPL1-silenced (TRV-RcSPL1) rose plants, but substantially higher in the RcSPL1overexpressed (sTRV-RcSPL1) rose plants ( Fig. 4c and f). Consistent with the RcSPL1 expression levels, the f lowering of TRV-RcSPL1 and sTRV-RcSPL1 plants was significantly delayed and accelerated, respectively (Fig. 4a, b, d, e). More specifically, f lowering was observed at 46.10 ± 2.25 days for the TRV control plants, which was ∼8 days earlier than the f lowering of the RcSPL1-silenced plants (54.05 ± 2.78 days) (Fig. 4b).
Conversely, the f lowering of sTRV-RcSPL1 rose plants (41.85 ± 1.69 days) occurred significantly earlier (∼5 days) than the f lowering of the control plants (46.25 ± 1.40 days) (Fig. 4e). Moreover, to determine the inf luence of RcSPL1 on vegetative phase transition of rose as in transgenic Arabidopsis, the expression levels of miR156 and SPL1 were further detected during the development of 'Mount Shasta' (CF type) and Rosa multif lora (OF type) seedlings from the emergence of leaves to f lowering by RT-qPCR. The results showed that the expression levels of rch-miR156 decreased, while  . RcUBI2 was used as the internal control. Asterisks represent statistically significant differences determined by Student's t-test ( * P < 0.05, * * P < 0.01, * * * P < 0.001, * * * * P < 0.0001).

Altered RcSPL1 expression affected the expression of flowering-related genes
To determine whether RcSPL1 affects the expression of the downstream genes, including AP1, LFY, and FRUITFULL (FUL), we analyzed gene expression levels in transgenic A. thaliana and rose plants. Compared with the control plants, the AtAP1, AtFUL, and AtLFY expression levels were significantly higher in the RcSPL1-OE lines (Fig. 3g-i). Similarly, the expression levels of their rose homologs (RcAP1, RcFUL, and RcLFY) increased in the RcSPL1-overexpressed plants (Fig. 3f), but decreased in the RcSPL1-silenced plants (Fig. 3c). In addition, similar to the trend in RcSPL1 expression, the RcAP1, RcFUL, and RcLFY expression levels in 'Old Blush' and R. chinensis var. spontanea increased continuously during the f loral development period (Supplementary Data Fig. S8). These results imply that RcSPL1 regulates rose f loral development by activating the expression of the f lowering-related genes RcAP1, RcLFY, and RcFUL.

RcSPL1 interacted with RcTAF15b, a key regulator in the autonomous pathway
To further examine the regulatory effects of RcSPL1 on f lower development, we searched for proteins that can interact with RcSPL1. Preliminary screening of a rose cDNA library using a yeast two-hybrid (Y2H) system resulted in the detection of 23 proteins that may interact with RcSPL1. Among these proteins, four were identified as interacting proteins at least twice (Supplementary Data Table S4). Specifically, TAF15b was identified four times as a protein capable of interacting with RcSPL1 (Fig. 5a). The phylogenetic and conserved domain analyses showed that RcTAF15b is closely related to A. thaliana protein AtTAF15b and F. vesca protein FvTAF15b, which contained one Zn-finger-like structural motif and an RNA recognition motif (Supplementary Data Fig. S9a and b).
To verify the interaction between RcSPL1 and RcTAF15b, a bimolecular f luorescence complementation (BiFC) assay was conducted using N. benthamiana. A strong reconstituted yellow f luorescent protein (YFP) signal was observed in nuclei coexpressing RcSPL1-nYFP and RcTAF15b-cYFP (Fig. 5b). To determine whether these two proteins function in the same subcellular location, a co-localization assay was performed. Tobacco leaves were co-infiltrated with RcSPL1-GFP and RcTAF15b-mCherry. Subsequent examination indicated that RcSPL1 and RcTAF15b were both localized in the nucleus (Fig. 5c). These findings suggest that RcSPL1 controls f lowering time possibly through its interaction with RcTAF15b, an important component of the autonomous pathway.

Altered RcTAF15b expression influenced rose flowering
The observed interaction between RcTAF15b and RcSPL1 compelled us to assess whether RcTAF15b is involved in rose f loral development. RT-qPCR analysis confirmed that the RcTAF15b transcript levels during the f loral development period of 'Old Blush' and R. chinensis var. spontanea were consistent with the trends in RcSPL1 transcript abundance (Supplementary Data Fig. S10), suggestive of a role for RcTAF15b during the f lowering of rose plants. The analysis of the RcTAF15b-silenced (TRV-RcTAF15b) and -overexpressed (sTRV-RcTAF15b) rose plants revealed that the RcTAF15b transcript level was much lower in the RcTAF15b-silenced plants than in the TRV control plants (Fig. 6a,  Supplementary Data Fig. S10), but it was significantly higher in the sTRV-RcTAF15b plants than in the sTRV control plants (Fig. 6d,   Supplementary Data Fig. S11). As expected, compared with the control plants, which f lowered at 44.70 ± 2.06 days, the RcTAF15bsilenced plants f lowered ∼5 days later (49.70 ± 1.64 days) (Fig 6b and c). The RcTAF15b-overexpressed rose plants f lowered earlier (41.85 ± 1.27 days) than the sTRV control plants (46.65 ± 1.42 days) (Fig 6e and f). These results ref lected the importance of RcTAF15b for the f lowering of rose plants.

Discussion
Rose, which is one of the most popular ornamental f lower species, can be divided into three types: recurrent or continuous f lowering (CF), occasional re-blooming (OR), and once-f lowering (OF). Because of its unique and diverse f loral habits, rose is a good model for investigating the molecular mechanism underlying the f lowering of perennial plants. Therefore, identifying the key f lowering regulators in rose may help researchers and breeders develop superior perennial plant varieties that can f lower continuously. In this study, we determined that the interaction between RcSPL1 and RcTAF15b facilitates the co-regulation of rose f loral development.
There are 15 SPL genes in rose (Supplementary Data Table S1), but in this study only RcSPL1 transcript levels were continuously upregulated during the rose f lower development period (Fig. 1a). Overexpressing RcSPL1 in A. thaliana promoted the emergence of abaxial trichomes and the serration of the leaf margins (Fig. 3b), suggesting that RcSPL1 might play a similar function by accelerating the juvenile-to-adult transition in rose. In A. thaliana, decreases in the miR156 level or the overexpression of its targets (SPL genes) can accelerate vegetative and reproductive phase transitions [41,44]. Similarly, the upregulated expression of SPL genes promotes the production of deltoid leaves (a marker of the vegetative phase transition) in poplar (Populus × canadensis) [45]. In addition, previous studies showed that the age pathway is mainly controlled by miR156 and its target transcription factors, SPLs; the expression of miR156 gradually decreases as SPL gene expression increases during the transition to f lowering [15,16]. Moreover, rch-miR156 and RcSPL1 exhibited opposite expression patterns during the development of rose seedlings ( Supplementary Data Fig. S7). Therefore, rch-miR156/RcSPL1 may be a master regulator of the age pathway in rose. Notably, the rch-miR156 transcript levels in 'Old Blush' and R. chinensis var. spontanea gradually decreased during the f lower development period, which was in contrast to the changes in RcSPL1 expression (Fig. 1a, Supplementary Data Fig. S12), indicating that rch-miR156 may also participate in rose f loral development. Furthermore, overexpressing RcSPL1 in A. thaliana and rose plants led to a marked decrease in f lowering time (Figs 3d and e and 4d and e). Conversely, the silencing of RcSPL1 resulted in a latef lowering phenotype ( Fig. 4a and b). Accordingly, RcSPL1 is a key factor inf luencing the timing of rose f lowering. In addition to regulating the vegetative phase transition and f lowering, SPL family members also have multiple functions affecting diverse processes related to growth and development, such as plant architecture and fruit ripening in various plant species [27,46]. These findings indicate that different SPL genes may encode functionally diverse proteins. The 14 other RcSPL genes will need to be functionally validated in future studies.
The AP1, FUL, and LFY genes have critical effects on f lowering induction, f loral organ development, inf lorescence, and f loral meristem identity [47,48]. In our study, RcAP1, RcFUL, and RcLFY expression was significantly activated by RcSPL1. The transcripts of these genes were more highly accumulated in the Figure 5. RcSPL1 interacts with RcTAF15b. a Interaction between RcSPL1 and RcTAF15b by Y2H. AD is empty vector of pGADT7 AD, BD is empty vector of pGBKT7 BD. Positive control was performed using pGBKT7-53 + pGADT7-T, while negative control was pGBKT7-Lam + pGADT7-T. Experiments were performed independently twice, with similar results. b Interaction of RcSPL1 and RcTAF15b in a BiFC assay. RcSPL1-nYFP was co-infiltrated with RcTAF15b-cYFP in N. benthamiana leaves. Infiltrated leaves were visualized by confocal microscopy 3 days after infiltration. Co-infiltrations of RcSPL1-nYFP with cYFP and of nYFP with RcTAF15b-cYFP were used as negative controls. pSuper:NF-YA4-mCherry was co-infiltrated as a nuclear marker. Scale bar, 50 μm. The experiment was performed independently three times, with similar results. Representative results from one experiment are shown. c RcSPL1 co-localization with RcTAF15b in tobacco leaves. RcSPL1-GFP and RcTAF15b-mCherry were transiently co-expressed in tobacco leaves. Co-localizations of RcSPL1-GFP with mCherry and of GFP with RcTAF15b-mCherry were used as negative controls. Scale bar, 50 μm. The experiment was performed independently three times, with similar results. Representative results from one experiment are shown.
transgenic A. thaliana lines and the RcSPL1-overexpressed rose plants than in their corresponding control plants, but the transcripts were accumulated at lower levels in the RcSPL1-silenced rose plants (Figs 3g-i and 4c and f). Therefore, altering the expres-sion of RcSPL1 disrupted the expression of RcAP1, RcFUL, and RcLFY. Previous research demonstrated that SPL proteins regulate the transcript accumulation of downstream genes through the binding of the SBP domain to the GTAC core motif [49]. Indeed, several GTAC-boxes exist in the RcAP1, RcFUL, and RcLFY promoters according to Plantpan3.0 (http://plantpan.itps.ncku.edu. tw/index.html), suggesting that RcSPL1 may bind directly to the promoters to regulate the f lowering time in rose.
The current reports suggested that SPL can link to multiple f lowering signals derived from photoperiodic, GA, and vernalization pathways to control f lowering. In Arabidopsis, SOC1 directly regulates the expression of SPL3, SPL4, and SPL5 by binding to their promoters in response to photoperiod signals [50]. In the GA f lowering pathway, DELLA proteins are recruited to the AP1 locus by SPL9, which induces the expression of AP1 and promotes the transition of lateral primordia to f lowers [51]. In Arabis alpina and Cardamine f lexuosa, age regulates the timing of sensitivity in response to vernalization, which is associated with the miR156-SPL-miR172 cascade [15,52,53]. In the current study, we revealed the interaction between RcSPL1 and RcTAF15b and determined that the resulting complex is localized in the nucleus (Fig. 5). Generally, SPLs, which are the key regulators of the age pathway, and TAF15b, which is a typical autonomous pathway protein, are involved in the regulation of reproductive processes [16,33]. Our results on the interaction between RcSPL1 and RcTAF15b may imply crosstalk of the age and autonomous pathways in f lowering time regulation. Transiently silencing or overexpressing RcTAF15b in rose via TRV resulted in late or early f lowering, respectively (Fig. 6), indicating RcTAF15b contributes to rose f loral transition. In A. thaliana, TAF15b promotes f lowering by repressing the transcription of FLOWERING LOCUS C (FLC), which is a f lowering repressor [33]. However, both 'Old Blush' and R. chinensis var. spontanea lack the FLC gene [1,54]. In rose, both RcSPL1 and RcTAF15b promote early f lowering. Thus, we speculate that RcTAF15b acts synergistically with RcSPL1 to coregulate the expression of the f loral meristem identity genes RcAP1, RcFUL, and RcLFY. Further studies will be required to clarify whether RcTAF15b directly regulates f lowering-related genes in rose.

Conclusions
The study provides new insights into the mechanisms regulating f lowering time. Specifically, RcSPL1, a member of the SPL family, exhibited continuous increased expression during the rose f lowering process and controls the expression of the f loral meristem identity genes, RcAP1, RcFUL, and RcLFY. Additionally, its upstream regulator of rch-miR156 and the interacting partner RcTAF15b were also identified. These results help to understand the mechanism underlying rose f lowering control via RcSPL1 and RcTAF15b synergistically regulating f lowering time, with potential implications for the breeding and cultivation of rose or other perennial plants.
WT and transgenic A. thaliana plants and N. benthamiana plants were grown at 25 • C with a 16-hour light/8-hour dark photoperiod and 60% relative humidity.

Rose morphological examination
'Old Blush' samples at stages 1-3 were fixed in FAA solution (formaldehyde:acetic acid:50% ethanol = 5:5:90 v/v) and then dehydrated in a graded ethanol series before being embedded in paraffin. Sections (0.5-0.8 μm) prepared using a microtome (Leica RM2265) were deparaffinized using xylene and rehydrated in a graded ethanol series. Finally, the sections were stained with toluidine blue and examined using an inverted microscope (Zeiss Observer Z1).
R. chinensis var. spontanea (stages 4-6) were carefully dissected using a binocular stereomicroscope and then fixed in 2.5% glutaraldehyde (v/v) for >2 h under vacuum conditions. The samples were rinsed with 0.1 ml phosphate-buffered saline (PBS), followed by a post-fixation treatment with 1% osmium tetroxide. The samples were rinsed again with 0.1 ml PBS and then dehydrated using a graded alcohol series and a critical point dryer (Leica EM CPD). Finally, samples were coated with gold (EIKO IB-3) and examined using a scanning electron microscope (Hitachi S-3400 N).

RNA extraction and RT-qPCR
Total RNA was isolated using the FastPure Plant Total RNA Isolation Kit (Vazyme, Nanjing, China) and then reverse-transcribed to cDNA using the HiScript III RT SuperMix for qPCR (+ gDNA wiper) kit (Vazyme). First-strand miRNA was reverse-transcribed using step-loop primers and the miRNA 1st Strand cDNA Synthesis Kit (Accurate Biology, China). The RT-qPCR analysis was performed using the ChamQ Universal SYBR qPCR Master Mix (Vazyme), with AtActin and RcUBI2 used as the reference genes for A. thaliana and R. chinensis, respectively. Additionally, 5.8S rRNA and U6 snRNA were used to normalize the abundance of miRNA in R. chinensis and N. benthamiana, respectively. All primers used in this study are listed in Supplementary Data Table S5. The significance of any differences in the data was evaluated using SPSS software (version 22.0 for Windows; Chicago, IL, USA) and GraphPad Prism (version 9.0). Student's t-test, ANOVA, or Duncan's multiple range test was used to analyze the experimental data.

Subcellular localization
The RcSPL1 coding sequence (CDS) was fused to the sequence encoding GFP and inserted into the pCAMBIA1300 vector containing the Super promoter to construct the pSuper:RcSPL1-GFP plasmid. pSuper:NF-YA4-mCherry was applied as a nuclear marker [39]. The recombinant plasmids were introduced into Agrobacterium tumefaciens strain GV3101 cells by electroporation. A. tumefaciens harboring different plasmids was resuspended in MMA infiltration buffer (10 mM MgCl 2 , 10 mM MES, and 100 μM acetosyringone, pH 5.6; OD 600 = 0.8). N. benthamiana leaves were coinfiltrated with the A. tumefaciens suspensions. After 3 days, f luorescent signals were detected using a laser confocal microscope (Leica TCS SP8). The primers used in this study are listed in Supplementary Data Table S5.

Prediction of RcSPL1 targeted by rch-miR156
The previously reported rch-miR156 sequence was obtained [56], whereas the rch-miR156 precursor sequence in rose was downloaded from the 'Old Blush' pre_miRNA database (https://lipmbrowsers.toulouse.inra.fr/pub/RchiOBHm-V2/). The RcSPL1 gene potentially targeted by miR156 was identified from the 3 UTR of RcSPL1 for sequences complementary to rch-miR156 using the default parameters of the psRNATarget server (https://www. zhaolab.org/psRNATarget/). The predicted cleavage site in RcSPL1 and the primary of rch-miR156 (pri-rch-miR156) transcript were amplified separately by PCR and cloned into the pCAMBIA1300 vector with GFP (130-GFP vector) or without GFP (pCAMBIA1300 vector) to construct the p35S:RcSPL1-sensor-GFP and p35S:pri-rch-miR156 recombinant plasmids, respectively. The plasmids were introduced into A. tumefaciens GV3101 cells by electroporation. Cultures of A. tumefaciens cells containing p35S:RcSPL1-sensor-GFP (RcSPL1-sensor-GFP), pCAMBIA1300, or 130-GFP (OD 600 = 0.5), and cultures of A. tumefaciens cells containing p35S:pri-rch-miR156 (prirch-miR156) (OD 600 = 0.2, 0.5, and 0.8) were prepared. Cultures of RcSPL1-sensor-GFP with pri-rch-miR156 or pCAMBIA1300 in a 1:1 (v/v) ratio were used for the infiltration of N. benthamiana leaves. After 2-3 days, the percentage of cells with GFP signal in the same-size microscopic field was determined. Specifically, GFP f luorescence was detected using FUSION FX EDGE SPECTRA and a laser confocal microscope (Leica STED). The abundance of rch-miR156 in N. benthamiana leaves was quantified on the basis of an RT-qPCR analysis. The RcSPL1 ORF with or without the 3 UTR was amplified by PCR and inserted into 130-DAS-His vector at the 3 end of the 6× His-tag encoding sequence to construct the p35S:His-RcSPL1-3 UTR and p35S:His-RcSPL recombinant plasmids, respectively. All of the recombinant constructs were expressed in N. benthamiana leaves. After 3 days, the Hisfused protein was detected. The experiments were repeated three times.

Arabidopsis thaliana transformation
The RcSPL1 CDS was amplified by PCR and inserted into the SacI and XbaI sites of the pBI121 vector to construct the p35S:RcSPL1 recombinant plasmid, which was subsequently introduced into A. tumefaciens cells for the transformation of A. thaliana via the f loral dip method [57]. Independent transgenic A. thaliana lines were screened on MS basal medium supplemented with 40 mg l −1 kanamycin. The putative transformants were confirmed via a PCR amplification. The T 2 generation plants were used for further analysis. To detect target gene expression, the rosette leaves of 3week-old A. thaliana plants were collected and immediately frozen in liquid nitrogen.

Virus-induced gene silencing and overexpression in rose
To silence RcSPL1 and RcTAF15b, gene-specific fragments of RcSPL1 (343 bp) and RcTAF15b (305 bp) were cloned into separate pTRV2 vectors to obtain the pTRV2-RcSPL1 and pTRV2-RcTAF15b recombinant plasmids. The RcSPL1 and RcTAF15b genes were silenced in rose plantlets as previously described [39]. Brief ly, cultures of A. tumefaciens cells transformed with pTRV1, pTRV2-RcSPL1, pTRV2-RcTAF15b, or pTRV2 (negative control) were prepared (OD 600 = 1.2), after which the cells were resuspended in MMA buffer supplemented with 0.001% (v/v) Silwet-L77 (OD 600 = 1.0). Finally, solutions comprising cells containing pTRV1 and cells containing pTRV2-RcSPL1, pTRV2-RcTAF15b, or pTRV2 (equal ratio; v/v) were prepared. Before infiltrating plants, the mixtures were kept in darkness at room temperature for 3-5 hours. Whole plants were submerged in infiltration buffer and exposed to vacuum conditions (−20 kPa) twice, each for 1 minute, and then the plants were rinsed with distilled water and transferred to pots. After 3-4 weeks, new leaves were collected from the treated plants and immediately frozen in liquid nitrogen. The transcription of the target genes (RcSPL1 or RcTAF15b) in each plant was detected by performing an RT-qPCR analysis. Approximately 20 plants in which the expression levels of the target genes were downregulated were selected for phenotypic examination.
Rose plants overexpressing the target genes were created using psTRV1 and psTRV2 plasmids [43]. RcSPL1 and RcTAF15b ORF fragments containing a KpnI restriction site were inserted into psTRV2 to obtain psTRV:RcSPL1 and psTRV:RcTAF15b recombinant plasmids, respectively. The RcSPL1 and RcTAF15b genes were transiently overexpressed in rose plantlets via the above-mentioned operation of the VIGS experiment. Plants infiltrated with the empty psTRV vectors (psTRV1 + psTRV2) were used as the sTRV control plants. After 3-4 weeks, new leaves were collected from each treated plant and immediately frozen in liquid nitrogen. RT-qPCR analysis was completed to confirm that the target genes were more highly transcribed in the gene-overexpressing rose plants than in the control plants. Finally, ∼20 plants in which the target gene expression levels were upregulated were selected for phenotypic examination. Both TRV-mediated gene silencing and overexpression experiments were performed three times using different batches of rose plants, with each batch comprising at least 70 plantlets.

Yeast two-hybrid assay
A Y2H assay was performed to screen a rose f loral bud cDNA library for RcSPL1-interacting proteins [58]. The RcSPL1 ORF was cloned into pGBKT7 BD to construct the bait vector, whereas the RcTAF15b ORF was inserted into pGADT7 AD to generate the prey vector. The Y2H assay was completed using the Matchmaker™ Gold Yeast Two-Hybrid system. As positive and negative controls, Y2HGold yeast cells were co-transformed with pGADT7-T and pGBKT7-53 or pGBKT7-Lam. The transformants were grown on SD/−Trp/−Leu and SD/−Trp/−Leu/−His/−Ade media for 3 days at 30 • C.

Bimolecular fluorescence complementation assay
For the BiFC assay, the RcSPL1 ORF was inserted into pSYNE-35S, which contains the sequence encoding the N terminus of YFP (nYFP), while the RcTAF15b ORF was inserted into pSPYCE-35S, which contains the sequence encoding the C terminus of YFP (cYFP). pSuper:NF-YA4-mCherry was used as a nuclear marker [39]. A mixture comprising A. tumefaciens cells harboring RcSPL1-nYFP and RcTAF15b-cYFP was used for the co-infiltration of N. benthamiana leaves. The following combinations were used as negative controls: RcSPL1-nYFP + cYFP and nYFP + RcTAF15b-cYFP. After 3 days, the leaves were examined for YFP and mCherry signals.