KSN heterozygosity is associated with continuous flowering of Rosa rugosa Purple branch

Rose (Rosa spp.) plants flower via two contrasting methods: once flowering (OF) and continuous flowering (CF). Purple branch is a rare continuously flowering variety of Rosa rugosa that is extensively cultivated in China. However, the genetic basis of its CF behavior is unknown. We demonstrated that Purple branch is heterozygous for the TFL1 homolog KSN. One KSN allele with a 9 kb Copia insertion was found to be identical to that from continuously flowering Rosa chinensis Old blush. The other allele was found to be a functional wild-type allele. The overall expression of KSN was closely linked to the floral transition, and it was significantly repressed in continuously flowering Purple branch compared with OF Plena. The promoter region of the normal KSN allele was hypermethylated, and histone methylation at H3H4, H3K9, and H3K27 of the KSN gene locus was modified in continuously flowering Purple branch. Silencing of the DNA methyltransferase genes MET1 and CMT3 and the histone methyltransferase gene SUVR5 in Purple branch led to enhanced KSN expression, but silencing of the histone demethylase gene JMJ12 suppressed KSN expression. Therefore, the CF habit of Purple branch may be due to reduced expression of KSN caused by the halved dose and may be associated with epigenetic modifications together with retrotransposon insertions along the chromosome. Our study revealed a novel mechanism underlying the CF behavior of rose plants.


Introduction
Rosa rugosa Thunb., a member of the Rosaceae family, which is indigenous to Eastern Asia, was introduced into Europe and North America in the middle of the nineteenth century 1,2 . In addition to producing fragrant flowers, this popular and economically valuable ornamental species has medicinal properties. R. rugosa flowers are widely used in traditional and folk medicine in China, Japan, and Korea due to the presence of secondary metabolites that exert pharmacological activities 3 . Essential oil from petals, known as "liquid gold", is a raw material used in perfumes, cosmetics, aromatherapy, spices, and the nutrition industry 4 . Unlike modern roses (Rosa hybrida), which flower continuously throughout the year, most R. rugosa varieties flower once a year during the spring. Efforts to produce flowers year-round rely on screening and breeding new continuously flowering R. rugosa varieties.
Flowering, which involves the transition from vegetative growth to reproductive development, is controlled by external environmental cues and endogenous signals through six genetic pathways in Arabidopsis: the photoperiod, vernalization, autonomous, gibberellin, age, and temperature pathways 5 . Rose plants have three flowering methods: an once flowering (OF) period in the spring; continuous flowering (CF) during favorable growth seasons; and occasional reblooming 6,7 .
Rose KSN, a homolog of TFL1 of Arabidopsis thaliana, acts as a flowering repressor to control OF/CF. In OF rose cultivars, the expression of KSN is repressed in the winter and early spring, so the plants bloom only in late spring. After blooming, KSN expression is activated and represses further flower formation in the summer 6,8,9 . Continuously flowering Rosa chinensis Old blush contains a 9 kb Copia retrotransposon insertion in the second intron of KSN, and this insertion blocks KSN expression, enabling continuous flowering 9 . Old blush is hybridized with European rose plants to generate modern rose plants, in which the recurrent flowering trait has presumably been transferred into the modern rose plants 6 . Therefore, most continuously flowering rose cultivars are expected to harbor this mutated allele of KSN from Old blush. However, the presence of a null allele has been suggested in the haploid Old blush genome sequence and possibly contributes to its CF behavior 10 , and some continuously flowering rose plants have no Copia insertion in their KSN gene 11,12 ; thus, the mutation of KSN may not be the only reason for the rose CF trait.
In Arabidopsis, TFL1 maintains the indeterminate growth of the inflorescence meristem by inhibiting AP1 and LFY expression 13,14 . In tfl1 mutants, indeterminate meristems are rapidly converted into determinate meristems, which produce terminal flowers 13,15,16 . Functional characterization of FvTFL1 has confirmed its role as a floral repressor that causes OF: mutations in this gene lead to CF in strawberry 9,17 . In perennial woody plant species, silencing the TFL1 homolog shortens the juvenile period and causes continuous flowering. For example, silencing MdTFL1 in apple (Malus domestica) reduces the juvenile period from 5 years to several months 18,19 , whereas silencing PcTFL1 in pear (Pyrus communis L.) accelerates flowering by 1-8 months 20 . In Populus spp., silencing of the TFL1 orthologs PopCEN1/PopCEN2 hastens the first onset of flowering, maintains the indeterminate growth of axillary meristems and accelerates bud dormancy release upon chilling 21 . In perennial Arabis alpine plants, AaTFL1 regulates the duration of agedependent vernalization required for AaLFY expression and sets a threshold for flowering 22 . Therefore, TFL1 (KSN) is functionally conserved as a flowering repressor.
Purple branch is a rare continuously flowering variety of R. rugosa that is cultivated widely in China. R. rugosa Purple branch provides raw materials for the production of essential oils, jams, teas, pies, beverages, and other derivatives, integrating the primary, secondary and tertiary industries. To date, the origin of Purple branch remains unclear, but it is generally considered to have derived from an interspecific cross between R. rugosa Plena and Rosa davurica 23,24 . Intriguingly, its presumptive parents are OF rose species (Fig. 1); thus, its CF behavior remains unexplained. Elucidating the molecular mechanism of the CF trait in Purple branch will provide a theoretical basis for breeding new varieties with different flowering phenotypes.
In this paper, we demonstrate that the continuously flowering R. rugosa Purple branch is heterozygous for KSN: it contains one normal transcribed wild-type allele and another transcriptionally blocked allele with a 9-kb Copia insertion, identical to that in R. chinensis Old blush. We show that the CF behavior of Purple branch is associated with reduced KSN expression. This reduction is due to the halved dose of the wild-type KSN allele and is linked to promoter hypermethylation and histone modification at the KSN locus. We thus present a novel mechanism for the production of the CF habit of rose plants.

Gene cloning and expression analysis
Genomic DNA from Purple branch, Plena, R. davurica and Old blush leaves was isolated by the CTAB method 25 . The different fragments of the KSN gene were isolated by PCR using the primers listed in Table S1. For expression analysis, shoot apices of field-grown Purple branch, Plena and R. davurica were collected on 29 March, 11 April and 30 September 2019, flash frozen in liquid nitrogen, and then stored at −80°C. Total RNA was extracted using a Vazyme FastPure Plant Total RNA Isolation Kit (Polysaccharides & Polyphenolics rich, Vazyme Biotech, Nanjing, China). One microgram of high-quality total RNA was reverse transcribed using TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix (Transgen Biotech, Beijing, China) according to the manufacturers' instructions. The cDNA was amplified via RT-PCR, and quantitative reversetranscription PCR (qRT-PCR) of KSN and FT was carried out with ABI QuantStudio 5 instrument (ABI Life Technologies, Carlsbad, USA) using the primers listed in Table  S1; RcGAPDH was used as an internal reference 26 . Each experiment was conducted for three biological replicates, with three technical repeats each.

Methylation-specific PCR (MSP)
Methylation of the KSN promoter was studied using MSP. Less than 500 ng of genomic DNA was treated with sodium bisulfite using an EZ DNA Methylation Kit (Zymo Research, Irvine, USA). The treated DNA was subsequently used as a template for PCR. Approximately 1 kb of the KSN promoter region was amplified using two pairs of primers (Table S1) that were designed via Methyl Primer Express v1.0. The total volume of the PCR was 25 μl, and PCR was performed using TaKaRa EpiTaq HS (for bisulfite-treated DNA) according to the following protocol: 35 cycles of 98°C for 10 s, 55-60°C for 30 s, and 72°C for 30 s, followed by 72°C for 10 min. The resulting PCR fragments were ligated into a pMD-18-T vector (TaKaRa, Dalian, China), which was then transformed into Escherichia coli (Trans-T1) (Transgen Biotech, Beijing, China), and 10 clones per sample were sequenced using a bioanalyzer. Each experiment was performed for three biological repeats. The methylation percentages of cytosine (C) in CG, CHH and CHG (H=A, C or T) were analyzed using online software (http://katahdin.mssm.edu/ kismeth/revpage.pl), and the running parameters were as follows: minimum fraction of positive matches, 0.8; minimum fraction of length, 0.5.

Chromatin immunoprecipitation PCR (ChIP-PCR)
ChIP-PCR was performed as described previously 27 , with slight modifications. Shoot apices of field-grown Purple branch and Plena were collected in September 2020 and ground in liquid nitrogen in M1 buffer (10 mM phosphate buffer [pH 7.0], 0.1 M NaCl, 10 mM mercaptoethanol, 1 M hexylene glycol, 1× protease inhibitor cocktail [Roche], 5% PVP, and 1 mM PMSF). The suspension was filtered through four layers of Miracloth, after which the filtrate was centrifuged at 12,000 rpm for 10 min. The pelleted chromatin was washed thrice with M2 buffer (M1 buffer plus 10 mM MgCl 2 and 0.5% Triton X-100) and once with M3 buffer (10 mM phosphate buffer [pH 7.0], 0.1 M NaCl, 10 mM mercaptoethanol, 1× protease inhibitor cocktail [Roche], and 1 mM PMSF). The chromatin was resuspended in nuclear lysis buffer and sonicated to generate DNA fragments of approximately 500 bp. The lysate was precleared by incubation together with 50 µl of protein-A agarose beads/salmon sperm DNA (Millipore, Billerica, USA) for 1 h. It was then incubated together with IgG, anti-H3K4me3, anti-H3K9me3, and anti-H3K27me2 antibodies (Abcam, Cambridge, UK) overnight. The bound DNA fragments were recovered and purified using columns from a plasmid extraction kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Quantitative real-time PCR was subsequently performed using the bound and input DNA as templates in conjunction with the primers listed in Table S1.

Transient transformation system
The coding regions of MET1, CMT3, JMJ12, and SUVR5 of R. rugosa Purple branch were cloned using the primers listed in Table S1 and then sequenced. The resulting sequences were used for phylogenetic analysis with Arabidopsis thaliana to validate their reliabilities and then submitted to the NCBI database.
To analyze the RNAi effects, young shoots of Purple branch were transiently transformed as previously described 28 . Briefly, young shoots were placed into a 50-mL Agrobacterium solution carrying the gene fragments of interest and vacuum infiltrated for 5 min, and an Agrobacterium solution carrying the empty vector served as a control. After being released from the vacuum, the shoots were washed with deionized water, and the leaves were collected for RNA extraction and q-PCR analysis after 3 days.

Statistical analyses
Statistical analysis of the data was performed via SPSS 17 statistical software. Two groups of data were compared using Kolmogorov-Smirnov (KS) tests (*P < 0.05; **P < 0.01). Multiple groups of data were compared using the Kruskal-Wallis (KW) test, with P < 0.05 considered significant.

Purple branch contains two KSN alleles
We cloned the KSN promoter and coding region from continuously flowering Purple branch, OF Plena, and R. davurica using the same primers that were previously used to amplify this region in R. chinensis Old blush 29 . We used the primer pair F1/R1 to amplify the promoter fragment and the primer pair F1/R2 to clone the fulllength KSN gene, including both the promoter and coding areas (Fig. 2a). Based on the results of electrophoresis analysis of the PCR products on agarose gels, we obtained two bands from the Purple branch, with primers F1/R1 or F1/R2, and only one band of KSN from Plena and R. davurica (Fig. 2b, d).
Subsequent sequence alignment showed that the sequence of one of the KSN alleles from Purple branch was almost the same as that from Plena, except for a few SNPs and a 44 bp insertion in the second intron. The other allele was similar to that from CF R. chinensis Old blush and contained a Copia retrotransposon insertion in the second intron (Figs. 2c, e and S1). These results indicate that the diploid Purple branch is heterozygous for KSN (KSN-Copia/KSN-Wt-Pb) (Fig. 2f). Because neither allele was similar to that of R. davurica, our work contradicts the notion that Purple branch was derived from an interspecific cross between R. rugosa Plena and R. davurica.
KSN expression is lower in Purple branch than in Plena and R. davurica Because one of the KSN alleles contains a 9 kb Copia insertion in the second intron in Purple branch, we tested whether the other allele contains a 44 bp insertion in the second intron that was transcribed normally. We cloned the coding regions of KSN from the cDNA of Purple branch, Plena, and R. davurica. We obtained a single band from the three different genotypes and sequenced the amplicons (Fig. 3a). Purple branch was identical to Plena (Fig. 3d, e) and contained a few bases that were different To further compare the expression levels of KSN among Purple branch, Plena, and R. davurica and determine the link between KSN and flowering modes, we conducted qRT-PCR on samples collected in March (before blooming), April (during blooming), and September (continuously flowering Purple branch blooms, but OF R. rugosa Plena and R. davurica do not bloom) from all three genotypes. KSN expression was inhibited before flowering, maintained at relatively low levels during flowering, but obviously increased after flowering in September not only in OF R. rugosa Plena and R. davurica but also in continuously flowering Purple branch (Fig. 3b). However, the KSN expression in continuously flowering Purple branch being dramatically lower than that in OF Plena and R. davurica at all three time points may be caused by halved doses, leading to a CF habit. We also analyzed FT expression at the same time points. The expression levels of FT in Purple branch, Plena, and R. davurica were all higher in April than in March, and the levels still increased in September in continuously flowering Purple branch but declined significantly in OF R. rugosa Plena and R. davurica (Fig. 3c). Therefore, KSN expression was negatively associated with the flowering transition in the spring, and the flowering of Purple branch in autumn may be due to the integrative effects of both KSN and FT.
KSN from Purple branch rescues the early-flowering tfl1 mutant phenotype To determine the functional conservation of the KSN transcript from Purple branch, we expressed this gene under the constitutive 35S promoter in the Arabidopsis tfl1 mutant background. We then examined the flowering phenotype and gene expression levels in the transgenic plants.
The heterologous expression of KSN from Purple branch rescued the early flowering and determinate growth of the inflorescence of tfl1 mutants (Fig. 4a, b). The tfl1-14 mutant produced 5.8 rosette leaves in contrast to the 9 rosette leaves produced by wild-type Arabidopsis at flowering. The flowering of the KSN-OE-1 and KSN-OE-2 lines was delayed to occurring when there were more than 20 rosette leaves. This delay was associated with increased KSN mRNA levels. Seedlings of the KSN-OE-3 line, with relatively high KSN expression levels, were more delayed and produced more than 30 rosette leaves before flowering (Fig. 4b, c).
In addition, some transgenic seedlings produced leaflike floral organs that generated no seeds. These results corroborated those from studies of overexpression of TFL1 from Vitis vinifera, Lotus japonicus and Rosa wichurana in Arabidopsis [30][31][32] . Thus, the KSN of Purple branch is the functional homolog of TFL1 in Arabidopsis and other species and acts as a floral repressor.

The KSN promoter in Purple branch is hypermethylated
We hypothesized that the heterozygosity-induced dose decrease may not be the only reason for the inhibition of KSN. The 9 kb insertion may alter the chromatin structure and induce epigenetic suppression or activation. Using MSP, we tested the methylation status of CpG islands in the promoter of the normally transcribed KSN allele in Purple branch. The B fragment (-984 to -502 bp) rather than the A fragment (-501 to -92 bp) was identified as the CpG island for DNA methylation (Fig. 5a, b). The total percentages of the three types of methylation (CG/CHG/ CHH) within the B fragment were slightly higher in continuously flowering Purple branch than in OF Plena at all three time points (Fig. 5d), while it was higher only in April within the A fragment in continuously flowering Purple branch (Fig. S2a and S2c). Of the three types of methylation, CG and CHG methylation were most common (Figs. 5c and S2b). The methylation intensity within the promoter was negatively associated with the corresponding level of the KSN transcript presented in Fig. 3. Thus, hypermethylation may have partially contributed to the lower expression of KSN in Purple branch.

Histone methylation of the KSN locus in Purple branch is modified
To investigate how histone modification regulates KSN expression, we collected shoot apices from fieldgrown Purple branch and Plena and analyzed KSN expression and the chromatin environment at the KSN locus in continuously flowering Purple branch and OF Plena via ChIP using H3K4me3-specific antibodies (to determine active loci) and H3K9me3-and H3K27me2specific antibodies (to determine silent loci). We designed specific primers throughout the promoter and the second intron regions (Fig. 6a and Table S1) and performed ChIP-qPCR. KSN expression was obviously lower in Purple branch than in Plena (Fig. S3), which is in line with the previous results shown in Fig. 3. The histone methylation marks H3K9me3 and H3K27me2 increased in continuously flowering Purple branch but remained unchanged or decreased in OF Plena. The H3K4me3 mark decreased in continuously flowering Purple branch but remained unchanged in OF Plena (Fig. 6b). Thus, the enhanced histone methylation at H3K27 and H3K9 and the reduced methylation at H3K4 may play important roles in maintaining the lower expression of the normal KSN allele in Purple branch, leading to its CF behavior.
KSN expression increases upon silencing MET1, CMT3, and SUVR5 In Arabidopsis, DNA methylation is mainly performed by members of the C5-MTase families. MET1, which belongs to the METHYLTRANSFERASE (MET) family, maintains CG methylation. CMT3, which belongs to the CHROMOMETHYLTRANSFERASE (CMT) family, maintains CHG methylation 33 . The cmt3 mutant shows a near-complete loss of CpXpG methylation, and met1 shows a marked reduction in CpG methylation; both types of methylation in turn cause gene silencing 34,35 . Histone modifications affect various changes in chromatin structure, leading to the promotion or suppression of gene expression 36 . JMJ12 (REF6), which belongs to the KDM4/JHDM3 group of the JmjC family 37 , encodes a histone H3 lysine 27 demethylase 38 , and silencing of RcJMJ12 induced late flowering in Rosa chinensis 39 . The SET family gene SUVR5 regulates flowering time through H3K9me2 and H3K27me3 and is independent of the vernalization pathway [40][41][42] .
To examine the epigenetic regulation of KSN expression in Purple branch, we cloned the homologs of Arabidopsis MET1, CMT3, JMJ12, and SUVR5 from continuously flowering Purple branch and constructed phylogenetic trees together with their homologs in Arabidopsis to verify their correctness before submitting the sequences to the NCBI database (Fig. S4). We then designed primers (listed in Supplemental Table S1) specific to their conserved regions and silenced the genes in young rose shoots using a transient transformation system described previously 28 . The subsequent qRT-PCR results showed that silencing MET1, CMT3, JMJ12, and SUVR5 significantly altered the DNA methylation/histone acetylation status at the KSN locus (Figs. S5 and S6), which was followed by a significant increase in KSN transcription in Ri-MET1, Ri-CMT3 and Ri-SUVR5 seedlings compared with a decrease in KSN expression in Ri-JMJ12 seedlings of continuously flowering Purple branch (Fig. 7). Thus, DNA methylation and histone methylation are potentially partially responsible for the repression of KSN in Purple branch.

Discussion
Previous studies have genotyped KSN in a wide range of rose cultivars and established that some Damask-related roses (e.g., Damask, Moss, Hybrid musk, and Bourbon roses) and Asian roses (e.g., Hybrid rugosa and Hybrid bracteate roses) were CF types but contained no mutated alleles of KSN from R. chinensis 12 . Similarly, it was also found that continuously flowering R. rugosa Hamanasu from Japan had only a wild KSN allele without any insertion or obvious mutations 11 . Therefore, the CF behavior of rose plants may not be simply explained by the mutation or insertion of the KSN gene. Interestingly, the present results clearly showed that the KSN allele with the Copia insertion from R. chinensis was substantial in R. rugosa Purple branch (Figs. 2 and 3), which is heterozygous (KSN-Copia/KSN-Wt-Pb) at the KSN locus. The wild-type KSN allele was proven to be functional via heterologous expression in Arabidopsis (Fig. 4). This discrepancy may be caused by different experimental materials or methods, as there are different cultivars and varieties of R. rugosa from different origins. Additionally, the KSN allele carrying the 9 kb insertion was not easily amplified with the normal DNA polymerase and general PCR protocol.
Next, KSN expression was suppressed during the floral initiation period in early spring (April) and dramatically upregulated in the shoot apices after flower initiation in all three species, although KSN expression was still much lower in continuously flowering Purple branch than OF Plena and R. davurica in September (Fig. 3b). In contrast, expression of the floral activator FT significantly increased in April in all three species, decreased in OF Plena and R. davurica but continued climbing in continuously flowering Purple branch (Fig. 3c). Consequently, continuously flowering Purple branch exhibited high expression of both FT and KSN in September. Plant flowering is precisely determined by the equilibrium between floral promoters and repressors. FT, a member of the phosphatidylethanolamine-binding protein (PEBP) family, is a key florigen that integrates the internal and external signals for floral transition 43,44 . In the shoot apical meristem (SAM), FT interacts with FD to activate floral identity genes such as AP1 and LFY and initiates floral bud formation 45,46 . Like the flowering promoter FT, TFL1 is also classified as a PEBP member 47 but with a function opposite that of FT, acting as a floral repressor. TFL1 interacts with FD to form a TFL1-FD complex, opposing the function of the FT-FD complex in regulating downstream gene expression 48 . Thus, the floral activator FT may antagonize or suppress the repressor KSN in continuously flowering Purple branch, enabling its flowering in autumn.
Nevertheless, the overall expression of KSN was clearly linked to the alternation of vegetative/reproductive stages of rose plants in the present paper and in six other nonrecurrent flowering species and nine recurrent flowering cultivars 49 . Furthermore, the expression level of KSN in continuously flowering Purple branch was obviously lower than that in OF Plena and R. davurica at all three time points, which was convincingly associated with the CF trait of Purple branch. Although other factors may be involved in controlling the CF trait, the roles of reduced KSN expression should be predominant and can be mainly attributed to the decreasing dose associated with the heterozygosity of KSN loci in Purple branch (Figs. 2 and 3).
All plants have various transposons (TEs) that can disrupt local gene structure, affect the expression of nearby genes, and induce chromosomal instability. Most TEs are silenced and immobilized under normal conditions. DNA methylation and histone methylation are reversible epigenetic modifications that control transposon activity [50][51][52] . The 9 kb Copia-like retrotransposon insertion in R. chinensis Old blush not only blocks transcription of the host KSN allele but also results in a large rearrangement at the CF locus, leading to the complete deletion of the other KSN allele 11 . Therefore, the diploid Old blush is hemizygous: RoKSN Copia /RoKSN null . This allowed us to test the DNA and histone methylation status of the KSN locus. We observed hypermethylation of the KSN promoter and enhanced methylation levels of H3K9 and H3K27 at the KSN locus in continuously flowering Purple branch compared with OF varieties without Copia insertions (Figs. 5 and 6). These results implied that the presence of Copialike retrotransposon insertion may potentially affect the epigenetic status of the KSN locus in continuously flowering Purple branch. Further silencing the DNA methyltransferase genes MET1 and CMT3 and histone methyltransferase gene SUVR5 in young shoots significantly increased the expression of KSN in the Purple branch (Fig. 7), highlighting the importance of epigenetic modification in KSN expression. Taken together, these data suggested that the reduction in KSN expression may be associated with epigenetic modifications, except for the halved dose in continuously flowering Purple branch.
In conclusion, the R. rugosa Purple branch was found to be heterozygous at KSN, and its CF trait was associated with the lower expression of KSN. The reduction in KSN was caused by the halved dose and was also associated with hypermethylation of the promoter region and histone modification at the KSN locus following the 9 kb Copia insertion in the other allele. This suggests a novel mechanism for the production of the CF habit in rose plants.