Low temperature-induced DNA hypermethylation attenuates expression of RhAG, an AGAMOUS homolog, and increases petal number in rose (Rosa hybrida)

Flower development is central to angiosperm reproduction and is regulated by a broad range of endogenous and exogenous stimuli. It has been well documented that ambient temperature plays a key role in controlling flowering time; however, the mechanisms by which temperature regulates floral organ differentiation remain largely unknown. In this study, we show that low temperature treatment significantly increases petal number in rose (Rosa hybrida) through the promotion of stamen petaloidy. Quantitative RT-PCR analysis revealed that the expression pattern of RhAG, a rose homolog of the Arabidopsis thaliana AGAMOUS C-function gene, is associated with low temperature regulated flower development. Silencing of RhAG mimicked the impact of low temperature treatments on petal development by significantly increasing petal number through an increased production of petaloid stamens. In situ hybridization studies further revealed that low temperature restricts its spatial expression area. Analysis of DNA methylation level showed that low temperature treatment enhances the methylation level of the RhAG promoter, and a specific promoter region that was hypermethylated at CHH loci under low temperature conditions, was identified by bisulfite sequencing. This suggests that epigenetic DNA methylation contributes to the ambient temperature modulation of RhAG expression. Our results provide highlights in the role of RhAG gene in petal number determination and add a new layer of complexity in the regulation of floral organ development. We propose that RhAG plays an essential role in rose flower patterning by regulating petal development, and that low temperatures increase petal number, at least in part, by suppressing RhAG expression via enhancing DNA CHH hypermethylation of the RhAG promoter.


Background
Floral patterning, which is essential for angiosperm reproduction, involves the arrangement of four organs types in concentric whorls: the sepals and petals, which comprise the perianth and form the outer two whorls; and the stamens and carpels, which are the male and female reproductive organs, respectively, and form the inner two whorls. The ABCE model, a broadly accepted model of flower development that was first proposed two decades ago, describes how the combinatorial activity of four classes of homeotic genes determines floral organ identity [1][2][3][4][5]. According to the ABCE model, sepals are specified by A-and E-class genes, petals by A-, B-and E-class genes, stamens by B-, C-and E-class genes, and carpels by C-and E-class genes.
In Arabidopsis thaliana, several homeotic genes have been identified, including two A-class genes, APETALA1 (AP1) and APETALA2 (AP2), two B-class genes, APE-TALA3 (AP3) and PISTILLATA (PI), one C-class gene, AGAMOUS (AG), and four E-class genes, SEPALLATA1 (SEP1), SEP2, SEP3, and SEP4 [4]. All of the four classes of genes, with the exception of AP2, encode MADS-box family transcription factors. These proteins have been proposed to form higher-order complexes that are required for the correct transcription of organ-specific genetic programs [3,[6][7][8][9]. Genetic and molecular studies with several model plant species have shown that mutations in A-, B-, C-, and E-class genes all result in abnormal flowers, due to the replacement of one organ type by another [9][10][11].
A number of ornamental plants, including rose (Rosa hybrida), peony (Paeonia suffruticosa), carnation (Dianthus caryophyllus), and camellia (Camellia japonica), have flowers with greater numbers of petals (termed double flowers) and consequently are popular garden plants, due to their attractive appearance. Efforts have been made to characterize the genetic mechanisms involved in the formation of double flowers and studies of various species, including A. thaliana, have shown that loss of expression of AG results in the conversion of reproductive organs to perianth organs, as well as indeterminacy of the floral meristem, leading to showy double flowers [12][13][14][15]. In the ranunculid, Thalictrum thalictroides, down-regulation of the AG homolog ThtAG1 has been shown to result in homeotic conversion of stamens and carpels into sepaloid organs, as well as a loss of flower determinacy. Moreover, it was reported that a mutant ThtAG1 protein with Kdomain deletions, which was identified in a doubleflower ornamental cultivar, cannot interact with the putative E-class protein ThtSEP3, suggesting a deep conservation of the dual function of C-class genes, and of the interactions between C-and E-class proteins in floral patterning [9]. Genetic mapping studies in rose uncovered that the simple versus double corolla phenotype is associated with a single dominant locus, namely Blfo or d6 [16][17][18], and several QTLs [19][20][21]. The orthologue of AGAMOUS (RhAG) was also identified and proved to play an important role in petal doubling in rose, through RhAG do not colocalize with Blfo or the QTL for petal number [18,22]. The spatial restriction of the RhAG expression domain may result in a homeotic conversion of organ identity from stamens to petals, and is a key factor for selection of double flowers in both the Chinese and peri-Mediterranean centers of domestication [23]. The role of AG in the transformation of stamens into petals has been shown to be associated with the A-class gene AP2 in A. thaliana, and the mutual antagonism of AG and AP2 is the central tenet of the ABC model of floral patterning [1]. However, recent research has revealed that the microRNA miR172, which might be AG-independent, is a major factor in restricting AP2 activity, and that whether stamens or petals develop relies on the balance between AP2 and AG activity, rather than a mutual exclusion of the two genes [15].
In addition to genetic determination, petal number in angiosperms is also regulated by phytohormones, including auxin and gibberellic acid [24][25][26], and by environmental cues, such as light and temperature [24,27]. For example, early reports demonstrated that either excessively high or low temperatures can cause the malformation of floral organs, especially petals and stamens [24]. Cultivating carnation at a low temperature (5°C) promoted the formation of secondary growing centers within the flower, and the marked increase in petal number was attributed to the additional petals produced from these centers [24]. In rose, reduced temperature could cause the so-called 'bullhead' phenotype, which was accompanied by an increased number of petals and a decreased number of stamens [28][29][30]. To date, however, little is known about the mechanisms involved in the temperature dependent regulation of petal number. Here, we propose a hypothesis where AG genes are involved in the temperature regulated control of petal number and thus in the formation of double flowers.
Roses have been one of the most economically important ornamental plants in the floriculture industry for centuries. As a widespread cut-and cultivated garden flower, the floral pattern is a key trait that determines its ornamental value. In this current study we found that low temperature treatments result in abnormal flowers with more petals than control flowers. We identified a rose C-class gene, RhAG, the silencing of which caused an increase in petal number. The overall spatial distribution of RhAG transcript in the floral bud was clearly decreased under low temperature conditions and further studies suggested that low temperature exposure caused DNA hypermethylation of the RhAG promoter. We conclude that RhAG plays an important role in flower patterning and that low temperatures increase the petal number of rose flowers, at least partially, by restricting the expression of RhAG.

Effects of low temperature on petaloidy of rose stamens
To investigate the effect of low temperature on rose petal development, two-year-old plants were subjected to different temperature treatments: 25/15°C (day/night, hereafter), 20/10°C and 15/5°C. We found that, compared with the control treatment (25/15°C), a lower temperature regime of 20/10°C or 15/5°C resulted in the formation of deformed flowers, which had irregular shape in flower center (Fig. 1d). As shown in Fig. 1a-d, the typical phenotype of these flowers was two growing centers and a larger flower size than the control (Additional file 1: Figure S1). Longitudinal sections of the flowers showed that the petal shape of the deformed flowers was clearly irregular, unlike that of the normal flowers (Fig. 1e, f).
The proportion of deformed flowers was only approximately 2 % at the control temperature, but was 36 % at 20/10°C and 88 % at 15/5°C, indicating that the occurrence of deformed flower formation increased markedly at lower temperatures (Additional file 1: Figure S2). Generally, the modern cultivated rose has double flowers and we investigated whether the larger flower size resulting from the treatments was a result of the petal size and/or number. Rose flowers at stage 7 (completely opened bud; Please see Methods for details) exposed to each treatment were dissected and floral organs (sepals, petals, stamens and carpels) were counted. Neither sepal nor carpel number was significantly different between the three treatments ( Table 1), but the number of petals showed a marked increase under low temperature conditions. At the control temperature, the average number of petals was 28, indicating that the flowers are semi-double flowers (8 to 40 petals) [23]. However, at a lower temperature of 20/10°C, the petal number significantly increased to 33, and at the lowest temperature to 49, turning the flowers from semi-double flowers into double flowers (i.e. >40 petals). Interestingly, the decrease in temperature also reduced the number of stamens as the average stamen number was 126 at the control tempe rature and decreased to 121 and 106 following 20/10°C and 15/5°C treatments, respectively. Generally, rose flowers have a few internal petals that have a slender base and an irregular shape, suggesting a homeotic conversion from a stamen into a petal ( Fig. 1g-i), and it is therefore noteworthy that the number of petal/stamen chimeras in low temperature treated flowers was higher than in the control flowers, thereby contributing to an increase in total petal number (Fig. 1g, h). These results indicated that there might be a gradual transition of petaloid stamens from outer whorls to inner whorls  even in normal cultivated rose flowers, but low temperature could enhance the formation of petaloid stamens (Fig. 1g, h). It is also worth noting that total number of floral organs is not significantly different among treatments (Table 1). Therefore, low temperature has not indeterminacy effect on flower development. Taken together, these results suggest that low temperature exerts a strong influence on petaloidy of rose stamens.

Expression analysis of RhAG gene in response to low temperatures during flower development
It has been reported that AG gene plays a dual role in specifying reproductive organ identity and floral meristem determinacy [9], as well as being involved in the determination of petal number [23]. To understand whether AG gene is associated with the low-temperature-regulated petal doubling in rose, we analyzed the effect of low temperature on the expression of RhAG, a rose homolog of AG [23], during rose flower development through quantitative RT-PCR. Results showed that the expression level of RhAG was low at stages 1 and 2, when the sepal and petal primordia form, before rising markedly at stages 3 and 4, when the stamen and carpel primordia form (Fig. 2). This expression pattern correlates with its classification as a C-class gene. Furthermore, low temperature exposure significantly decreased its expression level at the stamen and carpel formation stages, suggesting that RhAG may be involved in the low temperature induced homeotic conversion of stamens into petals.

Flower patterning of RhAG-silencing flowers phenocopies low temperature treated flowers
Since the expression of RhAG was reduced under low temperature conditions, we looked for evidence of causal relationship between RhAG function and the low temperature induced increase in petal number, by silencing RhAG in three month old rose plants using virusinduced gene silencing (VIGS). Floral phenotypes indicative of gene silencing were observed in fully opened infiltrated plants (Fig. 3a, b). At stage 10 the yellow anthers of the TRV-infected flowers were visible. In contrast, the anthers of the RhAG-silenced flower were not exposed as they were covered with the additional internal petals. We counted the number of petals and stamens of flowers from both groups and found that the petal number of the silenced flowers was substantially higher, while the stamen number decreased compared with the TRV controls ( Fig. 3c, d). TRV flowers had an average of 19 petals, while RhAG-silenced flowers had 25. Conversely, the average stamen number in RhAG-silenced flowers was 25 % lower than that of the TRV controls. Furthermore, compared to TRV-treated flowers, the RhAG-silenced flowers had more petaloid stamens (Fig. 3c), as was observed in the low temperature treated flowers. Interestingly, we found that partial gynoecia in RhAG-silencing flower were converted to sepal-like organs, indicating that RhAG was also involved in identity of gynoecia (Additional file 1: Figure S3). These results suggested that RhAG plays an important role in rose flower petal number control and homeotic conversion of stamens into petals.
Expression pattern of RhAG in floral primordia during low-temperature-responsive rose flower development To further confirm the proposed role of RhAG in the regulation of flower development, we examined its expression pattern in the floral primordia of low temperature treated flowers by in situ hybridization. As expected, irrespective of the ambient temperature regime or development stage, the expression of RhAG was undetectable in whorls 1 and 2 ( Fig. 4a-d and g-j), while a persistent signal was detected in whorls 3 and 4 once they emerged (Fig. 4c, d, i, and j), consistent with previous reports [23]. Interestingly, we observed that the expression pattern was different at stage 4 between lowtemperature treated buds and controls. In lowtemperature treated buds, which developed more petal primordia, the expression area of RhAG was restricted towards the center of the meristem, which might give rise to the fourth whorl, and extended slightly to the lateral area where only a few stamen primordia emerged (Fig. 4j). In control flowers, the RhAG signal extended to a wider domain of whorl 3, from which many stamen primordia emerged (Fig. 4d). The reduced expression . Asterisks indicate significant differences calculated using the t test (**p < 0.01; *p < 0.05) area supported the data obtained by quantitative RT-PCR and further suggested that the restricted expression of RhAG caused by the low temperature treatments might promote the petaloidy of stamens, resulting in the formation of double flowers.

Low temperate induces DNA hypermethylation in the RhAG promoter
Over the last decade, there has been growing evidence that DNA methylation is involved in plant responses to various environmental stimuli [31]. Previously, AG gene of A. thaliana was reported to be hypermethylated in an antisense-MET1 transgenic line [32].
We hypothesized that the observed reduction in RhAG expression at low temperatures might be related to a change in DNA methylation status. To test this we firstly cloned the genomic sequence and a 1,182 bp-DNA sequence upstream from the start codon of RhAG from rose flowers. The gene structure consists of eight exons and seven introns, with a total length of 4,942 bp (Additional file 1: Figure S4). Then the 1,182 bp-fragment upstream from RhAG was digested by the McrBC restriction enzyme that cleaves methylated DNA, no matter it is CG, CHG or CHH methylation [33]. Three regions were then chosen for PCR amplification: −1,149 to −826 bp (F1 + R1); −446 to −48 bp (F2 + R2); and +4493 to +5058 bp (F3 + R3) (Fig. 5a). Figure 5b shows that the F1 + R1 region of the RhAG promoter was heavily methylated when the rose plants were exposed to the low temperature condition 15/5°C, while the two other Fig. 3 Silencing of RhAG in rose flowers. Three month old rose plants were infiltrated with A.tumefaciens containing a TRV control (TRV, pTRV1 + pTRV2), or TRV carrying an RhAG fragment (TRV-RhAG, pTRV1 + pTRV2-RhAG). a Phenotype of the TRV control (left) and RhAG-silenced (right) flowers at stage 10. b Quantitative RT-PCR analysis of RhAG expression in TRV control and RhAG-silenced petals. The expression level of RhAG in the TRV19 control was set to 1.0. RhTCTP was used as the internal control. Floral buds at stage 3 were used. Values are means ± SD (n = 3). c Petals from TRV flowers (left) and RhAG-silenced flowers (right). d Petal (left) and stamen (right) numbers of TRV flowers and RhAG-silenced flowers (n = 5). Asterisks indicate significant differences calculated using the t test (**p < 0.01; *p < 0.05). Scale bars = 1 cm regions did not exhibit obvious differences in DNA methylation between growth at a normal temperature (25/15°C) and the low temperature (15/5°C). Since several digestion sites of HaeIII, a CHH (H represents A, T, or C) locus methylation-sensitive enzyme, were identified in the RhAG promoter, we further tested its methylation level by digestion with HaeIII, followed by PCR (Chop-PCR) [33]. We found that the low temperature condition led to CHH DNA hypermethylation in the tested F1 + R1 region, which was thus resistant to HaeIII cleavage (Fig. 5c).
We next examined the methylation status of the −1,149 to −826 bp region by bisulfite sequencing using sequencing primers designed with MethPrimer (http://www.urogene.org/cgi-bin/methprimer/methprimer.cgi) [34]. According to sequencing results, we found that methylation status was changed under low temperature in a 180 bp-inlength fragment ( −1,042 to −863 bp). In this region, almost all CG loci were methylated and the methylation level was nearly the same in plants grown under normal or low temperature conditions ( Fig. 6; Additional file 1: Figure S5A). All four CHG (H represents A, T, or C) loci were non-methylated under normal temperature conditions, while two CHG loci were methylated when plants were grown at low temperatures, although the methylation percentage was only~20 % (Fig. 6; Additional file 1: Figure S5B). As expected, the largest difference in DNA methylation was found in the CHH loci, which were almost all non-methylated at the normal temperature, while 20 out of 34 were hypermethylated at the low temperature. In addition, all CHH loci that showed a change in DNA methylation status were clustered ( Fig. 6; Additional file 1: Figure S5C). Given that DNA hypermethylation of promoters usually correlates with repression of gene expression [31], these results support the hypothesis that the low temperature induced reduction in the

Discussion
Effects of low temperature on petal doubling through stamen petaloidy Flowering is an essential phase of angiosperm reproduction and continuation of the species. The flowering process is controlled by various environmental factors, such as ambient temperature, which is known to influence the rate of flower development and flower quality. For example, low temperatures during the early stages of flower development can delay flower bud initiation and development in some cultivars of rose (Rosa hybrida), lily (Lilium hansonii), and chrysanthemum (Dendranthema morifolium) [35][36][37]. Temperature also regulates floral organ identity, resulting in homeotic transformation between different whorls and changes in organ numbers. In rose, high temperatures can cause the transformation of reproductive organs into leaf-like organs [27], and petal number has also been reported to be regulated by either high or low temperatures in several species. One example is from carnation, where Holley and Baker [38] reported that the petal number of some cultivars was reduced at high temperatures. This contrasts with the findings of Garrod and Harris [24], who suggested that a low temperature (5°C) promotes the formation of secondary growing centers, producing additional petals and hence a marked increase in total petal number. This latter result is consistent with studies of rose flowers, where it was found that the "bullhead" phenotype with increased number of petals and decreased number of stamens was induced by low temperatures and substantially decreased when the temperature was increased [28][29][30]39].
In this study, we demonstrated that flowers of rose R. hybrid cv. Vendela exposed to low temperatures have a greater total number of petals and a larger flower bud than flowers grown at normal growth temperatures, in agreement with a previous study [39]. Moreover, we found that the increase in the number of petals was accompanied by a decrease in stamen number, while the total number of both petals and stamens was similar between the flowers from plants grown at different temperatures. We also observed that more petaloid stamens were formed in the low temperature treated flowers. One explanation for these phenotypes is that low temperatures cause increase of petal numbers, at least in part through stamen petaloidy. In addition, we observed that some plants had sepal-like organs in the center of the flower at low temperatures, suggesting a homeotic transformation of carpels to sepals, which is consistent with previous reports in A. thaliana [10].
Interestingly, a previous report showed that in Rosa hybrida cv. Motrea, high temperature regime (26/21°C, day/night) elevated the proportion of flowers exhibiting phyllody phenotype to four times higher than low temperature (21/15°C). The petal number of phyllodyphenotype flowers was higher than normal flowers,   [58] under either 26/21°C or 21/15°C condition, and the increase of petal number was positively correlated with the extent of phyllody phenotype, more severe phenotype with more petals [27,40]. Reduced cytokinin content was considered to be responsible to the phyllody phenotype [27]. These reports implied that petal number might be controlled by several pathways and more studies should be conducted to clarify it in the future.

Involvement of RhAG in petal number increasing at low temperatures
The floral homeotic C-class gene AG has a dual role in regulating floral meristem determinacy and reproductive organ identity, which is strongly conserved even in distantly related angiosperms, including the model plants A. thaliana, Antirrhinum majus, and the ranunculid T. thalictroides [9,41,42]. These studies suggest that plants with reduced AG function convert reproductive organs into perianth organs and develop indeterminacy of the floral meristem, causing double-flowers with excess petals. In rose, expression of RhAG was reported to be associated with selection of cultivars with higher petal numbers during domestication in Europe/Middle East and in China [23]. Among roses with similar genetic backgrounds, the expression domain of RhAG is restricted toward the center of the flower and is clearly narrower in double-flowered roses than in simple flowered cultivars. Moreover, this border of the RhAG expression domain is labile, allowing the selection of rose flowers with increased petal number [23].
Based on the above studies, we hypothesized that the rose flower with extra-petals-phenotype induced by low temperatures is related to AG expression, and so we examined the expression of RhAG and other homeotic genes in rose plants grown at low temperatures. Expression analysis showed that, as expected, the expression of RhAG was significantly down-regulated by low temperature treatments at the stamen and carpel formation stages. To obtain additional genetic evidence, we suppressed the expression of RhAG in rose plants using VIGS. This resulted in additional petal formation in the flower buds, a reduced number of stamens and an increased number of petaloid stamens. We noticed that the average petal number in the TRV control flowers in VIGS experiment (Please see Fig. 3d) was substantially lower than in the control flowers in low temperature treatment experiment (Please see Table 1). This phenomenon is likely due to the variability of flowers during different seasons, batches, ages of seedlings as well as viral effects, since it has previously been reported in ranunculid species that TRV2-empty plants showed asymmetric reduction in sepal size, occasional brown, necrotic spots on sepal, as well as stunted growth. However, the stunted growth did not affect subsequent growth [9,43]. Whether such virus-induced effects existed in the TRV2-empty rose plants in this current study is unclear. In addition, the total number of petal plus stamen was less in RhAG-silenced flowers than in TRV controls, while it was not influenced in the low temperature test. This could be an environmental effect due to different growth conditions, such as seasons. Regardless, the difference in petal and stamen numbers between RhAG-silenced flowers and TRV controls indicates that low temperature can regulate petal numbers at least in part by a homeotic conversion of stamens into petals, via suppressing the expression of RhAG. Finally, we also found sepaloid organs in the center of RhAG-silenced flower, where carpels normally form. This phenotype was not observed in the TRV control flowers and was similar to the phenotype of low-temperature treated flowers, further supporting the proposed role of AG in determining reproductive organ identity and the reliability of the VIGS experiments.
We also conducted an in situ hybridization study of rose buds to determine the expression pattern of RhAG in response to growth at low temperatures, and observed that under such conditions RhAG was expressed in a reduced area toward the center of the flower. The change of the spatial pattern of expression of RhAG in response to low temperatures might promote the formation of double-flowers, as has been proposed as a mechanism underlying the selection of double-flowers during rose domestication [23].
Since determination of floral organ identity requires the involvement of multiple floral homeotic genes and other regulatory factors during floral development, it is possible that low temperature also influence the expression of other homeotic genes, which might contribute to double-flower formation. The identity of other genes involved in this process, especially those related to petal and stamen primordia formation, is an interesting subject for future research.

Involvement of epigenetic DNA methylation in RhAG regulation
It has been well documented that epigenetic DNA methylation is involved in many aspect of plant development, as well as in responses to endogenous and exogenous cues [31,[44][45][46]. In A. thaliana, DNA methylation could occur in the contexts of CG, CHG, and CHH (H = A, C, or T) in plants. Once established, CG and CHG methylation is maintained by MET1 and CMT3, respectively, whereas CHH methylation needs to be established de novo by DRM2 and CMT2 during every cell cycle [47,48]. For gene expression, the most important factor is the region where DNA methylation occurs instead of DNA context. If DNA methylation occurs in the promoter region of a certain gene, no matter it is CG, CHG and CHH methylation or both, it will cause gene silencing. However, DNA hypermethylation in gene body usually represent a feature of transcribed genes [47,[49][50][51]. Here, we report that DNA methylation of the promoter of rose RhAG, a homolog of A. thaliana AG, is regulated by ambient temperatures. McrBC and HaeIII-digestion indicated that growth at low temperatures (15/5°C) resulted in heavier methylation in the RhAG promoter when compared to the plants grown at normal temperatures (25/15°C). Bisulfite sequencing further indicated that CHH DNA methylation was highly induced by low temperatures. Interestingly, only~50 % of the CHH loci in the tested region were methylated, suggesting that low temperature induced CHH methylation is site-specific. Strikingly, AG gene could be hypermethylated in antisense-MET1 transgenic A. thaliana lines. In addition, most methylated sites in tested regions were CHH, while MET1 is considered to be responsible for CG loci methylation [32]. Thus, epigenetic regulation of AG gene might be a conserved pathway. Generally, DNA methylation near gene promoters is considered to correlate with repression of gene expression [47,[49][50][51]. In tobacco, cold stress activated gene expression of NtGPDL via an induction in DNA demethylation in the coding region [52]. Thus, our results suggest that ambient temperature triggers complex changes in the DNA methylation status of RhAG to regulate its expression level in a conserved manner. The nature and dynamics of these modifications, as well as the identification of possible regulators, will be the subject of future studies.

Conclusions
In present work, we found that RhAG, an AG homolog in rose, regulated petal number in an ambient temperaturedependent manner. This is based on several lines of evidence. First, low temperature treatment significantly increases petal number in rose through the promotion of stamen petaloidy. Second, quantitative RT-PCR analysis revealed that the expression pattern of RhAG is associated with low temperature regulated flower development, and the silencing of RhAG caused an increase in petal number through an increased production of petaloid stamens. Third, in situ hybridization studies showed the overall spatial distribution of RhAG transcript in the floral bud was clearly decreased under low temperature conditions. Fourth, analysis of DNA methylation level showed that low temperature treatment enhances the methylation level of the RhAG promoter, and a specific promoter region that was hypermethylated at CHH loci under low temperature conditions was identified by bisulfite sequencing. In summary, our results provided new insights into the underlying mechanism of ambient temperatureregulated flower patterning. And we demonstrated that low temperature probably attenuated RhAG expression at least partially via enhancing DNA CHH hypermethylation of the RhAG promoter.

Plant materials
Rose (Rosa hybrida) cv. Vendela plants were grown in a greenhouse at the Shenzhen Polytechnic. Flower development in roses includes early development stages [23] and opening stages [53]. We divided the whole process of flower development into

Low temperature treatment of flowers
Two-year old rose plants were pruned uniformly before treatment. All plants were then cultivated in three controlled environment chambers (Thermoline TPG-6000-TH) with day/night temperature regimes of 25/15°C (control), 20/10°C and 15/5°C. For each treatment, around 30 plants were used. All the three treatments were under a 12/12 h light/dark cycle, 70-80 % relative humidity and 850 μmol m −2 s −1 light intensity. For the plants subjected to 20/10°C and 15/5°C, once the flower bud differentiation was complete the chamber temperature was changed to the normal growth temperature of 25/15°C.

Observation of flower phenotypes and floral organ counts
Flower phenotypes were observed after flower development stage 6 and floral organs were counted at stage 7. Fifty flowers from each treatment were selected randomly (only one or two flowers were used from each plant), and the proportion of deformed rose flowers was recorded. Fifteen flowers from each treatment were dissected and 4 whorls of organs (sepals, petals, stamens and carpels) were counted. And petal/stamen chimera was counted as petal. The statistical significance was analyzed using a one-way ANOVA or Student's t test.

RNA extraction
For quantitative RT-PCR analysis, floral buds of plants subjected to a cold treatment (15/5°C) or grown in control conditions (25/15°C) were sampled at stages 1, 2, 3 and 4. Whole buds were harvested and frozen in liquid nitrogen before storage at −80°C. Total RNA was extracted using the Trizol reagent (Life Technologics, Ohio, USA) according to the manufacturer's instructions.

Quantitative RT-PCR
For quantitative reverse transcriptase polymerase chain reaction (quantitative RT-PCR), complementary DNAs (cDNAs) were synthesized from 2 μg total RNA using the SuperScript™ II RNase H reverse transcriptase (Invitrogen). One μl cDNA samples were used as the template in a 20 μl PCR reaction using the ABI PRISM® 7500 Sequence Detection System in standard mode with the Platinum SYBR Green qPCR Supermix-UDG (Invitrogen). The primers used are listed in Additional file 1: Table S1. RhTCTP was used as an internal control. The expression levels were analyzed using the relative standard curve method normalized to RhTCTP. All reactions were performed for three biological replicates.

RhAG promoter isolation
For isolation of 5' upstream sequences, including promoter and 5' UTR of RhAG, hi-TAIL PCR was performed as in Liu and Chen [54].

RhAG silencing in rose flowers
Since the efficiency of original TRV vector was relative low in rose plants, we conducted the VIGS experiments by using a modified and GFP-labeled TRV-GFP system [55]. The silencing of RhAG in rose flowers by VIGS was performed as previously described with some modifications [55,56]. For silencing of RhAG, a 332-bp fragment at the 3' UTR was chosen and inserted into the pTRV2 vector. For vacuum infiltration, three-month-old rose plants were immersed in the bacterial suspension solution and infiltrated under vacuum at 0.9 atm. Then plants were washed and cultivated in MS liquid medium at 23/18°C. After the plants had rooted they were transplanted into pots and grown in controlled environment chambers at 25/15°C, 70-80 % relative humidity and 850 μmol m −2 s −1 light intensity, under a 12/12 h light/ dark cycle. When floral buds emerged, stage 3 buds were pre-screened by detecting GFP fluorescence, and the positive ones were sampled for RNA extraction. The expression level of RhAG was determined by quantitative RT-PCR and floral phenotypes of the selected plants were also observed.

In Situ hybridization
Tissue fixation and in situ hybridization were performed as in Zhang et al. [57], Ma et al. [56] and Dubois et al. [23]. To synthesize the in situ probes, PCR amplification of cDNA was conducted using gene-specific primers containing the T7 and SP6 RNA polymerase binding sites [56]. T7 RNA polymerase was used to generate the antisense probes and SP6 RNA polymerase for sense probes. The primer pairs are listed in Additional file 1: Table S1.
Chop-PCR and bisulfite sequencing PCR DNA methylation status was analyzed by Chop-PCR and bisulfite sequencing [33]. For Chop-PCR, genomic DNA (500 ng) was digested with the methylation-sensitive restriction enzyme HaeIII or the methylated DNAdigesting enzyme McrBC for 3 h. The digested DNA was used as a template to amplify the RhAG promoter and undigested genomic DNA was amplified as a control. For bisulfite sequencing, the primers were designed by MethPrimer (http://www.urogene.org/cgi-bin/methprimer/ methprimer.cgi) [34]. Genomic DNA was isolated from a pool of ten flower buds grown under normal temperature (25/15°C) and low temperature (15/5°C) conditions, before being subjected to bisulfite treatment using an EZ DNA Methylation-Gold kit (Zymo Research, Orange, CA, USA) according to the manufacturer's instructions. For each treatment, 17 clones were chosen randomly and sequenced. The sequences were aligned using the ClustalW server (http://www.ch.embnet.org/software/ClustalW.html) and then analyzed using the CyMATE program (http:// cymate.org/cymate.html) [58].

Availability of supporting data
The complete coding sequence and the promoter of RhAG is available at NCBI under accession number U43372 and BankIt1849645 Seq1 KT429820, respectively.

Additional file
Additional file 1: Figure S1. Effect of low temperature on flower diameter (mm) at different flower developmental stages. Data represents means from 5 replicates ± SD. RT, regular temperature (25/15°C; day/night); LT, low temperature (15/5°C; day/night). Figure S2. The proportion of deformed rose flowers from plants grown under different temperature conditions. Two year old plants were pruned uniformly and exposed to day/night temperature regimes of 25/15°C (control, left), 20/10°C (middle) or 15/5°C (right). For plants grown under low temperature conditions, once bud differentiation was complete, the temperature was returned to the normal growth temperature. When flowers were at stage 7, the proportion of deformed rose flowers was recorded (n = 50). Figure S3. Silencing of RhAG gene caused conversion of gynoecia to sepal-like organs. Top panel, The RhAG-silenced and TRV control flowers without petals. The sepal-like gynoecia and normal gynoecia are indicated. Middle and Bottom panels, the sepal-like gynoecia from RhAG-silenced flowers. Scale bar, 10 mm in Top panel; 10 mm (left) and 5 mm (right) in Middle panel; 10 mm in Bottom panel. Figure S4. Schematic view of the RhAG gene structure. Lines represent introns, black boxes represent exons. The first base pair of start codon was defined as 0.The numbers indicate the 3' ends of the DNA fragment. Bp, base pair. Figure S5. Percentage of CG, CHG, and CHH methylation in the RhAG promoter. Percentage of CG (A), CHG (B), and CHH (C) methylation was determined by bisulphite sequencing analysis of the RhAG promoter, using genomic DNA extracted from flower buds grown at the indicated temperature day/night temperatures. Seventeen clones were analyzed for each treatment. Table S1. Oligonucleotide primer sequences. (PDF 645 kb)