Auxin controls circadian flower opening and closure in the waterlily

Background Flowers open at sunrise and close at sunset, establishing a circadian floral movement rhythm to facilitate pollination as part of reproduction. By the coordination of endogenous factors and environmental stimuli, such as circadian clock, photoperiod, light and temperature, an appropriate floral movement rhythm has been established; however, the underlying mechanisms remain unclear. Results In our study, we use waterlily as a model which represents an early-diverging grade of flowering plants, and we aim to reveal the general mechanism of flower actions. We found that the intermediate segment of petal cells of waterlily are highly flexible, followed by a circadian cell expansion upon photoperiod stimuli. Auxin causes constitutively flower opening while auxin inhibitor suppresses opening event. Subsequent transcriptome profiles generated from waterlily’s intermediate segment of petals at different day-time points showed that auxin is a crucial phytohormone required for floral movement rhythm via the coordination of YUCCA-controlled auxin synthesis, GH3-mediated auxin homeostasis, PIN and ABCB-dependent auxin efflux as well as TIR/AFB-AUX/IAA- and SAUR-triggered auxin signaling. Genes involved in cell wall organization were downstream of auxin events, resulting in the output phenotypes of rapid cell expansion during flower opening and cell shrinkage at flower closure stage. Conclusions Collectively, our data demonstrate a central regulatory role of auxin in floral movement rhythm and provide a global understanding of flower action in waterlily, which could be a conserved feature of angiosperms. Electronic supplementary material The online version of this article (10.1186/s12870-018-1357-7) contains supplementary material, which is available to authorized users.


Background
In flowering plants, many flowers open during the day and close at night, exhibiting a 24-h circadian rhythm, which is known as Linné's floral clock [1]. This circadian movement pattern serves as an environmental adaptation characteristic for creating an opportunity for pollination [1]; in Ipomoea nil, the flower-opening time is determined by the length of dark period [2]. However, without a light-dark switch period, flowers continue to display a rhythm in opening and closing when in a complete light or dark environment using their "internal clock", i.e. circadian clock [3]. Yon et al. found that silencing the core circadian clock genes Late Elongated Hypocotyl (LHY) and ZEITLUPE (ZTL) in Nicotiana attenuate strongly altered the floral rhythms [4]. Therefore, the internal circadian clock can recognize a 24-h rhythmic oscillation, establishing an endogenously sensing system to adapt to seasonal environmental cycles. Due to the daily changes in solar direction and irradiation, light becomes the key connection between the circadian clock and other environmental signaling pathways. So far, sophisticated light responsive networks have been intensively established in model plant developmental processes (summarized in [5]), and the output models are conceptually useful for explaining the connections between light and circadian rhythm in flowers. For example, sunflowers track the sun from east to west, reorienting during the night so that their leaves and apices face east before sunrise [6]. We term this phenomenon as phototropism, which is a typical plant response to light identified during floral rhythms [7]. Phototropism has been well characterized as being controlled by an asymmetric auxin distribution [8]. Correspondingly, application of the auxins indole acetic acid (IAA) and naphthaleneacetic acid (NAA) strongly promotes the opening of iris flowers [9,10]. Thus, auxin is also involved in the regulation of circadian floral movement.
Auxin, one of the major plant hormones, is essential to regulate diverse processes of plant development by coordination of cell expansion and cell division, such as embryogenesis, tropic response, shoot branching, vascular development, apical dominance, flowering and fruit ripening [11,12]. The effects of auxin on plant development have been largely attributed to the transcriptional and translational regulations governed by the tryptophan aminotransferase (TAA) and flavin monooxygenase (YUC)-dependent synthesis pathway, Transport Inhibitor Response1/ Auxin-Related F-BOX (TIR1/AFB)-Aux/IAA-regulated signaling pathway and PIN-FORMED (PIN)-mediated transport pathway [13][14][15][16][17][18]. Cell wall biogenesis has been implicated as a core event downstream of auxin regulations that determines anisotropic cell expansion during plant development including floral circadian movement. In Ipomoea nil, transcripts of xyloglucan endotransglucosylase/ hydrolases (XTHs) encoding the genes InXTH1-InXTH4 in petals, which are involved in cell wall modification, were closely correlated with the rate of flower opening, further controlling plastic petal growth [2]. In the 'Mitchell' petunia, knockdown of cell-wall-associated β-galactosidases, which determine the galactose level among cell wall polysaccharides, severely reduces the flower-opening angle due to the disruption of petal integrity [19]. Hence, a combination of transcriptional, translational and post-translational regulations leads to coordination among various processes including the circadian clock, hormone biosynthesis, light signaling, and cell wall modification to establish a proper opening-closing period in flowers. Although the phenomenon of floral movements have been broadly described during the past 20 years' research [9,20], the underlying regulatory mechanisms are largely elusive.
Flower is a characteristic feature of the angiosperms, and floral rhythm has long been discussed regarding the evolutionary origin of flower and subsequent diversification. Nymphaeales (also called waterlily), one of the most ancient angiosperm lineages, represents an early-diverging grade of flowering plants [21]. Nymphaeales consists of five genera with approximately 60 species [22,23], and most waterlily species have flowers that open during the day and close at night. Thus, owing to the special phylogenetic position of Nymphaeales, evaluating flower actions in Nymphaeales is highly useful to reconstruct the molecular characteristics of angiosperms.
In this study, Nymphaeales flowers were used as a model to study the rhythm of floral movement. We observed that the intermediate segment of petal cells of the Nymphaeales flower are severely flexible and respond to the photoperiod. Exogenous application of auxin pronouncedly blocks flower closure; whereas, treatment of auxin inhibitor suppresses flower opening. Transcriptome analysis generated from the intermediate segment of petals at different day-time points provide a global understanding that auxin metabolism, signaling and transport coordinate the movement rhythm formation of flowers. In addition, cell wall modification which acts as a downstream cascade of auxin is synchronously oscillating with auxin events, resulting in circadian cell expansion to accomplish floral opening and closing. This study offers an integrated view for understanding floral circadian rhythms and provides useful information for studying the traits of flowering plants across angiosperms.

Results
The petal is responsible for sensing the circadian signal for movement Plants keep track of the day-night photoperiod, ensuring reproductive success. To study this typical floral rhythm process and understand the mechanism by which the flower opening period is initiated, we adopted flowers from one of the most ancient angiosperm lineages, Nymphaeales (henceforth called waterlily), as the study model system. In the late summer of 2016 (photoperiod: 6:00-18:00 gradual light/18:00-6:00 gradual darkness; temperature: 25-32°C), we tracked the flower movement of Nymphaea colorata via time-lapse photography for every 0.5 h. As shown in Additional file 1: Figure S1A, the flower remained closed at 7:30 and rapidly opened 0.5 h later (8:00); at 9:00, the opened angle of the flower reached the maximum value; and after 16:30, the flowers started closing continuously. Thus, the flowers of Nymphaea colorata showed a natural opening-closure pattern. Due to the shortage of Nymphaea colorata samples, we purchased the fresh-cut flowers of the waterlily cultivar Nymphaea nouchali as experimental materials. To test if the cut flowers exhibit a similar circadian movement pattern as natural flowers, the cut flowers were cultivated and tracked in a growth chamber at a controlled temperature at 25°C and a 16-h light/8-h dark photoperiod (7:00-23:00 light/23:00-7:00 dark). Accompanied with the light exposure at 7:00, the flowers gradually opened. At 10:00, the opening angle rapidly increased until 14:00, when the opened angle reached the maximum of 75°. After 14:00, the flowers gradually closed, and at 23:00, the flowers were almost completely closed (Fig. 1a). Thus, with the manual manipulation of growth conditions, the cut flowers also displayed a stable opening-closure rhythm movement pattern, as shown by the key time points: 7:00 (initiation), 10:00 (rapid opening), 14:00 (maximal opening) and 18:00 (closure).
To explore the most sensitive part of flower that is responsive for period regulation, we first cut off the sepal of the flowers and tested the responsiveness between flowers with and without sepals. Without the sepals, the flowers continued to display an opening-closure pattern following period of (7:00-10:00-14:00-18:00) identical to that of the flowers with sepals. Moreover, without the sepals, the flowers exhibited a greater opening angle (maximum angle = 130°), suggesting that the sepals might act as a shield to restrict petal movement while protecting the internal flower organs (Fig. 1b, e). The sepals did not influence the floral rhythm; thus, we removed the sepals for the following experiments, owing to the convenience of study on internal flower organs.
To further evaluate the contribution of petals, the petals were divided into three sections: the top segment (> 1 cm distance to the bottom end), the intermediate segment (0.5-1 cm distance to the bottom end) and the bottom segment (< 0.5 cm to the bottom end). Petals without the top sections continued to response to the photoperiod and displayed a floral movement pattern similar to that of a complete flower (Fig. 1c, e), whereas the petals could not expand properly when only the bottom section remained (Fig. 1e, Additional file 1: Figure   S1B, according to measurement standard of Additional file 1: Figure S1C). Meanwhile, the stamens exhibited an irreversible opening action (Fig. 1c, Additional file 1: Figure S1B). These observations suggest that the intermediate segment of petals is very sensitive to receive environmental signals and reacts. Furthermore, we even removed the stamens and left only 1 cm of the bottom segment of petals to verify if the stamens provide a thrust to push petal opening. However, without the stamens, the basal 1 cm segment of the petal was still able to respond to the photoperiod and display the 7:00-10:00-14:00-18:00 movement pattern (Fig. 1d, f ). Thus, the physiological data indicated that the petal is the most important organ for receiving circadian signals and displays plastic curvature, and the 1 cm segment of the basal petal that connects to the sepal and the receptacle is primarily responsibility for floral movement. Intermediate petal determines flower opening and closure movement rhythm. a Floral opening and closure rhythm in Nymphaea nouchali cut flowers was tracked from 7:00 to 23:00. b Floral opening-closure rhythm without sepal. c Without sepal, the floral opening and closure rhythm was tracked when only 1 cm of the petal was left. White arrows highlighted the movement rhythm of the petals. d Without the sepal and stamen, the floral opening and closure rhythm was tracked when only 1 cm of the petal was left. White arrows highlighted the movement rhythm of the petals. e-f Floral opening-closure angles were measured according to the above pictures. Blue rectangular boxes below the curves highlight the maximum opening period. Scale bar, 25 mm (a-d). Error bar = SEM (e, f)

Cell morphogenesis of the intermediate petal is circadiandependent
To determine whether the intermediate segment of petal is responsive for circadian rhythm, we tracked longitudinal sections of waterlily flowers at different time points (7:00, 10:00 and 14:00 were the flower-opening points, 18:00 and 21:00 were the flower-closure points). As shown in Fig. 2a, the intermediate segments of all four petal layers responded to the photoperiod, exhibiting gradually expanded curvatures. Particularly, the intermediate segment showed a strikingly plastic change accompanied with the circadian petal movement (Fig. 2a). To better visualize the cellular change, we used Scanning Electronic Microscopy (SEM) to track the adaxial and abaxial epidermis of the intermediate petal. In both the adaxial and abaxial epidermis, long and narrow cells quickly expanded or shrank along the horizontal direction at different time points. As we observed, from 7:00 to 14:00, the cells were continuously swelling, reaching a maximum cell width of~22 μm; later, from 14:00 till 21:00, the expanded cells shrank back to 19 μm ( Fig.  2b-c). Obviously, the amplitude of cell length alteration in adaxial side was higher than abaxial side. Moreover, flower opening (at 10:00) elevated adaxial cell width whereas did not influence abaxial cells, implying that adaxial cells were more flexible than abaxial ones. Hence, the cellular change of intermediate segment of petal was coordinated by the alteration of cell width and cell length (Fig. 2c). We thereby calculated a ratio to compare the cell width with its individual cell length; the abaxial cells apparently exhibited a circadian oscillation of width/length ratio, as shown by the highest at 10:00 and 14:00, when the flowers rapidly opened, and the lowest at 18:00 and 21:00, when the flowers were closing (Fig.  2c). This oscillation pattern of cellular morphogenesis is tightly correlated with petal curvature. Therefore, the rapid remodeling of cells confirmed the above speculation that the intermediate petal cells receive and respond to circadian movement signals, subsequently bending via anisotropic cell expansion.

Flower opening-closure pattern is influenced by exogenous auxin
The plant hormone auxin is well-known to stimulate cell expansion and division via the promotion of cell wall flexibility and extensibility, which depends on the cell wall modification [24,25]. Cell expansion at flower opening stage while shrinkage at flower closure stage are tightly correlated with the dynamic property regulated by auxin. It intrigues us to further understand if auxin is required for floral rhythm movement. To verify our speculation, we sprayed the synthetic auxin analog NAA and the auxin transport inhibitor 2,3,5-triiodobenzoic acid (TIBA) [26] on the intermediate petals before the phototracking. NAA significantly enhanced the rate and angle of flower opening (Fig. 3). Moreover, compared with the DMSO-control samples, whose closed angle was approximately 30°, the flowers could not be completely closed after NAA application, as shown by 70°closure angles at 20:00 and 80°at 23:00 (Fig. 3). It appears that NAA triggered the second rhythm of flower movement after 20:00. In contrast, TIBA suppressed flower opening. The maximum opening angle was significantly smaller than that in control, and the opening schedule was delayed for 1 h following TIBA treatment (Fig. 3). Thus, consistent with the role of auxin in the opening of the iris flower [10], our finding further unveils that auxin participates in the floral rhythm of opening and closure.
Flower opening-closure pattern is coordinated by endogenous auxin metabolism In order to further understand the mechanism of auxin regulation and global transcription-based regulatory network occurred during the flower opening-closure processes, we performed a de-novo RNA-seq analysis based on the above described key time points of intermediate segment of petals. Samples at 6:00 (T1) were used as the control when the flowers were still under the dark period; and then, petal samples were individually collected at 7:00 (T2), 10:00 (T3), 14:00 (T4) and 18:00 (T5) (Fig. 4a). The differential transcriptome profiles were analyzed based on the standard of q-value< 0.05, fold change> 1.5, and FPKM (expected number of Fragments Per Kilobase of transcript sequence per Millions base pairs sequenced) ≥ 1 by comparison of different couples (such as T2 vs. T1). For annotation, in total 33,333 waterlily unigenes were searched and identified against the NCBI non-redundant (Nr) database using BlastX. According to GO enrichment and KEGG pathway analysis, in the T2 vs. T1 comparison group when the flowers started to receive the light signal, most of the downregulated genes (in total 422) were involved in oxidative stress and cell wall modification, and most of the upregulated genes participated in photosynthesis or lipid biosynthesis pathways (in total 166) (Additional file 2: Figure S2). At T3, the genes related to cell wall biogenesis started to be upregulated, corresponding to rapid flower opening action. At T4 when the flower opened maximally, the expression level of photosynthesisand lipid-related genes decreased. When the flower was closing at T5, peroxidase-related genes were activated (Additional file 2: Figure S2). Thus, cells from the intermediate area of petals undergo a series of cellular changes in conjunction with the transcriptional regulation of photosynthesis, cell wall reorganization, oxidation, and lipid biosynthesis, among others.
To further understand the biological correlation during floral movement rhythm, we clustered all differential expressed genes into 8 clusters based on their expression pattern ( Fig. 4b-i). At the onset of light stimulation at 7:00, the responsive genes of cluster 1 were immediately upregulated compared with their expression at 6:00 (Additional file 3: Figure S3A). According to a false discovery rate (FDR) < 0.05, 70 GO items of cluster 1 were found, and these items mainly belonged to photosynthesis and transmembrane transporters. Photosynthesis is a fundamental process used by plants to convert light energy to sugar and organic compounds [27]; therefore, photosynthesis I/II reaction center subunits and chlorophyll a-b binding proteins were greatly stimulated upon light stimulation (Additional file 3: Figure S3B, Additional file 4 Table S1).
Based on the physiological data, opening of flower is regulated by exogenous auxin or auxin inhibitor treatment (Fig. 3), we try to understand if endogenous auxin level fine-tunes the regulation of floral movement rhythm. Indole-3-acetic acid (IAA) is the major natural auxin in plants and primarily synthesized from the substrate, Tryptophan (Trp). Trp is firstly converted to indole-3-pyruvate (IPA) by the Tryptophan aminotransferase (TAA) enzymes, subsequently IAA is produced from IPA by flavin monooxygenase (YUC) proteins [16,28,29]. Cytochrome P450 enzymes such as CYP72B2, CYP79B3 and CYP83B1 are also involved in auxin synthesis [15,16]. Using annotated genes as query input, we tracked the DEGs of auxin metabolism events which were mainly present in cluster 3 ( Fig. 5a-b). In the transcriptome profile, YUC homologues of YUC8, YUC9 and YUC10 were pronouncedly upregulated at the rapid opening stage of flowers and YUC5 was activated at flower closure stage (Table 1, Fig. 5b). Similar as the oscillation of YUCs, CYP83B1 was also induced at rapid opening stage. However, TAA family showed a different expression pattern, as shown by upregulation of TAA3 but downregulation of TAA2 during flower opening (Table 1, Fig. 5b). Nevertheless, most of auxin synthesis-related genes are activated at flower opening stage, indicating that floral movement rhythm requires the participation of new synthesized auxin (Fig. 5b).
Free IAA comprises no more than 25% of the total amount of IAA, depending on the plant tissue and species. Thus, in addition to free auxin, IAA is present in a variety of modified forms, including ester-linked indole-3-acetic acid (IAA)-sugar conjugates and amide-linked IAA-amino acid conjugates [30]. Free IAA is biologically active in diverse plant developmental processes; however, most IAA conjugates are inactive, and they serve as a reservoir to maintain appropriate balance of free IAA and conjugated IAA level [30]. The conversion of IAA to IAA-amino acids is determined by the Gretchen Hagen3 (GH3) family of auxin-inducible acyl amido synthetases [31]. In contract, IAA-Leu-resistant1/ILR1-like (ILR1/ILL) encodes an amidohydrolase that cleaves IAA-amino acid conjugates to release active IAA [32]. The transcriptome profile showed that ILL5 and ILL6 were downregulated along flower opening, whereas 11 out of 16 GH3 homologues were particularly upregulated at flower opening stage (Table 1, Fig. 5b), suggesting that flower opening action triggers the frequent conversion of free IAA to conjugated IAA, which serves a storage of auxin for the further reuse. Hence, the influence of auxin synthesis and auxin homeostasis together demonstrated that flower opening triggers the transcriptional regulation of endogenous auxin metabolism, which coordinates the opening-closure rhythm formation of flower.

Auxin flow toward the cell wall determines the plastic cell expansion
Auxin is synthesized in the tissues with active cell division and expansion, and is directionally transported via a cell-to-cell manner to the entire plant body [18]. The transportation of auxin is controlled by auxin influx and efflux carriers. PINFORMEDs (PINs) are the well-known auxin efflux transporters, which are constitutively cycling and recycling between the plasma membrane (PM) and endosomal compartments to control the polar auxin flow out of the cell [17]. Expression of PIN1 homologues was relatively high at the flower opening stage while decreased when the flowers were closing (Fig. 5c, Table 2), suggesting that PIN-mediated auxin efflux is required for the rapid opening action. PP2A phosphatase is an important regulator for polar-targeting of PINs via dephosphorylation [33]. One PP2A homologue was downregulated while two PP2A homologues were upregulated during flower opening (Fig. 5c, Table 2), indicating that a fine-tune regulation of PIN phosphorylation occurs along the opening action. In addition, subcellular  dynamics of PINs is coordinated by the endosomal regulators of guanine-nucleotide exchange factors for ADPribosylation factor GTPases (ARF-GEF) and Brefeldin A-inhibited guanine nucleotide-exchange protein (BIG) which mediate vesicle budding process [34][35][36], as well as Sorting Nexin1 (SNX1) and Vacuolar protein sorting proteins (VPS) which control vacuolar sorting process [37][38][39]. Expression of BIGs was stimulated associated with opening action, whereas SNX1 and VPSs were downregulated at maximal opening stage (Fig. 5c, Table 2), implying that the opening of flower triggers frequent endocytosis events but suppresses degradation of PINs. Altogether, compared with flower closure stage, opening action induces PIN cycling between PM and endosomal compartments, thereby resulting in more auxin flow toward the cell wall.
P-glycoprotein (PGP/ABCB) proteins are defined as another type of auxin efflux transporters, which interact with PINs and maintain the stability of PINs at the PM [40]. Ten ABCB homologues were found in DEGs, and 6 of them were upregulated at rapid opening stage (Fig. 5c, Table 2), suggesting that ABCBs might assist PINs to maintain high auxin flowing out of the cell during rapid flower opening. Due to the sequence similarity with PIN proteins, PIN-LIKES (PILS) proteins are defined as the putative auxin carriers which regulate intracellular auxin  Tables 1, 2 and 3 according to the labelled number "1" to "21" in each cluster) (modified from [89]) accumulation and homeostasis at the endoplasmic reticulum (ER), subsequently affecting nuclear auxin signaling [41]. PILS2 and PILS6 were suppressed at maximal opening of flower, while PILS3 and PILS7 were promoted along flower opening (Fig. 5c, Table 2), indicating that floral movement stimulates a feedback auxin flow among the intercellular compartments.
Among all auxin transporters, the PIN and ABCB proteins are the major auxin efflux carriers, whereas AUXIN1/LIKE-AUX1 (AUX/LAX) proteins represent the major auxin influx carriers, which are responsive for auxin uptake within the cells [42]. Interestingly, 5 out of 7 AUX1/LAX homologue was upregulated at 14:00 and 18:00 of flower closing stage (Fig. 5c, Table 2), implying that auxin uptake is required for closure action. It is apparent that flower opening promotes auxin efflux while flower closing enhanced auxin influx. The above transcriptome data supported a scenario that flower opening action pumps more auxin to the cell wall, presumably leading to the further cell wall remodeling; while flower closing action keeps auxin stay within the cell, ready for the second round of floral movement.
Auxin signaling is highly active throughout the entire floral movement rhythm Auxin is extraordinarily multifunctional, with different cells responding very differently to changes in auxin levels, which is required for the coordination of auxin signaling at transcriptional level [43]. The core components of the nuclear auxin signaling machinery consist of three protein families: the F-box Transport Inhibitor Response 1/Auxin Signaling F-box Protein (TIR1/AFB) auxin co-receptors, the Auxin/Indole-3-Acetic Acid (Aux/IAA) transcriptional repressors, and the Auxin Response Factor (ARF) transcription factors [44]. Briefly, the small molecule auxin targets to its nuclear receptor Transport Inhibitor Response 1/Auxin Signaling F-BOX (TIR1/AFB) proteins, bringing TIR1/AFBs and transcriptional repressors Aux/IAAs together and subsequently leading to the ubiquitination and degradation of this complex. In the transcriptome profile, expression of TIR1 and AFBs was gradually upregulated associated with floral movement rhythm, and reached to the maximal level at 18:00 (Fig. 5d, Table 3), suggesting that auxin signaling is required throughout the entire processes of flower opening and closure. Aux/IAAs act as transcriptional repressors to dimerize with Auxin Response Factor (ARF), allowing the rapid changes of transcription in response to auxin [28,43]. In DEGs of waterlily Aux/IAA homologues, 13 out of 17 IAAs maintained high transcript during the opening process, but was suppressed at closing stage (Fig. 5d, Table 3). TOPLESS  (TPL) which physically interacts with IAA12 and acts as a co-repressor to regulate auxin transcription [45], was consistently oscillating with TIR1/AFBs (Fig. 5d, Table 3). Moreover, the downstream components of auxin signaling, ARFs were also generally high expressed at flower opening stage (Fig. 5d, Table 3). Thus, expression of auxin signaling regulators was generally induced associated with flower opening, suggesting that opening action requires very active auxin perception network. Particularly, DEGs of small auxin up RNA (SAUR) family are greatly found in our transcriptome profile, which represent the largest family of early auxin response genes with elusive auxin-related functions. Recently studies have implicated the key roles of SAURs which integrate auxin and environmental signals in a wide range of plant development, including auxin-dependent tropic growth, apical hook development, leaf senescence, root growth, seed germination etc. [46]. There are 81 SAURs in Arabidopsis, 58 SAURs in rice, 18 SAURs in moss, 134 SAURs in potato and 79 SAURs in maize [46]. Due to the gene evolution and differentiation, tandem and segmental duplication events are usually happened which contributed to the expansion of the SAUR family genes. Interestingly, 25 SAUR homologues were differentially expressed during floral opening-closure process: SAUR32, SAUR36 and SAUR72 were highly expressed at the onset of flower opening (7:00), while being suppressed at the later flower movement processes; 13 out of 25 SAURs were particularly induced associated with rapid opening process (10:00); and the rest of SAURs were stimulated at the closing stage (18:00) (Fig. 5d, Table 3). Thus, different SAUR members displayed distinct responsiveness to flower opening or closing actions, supporting the underlying roles of SAURs involved in auxin-mediated anisotropic cell expansion during floral movement.
In summary, through the regulation of auxin metabolism, auxin transportation and auxin signaling, auxin serves as a key phytohormone to determine the circadian movement rhythm of flowers.

Cell wall remodeling is downstream of auxin events and controlled by auxin oscillation during floral movement
In addition to auxin regulations, plenty of vesicle trafficking-related genes were found in cluster 3, which maintained high expression levels from 7:00 till 10:00 (Figs. 4d, 5a, Additional file 5: Figure S4A), including multiple COPI vesicle coat genes such as Coatomer Subunits, Conserved oligomeric Golgi complex subunits (COG) and Protein transport Sec genes (Additional file 4: Table S2), suggesting that vesicle trafficking is the earliest responding process during the rapid opening stage. Vesicle trafficking is a dynamics and rapid movement process, including endocytosis and exocytosis, packaging the signal molecules from the synthesis locations, such as endoplasmic reticulum (ER) to Golgi body; the molecules are finally released to their locations either on the plasma membrane, secreted to the extracellular space or targeted to the vacuoles for degradation [47]. Particularly, vesicle flow is crucial for cellular morphogenesis as it delivers membrane cargoes and materials for cell wall formation. Thus, promotion of vesicle trafficking might activate reorganization of the cell wall components [48].
Plant cell walls are rigid, flexible and extensible layers composed of polysaccharides, proteins, and lignin. Among Table 3 Auxin Signaling (corresponding to Fig. 5d)  Transcriptome profile of auxin signaling-regulators, corresponding to Fig. 5d the wall polysaccharides, cellulose, made of β-1,4-linked glucan microfibril, is the major wall component synthesized by cellulose synthase (CesA) complexes [49]. In the transcriptome profiles, multiple cell wall synthesis-or modification-related genes were differently responding to floral movement at distinct stages. For example, CESAs were gradually upregulated at 7:00 till 18:00, suggesting that cell wall is continuously synthesized along the entire floral movement process (Fig. 6c, Table 4). The synthesized cell wall components can be loosened either by cell wall-loosening proteins Expansin (EXP) or xyloglucan endotransglucosylase (XTH/XET). EXPs promote cell wall acidification to increase wall extensibility, and XETs break the xyloglucan cross-links between cellulose microfibrils to loosen cell wall [50,51]. Several EXPs were highly expressed from 6:00 to 10:00 of the early stage of flower opening, whereas XETs showed upregulation from 10:00 till 18:00 (Fig. 6c, Table 4). Regarding the previous study showed that auxin triggers rapid cell wall acidification [52], the complementary responsiveness of EXPs and XETs to floral movement rhythm in our study also implies that EXPs-mediated wall loosening is important for early flower opening, probably in conjunction of auxin-induced rapid wall acidification; while the later stage of floral movement requires XETs-dependent cellulose microfibrils rearrangement. In addition, pectin is another major component of the plant cell wall, and it contains a highly heterogeneous group of polymers including homogalacturonans and rhamnogalacturonans I and II [53]. Pectinesterase (PME) catalyzes the de-esterification of pectin into pectate and methanol to alter the integrity of the cell wall [54]. Consistent with XETs expression, PMEs were additionally activated from 10:00 to 18:00 (Fig. 6c, Table 4), implying that a dramatic degradation and re-establishment of the cell wall occurred during the floral movement. Apparently, activation of these genes involved in cell wall remodeling was sustained throughout the entire process of floral movement, suggesting the importance of cell wall organization during flower opening and closure (Fig. 6c, Table 4).
Strikingly, a large number of plasma membraneassociated kinases, such as Serine/threonine-protein kinase, Brassinosteroid-Signaling Kinase 3 (BSK3) and NIMA (never in mitosis gene a)-related kinase 5 (NEK5) [55,56]; Leucine-rich repeat receptor (LRR)-like serine/ threonine-protein kinase, Receptor-like protein kinase 2 (RPK2) and Barely Any Meristem 1 (BAM1) [57,58]; Receptor-like protein kinase FERONIA and transmembrane kinase (TMK) [59,60], were found in cluster 8 (Additional file 5: Figure S4B, Additional file 4: Table S3). Most of them belong to plasma membrane-localized receptor-like kinases (RLKs). Ligands-activated RLKs are involved in a wide range of plant growth processes, such as plant immunity, male-female interaction and cell wall modification, which perceive and transmit extracellular signals into the cytosolic, leading to signaling transduction by phosphorylation [61][62][63]. For example, THESEUS1 (THE1) is a cell wall integrity sensor to mediate the response of plant cells to the perturbation of cellulose synthesis [64]. FERONIA (FER) elicits calcium transients to maintain cell wall integrity during salt stress [65]. In the transcriptome profile, THE1 and FER were upregulated from 10:00 till 18:00 (Fig. 6c, Additional file 4: Table S3), which raised a possibility that RLKs might be the signal sensors of cell wall to coordinate the floral movement rhythm.
It has been well described that auxin induces cell elongation and expansion through its effect on modification of the cell wall [24]. The above transcriptome data further confirm that auxin signaling and cell wall remodeling were synchronously oscillating, leading to the output phenotype of cell expansion at flower opening stage and cell shrinkage at flower closure stage. Thus, we would speculate a scenario of floral movement rhythm: I. at flower opening stage when cells anisotropically expand, auxin is highly synthesized and transported to the cell wall, subsequently leading to the dramatic loosening and reorganization of cell wall; II. At flower closure stage when cell expansion are restricted, auxin synthesis is relatively delayed with low auxin flow to the cell wall (Fig. 6c). Thus, our finding further unveils that auxin-mediated cell wall remodeling is essential for the movement rhythm formation of flower opening-closure.

Discussion
For the purpose of pollination, flowers open for a considerable period, and then abscise immediately after pollination [66]. Several factors are involved in this process, including transcriptional factors, hormonal regulation, cell wall modification and circadian rhythm control [9,20]. In our study, one of the most ancient flowering plants, Nymphaeales, is used to explore the mechanism of floral rhythm. This work globally described auxin-dependent transcriptional signatures of flower opening-closure movement in Nymphaeales, which could be conserved in most of flowering plants.
Van Doorn et al. have summarized the major mechanisms during flower opening and closure [9,20]. They concluded that turgor pressure and water transport are the major restrictive factors to limit fast movement. During the flower opening of the Asiatic Lily, the most influential factor is the change in angle between the midribs and the pedicel, where an increasing curvature of the midribs pulls the tepals apart [67]. Additionally, Liang and Mahadevana indicated that the edges of lily tepals are wrinkled as the flower opens [68]. Plant structures overcome these limits of turgor pressure, such as sepals of the waterlily flower, resulting in a very rapid cell expansion. Our research revealed that the intermediate petal is the major tissue responding to endogenous change and environmental stimuli. The realistic petal curvature determines the efficiency of insect-propagated pollination. Thus, plastic intermediate cells of petal offer a good platform to study the global signaling events during the circadian floral movement.
Consistent with the previous studies on light signaling which plays fundamental roles in diverse aspects of plant development, including floral movement rhythm, the crucial light signaling components were rapidly responding at the onset of flower opening (Additional file 3: Figure  S3C). Using Arabidopsis homologues as input, the transcript oscillations of waterlily homologues of red/ far-red light receptors Phytochromes (PHYs) [69], blue light receptor Cryptochromes (CRYs) [70] and  Table 4 according to the labelled number "22" to "27" in each cluster) Phototropins (PHOTs) [71], Ultraviolet-B receptor UV-B resistance 8 (UVR8) [72], and their downstream effectors including Phytochrome-Interacting Factor 3 (PIF3), Constitutive Photomorphogenic 1 (COP1), Suppressor Of Phytochrome A1 (SPA1), Elongated Hypocotyl 5 (HY5) [73] were depicted in a light signaling network (Additional file 3: Figure S3C, Additional file 4 Table S4). Moreover, the transcript level of the core circadian clock regulators, downstream of light signaling, such as Pseudo-Response Regulators (APRR) [74], Late Elongated Hypocotyl (LHY) [75] and Casein kinase (CKL) [76], were additionally triggered associated with flower opening (Additional file 3: Figure S3C, Additional file 4: Table S4). Therefore, the consistent upregulation of photosynthesis, light signaling and circadian clock events support the fundamental role of light at the beginning of flower opening. Beside of the regulations of light signaling, auxin and cell wall, meanwhile other signaling pathways are significantly influenced associated with floral movement, as seen by downregulations of transcription factors and ribonucleoproteins in cluster 2, 4 and 6 (Additional file 6: Figure S5, Additional file 4: Tables S5-S7). In model plants, increasing numbers of TFs are shown to participate in flower development. For example, the expression of ERF TFs is highly correlated with flower senescence, because increased ethylene is produced during floral organ senescence [77,78]. WRKY12 and WRKY13 oppositely regulate flowering under a short-day condition by controlling the plant hormone gibberellin [61]. Interestingly, RhNAC100 transcriptional factor was identified in rose petal, which is rapidly induced by ethylene in the petals. Silencing of RhNAC100 significantly increased the petal size and promoted cell expansion in the petal abaxial sub-epidermis [79]; this result is consistent with our finding that NAC TFs were downregulated during flower opening and petal cell expansion. Therefore, floral opening was associated with lower transcriptional cascades, suggesting that these transcriptional factors might act as negative regulators of cell expansion, which deserve further investigation.

Conclusion
Similar to all touch responses, which can turn plants away from aggressors or enable flowers to move toward sunrise, morphogenesis during the flower opening-closure process is also influenced by a combination of endogenous and exogenous perturbations [80]. In our study, transcriptome profiles of the intermediate cells of petal were unearthed in the waterlily, which confirm that the signaling cascades of auxin metabolism, signaling and transport play center roles in flower opening and closure, and cell wall remodeling event regulated by auxin realizes circadian cell expansion and shrinkage associated with flower movement rhythm. Additionally, photosynthesis, light signaling and transcriptional regulations are individually incorporated into different stages of floral movement. Our study provided a preliminary understanding of the opening-closure processes of the waterlily flower by dividing them into three stages (Fig. 7).
Stage I (7:00-10:00): rapid-response stage, when the flowers are initially triggered for opening. At this stage, photosynthesis and light perception are activated, while gene transcription is blocked. Photosynthesis is the first reactive event once the flowers are exposed to light, which utilizes light energy to synthesize carbohydrate molecules (sucrose) for consumption or storage. Photoreception-mediated extension of the cell volume of the petal intermediate segment may be initiated by an unloading of sucrose from the phloem, possibly resulting in the lower apoplastic water potential [80], subsequently causing a change in cell turgor pressure. Active auxin perception system creates a complex but flexible environment, being able to re-arrange cell wall components (Fig. 7).
Stage II (10:00-14:00): flower-opening stage, when the flowers are quickly blooming until the maximal angle. At this stage, vesicle trafficking, auxin synthesis, auxin signaling and efflux, and cell wall remodeling processes occupy dominant roles during flowers blooming, accompanied with downregulation of transcription. Frequent vesicles trafficking events enhance the endocytosis and exocytosis of PM-located cargoes, such as auxin transporters PIN proteins, facilitating active auxin flow during organogenesis [81]. Apparently, auxin-related genes, such as GH3 family, AUX/IAAs, ARFs, SAURs are enriched in cluster 3 which particularly responded to rapid flower opening. YUC-dependent auxin synthesis, GH3-regulated auxin homeostasis, TIR1/AFBs and SAUR-mediated auxin signaling, as well as PIN-and PGP-regulated auxin efflux are primarily induced during stage II. A high level of auxin is synthesized and then being pumped to the cell wall; consequently, cell walls are acidified by high auxin and loosened by expansins [82]. Stage III (14:00-18:00): late response stage, when flowers start to be closed. The waterlily represents a typical flowering plant with a cross-pollination trait. The opened flower waits for insects that bring pollens from other flowers, crawling down towards its stigma for further pollination [80]. Apparently, the completely opened flowers with expanded stamens facilitate insects in touching the bottom of the stigma. Therefore, the fertilization event occurs usually after the maximum opening (after 14:00). In our transcriptome profiles, several RLKs are activated during stage III, possibly leading to fertilization. On the other hand, RLKs might function as cell-wall signaling receptors, in conjunction with cell-wall modification enzymes, to coordinate cell wall reorganization during flower closure. Auxin influx and cell wall modification events are still active in stage III, which confirms the crucial contributions of auxin and cell wall throughout the entire processes of flower opening and closure.
Growth condition and chemical treatment NAA (Biosharp) and TIBA (Macklin) were dissolved in dimethyl sulfoxide (DMSO). For the chemical treatment, 10 μM NAA and 10 μM TIBA were pre-sprayed on the intermediate and bottom section of the petals for 8 h before phototracking, and then the floral movement was tracked for the following 24 h by capturing images using an automatic camera.

Time-lapse photography and measurement of flowering angles
The flower opening-closure patterns were tracked using time-lapse photography every 1 h for a total of 24 h. Deviated angles of petal opening and closure were measured using ImageJ (Additional file 1: Figure S1C). In each experiment, at least 10 flowers were analyzed, or at least 85 cells were measured. The experiments were repeated three times. Statistical analyses were performed using unpaired two-tailed t-test, where ns, *, **, *** and **** correspond to no significant difference, p-value < 0.05, < 0.01, < 0.001, < 0.0001, respectively.

RNA-seq, expression annotation, GO and KEGG pathway enrichment analyses
The sepals of each flowers were removed, and the extracted samples were collected from the intermediate and bottom segments of petal (≤1 cm distance to the bottom end) at different times of 6:00, 7:00, 10:00, 14:00 and 18:00 (Fig. 3a). Each sample was collected from at least three flowers, and the samples at each time point were collected in triplicated and ground into fine powder with liquid nitrogen for RNA extraction. The total RNA of the waterlily flowers was extracted using the RNAprep Pure Plant Kit (Tiangen). RNA-Seq libraries were sequenced on an Illumina HiSeq X platform. After demultiplexing, the adapter sequences were trimmed, and poly-N and low-quality reads were removed to obtain the high-quality, clean reads. The Next Generation Sequencing (NGS) reads for the samples were listed in Additional file 4: Table S8. Trinity V2.06 was used to assemble the Paired-end clean reads and the assembled results indicated the contig Nx statistics (eg. the contig N50 value) (Additional file 4: Table S9). TransDecoder V2.0.1 was applied to predict the protein-coding regions of the transcripts. The NCBI/nr, SWISS Prot, and Phytozome V12 databases were searched to annotate the peptide functions [83]. To quantify the expression levels of genes, Trinity reference was annotated, and paired-end reads were mapped to the assembled transcripts and assigned to genes using RSEM V1.2.20. Differential expression analysis was performed using R package DEseq2 V1.18.1. The FPKM (expected number of Fragments Per Kilobase of transcript sequence per Millions base pairs sequenced) of each gene was calculated based on the length of the gene and the reads counts mapped to this gene. Genes with a fold change > 1.5 and a q value< 0.05 were assigned as differentially expressed genes (DEGs). The DEGs were mapped to GO terms in the GO database (http://www.geneontology.org/) in order to calculate the gene numbers for every term. KEGG (http://www.genome.jp/kegg/) is used to perform the pathway analysis of the DEGs. Statistical enrichment of the DEGs for the GO terms implemented by the agriGO v2.0. Significantly enriched GO terms and KEGG pathways (corrected p < 0.05) were identified based on a hypergeometric test [84,85].