OsFLZ2 interacts with OsMADS51 to fine-tune rice flowering time

Rice is one of the most important crops worldwide. Appropriate flowering time (also known as heading date) is crucial for rice regional adaptation and optimal yield determination (Hayama et al., 2003; Izawa, 2007; Tsuji et al., 2011; Zhao et al., 2015). Precocious flowering shortens vegetative development available for carbon assimilation and causes yield reduction. On the other hand, delayed flowering could lead to reduction of seed setting in late growing season and trigger yield loss. Thus, identifying flowering time regulatory genes and exploring the underlying molecular mechanisms is crucial for rice genetic improvement (Zhou et al., 2021).

Flowering time in rice is strongly affected by exogenous factors, among which photoperiod is a key factor. As a typical facultative short-day (SD) plant, rice harbors at least two independent photoperiodic flowering pathways (Zhou et al., 2021). The first one is the Heading date 1 (Hd1; SE1) pathway (Yano et al., 2000), which is conserved both in rice (OsGI-Hd1-Hd3a) and in Arabidopsis (GI-CO-FT) (Tsuji et al., 2011; Shrestha et al., 2014). Arabidopsis GI-CO-FT signaling flow is active under long-day (LD) conditions, whereas rice OsGI-Hd1-Hd3a is active under SD conditions to promote flowering (Tsuji et al., 2011; Shrestha et al., 2014). Another photoperiodic flowering pathway in rice is the Early heading date 1 (Ehd1; EF1) pathway (Doi et al., 2004). Ehd1 was originally cloned from a major flowering time quantitative trait locus and encodes a B-type response regulator (Doi et al., 2004) that can promote flowering by inducing expression of the florigenic genes Heading date 3a (Hd3a) and RICE FLOWERING LOCUS T1 (RFT1), regardless of Hd1 function or day-length conditions (Doi et al., 2004; Nemoto et al., 2016). The precise regulation of Ehd1 expression is crucial for the appropriate heading date and several proteins involved in this process have been identified (Lee et al., 2004; Kim et al., 2007; Xue et al., 2008; Wei et al., 2010; Yan et al., 2011; Nemoto et al., 2016; Zhou et al., 2021). For example, OsMADS51, an ortholog of Arabidopsis FLOWERING LOCUS C (FLC) (Ruelens et al., 2013), functions as an upstream transcriptional activator of Ehd1 (Kim et al., 2007). The transcript level of Ehd1 is increased in OsMADS51 overexpression plants, but decreased in osmads51 mutants. Further study has revealed that OsMADS51 acts downstream of OsGI and can transmit a promotional signal from OsGI to Ehd1 under SD conditions, thus promoting flowering (Kim et al., 2007). The outline of OsMADS51-Ehd1-Hd3a flowering pathway has been established, but the detail regulatory mechanisms remain to be further characterized.

FCS-LIKE ZINC FINGER (FLZ) proteins are plant-specific regulatory proteins containing a conserved FLZ domain (also known as Domain of Unknown Function 581) (Jamsheer and Laxmi, 2014). Studies in Arabidopsis thaliana, maize (Zea mays) and wheat (Triticum aestivum) have demonstrated that FLZ proteins are implicated in the plant stress response (Jamsheer and Laxmi, 2015; Jamsheer et al., 2019; Chen et al., 2021; Qin et al., 2021). For example, ectopic expression of TaFLZ2D confers better tolerance of transgenic plants to salt stress (Qin et al., 2021). Recently, we systematically identified 29 OsFLZ genes in the rice genome and demonstrated that OsFLZ18 negatively regulates rice submergence tolerance by interacting with Snf1-related protein kinase 1A (SnRK1A) and inhibiting the transcriptional activation activity of SnRK1A in modulating the expression of its target gene αAmy3 (Ma et al., 2021). However, the roles of FLZ proteins in rice growth and development, especially in flowering transition, have not yet been well clarified. Here, we reported that OsFLZ2 (LOC_Os01g41010) negatively regulates rice flowering time.

Most OsFLZ genes are ubiquitously but differentially expressed in various tissues or at different developmental stages, implying their important roles in rice growth and development (Ma et al., 2021). To investigate whether and how OsFLZ genes regulate rice growth and development, we used the CRISPR/Cas9 gene editing system to generate OsFLZ mutants and observed the phenotypical difference from wild-type Nipponbare (Nip) plants. Among the obtained mutants, we found two OsFLZ2 Crispr lines (Crispr-4 and Crispr-11, with, respectively, 43- and 1-nucleotide deletions, both resulting in a predicted truncated OsFLZ2 protein, Fig. 1A) that flowered earlier than Nip plants (Fig. 1C,D and Fig. S1). Under natural LD conditions (NLD, from mid-April to July, Guangzhou, China), OsFLZ2 Crispr lines flowered ∼4 days earlier than Nip. Under natural SD conditions (NSD, from mid-July to October, Guangzhou, China), they flowered ∼3 days earlier than Nip (Fig. 1C,D and Fig. S1). These observations indicated that OsFLZ2 negatively regulates rice heading date. To further confirm this notion, transgenic rice plants overexpressing OsFLZ2 (OE-2 and OE-7, Fig. 1B) were generated in the Nip background. As expected, the OE transgenic plants flowered 4-5 days later than Nip under NLD conditions and 6-7 days later under NSD conditions (Fig. 1C,D and Fig. S1). These results indicate that OsFLZ2 is a negative regulator of rice flowering time.

To elucidate the roles of OsFLZ2 in regulating heading date, we examined the expression levels of four key flowering-related integrator genes (Hd1, Ehd1, Hd3a and RFT1) using quantitative real-time PCR assay (qRT-PCR, Fig. 2). Diurnal gene expression patterns were investigated using 14- and 30-day-old Nip and OsFLZ2 transgenic plants grown under SD (10 h light) and LD (14 h light) conditions, respectively. Under SD conditions, the transcript levels of Hd1, Ehd1 and Hd3a were greatly increased in the Crispr-4 line, particularly at the time points of peak expression in Nip, whereas they were considerably decreased in the OE-2 line when compared with those in Nip (Fig. 2A). In addition, the expression of RFT1 was comparable between Crispr-4 and Nip, whereas it was lower in OE-2 lines (Fig. 2A). Under LD conditions, all these four genes generally showed increased expression in the Crispr-4 line but decreased expression in the OE-2 line at the peak time points, when compared with those in Nip (Fig. 2B). These results indicate that OsFLZ2 can downregulate the expression of key flowering-promoting genes to delay rice heading date.

To date, no evidence showed FLZ proteins can function as transcription factors to directly modulate gene expression. To explore how OsFLZ2 regulates flowering-time gene expression, we explored OsFLZ2-interacting proteins using immunoprecipitation followed by a mass spectrometry (IP-MS) assay. In this assay, we obtained 353 potential OsFLZ2-binding proteins, including the previously described proteins SnRK1A and SnRK1B (Table S1; Ma et al., 2021). Interestingly, one high-scored peptide of OsMADS51 was identified (Fig. 3A and Table S1). OsMADS51 is a type I MADS transcription factor that functions as a floral activator by promoting the expression of Ehd1, Hd3a and RFT1 (Kim et al., 2007). Therefore, we proposed that OsFLZ2 might interact with OsMADS51 to regulate flowering gene expression. To test this, we first confirmed the OsFLZ2-OsMADS51 interaction in detail (Fig. 3B-D). In glutathione S-transferase (GST) pull-down assay, His-OsMADS51 protein could be effectively pulled down by GST-OsFLZ2, but not by GST alone (Fig. 3B), demonstrating a direct interaction between OsFLZ2 and OsMADS51. The interaction was further corroborated by a co-immunoprecipitation (co-IP) assay. Tobacco (Nicotiana benthamiana) leaves expressing OsFLZ2-GFP/OsMADS51-6HA and GFP/OsMADS51-6HA were harvested to isolate total proteins followed by an immunoprecipitation assay. Western blotting analysis showed that OsMADS51-6HA could be successfully co-precipitated by OsFLZ2-GFP, but not by GFP (Fig. 3C). This interaction was also assessed by firefly luciferase complementation imaging (LCI) assay. As shown in Fig. 3D, the co-expression of FLZ2-nLUC with cLUC-MADS51 led to strong LUC signals in tobacco leaves, whereas no obvious signal was observed from the co-expression of the negative controls, further supporting the association of OsFLZ2 and OsMADS51 in plant cells. Other members of MADS family, such as OsMADS8, OsMADS15 and OsMADS50, have also been reported to modulate rice heading date (Kang et al., 1997; Lee et al., 2004; Bian et al., 2011; Lu et al., 2012), so we checked the interaction of OsFLZ2 with them. The obtained LCI results showed that OsFLZ2 can interact with OsMADS50, but not with OsMADS8 and OsMADS14 (Fig. S2).

OsMADS51 functions early in the first 2 months of growth in rice leaves to activate flowering under SD conditions (Kim et al., 2007). To examine the expression profile of OsFLZ2 in leaves following rice growth, we harvested penultimate leaves from Nip plants every week from seedling (3 weeks old) to heading stage (9 weeks old) under NSD conditions to perform a qRT-PCR assay. Consistent with the previous results (Kim et al., 2007), OsMADS51 showed declining expression after rice growth (Fig. 4A). Similarly, OsFLZ2 also displayed its highest expression in leaves at the seedling stage (3 weeks old), followed by a gradually decreased expression to tillering stage (5 weeks old) and remained low until heading stage (9 weeks old) (Fig. 4A). The flowering time of Nip is roughly 7 weeks under NSD conditions in Guangzhou and the flowering transition occurs before 5 weeks. Thus, the expression of OsFLZ2 was higher before the flowering transition, indicating that OsFLZ2 functions in flowering determination at the early stage. We then assayed the diurnal expression of OsFLZ2. To achieve this, we harvested the penultimate leaves over a 48 h time course from 14-day-old Nip plants under SD conditions (Hori et al., 2013) to measure the transcript level of OsFLZ2. The qRT-PCR results showed that OsFLZ2 has an obvious diurnal rhythmic expression pattern, showing relatively low expression during light periods, high expression at night and a peak at midnight (Fig. S3). Collectively, these data indicate that OsFLZ2 expression overlapped that of OsMADS51 after rice growth and had an obvious diurnal rhythm.

To dissect the functional connection between OsFLZ2 and OsMADS51 in the regulation of flowering gene expression, a colocalization analysis was performed (Fig. 4B). The result showed that the fluorescence signal of OsFLZ2-GFP overlapped that of mCherry-OsMADS51 in the nucleus (Fig. 4B), in agreement with their roles in modulating transcription. OsMADS51 serves as a transcriptional activator of Ehd1 (Kim et al., 2007), whereas OsFLZ2 represses Ehd1 expression (Fig. 2); thus, we hypothesized that OsFLZ2 might be able to influence the transcriptional activation of Ehd1 by OsMADS51. To test this possibility, a dual-luciferase analysis was performed using Ehd1 promoter to drive a LUC reporter gene (Fig. 4C). In line with the previous report (Kim et al., 2007), the expression of the pEhd1::LUC reporter gene was upregulated about fourfold by addition of OsMADS51 (Fig. 4D), suggesting activation of Ehd1 by OsMADS51. However, co-expression OsFLZ2 (62-SK-OsFLZ2) but not the negative control (62-SK empty vector) with OsMADS51 could significantly downregulate the expression of the pEhd1::LUC reporter gene (Fig. 4D), demonstrating that OsFLZ2 can weaken the transcriptional activation of Ehd1 by OsMADS51. To clarify how OsFLZ2 involves in this process, we checked the protein levels of OsMADS51 with or without OsFLZ2 co-expression. The data showed that co-expression of OsFLZ2 decreased the protein accumulation of OsMADS51 dramatically (Fig. 4E), indicating that OsFLZ2 can destabilize OsMADS51.

Based on these results, we propose a molecular model regarding the function of OsFLZ2 in rice flowering time control (Fig. 4F). OsFLZ2 interacts with OsMADS51 in the nucleus to destabilize and attenuate the transcriptional activation of Ehd1 by OsMADS51, in order to decrease the expression of the downstream florigenic genes Hd3a and RFT1, thus delaying heading (Fig. 4F). Of note, OsMADS51 is a photoperiod-sensitive flowering activator; its overexpressors exhibits an obvious early flowering only under SD conditions (Kim et al., 2007). OsFLZ2 functions similarly in flowering transition under SD and LD conditions, whereas the transgenic plants exhibited less photoperiodic sensitivity (Fig. 1 and Fig. S1).

OsFLZ2 has a declining expression pattern in leaves following rice growth under SD conditions, which is highly similar to the action of OsMADS51 (Fig. 4A, Kim et al., 2007), meeting the requirement for their interaction in vivo. Additionally, OsFLZ2 colocalizes with OsMADS51 in the nucleus (Fig. 4B) and can attenuate the transcriptional activation activity of OsMADS51 to Ehd1 (Fig. 4C,D), in strong agreement with the repression of Ehd1 by OsFLZ2 (Fig. 2). Interestingly, co-expression of OsFLZ2 can decrease the accumulation of OsMADS51 protein (Fig. 4E), indicating that OsFLZ2 can destabilize OsMADS51. In Arabidopsis, FLZ proteins serve as scaffolding proteins and can broadly interact with protein kinases, transcription factors and other types of regulatory proteins (Nietzsche et al., 2016). Therefore, it is possible that OsFLZ2 can mediate the interaction of OsMADS51 with its regulatory proteins, such as E3 ubiquitin ligases, to promote its turnover. Beyond OsMADS51, OsFLZ2 also interacts with OsMADS50 (Fig. S2), which exerts a similar function in regulating flowering by promoting Ehd1 expression (Lee et al., 2004; Bian et al., 2011). It will be interesting to test whether OsFLZ2 also enables the interference functions of OsMADS50 in the same manner.

The transcription of Hd1 is obviously upregulated in OsFLZ2 Crispr line under both SD and LD conditions (Fig. 2), indicating that OsFLZ2 also affects Hd1-mediated flowering pathway. Hd1 is a major photoperiod sensitivity flowering regulator in rice that induces flowering under SD conditions and delays it under LD conditions (Yano et al., 2000; Kojima et al., 2002; Hayama et al., 2003; Ishikawa et al., 2011; Nemoto et al., 2016). Previous studies have shown that OsMADS50 and OsMADS51 mainly function under SD conditions to promote flowering, and the expression of Hd1 is not changed in their mutants or in overexpression plants (Lee et al., 2004; Kim et al., 2007), indicating that OsFLZ2 regulates Hd1 expression independently on these two transcription factors.

Collectively, our current study functionally characterizes OsFLZ2 as a novel flowering regulator in rice. We also reveal a hitherto unknown function of FLZ proteins in regulating plant growth, and highlight the functional importance and versatility of FLZ proteins. These findings will broaden our understanding of the roles of FLZ proteins and provide a new gene resource for the regulation of rice flowering.

The Nipponbare (Nip) was used for rice transformation. For phenotypic observation, OsFLZ2-OE, OsFLZ2-Crispr and Nip plants were grown at the Rice Research Institute, Guangdong Academy of Agricultural Sciences, Guangzhou (23.13°N, 113.27°E), China, where the normal early growing season with natural long-day (NLD) conditions is from mid-April to July and the late growing season with natural short-day (NSD) conditions is from mid-July to October. The heading date was recorded as the day when the first panicle emerged from the leaf sheath. For diurnal expression analysis, OsFLZ2-OE, OsFLZ2-Crispr and Nip plants were grown in climate chambers under short-day (SD) conditions (10 h light, 28°C/14 h darkness, 25°C) or long-day (LD) conditions (14 h light, 28°C/10 h darkness, 25°C). The penultimate leaves of 14-day-old plants in SD conditions or 30-day-old plants in LD conditions (Hori et al., 2013) were collected at the indicated Zeitgeber time (at the points of light on, light off and the mid-point) for RNA isolation.

The full-length coding sequence of OsFLZ2 was PCR amplified and cloned into the pCAMBIA1300-GFP (Ma et al., 2021) to generate pUBQ10:OsFLZ2-GFP construct to express OsFLZ2. To generate OsFLZ2 knockout plants, a 20 bp gene-specific spacer sequence (5′-TGCTCCTGTCGAAGCTGGGG-3′) of OsFLZ2 was inserted into the sgRNA/Cas9 construct (Ma et al., 2015). Primers are listed in Table S2. The resulting vectors were confirmed by sequencing and transformed into the Agrobacterium tumefaciens strain EHA105, and then introduced into Nip. T₀ transgenic lines of CRISPR/cas9 were analyzed by sequencing to confirm the gene editing. Homozygous OsFLZ2-OE (OE-2 and OE-7) and OsFLZ2-Crispr (Crispr-4 and Crispr-11) transgenic lines were used for phenotypic observation.

Total RNA was extracted from the collected samples using the Hipure plant RNA Mini Kit (Magen) according to the manufacturer's instructions. The first strand of cDNA was synthesized from 1 μg total RNA using the Primescript RT reagent Kit (Takara) following the manufacturer's instructions. qRT-PCR was performed in a 96-well plate using SYBR Premix ExTaqTM (Takara) on a CFX Connect real-time PCR detection system (Bio-Rad). The following PCR program was used: 95°C for 2 min, 50 cycles of 95°C for 10 s and 60°C for 15 s, followed by a melting-curve program. The rice Ubiquitin (UBQ) gene was used as an internal control. Relative expression levels of genes were calculated using the 2^−△△CT method (Livak and Schmittgen, 2001). Primers used for qRT-PCR are listed in Table S2.

Leaves of 3-week-old OsFLZ2-GFP (OE-2) plants were collected and ground in liquid nitrogen and then homogenized in ice-cold immunoprecipitation buffer [50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM MgCl₂, 10% glycerol, 0.2% NP-40, 5 mm DTT, 0.1 mM PMSF and proteinase inhibitor cocktail (Roche)]. The suspensions were centrifuged twice at 16,000 g for 10 min at 4°C to remove debris. The supernatant was incubated with GFP-Trap_MA (gtma-20, ChromoTek) at 4°C for 2 h and then the beads were washed five times with ice-cold immunoprecipitation buffer. After washing, the beads were re-suspended in 30 μl 1×SDS sample loading buffer and boiled at 95°C for 10 min. Finally, all the samples were loaded in 8% stacking gel and ∼1 cm-length gel from the loading pore was cut after running for 15 min at 100 V. The proteins were digested in-gel with trypsin (Guangzhou Huijun Biosciences). Mass spectrometry measurements were performed on a Q Exactive Mass Spectrometer (Thermo Fisher Scientific). The resulting spectra were searched against the RGAP database (http://rice.uga.edu/).

For pull-down assays, the coding sequence of OsFLZ2 was PCR amplified and cloned into the vector pGEX4T-3, while the coding sequence of OsMADS51 was PCR amplified using KOD polymerase (Toyobo) and cloned into the vector pRSETA-His-SUMO (Ma et al., 2021). Primers are listed in Table S2. The resulting plasmids were separately transformed into E. coli BL21 and the proteins were induced with 0.1 mM IPTG. For GST pull-down assay, equal amounts of GST or GST-OsFLZ2 proteins were mixed with His-OsMADS51 at 4°C for overnight in 1 ml binding buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA and 1% NP-40). The beads were then washed three times with binding buffer and boiled at 100°C for 10 min. The elution was subjected to immunoblotting analysis using anti-His (HT501,TransGen; 1:6000) and anti-GST (HT601, TransGen; 1:6000) antibodies.

The full-length coding sequence of OsFLZ2 was inserted into the pCAMBIA13000-GFP (Ma et al., 2021) vector, while the coding sequence of OsMADS51 was inserted into pGreen-35S-6HA vector (Qian et al., 2021) to express OsFLZ2-GFP and OsMADS51-6HA, respectively. Primers are listed in Table S2. The resulting plasmids were transformed into Agrobacterium tumefaciens (strain GV3101) and then infiltrated into 5-week-old N. benthamiana leaves. The total proteins were extracted from leaves expressing OsFLZ2-GFP/OsMADS51-6HA or GFP/OsMADS51-6HA constructs with protein extraction buffer [50 mM Tris-HCl (pH 7.5), 1 mM EDTA, 150 mM NaCl, 10% glycerol and 1% NP-40] containing 1×protease inhibitor cocktail. An anti-GFP antibody was added to the protein extracts for immunoprecipitation for 3 h, then protein A magnetic beads (36403ES03, Yeasen) were added and incubated for another 2 h. After washing three times with the protein extraction buffer, the co-immunoprecipitated products were separated by SDS-PAGE gel and detected with anti-GFP (HT801,TransGen; 1:5000) and anti-HA (ab9110, Abcam; 1:5000) antibodies.

The full-length coding sequence of OsFLZ2 was PCR amplified and cloned into the pCAMBIA-nLUC (Chen et al., 2008) to generate OsFLZ2-nLUC vector. The full-length OsMADS8/15/50/51 were PCR amplified and separately cloned into the pCAMBIA-cLUC (Chen et al., 2008) to generate cLUC-OsMADS vectors. Primers are listed in Table S2. These plasmids were transformed into Agrobacterium tumefaciens GV3101. A. tumefaciens containing OsFLZ2-nLUC, A. tumefaciens containing cLUC-OsMADS and A. tumefaciens containing P19 were co-injected into fully expanded leaves of 5-week-old N. Benthamian plants at a ratio of 2:2:1. After infiltration for 3 days, the infiltrated regions were injected with 1 mM of luciferin (BioVision). The luminescence activity was captured with a chemiluminescent imaging system (Tanon).

The full-length coding sequence of OsFLZ2 was inserted into pCAMBIA13000-GFP vector to fuse with GFP. The full-length coding sequence of OsMADS51 was inserted into pCAMBIA13000-mCherry (Liu et al., 2021) vector to fuse with mCherry. Primers are listed in Table S2. The resulting plasmids were co-transfected into Arabidopsis protoplasts following a previously described method (Yang et al., 2018). After incubation in darkness for 16 h, the fluorescence signal of OsFLZ2-GFP and mCherry-OsMADS51 was observed and pictures taken with a confocal fluorescence microscope (LSM 710, Zeiss).

The dual-luciferase reporter assay was performed following our previously established method (Yang et al., 2018; Ma et al., 2021). The coding sequence of OsFLZ2 and OsMADS51 were separately cloned into the pGreenII 62-SK vector (Ma et al., 2021) to generate the effector vectors. The Ehd1 promoter fragment (2 kb upstream of the start codon) was amplified and cloned into the pGreenII 0800-LUC (Ma et al., 2021) to generate the reporter vector. The indicated plasmids (10 μg for each) were transfected into Arabidopsis protoplasts. After incubation in darkness for 16 h, protoplasts were harvested to extract the total proteins for luciferase activity detection using a Dual-Luciferase Reporter Assay System (E1910, Promega) in accordance with the manufacturer's instructions. The relative luciferase activity was calculated as the ratio of Firefly luciferase activity to Renilla luciferase activity. The experiments were carried out in three independent replicates. Primers used in this assay are listed in Table S2.

To check the effect of OsFLZ2 on the protein stability of OsMADS51, the coding sequence of OsFLZ2 was cloned into pSAT6-YFP-N1 (Yang et al., 2018) to generate the OsFLZ2-YFP vector. Primers are listed in Table S2. OsMADS51-6HA (10 μg) was co-transfected into Arabidopsis protoplasts with 10 μg YFP or OsFLZ2-YFP. After incubation in darkness for 16 h, protoplasts were harvested to extract the total protein with extraction buffer [50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM MgCl₂, 10% glycerol, 1% NP-40, 5 mm DTT, 0.1 mM PMSF and proteinase inhibitor cocktail (Roche)]. The protein samples were separated by SDS-PAGE gel and detected with anti-GFP (HT801,TransGen; 1:5000) and anti-HA (ab9110, Abcam; 1:5000) antibodies. Ponceau S (ab146313, Abcam) staining of the nitrocellulose membrane indicates the protein loading.

All chemicals were purchased from Sangon Biotech unless stated otherwise. Statistical analyses were performed using an unpaired two-tailed Student's t-test in Microsoft Excel. Statistical difference is shown as *P≤0.05 and **P≤0.01. The n values represent the number of independent biological replicates. Graphs were produced using GraphPad Prism 5 software.

The authors are grateful to Prof. Yao-Guang Liu (South China Agricultural University) for his kind provision of the CRISPR/Cas9 vector.