Introduction

Organ development in plants mainly occurs post-embryonically, with both external and internal inputs influencing the final shapes of the organs. Meristems possess a small self-maintaining group of pluripotent stem cells that give rise to all organs. These stem cells are located in the central zone of a dome-shaped meristem. Cells in the peripheral zone of the meristem divide faster than those in the central zone and become organ initial cells, also known as founder cells (Chandler 2011). The balance between cell proliferation and differentiation in these two functional zones is determined as a result of the integration of external and internal inputs. After the integration, interactions of multiple factors are thought to act as developmental outputs.

Flower development depends on the presence of determinate floral meristems. Floral meristem activity is necessary to produce floral organ primordia, but the meristematic activity must be fully terminated after sufficient cell proliferation has occurred (Sun and Ito 2015). In Arabidopsis thaliana, a key determinant gene for floral meristem establishment is WUSCHEL (WUS), which encodes a homeodomain protein (Laux et al. 1996; Mayer et al. 1998). WUS activity is precisely controlled at both the transcriptional and translational levels (Perales et al. 2016; Rodriguez et al. 2016), and mutations in this gene cause striking defects in the floral meristems; for example, wus-1 flowers lack carpels (Laux et al. 1996). WUS acts as both an activator and a repressor of gene expression (Ikeda et al. 2009) by interacting with various partners. In addition to forming a transcriptional repression complex with TOPLESS and HISTONE DEACETYLASE19 (Kieffer et al. 2006; Szemenyei et al. 2008), or with HAIRY MERISTEM (Zhou et al. 2015), WUS provides a different function. The stability of the WUS protein is maintained by DNA-dependent homodimerization (Rodriguez et al. 2016).

Multiple regulators of WUS mRNA levels have been identified. The WUS mRNA expression domain is largely restricted to the organizing center located below the shoot and floral meristems, which is necessary to maintain the stem cell population (Laux et al. 1996). A negative feedback loop controlled by a CLAVATA (CLV) signaling pathway governs stem cell homeostasis. The CLV pathway determines the level of WUS expression and prevents the accumulation of excess stem cells (Fletcher et al. 1999; Brand et al. 2000; Schoof et al. 2000; Kondo et al. 2006; Gruel et al. 2016). A leucine-rich repeat receptor kinase, CLV1, and a secreted peptide, CLV3, function as a ligand–receptor pair (Ogawa et al. 2008). CLV3 is expressed in the stem cell domain of floral meristems (Fletcher et al. 1999). WUS directly binds to a group of tightly clustered cis-elements in the CLV3 promoter to control its transcription (Perales et al. 2016), while the CLV3 signaling pathway in turn negatively regulates WUS expression. Mutants of CLV3 possess enlarged floral meristems and have higher levels of WUS expression due to the absence of this negative feedback loop (Clark et al. 1995; Brand et al. 2000; Szczesny et al. 2009).

Another negative feedback loop functions to terminate stem cell activity during flower development (Doerner 2001). The floral determinacy regulator gene AGAMOUS (AG) is synergistically activated by WUS and the floral meristem identity regulator LEAFY (LFY), beginning at stage 3 of flower development (Smyth et al. 1990; Yanofsky et al. 1990; Weigel et al. 1992; Lenhard et al. 2001; Lohmann et al. 2001). AG controls many reproductive developmental processes by activating or repressing its ~ 2000 direct targets (Ito et al. 2004, 2007; Gomez-Mena et al. 2005; Sun et al. 2009; Liu et al. 2011; Ó’Maoiléidigh et al. 2013). AG is required for determinacy; termination of floral meristem activity is delayed in flowers of the ag mutant, which results in flowers within flowers (Yanofsky et al. 1990). In certain conditions, such as short-day photoperiods or treatments with gibberellic acid, ag flowers show an inflorescence-like reversion (Okamuro et al. 1996). AG both directly and indirectly represses WUS expression through the precise control of histone modifications, and fully inhibits WUS expression during stage 6 of floral development (Sun et al. 2009, 2014; Liu et al. 2011). Another regulator of WUS expression in flowers is the transcriptional repressor SUPERMAN (SUP), which contains a C2H2 zinc-finger DNA-binding domain and an EAR repression domain (Bowman et al. 1992; Sakai et al. 1995). Since SUP is expressed in the cells surrounding the floral meristem at stage 3, the effect of SUP on WUS expression is largely non-cell-autonomous (Ito et al. 2003; Prunet et al. 2017).

The genetic interactions between the AG and SUP, the AG and CLV and the SUP and CLV pathways have previously been examined (Bowman et al. 1992; Meyerowitz 1997; Breuil-Broyer et al. 2016). Bowman et al. (1992) revealed that ag sup double mutants indeterminately produce additional whorls of petals from undifferentiated cells, while Meyerowitz (1997) reported that ag clv1 double mutants have an excessive number of organs and whorls. Breuil-Broyer et al. (2016) showed that sup alleles crossed to the clv1 resulted in synergistic enhancement of stamen number increase and carpelloidy. These findings suggest that these signaling pathways are largely distinct in their control of meristematic cell proliferation and differentiation in floral development; however, previous research has not investigated the potential genetic interactions among AG, SUP, and CLV3.

To further understand the coordinated actions of AG, SUP, and CLV3 in controlling the determination of the floral meristem, we generated double and triple mutants of these three genes and quantitatively analyzed their meristematic phenotypes. Our analysis revealed specific spatial and temporal roles for CLV3, AG, and SUP in floral meristem determinacy, enabling a coordinated modulation of floral meristem sizes in parallel pathways. Furthermore, an RNA-seq analysis was conducted to compare the transcriptomes of floral buds from the ag sup clv3 triple mutant with those of the double mutant with functional AG activity (sup clv3), enabling the systematic identification of genes and processes that depend on AG function in the floral meristem. Our findings suggest that AG functions to maintain the irreversible state of reproductive development through the negative regulation of floral meristem identity genes and genes involved in organ initiation.

Materials and methods

Genetic stocks and growth conditions

This study used the A. thaliana mutants ag-1, clv3-2, and sup-1, and pWUS::GFP-ER lines, all of which were in the Landsberg erecta genetic background, and which have all been described previously (Bowman et al. 1989, 1992; Clark et al. 1995; Gordon et al. 2007). The double and triple mutants were generated by genetic crosses and were identified in the F2 or later generations using PCR genotyping. Primer sequences are provided in Supplementary Table 1. Mutants containing transgenes were identified in the F2 or later generations by selection with antibiotics and PCR genotyping. All plants were grown on soil under continuous light conditions at 22 °C.

Phenotypic and statistical analyses

To quantify the size of flowers, at least 20 flower images were taken from the wild type, single, double, and triple mutants, respectively, for analyses. To minimize the environmental differences in growth chambers, these plants were grown side-by-side at the same density in each pot. Images for each genotype at stage 13, when buds open (Smyth et al. 1990) were quantified by Image J (NIH). Student’s t test was conducted to evaluate the statistical significance.

SEM

Scanning electron microscopy (SEM) was performed as previously described (Yamaguchi and Komeda 2013), with minor modifications. Prior to floral tissue fixation, flowers older than stage 10 were removed with forceps. The floral tissues were fixed overnight in a solution containing 45% ethanol, 5% formaldehyde, and 5% acetic acid, then dehydrated through an ethanol and acetone series. The resulting tissues were critical-dried with liquid CO2 using a critical point dryer (EM CPD300; Leica Microsystems) and coated with gold using an E-1010 sputter coater (Hitachi) before SEM imaging. The tissues were imaged under an S-4700 SEM (Hitachi) with an accelerating voltage of 15 kV. More than five floral primordia for each genotype were observed.

Tissue sectioning

Sectioning was performed as previously described (Yamaguchi and Komeda 2013), with minor modifications. To minimize the environmental differences in the growth chambers, the plants used for sectioning were grown side-by-side at the same density in each pot. Flowers older than stage 8 were removed with forceps, and the remaining floral tissues were fixed overnight in 45% ethanol, 5% formaldehyde, and 5% acetic acid, and then dehydrated through a series of ethanol solutions. The 100% ethanol was replaced using a Technovit 7100 resin solution (Heraeus), and the floral tissues were incubated at room temperature overnight. The resin-containing tissues were then polymerized, and 10-µm longitudinal sections were made using a RM2255 microtome (Leica Microsystems). The sections were placed onto a microscope slide and stained with 0.05% toluidine blue (Wako Chemicals) and then observed under an Axio Scope A1 microscope (Carl Zeiss). From serial sections in the abaxial–adaxial axis, floral primordia or floral buds cutting in the centers were selected for the height and width measurements. The height and width of the floral meristems were determined from at least 13 floral primordia or floral buds from individual plants of each genotype. Images for each genotype at specific developmental stages were quantified in ImageJ (National Institutes of Health). A Student’s t test was conducted to evaluate the statistical significance of the data.

GFP observation

Confocal microscopy was performed as previously described (Yamaguchi et al. 2016), with minor modifications. After using forceps to remove flowers older than stage 8, the inflorescences were embedded into a 5% agar block. A Liner Slicer PRO7 vibratome (Dosaka) was used to obtain 35-µm sections, which were placed onto a microscope slide in a drop of water, covered with a cover glass, and observed under a confocal laser scanning microscope (FV1000: Olympus) with a UPlanSApo objective lens (Olympus). More than five floral buds at each floral stage were observed for each genotype.

RNA-seq

An RNeasy Plant Mini Kit (Qiagen) was used to extract total RNA from four biological replicates of sup-1 clv3-2 and ag-1 sup-1 clv3-2 floral buds up to stage 10 of flower development. DNA was removed using an RNase-Free DNase Set (Qiagen), and the mRNA was extracted from the total RNA using oligo-dT magnetic beads (New England Biolabs). An RNA library was prepared using the Breath Adapter Directional sequencing method for strand-specific 3′ Digital Gene Expression (Townsley et al. 2015). Briefly, the mRNA was fragmented using magnesium ions at elevated temperatures, after which the polyA tails of mRNA were primed using an adapter-containing oligonucleotide for cDNA synthesis with DNA Polymerase I (Thermo Fisher Scientific). The 5′ adapter addition was performed using breath capture to generate strand-specific libraries. The final PCR enrichment was performed using oligonucleotides containing the full adapter sequence with different indexes and Phusion High-Fidelity DNA Polymerase (New England Biolabs). The cleanup and size selection of the resulting cDNA was performed using AMPure XP beads (Beckman Coulter). The size distribution and concentration of the library were measured using agarose gel electrophoresis and a microplate photometer, respectively, to enable the pooling of libraries for Illumina sequencing systems. The libraries were sequenced by NextSeq 500 (Illumina). The produced bcl files were converted to fastq files by bcl2fastq (Illumina). The data were deposited into the DNA Data Bank of Japan (DRA006355).

Data analysis

AG ChIP-seq data, organ-initial-cell transcriptomic data, and floral-organ-specific transcriptomic data were obtained from Ó’Maoiléidigh et al. (2013), Frerichs et al. (2016), and Jiao and Meyerowitz (2010), respectively. Genes of interest were identified and characterized as previously described (Winter et al. 2015), with minor modifications. An overlap between differentially expressed genes in sup clv3 and ag sup clv3 and the AG-bound genes was examined and visualized using VENNY 2.1 (http://bioinfogp.cnb.csic.es/tools/venny/index.html). To examine the overlap of the two datasets, a Chi-square test was performed in R (https://www.r-project.org/). A gene ontology (GO) term enrichment analysis was conducted using the agriGO web-based tool and database (http://bioinfo.cau.edu.cn/agriGO/). The TreeMap view of GO terms and interactive graph view were generated with REVIGO (http://revigo.irb.hr/) after minimizing redundant enriched GO terms. MeV (http://mev.tm4.org/#/welcome) was used to generate a heatmap and perform k-mean clustering.

RT-PCR

RT-PCR was performed as previously described (Yamaguchi et al. 2014, 2017), with minor modifications. RNA was extracted using an RNeasy Plant Mini Kit (Qiagen). To minimize contamination by genomic DNA, an RNase-Free DNase Set (Qiagen) was used prior to cDNA synthesis, which was performed using a PrimeScript 1st strand cDNA Synthesis Kit (Takara). The resulting cDNA was quantified with a LightCycler 480 (Roche) using FastStart Essential DNA Green Master mix (Roche). The results of the RT-PCR experiments were normalized against the internal control gene, EIF4 (At3g13920). Two independent experiments were performed, and similar results were obtained. The RT-PCR primers are listed in Supplementary Table 1.

Results

Characterization of single, double, and triple mutant flowers of ag-1, sup-1, and clv3-2

Prior to our detailed genetic interaction assay of AG, SUP, and CLV3, we first confirmed the previously reported phenotypes of the single and double mutants. A wild-type Arabidopsis flower consists of four types of organ: four sepals, four petals, six stamens, and two carpels (Fig. 1a). Since AG controls the termination of floral meristem activity and the identity of the floral organs, the ag-1 flowers had more whorls of organs in a repeating sepal–petal–petal sequence (Fig. 1b; Bowman et al. 1989; Yanofsky et al. 1990). The sup-1 flowers produced approximately ten stamens, but had fewer, smaller carpels than the wild type (Fig. 1c; Bowman et al. 1992; Sakai et al. 1995). Although the numbers of its floral organs varied, the clv3-2 flowers often had increased numbers in each of the four whorls (Fig. 1d; Clark et al. 1995; Fletcher et al. 1999). The flowers of the single mutants were slightly larger than those of the wild type because of the increased numbers of whorls and/or organs (Fig. 1a–d) (WT vs. ag-1: p = 4.1 × 10−11, WT vs. sup-1: p = 1.3 × 10−8, WT vs. clv3-2: 2.7 × 10−9). As previously reported (Bowman et al. 1992), ag-1 sup-1 forms numerous whorls of petals after a single whorl of sepals (Fig. 1e), resulting in larger flowers than those of the single mutants (ag-1 vs. ag-1 sup-1: p = 1.5 × 10−8). At the center of the ag-1 sup-1 flowers is a mass of undifferentiated cells (Fig. 1f).

Fig. 1
figure 1

Comparison of flower size between the wild type and single, double, and triple mutants. a–l Top views of the flowers formed in wild type (a), ag-1 (b), sup-1 (c), clv3-2 (d), ag-1 sup-1 (e, f) ag-1 clv3-2 (g, h), sup-1 clv3-2 (i, j), and ag-1 sup-1 clv3-2 (k, l). f, h, j, l A mass of undifferentiated cells located in the center of some mutant flowers is shown. m Quantification of flower size. Error bars represent SD of at least 13 measurements. The p values were calculated using a Student’s t test. Bar = 1 mm in a–e, g, i, and k; 200 µm in f, h, j, and l

Full size image

To understand the genetic interactions between AG and CLV3 in flower development, we generated the ag-1 clv3-2 double mutant. As reported for the ag-1 clv1 double mutant (Meyerowitz 1997), ag-1 clv3-2 had increased numbers of organs and whorls, as well as larger flowers, compared with those of the single mutants (Fig. 1b, d, g) (ag-1 vs. ag-1 clv3-2: p = 8.3 × 10−12). In ag-1 clv3-2, numerous sepals and petals were produced around a mass of undifferentiated meristematic cells (Fig. 1h), which appeared to be much larger than that of the ag-1 sup-1 flowers (Fig. 1f, h).

To examine the effect of mutations in both SUP and CLV3, we generated the sup-1 clv3-2 double mutant. The sup-1 clv3-2 flowers had a clv3-2-like phenotype in terms of sepal and petal number (Fig. 1d, i). As reported in sup-1 clv1-6 (Breuil-Broyer et al. 2016), sup-1 clv3-2 exhibited an increased number of stamens compared with the parental lines (Fig. 1i, j), and the size of the stigmatic region of sup-1 clv3-2 at the tip of the gynoecium was larger than that of the clv3-2 single mutant (Fig. 1d, i, j); a mass of undifferentiated cells was surrounded by an enlarged stigma-like structure. Because the organ number defects were enhanced in the double mutant, the size of the flowers of sup-1 clv3-2 was slightly increased (clv3-2 vs. sup-1 clv3-2: p = 8.1 × 10−5).

To further investigate the genetic interactions among CLV3, AG, and SUP, we generated triple mutants. The ag-1 sup-1 clv3-2 triple mutant had a large increase in the numbers of petal whorls produced after a single whorl of sepals, which was much more pronounced than those of the single mutants or either double mutant combination (Fig. 1a–l) (ag-1 clv3-2 vs. ag-1 sup-1 clv3-2: p = 7.2 × 10−6). The region of undifferentiated cells in the center of the triple mutant flowers was similar to that of ag-1 clv3-2 (Fig. 1f, h, j, k).

Sizes of the mutant floral meristems at floral developmental stage 3

Changes in the size of flowers and/or the number of floral organs are often correlated with the height and width of the floral meristems (Clark et al. 1995; Laux et al. 1996; Sawa et al. 1999). We quantified the heights and widths of floral meristems at stage 3 of flower development using SEM and sectioning in the adaxial–abaxial axis (Fig. 2a–i). No significant difference was observed in the heights (from the groove between sepal primordia and inner parts to the top of the floral meristem) and the widths (between the two grooves along the lateral axis) of floral meristems between the wild type, ag-1, and sup-1 at stage 3 of flower development (p > 0.01; Fig. 2a–c, i, j). As previously reported (Clark et al. 1995), approximately 1.5-fold and 1.1-fold increases in the heights and widths of clv3-2 stage 3 floral meristems were observed relative to the wild type, respectively; both clv3-2 values were significantly greater than those of the wild type (p = 2.9 × 10−8 and p = 1.9 × 10−5), ag-1, and sup-1 (Fig. 2d, i, j).

Fig. 2
figure 2

Comparison of floral meristem height and width between the wild type and single, double, and triple mutant floral buds at stage 3. a–h Side and top views of the floral buds at stage 3 formed in wild type (a), ag-1 (b), sup-1 (c), clv3-2 (d) ag-1 sup-1 (e) ag-1 clv3-2 (f), sup-1 clv3-2 (g), and ag-1 sup-1 clv3-2 (h). (Left) Side views from longitudinal sections. (Right) Top views visualized using SEM. Section and SEM images are shown at the same magnification. Bar = 50 µm in a–h. i, j Quantification of floral meristem height (i) and width (j) in floral buds at stage 3. Error bars represent SD of at least 13 measurements. The p values were calculated using a Student’s t test

Full size image

We next characterized the heights and widths of the floral meristems in the double and triple mutants. No significant difference in height or width was observed between the ag-1 sup-1 double mutant and its parental lines (p > 0.01), nor did a mutation in the AG gene alter the clv3-2 phenotype of the stage 3 meristems (p > 0.01; Fig. 2b, c, e, i, j). By contrast, the sup-1 mutation had a statistically significant effect in the sensitized clv3 mutant background at stage 3, enhancing the defect in floral meristem height observed in clv3 (p = 2.2 × 10−3), but not in the meristem width (p > 0.01). The average heights of stage 3 floral meristems in clv3-2 and sup-1 clv3-2 were about 25 and 31 µm, respectively. In the ag-1 sup-1 clv3-2 triple mutant, the width of the floral meristems at stage 3 was similar to those of clv3-2, ag-1 clv3-2, and sup-1 clv3-2 (Fig. 2d, f–h, j). By contrast, the floral meristems of ag-1 sup-1 clv3-2 were taller than those of any of the parental lines (Fig. 2g–i), with a statistically significant difference between the floral meristem heights of ag-1 sup-1 clv3-2 and sup-1 clv3-2, which had the tallest floral meristems among the double mutants (p = 2.3 × 10−3).

Sizes of the mutant floral meristems at floral developmental stage 5

To further address the roles of AG, SUP, and CLV3 in floral meristem proliferation, floral meristems at stage 5 were characterized in detail (Fig. 3a–j). There was no difference in the heights and widths of the floral meristems between the wild type and ag-1 (p > 0.01), suggesting that the increased size of the ag-1 flowers mainly occurs after stage 5 (Fig. 3a, b, i, j). In the sup-1 mutant, an increase in floral meristem width (p = 6.6 × 10−4) relative to the wild type becomes apparent by stage 5 (Fig. 3a, c, i, j). The average floral meristem height and width of clv3-2 were significantly higher those of the wild type (p = 1.2 × 10−14 and p = 1.5 × 10−4, respectively).

Fig. 3
figure 3

Comparison of floral meristem height and width between the wild type and single, double, and triple floral buds at stage 5. a–h Side and top views of the floral buds at stage 5 formed in wild type (a), ag-1 (b), sup-1 (c), clv3-2 (d), ag-1 sup-1 (e), ag-1 clv3-2 (f), sup-1 clv3-2 (g), and ag-1 sup-1 clv3-2 (h). (Left) Side views from longitudinal sections. (Right) Top views visualized using SEM. Section and SEM images are shown at the same magnification. Bar = 50 µm in a–h. i, j Quantification of floral meristem height (i) and width (j) in flowers at stage 5. Error bars represent SD of at least 13 measurements. The p values were calculated using a Student’s t test

Full size image

Although no difference was observed in the floral meristem sizes of sup-1 and ag-1 sup-1 at stage 3, the mutation in the AG gene enhanced the defect in floral meristem height in the sup-1 mutant by stage 5 (p = 4.2 × 10−3; Fig. 3c, e, i, j). The introduction of the ag or sup mutations into the clv3-2 background also further increased the floral meristem heights (p = 6.4 × 10−3 and p = 1.9 × 10−5, respectively; Fig. 3f–j). Interestingly, the phenotypic enhancement of the floral meristem widths was observed in sup-1 clv3-2 (p = 5.8 × 10−6), but not in ag-1 clv3-2 (Fig. 3i, j). In the ag-1 sup-1 clv3-2 triple mutant, the floral meristems were taller than those of any of the parental lines (Fig. 3a–i), with a significant increase even observed between the triple mutant and sup-1 clv3-2 (p = 1.8 × 10−3), which had the tallest floral meristems among the double mutants (Fig. 3g–j). These results indicate that CLV3, AG, and SUP act in partially redundant pathways and coordinately modulate the size of the floral meristem in a spatial and temporal manner. Although AG mainly functions to restrict floral meristem proliferation after stage 5, it also plays a role in the earlier stages of development, as revealed in the sensitized sup-1 and/or clv3-2 mutant backgrounds.

Expression of the stem cell determinant WUS in the mutant floral buds

To visualize the stem cell niches in the floral meristems at stage 5, we introduced the pWUS::GFP-ER transgene (Gordon et al. 2007) into the single, double, and triple mutant backgrounds and observed the resulting GFP signal by confocal microscopy. Wild-type floral meristems at this developmental stage showed faint WUS expression in their organizing center (Fig. 4a), which was repressed by stage 6 (Supplementary Fig. 1a). In ag-1 and sup-1 floral meristems at stage 5, pWUS::GFP-ER expression levels were higher than those of the wild type (Fig. 4a–c). WUS reporter misexpression in clv3-2 was observed in its elongated floral meristem at stage 5 (Fig. 4d), which reflected its phenotypic abnormalities (Fletcher et al. 1999). Unlike the wild type, the ag-1, sup-1, and clv3-2 flowers still possessed dome-shaped floral meristems even after the inner structures of the flowers were completely covered by sepals and the WUS reporter was expressed in their presumed organizing centers (Supplementary Fig. 1c–d).

Fig. 4
figure 4

Location of WUS expression in the wild type and single, double, and triple mutants. a–h Longitudinal sections through the pWUS::GFP-ER floral buds in the wild type (a), ag-1 (b), sup-1 (c), clv3-2 (d), ag-1 sup-1 (e), ag-1 clv3-2 (f), sup-1 clv3-2 (g), and ag-1 sup-1 clv3-2 (h). (Left) Stage 5 floral buds. (Right) Higher magnification of floral meristems. Bar = 50 µm in a–h (left); 10 µm in a–h (right)

Full size image

We further investigated the WUS reporter expression pattern in the double and triple mutants. In the ag-1 sup-1 double mutant at stage 5, the WUS expression domain was longitudinally elongated compared with its parental lines, which meant that its expression continued into the rib meristem cells (Fig. 4e). By contrast, the WUS expression pattern in the ag-1 clv3-2 floral meristem was similar to that of clv3-2 (Fig. 4d, f). The morphological changes in the floral meristem heights and widths of sup-1 clv3-2 relative to clv3-2 were reflected in their differences in the reporter expression domain, which was elongated in both the longitudinal and vertical directions in sup-1 clv3-2 (Fig. 4d, g). Furthermore, the degree of GFP signal intensity in sup-1 clv3-2 seemed higher than that of the clv3-2 mutant. In the ag-1 sup-1 clv3-2 triple mutant, the pWUS::GFP-ER expression domain and signal levels were larger and stronger than those of any other genotype (Fig. 4h). Interestingly, the reporter was never expressed in the L1 and L2 layers of stage 5 floral meristems in any of the genotypes (Fig. 4a–h, right panels).

Identification of direct targets of AG with a likely role in floral meristem development

AG mainly regulates flower size during the later stages of development. The WUS reporter expression domain in the ag-1 sup-1 clv3-2 triple mutant was more than 10 times as wide as that of the sup-1 clv3-2 double mutant at developmental stage 10 (Fig. 5a, b), which largely coincided with the mass of undifferentiated cells observed in these flowers at later stages (Fig. 1j, l).

Fig. 5
figure 5

Identification of direct AG targets with a likely role in floral meristems. a, b Longitudinal section through the pWUS::GFP-ER floral buds in sup-1 clv3-2 (a) and ag-1 sup-1 clv3-2 (b). c A flowchart of the pipeline for identifying high-confidence AG targets. Target genes are likely to be regulated in floral meristems. d Venn diagram showing the number of genes differentially expressed in ag-1 sup-1 clv3-2 and sup-1 clv3-2, and their overlap with the direct targets of AG. e Gene ontology (GO) term enrichment analysis of 125 genes. The top 10 terms determined by their − log10-adjusted p values are shown. The false discovery rate (FDR) was lower than 1.0 × 10−7. f The TreeMap view of GO terms. A FDR correction was conducted and a FDR cutoff of less than 0.01 was implemented. The resulting GO terms that fulfilled this criterion were further minimized using REVIGO. g The interactive graph view generated with REVIGO. Bright and pale colors indicate lower and higher p values, respectively. The size of circles indicates the frequency of the GO term in the underlying GOA database

Full size image

To determine whether AG has activities in addition to repressing WUS expression during the termination of floral meristem activity, we performed an RNA-seq analysis using the two mutants with enlarged floral meristems (sup-1 clv3-2 and ag-1 sup-1 clv3-2) as genetic tools. More than 10 M reads per sample were obtained and mapped based on TAIR10 in each sample, and the number of reads per kilobase of transcripts per million sequence reads was calculated (Supplementary Table 2). This analysis identified 2105 differentially expressed genes in total (Fig. 5c, Supplementary Table 3). Among them, 96 and 2009 genes were downregulated and upregulated in sup-1 clv3-2, respectively. Next, we computationally identified the AG direct targets using a published ChIP-seq dataset (Ó’Maoiléidigh et al. 2013) and found that 5% of the upregulated genes and 31% of the downregulated genes (125 in total) were bound by AG (Supplementary Table 4, Supplementary Fig. 2). To examine the likely functions of these 125 genes, we tested for GO term enrichment among them using agriGO (Fig. 5d, Supplementary Table 2; Du et al. 2010; Tian et al. 2017). We identified 85 significantly enriched GO terms (false discovery rate < 0.01; Fig. 5e, Supplementary Table 5), of which the 10 most highly enriched terms were related to either transcription or flower development (Fig. 5e). After reducing the GO terms using REVIGO (Supek et al. 2011), the enriched terms included “floral whorl development,” “negative regulation of biological process,” “gene expression,” “reproduction,” “developmental process,” “multicellular organismal process,” and “response to UV” (Supplementary Table 6). The GO term “floral whorl development” includes “meristem development.” On the other hand, “negative regulation of biological process” includes processes that stop, prevent, or reduce the frequency, rate or extent of a biological process. Based on the REVIGO gene ontology web server (http://revigo.irb.hr/), “negative regulation of biological processes” contains 7 GO terms; “negative regulation of biological process,” “regulation of cellular process,” “regulation of biological process,” “nucleobase-containing compound metabolic process,” “RNA metabolic process,” “macromolecule biosynthetic process,” “regulation of transcription, DNA-templated,” and “regulation of metabolic process,” suggesting a negative role for AG on cellular, metabolic, or transcription process in the control of floral meristems. Two major networks were visualized; the first was related to both “floral whorl development” and “negative regulation of biological process,” while the second was related only to “negative regulation of biological process.”

Characterization of high-confidence AG targets in floral meristems

To further characterize the 125 AG target candidate genes during flower development, we used k-means clustering and publicly available gene expression datasets. Two transcriptome datasets, one from organ initial cells and one from floral organs (Jiao and Meyerowitz 2010; Frerichs et al. 2016), were used, and two gene clusters were identified. Cluster 1 contains 30 genes that were upregulated in wild-type organ initial cells but downregulated in sup-1 clv3-2 (thus were negatively regulated by AG; Fig. 6a, b). Real-time RT-PCR confirmed the downregulation of a selection of these genes in sup-1 clv3-2 compared with ag-1 sup-1 clv3-2 (Fig. 6c). This cluster contained a lot of the known regulators that promote the specification of organ initial cells and the formation of the floral meristem. Four of these genes are direct targets of auxin-dependent transcription factor MONOPTEROS (MP); ARABIDOPSIS HISTIDINE PHOSPHOTRANSFER PROTEIN6 (AHP6), TARGET OF MONOPTEROS3 (TMO3), AINTEGUMENTA-LIKE6 (AIL6), and LEAFY (LFY) (Yamaguchi et al. 2013; Besnard et al. 2014; Wu et al. 2015). Among the 30 genes negatively regulated by AG, the expression patterns of 10 genes (Fig. 6a asterisks) were previously examined using in situ hybridization or reporter genes and were found to be expressed in organ initial cells (Blazquez et al. 1997; Byzova et al. 1999; Samach et al. 1999; Nole-Wilson and Krizek 2006; Pastore et al. 2011; Yamaguchi et al. 2013; Besnard et al. 2014; Chandler and Werr 2014; Wu et al. 2015). In cluster 1, there were no clear gene expression changes that coincided with the expression domains of AP1, AP3, and AG between stage 4 and stage 6–7 (Fig. 6a, b).

Fig. 6
figure 6

Clustering of high-confidence direct AG targets in floral meristems. a A k-means clustering of genes directly regulated by AG. Heatmap displays the log2 expression changes of the 125 targets based on two public transcriptome datasets. Two large clusters were identified. Arrows indicate genes previously found to be expressed in floral primordia initial cells. b Gene expression from the two hierarchical clusters. One cluster contains 30 genes downregulated in sup-1 clv3-2. The other cluster contains the 95 genes upregulated in sup-1 clv3-2. c–e qRT-PCR verification of AG targets. Error bars represent SD of three PCR replicates of 3 biological samples. c Downregulated genes in sup-1 clv3-2. d Upregulated genes in ag-1 sup-1 clv3-2. e Meristem marker genes

Full size image

Cluster 2 contained 95 genes that were downregulated in the organ initial cells but upregulated in sup-1 clv3 (thus activated by AG; Fig. 6a, b). This cluster contained the well-known AG-activated targets CRABSCLAW (CRC) and SPOROCYTELESS (SPL), which validated our approach (Ito et al. 2004; Gomez-Mena et al. 2005). Furthermore, real-time RT-PCR analysis confirmed that a selection of these genes were upregulated in sup-1 clv3-2 in comparison with ag-1 sup-1 clv3-2 (Fig. 6d). Although we observed a clear negative relationship between the gene expression in the organ initial cells and the differentially expressed genes in sup-1 clv3-2, none of the genes in cluster 2 have yet been linked to organ initial cell specification or floral meristem formation, unlike the genes in cluster 1. Finally, the expression levels of the meristematic marker genes KNAT1 and BARELY ANY MERISTEM3 (Douglas et al. 2002; Depuydt et al. 2013) were similar between sup-1 clv3-2 and ag-1 sup-1 clv3-2 (Fig. 6e), thus confirming the specificity of our approach. Taken together, these findings suggest that AG negatively controls the production of organ initial cells, possibly through regulating the genes identified here.

Discussion

The roles of the AG, SUP, and CLV3 pathways in regulating floral meristem activity

Here, we revealed the genetic interactions between two flower-specific transcription factors, AG and SUP, and a secreted peptide, CLV3, which were originally identified using forward genetics approaches (Yanofsky et al. 1990; Clark et al. 1995; Sakai et al. 1995). Two pieces of evidence support the presence of this genetic interaction. First, phenotypic enhancement was observed when we introduced another mutation; the floral meristem of the ag sup clv3 triple mutant is bigger than that of the wild type or any other single or double mutant (Figs. 1, 2, 3). Second, the expression domain of the key stem cell determinant marker gene WUS corresponded with the size of the floral meristem (Fig. 4), and the triple mutant had the largest WUS reporter expression domain of any mutant used in this study. Thus, we concluded that AG, SUP, and CLV3 coordinately modulate the floral meristem activity in parallel pathways.

Although AG expression occurs throughout the floral meristem from stage 3 of flower development (Yanofsky et al. 1990), we did not observe an obvious difference between the wild type and the ag-1 single mutant in terms of the floral meristem height or width by stage 5 (Figs. 2, 3; Table 1). AG was previously reported to terminate floral meristem proliferation through WUS repression during stage 6 of flower development (Sun and Ito 2015), which explains why we found that the ag mutant flower at anthesis was clearly larger than that of the wild type (Fig. 1; Table 1). Thus, the ag mutation mainly affects the floral meristem and the resulting flower size from stage 5 onwards. Indeed, WUS was expressed at the center of the ag-1 flowers even after petals were formed (Fig. 5b). Our results indicate that AG mainly controls the timing of floral meristem termination, but does not contribute much to the size of the floral meristems in the early stages of flower development, or at least its role is complemented by other genetic pathways mediated by SUP and CLV. The effects of the ag mutation in floral meristem height at stages 3 and 5 were seen in the sensitized sup-1 clv3-2 double mutant background. This also supports the previous finding that AG directly binds to the WUS promoter and negatively regulates its expression at stages 4 and 5 (Liu et al. 2011).

Table 1 Summary of flower morphology the wild type and single, double, and triple mutants
Full size table

The SUP protein localizes at the boundary between the incipient stamens and carpels at stage 3 (Prunet et al. 2017). In stage 5 floral buds, the SUP expression domain expands to a groove between the developing stamen primordia and the floral meristem (Prunet et al. 2017). Since sup mutants have a wider floral meristem at stage 5 than the wild type, SUP plays a role in repressing the expansion of the floral meristem along the horizontal axis (Figs. 2, 3; Table 1). We observed a morphological difference between the wild-type and sup-1 floral buds at stage 5, but not at stage 3, suggesting that SUP inhibits floral meristem activity during stage 4 or 5. Unlike the ag-1 mutation, the introduction of the sup-1 mutation into a clv3-2 or ag-1 clv3-2 background was found to increase the height of the floral meristems even at stage 3, suggesting that SUP negatively controls the size of floral meristem from stage 3 together with AG and CLV3.

CLV3 is expressed in the stem cells of inflorescence meristems and floral meristems from stage 1 onwards (Gruel et al. 2016). As reported previously, the shape of the floral primordium in clv3-2 is different from that of the wild type even at stage 1 of development (Szczesny et al. 2009), and floral meristems in clv3-2 at stages 3 and 5 were wider and taller than those of the wild type (Figs. 2, 3; Table 1). Although the clv3-2 mutant has a huge floral meristem from the early stages of flower development, its open flower is no larger than that of the wild type, and the cells located at the enlarged floral meristems appear to turn into ectopic floral organs. Interestingly, we did not see an increased flower size in sup clv3 compared with its parental lines, demonstrating that AG prevents the enlargement of flowers in this mutant. This also indicates the presence of partially redundant pathways mediated by CLV3, SUP, and AG; both CLV3 and SUP spatially restrict the size of floral meristems, while AG mainly functions to temporally terminate floral meristem activities.

Possible roles of AG target genes in floral meristems

AG acts as a master regulator to execute its central role in the control of floral meristem activity, in large part by repressing the stem cell determinant gene WUS (Sun and Ito 2015). Our understanding of the precise function of floral-meristem-specific transcription factors is restricted by the very limited numbers of floral meristem cells in plants, which makes it difficult to isolate these cells without using specialized equipment such as laser microdissection or fluorescence-activated cell sorting. Our morphological analyses revealed that ag-1 sup-1 clv3-2 had a large number of floral meristem-like cells compared with sup-1 clv3-2 (Figs. 1, 2, 3, 4), and a total of 2105 differentially expressed genes were identified between these plants. The quantitative differences in the expression of upregulated genes in sup-1 clv3-2 are most likely due to tissue composition for the following reasons. (1) sup-1 clv3-2 formed more than 10 stamens per flower, while the ag-1 sup-1 clv3-2 mutant plants did not produce any stamens. (2) p value of each annotation in the top 10 GO terms using upregulated genes in sup clv3 (p < 1.0 × 10−35) is much lower than those using downregulated genes (p < 1.0 × 10−8). (3) Even though we identified more genes upregulated in sup clv3, only 95 out of 2009 (4.7%) are AG direct targets. By contrast, 30 out of 95 genes (31.5%) downregulated in sup clv3 are direct targets. These tissue composition issues could affect more than the loss of transcription factor activity; genes upregulated in sup-1 clv3-2 were involved in the later stages of flower development, mainly sporogenesis (Supplementary Fig. 3). By comparing these genes to the direct targets of AG, we narrowed down the number to 125 high-confidence AG targets likely acting in floral meristems (Fig. 5). Since a few of the known AG targets were identified using our filtering criteria (Ito et al. 2004, Gomez-Mena et al. 2005), our strategy seemed to be appropriate (Fig. 6).

A total of 35 of the 125 AG target genes are involved in transcription. This represents a significant enrichment (p = 2.0 × 10−19) of transcription factors in the AG target genes (28% of the AG target genes), as only 1533 (4.5%) of the 33,602 genes in the TAIR10 database encode transcription factors (Riechmann et al. 2000). We were not surprised that this list included four genes encoding MADS-box transcription factors, four homeodomain or homeodomain-like proteins, or four AP2-type transcription factors, since these families have important roles in flower development and/or meristem development (Ng and Yanofsky 2001; Tan and Irish 2006; Licausi et al. 2013). Although their mutant phenotypes have been characterized (Smith et al. 2004; Morita et al. 2006; Kumar et al. 2007; Hwang and Quail 2008; Magnani and Hake 2008), the roles of three homeobox genes (SAWTOOTH, KNATM, POUND-FOOLISH), SHOOT GRAVITROPISM5, and PHYTOCHROME INTERACTING FACTOR3-LIKE1 in the termination of floral meristem activity have not yet been examined. Furthermore, MYB DOMAIN PROTEIN110, AUXIN RESPONSE FACTOR11, REPRODUCTIVE MERISTEM11 (REM11) NAC DOMAIN CONTAINING PROTEIN79, VERDANDI, At1g26610, and At3g57370 are largely uncharacterized.

A link between AG and “floral whorl development” was found, as expected. The functions of all 21 AG target genes involved in flower development, i.e., PHABULOSA, SHATTERPROOF1 (SHP1), SHP2, HALF FILLED, SPL, HECATE1 (HEC1), HEC2, LFY, REM11, MIR167A, ROXY1, AGAMOUS-LIKE18 (AGL18), JAGGED, CRC, CUP-SHAPED COTYLEDON1, GA INSENSITIVE DWARF1B, FLOWERING LOCUS C, PROTODERMAL FACTOR 2 (PDF2), LMI2, STERILE APETALA (SAP), and AIL6, have already been examined (Weigel et al. 1992; Aida et al. 1997; Byzova et al. 1999; Michaels and Amasino 1999; Liljegren et al. 2000; Ito et al. 2004; Ohno et al. 2004; Gomez-Mena et al. 2005; Ueguchi-Tanaka et al. 2005; Adamczyk et al. 2007; Xing and Zachgo 2008; Crawford and Yanofsky 2011; Pastore et al. 2011; Kamata et al. 2013; Rubio-Somoza and Weigel 2013; Mantegazza et al. 2014; Schuster et al. 2015; Yamaguchi et al. 2016). However, the molecular links between AG and some of those genes are not fully understood. Interestingly, while it was reported that PDF2, LFY, UNUSUAL FLORAL ORGANS (UFO), AIL6, and SAP act upstream of AG (Byzova et al. 1999; Chae et al. 2008; Krizek 2009; Winter et al. 2011; Kamata et al. 2013), we found that these five genes also act as downstream targets of AG. This suggests that AG is involved in a regulatory feedback mechanism, which is reasonable considering that many master regulators are regulated by their own downstream targets (Kaufmann et al. 2010).

AG is required for floral determinacy, and ag flowers only occasionally give rise to an inflorescence structure. Our findings, combined with previous reports and the direct and indirect WUS repression by AG, suggest that AG represses not only the stem cell determinant genes, but also a large set of genes specifically expressed in organ initial cells, such as DRNL, AHP6, LFY, AIL6, TMO3, LMI2, and ROXY1 (Nag et al. 2007; Xing and Zachgo 2008; Pastore et al. 2011; Yamaguchi et al. 2013; Besnard et al. 2014; Wu et al. 2015). Four of these genes are direct targets of MP, which is key to produce floral organ founder cells. But the MP mRNA level was similar between sup-1 clv3-2 and ag-1 sup-1 clv3-2 based on RNA-seq; thus, this repression by AG is achieved without changing the transcriptional levels of MP. During floral meristem formation, the proper coordination of organ initial cell division is key for the creation of organ primordia. For the termination of floral meristem activity, the generation of organ initial cells around the floral meristem must be attenuated. Our expression profiling suggests that AG functions to maintain the irreversible state of reproductive development through the negative regulation of floral meristem identity genes and genes involved in organ initiation. Further studies are needed to elucidate the negative regulation of organ initial cell generation during floral meristem termination.