AGLF provides C-function in floral organ identity through transcriptional regulation of AGAMOUS in Medicago truncatula

Significance

Organogenesis is of great importance in both plant and animal development. Flowers have been studied exclusively to illustrate the mechanism underlying floral organ differentiation, which is controlled by the well-known ABC model. In this study, we identified a MYB-like gene in Medicago truncatula, AGAMOUS-LIKE FLOWER (AGLF), the null mutants of which developed homeotic transformation of stamens and carpel into petals and sepals. The relationships of AGLF with conventional ABC genes illustrated by the genetic and expression assays indicate that AGLF plays a C-function role in floral organ identity. Our work not only identifies a regulator of C-function genes, but also demonstrates the necessity of exploring the ABC models in other plant species with derived flower patterns.

Abstract

Floral development is one of the model systems for investigating the mechanisms underlying organogenesis in plants. Floral organ identity is controlled by the well-known ABC model, which has been generalized to many flowering plants. Here, we report a previously uncharacterized MYB-like gene, AGAMOUS-LIKE FLOWER (AGLF), involved in flower development in the model legume Medicago truncatula. Loss-of-function of AGLF results in flowers with stamens and carpel transformed into extra whorls of petals and sepals. Compared with the loss-of-function mutant of the class C gene AGAMOUS (MtAG) in M. truncatula, the defects in floral organ identity are similar between aglf and mtag, but the floral indeterminacy is enhanced in the aglf mutant. Knockout of AGLF in the mutants of the class A gene MtAP1 or the class B gene MtPI leads to an addition of a loss-of-C-function phenotype, reflecting a conventional relationship of AGLF with the canonical A and B genes. Furthermore, we demonstrate that AGLF activates MtAG in transcriptional levels in control of floral organ identity. These data shed light on the conserved and diverged molecular mechanisms that control flower development and morphology among plant species.

Organ differentiation is crucial in the development of living organisms, such as limb-bud growth in mammals, body-segment development in insects, and flower-organ identity in plants. Most flowers of higher plants consist of four concentric whorls of floral organs: sepals, petals, stamens, and innermost carpel(s). Floral organs have distinct morphologies and functions. The outer sepals and petals provide protection to the reproductive organs inside, i.e., the stamens and carpels that produce gametophytes. The developmental process of four whorls of flower is controlled by the organ identity genes in the well-known ABC model (1–4). In Arabidopsis thaliana, sepals are controlled by the class A genes APETALA 1 (AP1) and AP2; petals are controlled by the class A genes and class B genes AP3 and PISTILLATA (PI); stamens are controlled by the class B gene and class C gene AGAMOUS (AG); and carpels are controlled by the class C gene only (5–9). In another model plant, Antirrhinum majus, similar genes have been identified, such as SQUAMOSA, LIPLESS1, and LIPLESS2 as class A genes, DEFICIENS and GLOBOSA as class B genes, and PLENA and FARINELLI as class C genes (10–12). All of the ABC genes except AP2 and its homologs belong to the MADS (MCM1, AGAMOUS, DEFICIENS A, and SERUM RESPONSE FACTOR) homeotic gene family (13). The knockout mutants of the ABC genes display homeotic transformation of the floral organs in corresponding whorls in Arabidopsis. The mutation of AP1 results in the transformation of sepals into carpels and petals into stamens, while the pi mutant causes petals and stamens to be transformed into sepals and carpels. The ag mutant develops indeterminate flowers with stamens and carpel transformed into numerous whorls of petals and sepals.

The ABC model has been well adopted in various plant species. However, given the great variations of flower morphologies in higher plant species, new regulators have been constantly added to the classic ABC model. Hence, it is necessary to explore new regulatory genes in other plant species and expand the ABC model. As the model legume species, Medicago truncatula develops bilaterally symmetric flowers, which are significantly different from the radially symmetric flowers in Arabidopsis. The M. truncatula genome contains the canonical ABC genes (14). The knockout mutant of the class A gene MtAP1 caused mosaic organs from sepals and petals (15). The mtpi mutant had petals and stamens transformed into sepals and carpels (16). There are two AG-lineage members in M. truncatula, MtAGa and MtAGb (MtAGs). The mtaga or mtagb single mutant had only a mild defect in flower development, but the knockdown of both genes via RNA interference showed that the stamens and carpel transformed into petals and sepals (17).

In this study, we characterize a gene involved in floral organ identity in M. truncatula. The flowers in knockout mutants displayed the loss-of-C-function phenotype and showed enhanced indeterminacy compared with those in the mtag mutant. Therefore, we named this mutant agamous-like flower (aglf). Loss-of-function of AGLF led to the differential expression of canonical ABC genes and resulted in an addition of loss-of-C-function phenotype in mtap1 or mtpi mutants. Furthermore, AGLF activated the expression of the class C genes MtAGa and MtAGb. Our work discovered a regulator that acts upstream of the class C gene in control of floral identity in M. truncatula.

Results

Identification of the Mutants with Loss-of-C-Function Floral Defects.

Different from Arabidopsis’s radially symmetric flowers, M. truncatula develops bilaterally symmetric ones (18). M. truncatula’s flowers have four whorls of floral organs, including 5 sepals, 5 petals, and 10 stamens fusing into a staminal tube surrounding one carpel inside (Fig. 1A and SI Appendix, Fig. S1E). The bilateral symmetry in petals, which consist of one vexillum, two alae, and two merged keels, is apparent (SI Appendix, Fig. S1E). To study the mechanism underlying flower development in M. truncatula, we carried out a forward genetic screen of nearly 22,000 Tnt1 retrotransposon-tagged mutant collections (19). Four mutant lines with the defective flowers were isolated (Fig. 1B and SI Appendix, Fig. S1 B–D). The numbers of petals and sepals in mutants were increased significantly, compared with those in wild type. But the stamen and carpel were lost in mutants (Fig. 1C and SI Appendix, Fig. S1 E and F). Further observation showed that the flowers of the mutants had normal sepals in the outmost whorl. In the second whorl of flower, the petals lost bilateral symmetry and displayed abnormal shapes. The stamens and carpel in the inner whorl were replaced by the extra whorls of petals and sepals (Fig. 1 D–F). Such indeterminate flowers resembled the floral phenotype of the loss-of-C-function ag mutant. Therefore, we named them agamous-like flower (aglf-1 to aglf-4).

Fig. 1.

The aglf mutant displays an indeterminate flower phenotype. (A and B) The wild-type (WT) and aglf-1 flowers. (C) The number of sepals, petals, stamens, and carpels in WT and aglf-1. The data represent means ± SD (n = 25). (D and E) Longitudinal sections of the WT and aglf-1 flowers stained with toluidine blue. (F) A schematic illustration of floral organs in WT and aglf-1. The wild-type flower consists of four whorls of organs: sepal (green), petal (purple), stamen (orange), and carpel (blue). An aglf-1 flower consists of sepals in the first whorl, petals in the second and third whorls, and reiterations of this pattern in interior whorls. (G) SEM analysis of floral meristem in WT and aglf-1. CPab is marked in yellow, and petal primordia are marked in green. [Scale bars: (A and B) 2 mm; (D and E) 200 μm; (G) 100 µm.] A, alae; C, carpel; CP, common primordia; CPab, abaxial CP; CPad, adaxial CP; K, keel; P, petal; S, sepal; Sab, abaxial sepal; Sad, adaxial sepal; Sl, lateral sepal; ST, stamens; V, vexillum. Asterisk indicates the indeterminate primordium in aglf-1.

To investigate how aglf-1 developed such indeterminate flowers, we examined the mutant floral meristem (FM) with scanning electron microscopy (SEM) (Fig. 1G). The development of FM in the wild type and aglf-1 was the same until stage 3 when five sepal primordia initiated. The differences between wild type and aglf-1 started to emerge at stage 4. In wild type, one carpel primordium and four common primordia (CPs) initiated, and the abaxial CP (CPab) began to differentiate into stamens and petal primordia (indicated by the highlighted areas in Fig. 1G). In contrast, the initiation of carpel primordium and CPs was delayed in aglf-1. At stage 5, all of the CPs differentiated into stamen and petal primordia in the wild type, while the CPab and an indeterminate primordium (indicated by asterisks in Fig. 1G) started to initiate in aglf-1. At stages 6 and 7, all of the floral organ primordia had been formed in the wild type. In aglf-1, the abaxial side of floral primordia irregularly differentiated into numerous petals and sepal primordia, and the adaxial side stayed indeterminate. At stage 8, the anthers and the inflection of the carpel were visible in the wild type; however, there were still no stamens and carpel differentiated in aglf-1. The indeterminate primordium continued to differentiate into numerous petals and sepals, instead of carpel, leading to the indeterminate flowers with homeotic transformation in aglf-1 (Fig. 1G).

AGLF Encodes a Regulator Involved in Floral Identity.

To confirm that the mutant phenotype was caused by Tnt1 insertion in a single gene, the aglf-1 mutant was crossed with the wild-type plants. In a segregating F2 population, the wild-type–like and mutant plants showed a segregation ratio of 3:1 (96:35), suggesting that the mutant phenotype was controlled by a single recessive gene. To identify the gene responsible for the floral defects in aglf-1, thermal asymmetric interlaced–PCR was performed to recover the flanking sequence tags of Tnt1. The genotyping assays led to a single Tnt1 insertion that cosegregated with the defective floral phenotype in the aglf-1 segregating populations. The Tnt1 insertion was located in the first exon of a previously uncharacterized gene, Medtr0007s0220 (Fig. 2A). Further analysis showed that a Tnt1 was inserted in the first exon of the same gene in aglf-3 and aglf-4 and in the first intron in aglf-2, respectively, and the transcription of Medtr0007s0220 was abolished in all four aglf alleles (SI Appendix, Fig. S2A and Table S1). To complement the mutant phenotype, we cloned the coding sequence of Medtr007s0220 and introduced it into the aglf-1 plants. Phenotypic observation confirmed that complementary Medtr0007s0220 expression fully rescued the floral defects in aglf-1 (Fig. 2B and SI Appendix, Fig. S2 B and C). We therefore named Medtr0007s0220 as AGLF. AGLF encodes a 1,007-aa protein containing a MYB-like DNA-binding domain at its N terminus and a truncated kinase domain at the C terminus (SI Appendix, Fig. S3). AGLF did not belong to any known regulators of floral organ identity. Then, we searched the National Center for Biotechnology Information protein sequence database and found 60 proteins in dicotyledonous species that shared high similarities with AGLF. The phylogenetic analysis showed that AGLF and the homologs from legumes formed a distinct clade, which in turn grouped with those from Brassica species (SI Appendix, Fig. S4). To explore the function of AGLF in other species, a phylogenetic tree was constructed using all proteins with identities higher than 30% to AtAGLF and AGLF, including eight proteins in Arabidopsis and seven proteins in M. truncatula. (SI Appendix, Fig. S5A). The result showed that AGLF only has one putative ortholog, At5g51800, in Arabidopsis. However, the knockout mutant of this gene showed no phenotypes in floral organ identity (SI Appendix, Fig. S5 B–E).

Fig. 2.

Molecular characterization of AGLF. (A) The scheme of AGLF gene structure and Tnt1 insertions in the gene. Boxes represent exons, and lines represent introns. Four arrows indicate the positions where Tnt1 inserts in each of the mutant alleles. (B) Restoration of AGLF expression under a constitutive CaMV 35S promoter (35Spro:AGLF) complements the indeterminate flowers in aglf-1. (C) Nuclear localization of the AGLF-YFP fusion protein in tobacco mesophyll cells. (D) In situ hybridization detects AGLF expression in floral meristem at different stages. [Scale bars: (B) 2 mm; (C) 20 μm; (D) 50 µm.] C, carpel; CP, common primordium; FM, floral meristem; P, petal; SE, sepal; ST, stamens.

The possession of the MYB-like DNA-binding domain suggests a possible role of AGLF as a transcriptional regulator in flower development. A transient expression of YFP-tagged AGLF in Nicotiana benthamiana leaves showed that the recombinant AGLF-YFP protein was localized in nucleus (Fig. 2C). Then, we performed an assay to test if AGLF exhibited transcriptional activity using a previously established procedure (20). The AGLF-GAL4BD fusion protein was transiently expressed in Arabidopsis protoplasts along with a 35Spro:UAS-LUC reporter gene. The AGLF-GAL4BD displayed the transactivation activity to some extent, based on the activation of luciferase activity in the protoplasts (SI Appendix, Fig. S6). The nuclear localization and transactivation activity of AGLF indicate that it may function as a transcriptional activator in floral organ identity.

AGLF Is Expressed Specifically in the Inner Two Whorls of Floral Organs.

To explore the tempo-spatial expression pattern of AGLF, we conducted a series of expression assays. We first examined the expression of AGLF in various organs via qRT-PCR and found that it was ubiquitously expressed with the highest expression level in flowers (SI Appendix, Fig. S7). The tempo-spatial expression pattern of AGLF during floral organ initiation and differentiation was further examined by in situ hybridization analysis in the wild type (Fig. 2D). The expression of AGLF was observed in the center of the FM at the early stages (stages 1–3). Its expression was confined in the CP and the central carpel primordium at stages 4 and 5. The expression of AGLF was restrained in the stamen and carpel primordia when floral organ primordia completed differentiation at the late stages (stages 6–8). The preferable expression pattern of AGLF in the inner two whorls of floral organs aligns with the loss-of-C-function phenotype in the aglf mutants, implying that AGLF is probably involved in the regulation of the class C gene in M. truncatula.

Loss-of-Function of AGLF Leads to the Dysregulation of ABC Genes.

To understand the regulatory role of AGLF in flower development, we carried out a transcriptomic analysis of aglf-1 using RNA sequencing. To eliminate any potential confounding effects from other mutations in the aglf-1 background, the floral buds from heterozygous plants were used as the control. Compared with the wild-type–like flowers, a total of 112 genes were up-regulated and 1,383 genes were down-regulated in aglf-1 (Dataset S1). Furthermore, the MADS family was significantly enriched in the down-regulated genes (Dataset S2). Among these MADS transcription factors (TFs), canonical ABC genes displayed differential expression, which was confirmed by qRT-PCR (Fig. 3A). Compared with wild type, two class C genes, MtAGa and MtAGb, were both significantly down-regulated in the floral buds of aglf-1, while class A gene MtAP1 and class B gene MtPI were up-regulated in aglf-1. The homolog of UNUSUAL FLORAL ORGANS (UFO) in M. truncatula (MtUFO) that regulates B genes in Arabidopsis (21) also showed elevation in its transcript level in aglf-1.

Fig. 3.

The molecular and genetic relationship between AGLF and ABC genes. (A) The expression levels of MtAP1, MtPI, MtAGa, MtAGb, and MtUFO in wild-type (WT) and aglf-1 flowers determined by qRT-PCR. MtUBIQUITIN was used as the internal control. Error bars represent the SD from three biological replicates. Columns labeled with asterisks indicate significant differences from those in WT (**P < 0.01, t test). (B) In situ hybridization of genes listed in A in FM of WT and aglf-1. A petal primordium is indicated by a dashed curve line in B. (C and D) Flower phenotype (C) and SEM analysis of floral meristem (D) of WT and aglf-1, mtaga-1 mtagb-2, mtap1-1, mtap1-1 aglf-2, mtpi-4, and mtpi-4 aglf-1 mutants. [Scale bars: (B) 50 µm; (C) 2 mm; (D) 20 µm.] C, carpel; CP, common primordium; P, petal; SE, sepal; ST, stamens. Asterisk shows indeterminate primordium.

The canonical ABC genes were under stringent spatial regulation during flower development, while our expression data exhibited the dysregulation of ABC genes in aglf-1. Therefore, their tempo-spatial expression patterns were further examined in detail in the FM of the wild type and aglf-1 (Fig. 3B). MtAGa and MtAGb expressions were restrained in the inner two whorls of stamen and carpel primordia in the wild type, while their expression was barely detected in aglf-1. Moreover, MtAP1 was expressed in the outer whorls of sepal and petal primordia, and MtPI expression was detected in the middle whorls of petal and stamen primordia in the wild type. However, both genes were expressed aberrantly across the inner whorls of primordia in aglf-1. The expression of MtUFO that provides B-function, like MtPI, also spread out into the center of the FM in aglf-1. These results demonstrate that the knockout of AGLF leads to ectopic expression of MtAP1 and MtPI across the FM and to reduced MtAGs expression in the center.

AGLF Provides C-Function in Flower Development.

In a previous study, the expression levels of MtAGa and MtAGb were merely suppressed via virus-induced gene silencing (17). To better characterize the relationship between AGLF and class C gene MtAGs in M. truncatula, we identified the Tnt1-tagged mtaga-1 and mtagb-2 knockout mutants (SI Appendix, Fig. S8 A and B). The mtaga-1 and mtagb-2 single mutants developed mostly wild-type flowers with minor transformation defects on carpel and stamens (SI Appendix, Fig. S8C). In mtaga-1 mtagb-2 double mutants, the inner two whorls were transformed into whorls of sepals and petals in an indeterminate manner (Fig. 3C and SI Appendix, Fig. S8 C and D), indicating the conserved C-function roles of MtAGs in M. truncatula. Then the aglf-1 mtaga-1 mtagb-2 triple mutant was generated. The flower phenotype and floral organ formation in the triple mutant were indistinguishable from those in aglf-1 (SI Appendix, Fig. S9 A–E). Furthermore, the numbers of sepals and petals between aglf-1 mtaga-1 mtagb-2 and aglf-1 were comparable, which were significantly higher than those in mtaga-1 mtagb-2 (SI Appendix, Fig. S9F). These results indicate that AGLF is genetically epistatic to MtAGs in terms of flower development and that the loss-of-function of AGLF leads to the enhanced floral indeterminacy compared with that in mtag mutant.

To further understand the relationship between AGLF and other ABC genes, genetic analysis was performed. As described previously, the flowers of loss-of-A-function mtap1-1 showed loss of keels and sepal-petal mosaic organs (Fig. 3C and SI Appendix, Fig. S10 A–C). The aglf-2 mtap1-1 double mutant developed indeterminate flowers with mosaic organs of sepal and petal tissues (Fig. 3C and SI Appendix, Fig. S10 D–F), which reflected a combination of loss-of-A- and C-function. Moreover, we identified a mutant line of the class B gene MtPI, mtpi-4, in which a Tnt1 insertion in the second exon of MtPI completely knocked out the gene expression (SI Appendix, Fig. S11 A and B). The mtpi-4 flowers had the petals and stamens transformed into sepals and carpels, respectively, displaying a typical loss-of-B-function phenotype (Fig. 3C and SI Appendix, Fig. S11 C and D). The aglf-1 mtpi-4 double mutant flowers exhibited a loss-of-B- and C-function phenotype with all of the floral organs transformed into sepals (Fig. 3C and SI Appendix, Fig. S11 E and F). SEM analysis confirmed the phenotypic observations in the mutants (Fig. 3D). In both mtap1-1 and mtpi-4 single mutants, the carpel primordia were initiated and differentiated. However, the carpel primordia were replaced by the indeterminate primordia in double mutants, which differentiated extra whorls of transformed petals and sepals in aglf-2 mtap1-1 (SI Appendix, Fig. S10 G–R) and sepals in aglf-1 mtpi-4 (SI Appendix, Fig. S11 G–N), respectively. These genetic evidences demonstrate that loss-of-function of AGLF in the mutant background of class A or B genes leads to an addition of loss-of-C-function phenotype, suggesting that AGLF provides a C-function in floral organ identity.

LEAFY (LFY) is a key regulator in FM initiation and able to regulate the expression of ABC genes in floral organ primordia via forming a complex with other regulators (22, 23). Its homolog in M. truncatula, SINGLE LEAFLET1 (SGL1) (24), shares the same functionality as the sgl1 mutant has all of the floral organs transformed into sepals (SI Appendix, Fig. S12 A, B, and E–H). The expression of SGL1 was induced in aglf-1 (Dataset S1). Moreover, the flowers of aglf-1 sgl1-2 double mutant resembled those of the sgl1-2 single mutant, indicating that SGL1 has an epistatic role to AGLF (SI Appendix, Fig. S12 C, D, and I–L).

AGLF Functions in Floral Organ Identity Through Transcriptional Regulation of MtAGs.

To further investigate the roles of AGLF in floral identity, a series of assays were carried out. First, the expression levels of MtAP1 and MtPI were measured in mtaga-1 mtagb-2 and aglf-1 mtaga-1 mtagb-2 (SI Appendix, Fig. S13A). The results showed that their expressions were up-regulated in both double and triple mutants, which are similar to those in aglf-1, indicating the similar regulatory roles between MtAGs and AGLF. Second, the expression level and the pattern of AGLF were detected in mtaga-1 mtagb-2. In contrast to low expression of MtAGs in aglf-1, AGLF displayed a similar expression level in the mtaga-1 mtagb-2 double mutant as in the wild type (SI Appendix, Fig. S13A). Furthermore, the in situ hybridization assay showed that the AGLF transcript was detected in the central primordia in the double mutant (SI Appendix, Fig. S13B). These results suggest that the expression of AGLF is independent of MtAGs activity.

Given that AGLF functions as a transcription activator, transient expression assays were performed to investigate if AGLF is capable of activating MtAGs. A GFP-tagged AGLF effector and the luciferase reporter constructs driven by the 2-kb promoter of MtAGa or MtAGb were transiently coexpressed in Arabidopsis protoplasts, resulting in a significant increase in the luminescence intensity compared with the control. Another reporter construct driven by a 2.5-kb promoter of SMOOTH LEAF MARGIN1 (SLM1) (25) was used as a negative control since the SLM1 expression showed no obvious change in aglf-1. Coexpression of the AGLF effector with the SLM1 reporter construct showed no difference from the control (Fig. 4A). These results indicate that AGLF recognizes MtAGs promoters in protoplasts and activates their expression in a direct or indirect way. Then the expression levels of MtAGa and MtAGb were assessed by qRT-PCR in the aglf-1;35Spro:AGLF transgenic plants. The data showed that the expression levels of MtAGa and MtAGb were up-regulated by two- to four-fold in the rescue lines (Fig. 4B and SI Appendix, Fig. S2C), confirming that AGLF regulates MtAGs at the transcriptional level. Furthermore, the Chromatin Immunoprecipitation (ChIP) assay was carried out to determine if AGLF directly binds to MtAGs promoters in vivo. Using aglf-1;35Spro:AGLF-YFP transgenic plants, ChIP-PCR data showed that two MYB-binding regions in the promoter of MtAGa were enriched in AGLF-YFP chromatin, but the potential binding sites in the promoter of MtAGb were not (Fig. 4C and SI Appendix, Fig. S14 C and D). These results suggest that AGLF directly activates MtAGa expression in vivo, most likely by binding to the two GT motifs in its promoter. As for MtAGb, a different regulation mechanism is probably involved. For example, AGLF binds to unknown motifs of the MtAGb promoter, or the coregulator may be required for the transcriptional activation of MtAGb.

Fig. 4.

Overexpression of MtAGa and MtAGb partially restores determinacy in aglf-2 flowers. (A) The transcriptional activation by AGLF on the promoters of MtAGa and MtAGb. The LUC/REN luminescence was normalized to the value of the control of GFP. The negative control, SLM1, was included. Error bars represent the SD from three biological replicates. Columns labeled with asterisks indicate significant differences from those in WT (*P < 0.05, t test). (B) The expression levels of MtAGa and MtAGb in wild-type (WT) and two aglf-1;35Spro:AGLF transgenic lines, #7 and #8, determined by qRT-PCR. MtUBIQUITIN was used as the internal control. Error bars represent the SD from three biological replicates. Columns labeled with asterisks indicate significant differences from those in WT (**P < 0.01; *P < 0.05, t test). (C) Quantitative ChIP-PCR analysis of AGLF binding to the promoter of MtAGa. MtUBIQUITIN was used for normalization. Error bars represent the SD from three biological replicates. Columns labeled with asterisks indicate significant differences from those in WT (*P < 0.05, t test). (D–J) The flowers of wild type (WT, D), aglf-2 (G), 35Spro:MtAGa in aglf-2 (E), 35Spro:MtAGb in aglf-2 (F), and 35Spro:MtAGa 35Spro:MtAGb in aglf-2 (H–J). The dissected floral organs are shown in I. The Inset in I shows a small defective flower developed in the place of the carpel inside the staminal tube, which is further dissected in J. (Scale bars, 2 mm.) C, carpel; P, petal; S, sepal; ST, staminal tube.

To investigate if the aglf mutant phenotype was caused by the down-regulation of MtAGs, MtAGa and MtAGb were overexpressed in the wild type to generate the stable transgenic plants (SI Appendix, Fig. S15). Both 35Spro:MtAGa and 35Spro:MtAGb plants produced slim and curly sepals and petals (SI Appendix, Fig. S15 B and E). Then the 35Spro:MtAGa and 35Spro:MtAGb were introduced into the aglf-2 by crossing. Phenotypic observation showed that overexpression of either MtAGa or MtAGb had no effect on homeotic transformation of floral organs in aglf-2 except for slimmer sepals and petals (Fig. 4 D–G). However, simultaneous overexpression of MtAGa and MtAGb in aglf-2 partially rescued the defects in flowers (Fig. 4H). The flowers developed with restored determinacy, which consisted of sepals as the first whorl, petals as the second whorl, and a staminal tube as the third whorl (Fig. 4I). In addition, an ectopic determinate flower was developed inside of the staminal tube, which was composed of a modified carpel with stigmatic anthers (Fig. 4J). These results suggest that AGLF plays a role in floral organ identity at least partially through the regulation of both MtAGa and MtAGb at the transcriptional level.

Discussion

As an attempt to discover regulatory elements in flower development, we identified a MYB-like gene, AGLF. Four aglf mutant alleles developed the indeterminate flowers with homeotic transformation of stamens and carpel, similar to the phenotype of the mtag mutant. Knockout of AGLF in mtap1 or mtpi mutants led to the addition of a loss-of-C-function phenotype. Together with the expression pattern and molecular function, our data demonstrate that AGLF provides C-function in floral organ identity in M. truncatula (SI Appendix, Fig. S16).

In Arabidopsis, both A- and B-function are supplied by multiple genes, but AG is the only C-function gene identified to date. Therefore, characterization of the potential regulators of AG is critical for the elaboration of the ABC model in flower development. So far, only a few noncanonical genes are involved in the regulation of AG. For example, HUA1 and HEN4 encode RNA-binding proteins that function in the processing of AG pre-mRNA, and hua1 hua2 hen4 triple mutant displays strong loss-of-C-function floral defects (26). Different from HUA and HEN in Arabidopsis, the single-gene mutant of AGLF shows enhanced floral organ indeterminacy in M. truncatula. In our experiments, AGLF and MtAGs displayed the similar tempo-spatial expression patterns in the FM of wild type. The expressions of MtAGs were barely detected in the FM of aglf, while the transcriptional level of AGLF was unaffected in the mtaga mtagb double mutant (SI Appendix, Fig. S13). These data suggest that MtAGs function downstream of AGLF, implying that AGLF regulates these two genes at the transcriptional level (SI Appendix, Fig. S16).

Additionally, the transgenic plants overexpressing MtAGa or MtAGb produced slim and curly floral organs (SI Appendix, Fig. S15). However, the aglf-1;35Spro:AGLF transgenic plants showed normal flowers, although the expression of MtAGs was increased. The different floral phenotypes between MtAGs overexpression lines and aglf rescue lines are probably due to the different expression levels of MtAGs, implying the possible gene dosage effects of MtAGs. This hypothesis is also supported by a recent report (27). The stamens of mtaga (^+/−) mtagb (^−/−) were transformed into petaloids, while the carpels were frequently kept. The carpels of mtaga (^−/−) mtagb (^+/−) were transformed into sepaloids, while most stamen filaments were kept. These observations suggest that MtAGa and MtAGb may function redundantly and distinctly in floral-organ identity specification in a dose-dependent manner. In addition, another possibility is that AGLF and MtAGs antagonistically regulate the same downstream genes that are responsible for the slim floral organs in MtAGs overexpression lines.

Although the aglf mutant had characteristics of a canonical loss-of-C mutant, it displayed more homeotically transformed floral organs than the canonical mtaga mtagb double mutant, implying enhanced indeterminacy in aglf flowers. The mtaga mtagb aglf triple mutant developed a similar number of floral organs as aglf, reflecting an epistatic role of AGLF over MtAGs. Therefore, AGLF provides C-function by regulation of MtAGs, but might have a broader role in floral organ identity. According to these observations, the transcriptional analysis in this study reveals that numerous genes of the MADS family are affected by AGLF (Dataset S2). Among them, the class A gene MtAP1 and class B gene MtPI show the ectopic expressions, which is consistent with the homeotic transformation of floral organs in aglf (Fig. 3). In addition to the ABC genes, the genes involved in FM initiation, such as SGL1/MtLFY, were affected in aglf. These results suggest that AGLF acts as a key regulator to regulate a number of genes involved in flower development.

Taken together, our study discovered the gene AGLF, which regulates MtAGs in determination of stamen and carpel identity in M. truncatula. Our findings add a player to the conventional ABC model, suggesting the involvement of additional floral regulators among plant species.

Materials and Methods

Plant materials and growth conditions, phenotype analyses, plasmid construction, and the generation of transgenic plants are described in SI Appendix, SI Materials and Methods. The detailed procedures of subcellular localization, quantitative real-time PCR, RNA sequencing, in situ hybridization, SEM, transactivation assay, and ChIP assay are provided in SI Appendix, SI Materials and Methods.