The establishment and maintenance of organ boundaries are vital for animal and plant development. In the Arabidopsis flower, three microRNA164 genes (MIR164a, b and c) regulate the expression of CUP-SHAPED COTYLEDON1 (CUC1) and CUC2, which encode key transcriptional regulators involved in organ boundary specification. These three miR164 genes are expressed in distinct spatial and temporal domains that are crucial for their function. Here, we show that the C2H2 zinc finger transcriptional repressor encoded by RABBIT EARS (RBE) regulates the expression of all three miR164 genes. Furthermore, we demonstrate that RBE directly interacts with the promoter of MIR164c and negatively regulates its expression. We also show that the role of RBE in sepal and petal development is mediated in part through the concomitant regulation of the CUC1 and CUC2 gene products. These results indicate that one role of RBE is to fine-tune miR164 expression to regulate the CUC1 and CUC2 effector genes, which, in turn, regulate developmental events required for sepal and petal organogenesis.
The formation of boundaries between different cell populations is crucial for development. Such boundaries, in addition to delineating identity, often act as organizing centers to regulate downstream signaling events (Dahmann et al., 2011). In Arabidopsis, a subset of NAM-ATAF-CUC (NAC) domain transcription factors plays a central role in the establishment and maintenance of organ boundaries. These partially redundant gene products are expressed at boundaries, and their loss of function results in organ fusions and defects in both vegetative and floral development (Aida et al., 1997; Aida et al., 1999; Aida and Tasaka, 2006; Takada et al., 2001; Vroemen et al., 2003). Two such NAC domain genes, CUC1 and CUC2, are post-transcriptionally regulated by microRNA164 (miR164) (Baker et al., 2005; Laufs et al., 2004; Mallory et al., 2004; Rhoades et al., 2002; Sieber et al., 2007). Negative regulation by miR164 fine-tunes the levels, as well as patterns of expression, of the CUC1 and CUC2 transcripts. Precise regulation of CUC1 and CUC2 transcript accumulation is crucial for proper control of organ number and boundary formation throughout vegetative and reproductive development (Baker et al., 2005; Laufs et al., 2004; Mallory et al., 2004; Sieber et al., 2007).
The microRNA miR164 is transcribed from three loci: MIR164a, MIR164b and MIR164c (Bonnet et al., 2004; Jones-Rhoades and Bartel, 2004; Reinhart et al., 2002; Wang et al., 2004). These genes exhibit partial functional redundancy, particularly in flower development (Baker et al., 2005; Sieber et al., 2007). Loss of function of the MIR164c gene, also known as EARLY EXTRA PETALS1 (EEP1), results in the growth of extra petals due to the failure to properly repress CUC1 and CUC2 in the second whorl (Baker et al., 2005). In the mir164abc triple mutant, the floral defects exhibited by eep1 are enhanced. Nevertheless, the mir164a and mir164b single mutants do not show obvious floral phenotypes (Sieber et al., 2007). These observations indicate that the miR164 genes are functionally redundant, yet differences in their expression patterns and mutant phenotypes imply that they have unique roles as well. As a consequence, deciphering the precise regulation of the spatial and temporal domains of miR164 gene expression is pivotal to revealing the functions of these microRNAs in organ boundary formation.
The gene RABBIT EARS (RBE) has been postulated to be involved in defining second whorl boundaries, and is specifically expressed in petal primordia during early stages of petal development (Krizek et al., 2006; Takeda et al., 2004; reviewed in Irish, 2008). RBE encodes a putative C2H2 zinc finger transcriptional repressor and, thus, is likely to function as a negative regulator (Takeda et al., 2004). Loss-of-function rbe mutants exhibit loss of or aberrant petals, as well as fused sepals (Krizek et al., 2006; Takeda et al., 2004). Interestingly, these morphological defects resemble the phenotypes observed in cuc1 cuc2 double mutants (Aida et al., 1997) and in 35S:MIR164b transgenic plants (Baker et al., 2005; Laufs et al., 2004; Mallory et al., 2004; Rhoades et al., 2002; Sieber et al., 2007), implying that RBE might function in regulating miR164-mediated processes.
In this study, we examine the role of RBE in the miR164-CUC pathway in Arabidopsis sepal and petal development. We demonstrate that RBE binds to the promoter of MIR164c and represses its transcription. Genetic analyses also support the idea that RBE negatively regulates the expression of MIR164c. In addition, we also demonstrate that RBE regulates the expression of the other two miR164 genes. Thus, RBE coordinately regulates the appropriate expression of the miR164 genes, which in turn impacts the expression of the CUC1 and CUC2 transcription factors.
MATERIALS AND METHODS
Genetic stocks and growth conditions
Arabidopsis thaliana plants were grown at 22°C under 16-hour light/8-hour dark conditions. The rbe-2 (Salk_037010), rbe-3, mir164b-1 (Salk_136105), cuc1-1 and cuc2-1 seeds were obtained from the Arabidopsis Biology Resource Center (ABRC). mir164a-4 (GABI 867E03) was obtained from Nottingham Arabidopsis Stock Center (NASC). eep1 (also referred to as mir164c-1) was a gift from Dr Elliot Meyerowitz (California Institute of Technology, Pasadena, CA, USA). mir164a-4 and mir164b-1 were in the Columbia (Col-0) background. All the other mutants are in the Landsberg erecta (L er) background. Double mutants were confirmed by genotyping. Methods for genotyping mir164b-1 and eep1 have been described previously (Mallory et al., 2004). rbe-2 was genotyped using primers RBE5C2 and LBb1 to amplify a 400 bp genomic fragment, whereas the wild-type allele was identified by primers RBE5C2 and RBEO3 producing a 388 bp product. rbe-3 was identified by derived cleaved amplified polymorphic sequence (dCAPS) (Neff et al., 1998) using primers RBE3GF2 and RBEO3 and followed by the digestion of PCR products with SalI (New England BioLabs, Ipswich, MA, USA), producing two fragments (151 bp and 23 bp) in L er and one fragment (174 bp) in rbe-3. In the other crosses with rbe-3, except mir164a rbe-3, homozygous rbe-3 plants were selected by phenotype and subsequently genotyped. Methods to genotype cuc1-1 and cuc2-1 have been reported previously (Takada et al., 2001). Primers used in genotyping were (5′-3′): LBb1, CAAACCAGCGTGGACCGCTTGCTGCAACTC; RBE5C2, TTTATTTTGGCCGTTTAGGGAAGA; RBE3GF2, GCCACACCTCCTGAATGTGATCGTCGA; RBEO3, GACTTCACCATTGTCATGTTCATATCT; RBE, ATGATGGATAGAGGAGAATGCTTGATG; RBE2, AGCTAGTTAACCTTAGGCG. Transgenic pCUC1:CUC1, pCUC1:CUC1m, pCUC2:CUC2 and pCUC2:CUC2m are all in the L er background and obtained from the ABRC. These lines were individually crossed with rbe-3. rbe-3 plants with the CUC1/2 or CUC1m/2m transgene were identified in the F2 population by screening seedlings on BASTA and genotyping for rbe-3.
GUS reporter lines pMIR164a:GUS and pMIR164b:GUS (Raman et al., 2008) were kindly provided by Dr Patrick Laufs (Institut National de la Recherche Agronomique, France). These two lines were originally in Col-0 background and were backcrossed twice into L er. These backcrossed lines were crossed to rbe-3 and mutant plants containing the GUS reporter were identified from the F2 population. pMIR164c:GUS (Baker et al., 2005) and pMIR164c:3XVENUS-N7 (Sieber et al., 2007) were gifts from Dr Elliot Meyerowitz (California Institute of Technology, Pasadena, CA, USA).
To make 35S:GR-RBE, the RBE coding region was amplified from L er genomic DNA using primers RBE and RBE2 (see above). The PCR product was first cloned into the pCR8 vector (Invitrogen, Carlsbad, CA, USA) and subcloned into the pJAN33 plasmid (a gift from Dr Kathy Barton, Carnegie Institute, Palo Alto, CA, USA) using the GATEWAY cloning strategy (Invitrogen, Carlsbad, CA, USA). The resulting pJAN33-RBE plasmid was transformed into Arabidopsis plants using the floral dip method (Clough and Bent, 1998) and transformants were selected on BASTA. pJAN33 contains the CaMV 35S promoter upstream of a GATEWAY cassette fused to a GR tag and a FLAG tag at the N-terminus, so the pJAN33-RBE transgenic plants ubiquitously express a double-tagged RBE protein (GR-FLAG-RBE).
Dexamethasone (DEX) induction and total RNA extraction
pJAN33-RBE floral buds were treated with DEX (10 μM dexamethasone, 0.1% ethanol, 0.015% silwet) or mock (0.1% ethanol, 0.015% silwet) solution for four hours. After the treatments, floral tissues were harvested, frozen in liquid nitrogen and stored at –80°C. Total RNA was extracted with Trizol (Invitrogen) according to the manufacturer’s instructions, treated with TURBO DNase (Applied Biosystems, Carlsbad, CA, USA), purified using the Rneasy kit (Qiagen, Valencia, CA, USA), and used in subsequent analyses.
Eight sequencing libraries (two biological replicates, each with two technical replicates for four-hour DEX- or mock-treated floral tissues) were produced from purified total RNA samples by standard Illumina protocols. The libraries were run on eight lanes of an Illumina Genome Analyzer, generating an average of 27,650,720 single-end reads of 36 bp each. We used Tophat v1.2.0 (Trapnell et al., 2009) (http://tophat.cbcb.umd.edu/) to perform spliced alignments of the reads against the Arabidopsis thaliana TAIR10 reference genome (http://www.arabidopsis.org/). Only reads that mapped to a single unique location within the genome with a maximum of two mismatches in the anchor region of the spliced alignment were reported in these results. We used the default settings for all other Tophat options. To obtain a tally of the number of the reads that overlapped the exons of a gene, we analyzed the aligned reads with HTSeq v0.4.5p6 (unpublished; http://www-huber.embl.de/users/anders/HTSeq/doc/) and the gene structure annotation file for the reference genome (TAIR10 gff file; downloaded from http://www.arabidopsis.org/). The tally for each sample was then processed with LOX v1.6 (Zhang et al., 2010) to analyze gene expression levels statistically. Data have been deposited in the Gene Expression Omnibus database under accession number GSE36469.
Total RNA from L er, rbe-3 and DEX- or mock-treated pJAN33-RBE floral tissues were extracted as described above and purified using TURBO DNA-free Kit (Applied Biosystems). qRT-PCR of mRNA transcripts was performed using Multiscribe reverse transcriptase (Applied Biosystems) and Taqman gene expression assay (Applied Biosystems) according to the manufacturer’s instructions. For qRT-PCR of miR164a/b, reverse transcription was carried out using stem-loop primers (Chen et al., 2005) and miR164a/b was quantified using the Taqman miRNA assay (Applied Biosystems). Changes in gene expression were calculated from three biological replicates using 2–ΔΔCt method. For mRNA quantification, the relative RNA levels were normalized to the value of ACTIN 2 (ACT2). For miRNA quantification, expression level of snoR66 was used as the standard for normalization.
Histology, confocal imaging and scanning electron microscopy
β-Glucuronidase (GUS) staining was carried out as published previously (Nakayama et al., 2005) with minor modifications. Whole inflorescences were collected, vacuum infiltrated for 15 minutes and incubated at 37°C in GUS staining solution (2 mM X-glucuronic acid, 50 mM sodium phosphate buffer, pH 7.2, 10 mM EDTA, pH 8, 0.1% Triton X-100, 6 mM FeCN). owing to the different promoter activities of three miR164 genes, we used different incubation times in this step. In both L er and rbe-3, we used 12 hours for pMIR164a:GUS, 3 hours for pMIR164b:GUS and 24 hours for pMIR164c:GUS. For whole mounts, floral tissues were cleared using an ethanol series, mounted in 30% glycerol; for sectioned material, cleared floral tissues were embedded in Paraplast (Fisher Scientific, Pittsburgh, PA, USA) and sectioned at 8 μm. Whole mounts and sectioned material were examined using a Zeiss Axiophot microscope. For confocal imaging, floral tissues were stained with FM4-64 (Invitrogen), mounted in water and imaged using a Zeiss LSM 510 confocal microscope. For scanning electron microscopy, tissues were fixed overnight in formalin/acetic acid/alcohol fixative with subsequent dehydration in an ethanol series. Samples were critical-point dried and sputter coated in gold/palladium and imaged using an ISI SS40 scanning electron microscope.
Chromatin immunoprecipitation (ChIP)
pJAN33-RBE floral tissues treated with 4-hour DEX or mock solution were harvested, washed with de-ionized water and crosslinked with 1% formaldehyde. Crosslinking was quenched with 0.125 M glycine. Procedures for nuclear extracts and immunoprecipitation were adapted from Gendrel et al. (Gendrel et al., 2002), with modifications. Sonications of nuclear extracts were performed at 4°C for 15 minutes (power level 5; Diagenode, Denville, NJ, USA). After chromatin shearing, 6 μl anti-FLAG monoclonal antibody (Sigma-Aldrich, St Louis, MO, USA) was added to each sample to immunoprecipitate GR-FLAG-RBE proteins. After reversing crosslinks, DNA was purified by phenol:chloroform extraction, ethanol precipitated, and resuspended in 50 ml TE. Two microliters of immunoprecipitated DNA was used in qPCR (MIR164c promoter) or semi-quantitative PCR (MIR164a and MIR164b promoter). The exon region of APETALA3 (AP3) was used as the negative control. Primers used for PCR are listed in supplementary material Table S2. qRT-PCR utilized Power SYBR Green PCR Master Mix (Applied Biosystems) for amplification. Semi-quantitative PCR conditions were: 33 cycles, 94°C for 30 seconds, 55°C for 30 seconds and 72°C for 30 seconds. DNA band intensity was measured using ImageJ. The abundance ratio of a specific promoter fragment in DEX-versus mock-treated ChIP samples was first normalized to the negative-control gene AP3, then was divided by the normalized ratio of DEX- to mock-treated input values to obtain a fold enrichment. Three biological replicates were carried out for each ChIP experiment.
RBE binds to the MIR164c promoter and represses its expression in the flower
To investigate the function of RBE in floral organ and boundary formation, we constructed an inducible form of RBE by fusing its coding region to the glucocorticoid receptor ligand binding domain sequence (GR). Transgenic plants expressing GR-RBE under the regulation of the strong constitutive 35S promoter were morphologically normal (supplementary material Fig. S1). However, treatment with 10 μM dexamethasone (DEX) every other day for 20 days resulted in 35S:GR-RBE plants becoming stunted and having epinastic rosette leaves (supplementary material Fig. S1). Striking phenotypes were also found in the reproductive phase. DEX-treated 35S:GR-RBE plants had compact inflorescences with short underdeveloped sepals, petals and stamens (Fig. 1A-D). The carpels of such plants were short and swollen and fertility was highly reduced. These defects resembled the miR164-resistant CUC1m and CUC2m overexpression phenotypes (Baker et al., 2005), which suggests that RBE is a regulator of the miR164-CUC pathway. Furthermore, because RBE is a putative transcriptional repressor, it probably acts upstream of one or more of the miR164 genes and negatively regulates their expression.
To identify targets of RBE action, we carried out whole-transcriptome RNA-seq analyses to compare gene expression in four-hour DEX and mock-treated 35S:GR-RBE young floral buds. We focused on the 832 genes expression of which was significantly reduced (P<0.025) by twofold or more in DEX compared with mock-treated plants. In this analysis, we identified MIR164c (EEP1) as a candidate target of RBE. MIR164c expression was strongly repressed, with no matching reads after 4-hour DEX treatment (supplementary material Table S1). This result was confirmed by quantitative PCR in which MIR164c expression was ∼30-fold decreased in 4-hour DEX versus mock-treated samples (Fig. 1E).
To determine whether RBE directly binds to MIR164c promoter sequences, we performed chromatin immunoprecipitation (ChIP) on 4-hour mock- and DEX-treated 35S:GR-RBE floral tissues. After immunoprecipitation, semi-quantitative PCR and qPCR were utilized to amplify three fragments in the 1000 bp region upstream of the transcription start site of MIR164c (Nikovics et al., 2006). Both approaches resulted in an approximately threefold enrichment of the same fragment specifically in DEX-treated floral tissues (Fig. 1F,G), indicating that RBE is associated with MIR164c promoter sequences.
RBE regulates MIR164c expression patterns and levels in floral organs
To determine how RBE regulates MIR164c expression, we examined the expression of pMIR164c:GUS in both wild-type (L er) and rbe-3 floral buds. Flowers containing the pMIR164c:GUS transgene in the L er background displayed an expression pattern of MIR164c based on GUS staining that was similar to that reported in previous studies (Baker et al., 2005; Sieber et al., 2007). Prior to stage 5, MIR164c promoter activity was observed in the inflorescence meristem, floral meristem and the regions between adjacent sepals (Fig. 2C,M,N; supplementary material Fig. S2). The expression of a pMIR164c:3XVENUS-N7 construct (Sieber et al., 2007) also showed a similar but weaker pattern of MIR164c promoter activity in early stage floral primordia (supplementary material Fig. S3). MIR164c was also expressed in the gynoecium and petal primordia during stages 5-7 (Fig. 2O; supplementary material Fig. S2). In late stage flowers, MIR164c expression was detected primarily in the sepals and in the septum of the gynoecium (supplementary material Fig. S2). In transgenic flowers in the rbe-3 background, the overall expression pattern of pMIR164c:GUS was similar to that of wild type, particularly after stage 5 (Fig. 2F,P-R; supplementary material Fig. S2). However, GUS staining was more intense in young floral buds, especially during stages 3-5, and in stage 3 sepal primordia and stage 4-5 petal primordia (Fig. 2F,P-R). Although overall expression levels were weaker for the pMIR164c:3XVENUS-N7 construct, GFP signal was somewhat more intense in early stage rbe-3 floral primordia compared with WT (supplementary material Fig. S3). These results were consistent with the qRT-PCR results using whole inflorescence tissue, in that MIR164c expression was mildly but significantly increased in the rbe-3 relative to L er backgrounds (Fig. 3), and suggests that RBE represses MIR164c in a stage-specific manner.
RBE functions largely through regulating MIR164c expression
Plants mutant for rbe-3 displayed various defects in the first two whorls of flowers, such as extra and fused sepals, as well as loss of or aberrant petals (Krizek et al., 2006; Takeda et al., 2004) (Tables 1, 2; Fig. 4B,C; supplementary material Fig. S4). These phenotypes indicate that RBE controls multiple biological processes during sepal and petal development. To evaluate the role of MIR164c in RBE-dependent developmental processes, we examined the genetic interactions of RBE with MIR164c through double-mutant analyses. The most striking floral phenotype of the mir164c-1 (eep1) mutant was the presence of extra petals in early-arising flowers (Baker et al., 2005) (Fig. 4D,E). In addition, we also observed morphological abnormalities in the second whorl of eep1 flowers, including aberrantly formed petals and petal filaments (Fig. 4F,G). Plants mutant for eep1 also occasionally produced extra sepals (supplementary material Fig. S5; Table 2). In the eep1 rbe-3 double mutant, flowers exhibited complex phenotypes, mainly in the first and second whorls. In the first whorl, the extra sepal phenotype that occurs at a low frequency in both eep1 and rbe-3 flowers was enhanced in the double mutant. The eep1 rbe-3 flowers had significantly more sepals than either single mutant (supplementary material Fig. S5) with extra sepals observed in 31% of the flowers, whereas eep1 and rbe-3 have only 5% and 8% flowers with extra sepals, respectively (Table 2; P<0.005 for both, Fisher’s Exact Test). We also observed a novel phenotype, pale green to white sepals with a partial homeotic transformation to second whorl identity, in the first whorl of eep1 rbe-3 flowers, which was not found in either single mutant (Fig. 4H; Table 2; supplementary material Fig. S6).
The sepal fusion phenotype of rbe-3 was largely rescued by eep1 (Table 1), which is consistent with the observation that RBE represses MIR164c expression in sepal primordia (Fig. 2). A similar rescue of the rbe-3 phenotype was also observed in the second whorl of eep1 rbe-3. The total second whorl organ number of eep1 rbe-3 was nearly the same as that of eep1, although normal petal number in eep1 rbe-3 was intermediate to that of the single mutants (supplementary material Fig. S5). In addition, the petal abnormalities present in rbe-3 flowers were still found in eep1 rbe-3 (Fig. 4H,I). Together, these results indicate that upregulation of MIR164c can account for the absence of petals and sepal fusion defects in rbe-3, but might not be the cause of other morphological abnormalities in rbe-3 mutants.
RBE also regulates MIR164a and MIR164b expression
Mature miR164 is transcribed from three different loci: MIR164a, b and c (Baker et al., 2005; Sieber et al., 2007). However, MIR164a and MIR164b expression levels were not significantly affected in DEX-treated 35S:GR-RBE flowers as shown by both RNA-seq and qPCR (Fig. 1; supplementary material Table S1). Furthermore, we did not observe any enrichment for MIR164a or MIR164b promoter fragments in 35S:GR-RBE ChIP analyses (supplementary material Fig. S7), indicating that it is unlikely that RBE regulates MIR164a or MIR164b expression directly. However, given that the three miR164 genes exhibit redundant functions in flower development (Sieber et al., 2007), it remains to be seen whether RBE also regulates MIR164a and MIR164b expression indirectly. To test this hypothesis, we examined the expression of promoter-GUS reporters for these two genes (pMIR164a:GUS and pMIR164b:GUS) in rbe-3 and wild-type backgrounds. Consistent with previous studies (Sieber et al., 2007), we observed that MIR164a and MIR164b have distinct expression patterns in wild-type flowers. Specifically, MIR164a was expressed in the boundaries between the inflorescence meristem and floral buds, and in stage 2-3 flowers (Fig. 2A,G). Its expression then became largely restricted to the stamens and carpels in later stages (Fig. 2A; supplementary material Fig. S2). At stage 12, pMIR164a:GUS expression was observed in the style, the septum and in some cells adjacent to the septum (supplementary material Fig. S2). By contrast, pMIR164b:GUS expression was not apparent until stage 6 and was detected predominantly in the sepals and carpels in later flowers (Fig. 2B,H,I; supplementary material Fig. S2). In stage 12 flowers, MIR164b displayed a distinct expression pattern in the carpel, with no expression in the septum (supplementary material Fig. S2).
In rbe-3 flowers, both GUS reporter lines displayed stage-specific changes in their expression patterns. For pMIR164a:GUS, we detected reduced GUS expression in stage 2-3 rbe-3 flowers (Fig. 2D,J), but not in older flowers (supplementary material Fig. S2). This reduction of expression is consistent with qPCR assays (Fig. 3) and suggests that RBE positively regulates MIR164a in young floral buds. Positive regulation of MIR164a by RBE is also consistent with the mir164a rbe-2 and mir164a rbe-3 double-mutant phenotypes. The rbe-2 allele is a strong loss-of-function mutation in the Col-0 background, resulting from a T-DNA insertion within the coding region (Takeda et al., 2004). Plants that were mutant for rbe-2 displayed a severe phenotype, with frequent conversion of petals into filaments, or exhibiting a lack of petals altogether (Fig. 4M). Plants mutant for rbe-2 also had sepal fusion defects (Table 1). Double mir164a-4 rbe-2 or mir164a-4 rbe-3 mutant flowers largely resembled those of rbe-2 or rbe-3, which is consistent with the hypothesis that RBE acts upstream of and promotes the expression of MIR164a (Fig. 4L,N; Table 1). By contrast, the levels of MIR164b were found to be slightly increased in rbe-3 by qPCR (Fig. 3). The patterns of pMIR164b:GUS expression indicate that this increase was likely to be due to the precocious or increased expression of MIR164b in stage 4-6 flowers of rbe-3, compared with that in L er at comparable stages (Fig. 2E,K,L). We did not observe any obvious difference in pMR164b:GUS expression between rbe-3 and L er in older flowers (supplementary material Fig. S2). These results indicate that RBE also negatively regulates MIR164b at specific stages of flower development.
Because the partially redundant functions of miR164 genes are likely to be due to their overlapping patterns of expression in the flower (Sieber et al., 2007), we tested the extent to which the combined activities of the miR164 genes were responsible for different aspects of floral patterning. The mir164a-4 single mutant had serrated leaves, but showed no obvious floral defects (Nikovics et al., 2006; Sieber et al., 2007) (Fig. 4J,K). The mir164b-1 mutant displayed a wild-type phenotype in both leaves and flowers (Baker et al., 2005; Sieber et al., 2007) (Fig. 4O). The double mir164ab mutants formed normal floral organs as expected (Fig. 4P) because the MIR164a and MIR164b expression domains are largely exclusive of each other. The double mir164bc mutant resembled eep1 in that these flowers produced extra petals, but otherwise were similar to wild type (Baker et al., 2005) (Fig. 4Q), suggesting that these genes have discrete functions that are probably related to the considerable differences in their expression patterns.
By contrast, the spatial and temporal expression of MIR164a and MIR164c overlapped with each other in flowers (Fig. 2; supplementary material Fig. S2). Consistent with this overlap, we observed novel floral phenotypes in mir164ac double mutants. The mir164ac plants showed an enhanced extra sepal phenotype compared with eep1; 24% of mir164ac flowers had extra sepals (P=0.008) and 20% possessed pale green sepals reminiscent of the petal-like sepals observed in eep1 rbe-3 (Fig. 4R; supplementary material Fig. S5; Table 2; P<0.005). The similarity between the mir164ac double mutant and the eep1 rbe-3 double mutant further supports the conclusion that MIR164a is downregulated in rbe-3 mutants. Some of the mir164ac sepals also exhibited abnormal shapes (Fig. 4S). In the fourth whorl, mir164ac double mutants had unfused carpels (Fig. 4R,S). In the second and third whorls, the mir164ac double and eep1 single mutant flowers displayed similar phenotypes (Fig. 4; data not shown). In addition to these double mutants, we also analyzed the phenotypes of 35S:MIR164b overexpression lines (Baker et al., 2005). Ubiquitous expression of MIR164b resulted in fused sepals, and lack of or abnormal petals in the transgenic flowers (Fig. 4T; supplementary material Fig. S5; Table 1). These phenotypic defects emphasize the importance of maintaining specific expression patterns of each of the miR164 genes for normal floral development. As such, the action of RBE in differentially regulating the expression of all three miR164 genes is likely to be important in controlling the levels and distribution of mature miR164 to ensure normal floral organogenesis.
RBE modulates CUC1 and CUC2 functions in flower development
CUC1 and CUC2 genes are targets of miR164 and expression of miR164 cleavage-resistant CUC1 or CUC2 genes have demonstrated that miR164 function is largely mediated through CUC1 and CUC2 (Baker et al., 2005; Larue et al., 2009; Laufs et al., 2004; Mallory et al., 2004; Nikovics et al., 2006; Peaucelle et al., 2007; Sieber et al., 2007). As CUC1 and CUC2 function serves as a read-out for the combined activities of the miR164 genes, we examined the extent to which RBE functions were dependent on CUC activity. Although RBE has opposite effects on MIR164a and MIR164b expression (Fig. 3), RBE appears to affect MIR164b more profoundly than it affects MIR164a, because we found that the mature miRNA164 a/b level was upregulated in rbe-3 mutant flowers (Fig. 3). Thus, an increased level of miR164a/b and MIR164c leads to an overall upregulation of miRNA164 expression in the rbe-3 mutant. Consistent with this upregulation, we observed a decrease in mRNA abundance of CUC1 and CUC2 in rbe-3 relative to wild type (Fig. 3). Furthermore, in four-hour DEX-induced 35S:GR-RBE floral tissues, we detected a mild upregulation of CUC1 and CUC2 expression compared with mock-treated controls (Fig. 1E).
To determine the extent to which regulation of CUC1 and CUC2 activity reflected the biological role of RBE, we introduced rbe-3 into the cuc1-1 and cuc2-1 mutant backgrounds. Although with our sample sizes it was not yet statistically significant (P>0.05), both cuc1-1 rbe-3 and cuc2-1 rbe-3 exhibited enhanced sepal fusion and more extensive petal defects than in each of the single mutants (Fig. 5A-D; supplementary material Fig. S8; Table 1). Our observation of enhanced sepal fusion in the cuc2-1 rbe-3 double mutant is also consistent with previous reports (Krizek et al., 2006). The number of mis-shapen petals and the extent of sepal fusion were comparable to the phenotypes seen in 35S:MIR164b plants, in which CUC1 and CUC2 expression levels are simultaneously reduced (supplementary material Figs S5, S8; Table 1). These observations support our expression data that CUC1 and CUC2 levels are both reduced in rbe-3 (Fig. 2).
We also tested the effect of overexpressing CUC1 or CUC2 in rbe-3 by introducing miR164-resistant versions of CUC1 or CUC2 (CUC1m or CUC2m) driven by their native promoters into the rbe-3 mutant. We utilized miR164-sensitive pCUC1:CUC1 and pCUC2:CUC2 lines as controls. pCUC:CUC1 and pCUC2:CUC2 both displayed wild-type floral organs owing to post-transcriptional regulation by endogenous miR164 (Baker et al., 2005), whereas rbe-3 pCUC1:CUC1 and rbe-3 pCUC2:CUC2 flowers resembled those of rbe-3 in organ number and morphology (Fig. 5E-H; supplementary material Fig. S8; Tables 1, 2). Similar to what has been previously reported (Baker et al., 2005), we observed phenotypes in pCUC1:CUC1m plants that mimic those of the eep1 mutant. These phenotypes included extra petals in the second whorl, as well as extra sepals in the first whorl (Fig. 5I-K; supplementary material Fig. S8; Table 2). However, we also observed that the extra sepal phenotype appeared to be more pronounced in pCUC1:CUC1m flowers compared with eep1 (20% in pCUC1:CUC1m compared with 5% in eep1; Table 2; P=0.03). Furthermore, we also found that the pCUC1:CUC1m flowers frequently displayed pale green to white sepals that were reminiscent of those seen in mir164ac and eep1 rbe-3 flowers (Fig. 5K; Table 2; P=0.02). These observations suggest that the subtle differences between the pCUC1:CUC1m and eep1 flowers are due to an excess accumulation of CUC1 gene product in the pCUC1:CUC1m plants that occurs because of the lack of miR164a- and miR164b-mediated degradation. Furthermore, many of the rbe phenotypes were rescued in the pCUC1:CUC1m background. These included first whorl sepal fusion that is seen in rbe-3 mutants but is completely ameliorated in rbe-3 pCUC1:CUC1m plants (Table 1). In the second whorl, the total organ number in rbe-3 pCUC1:CUC1m was equivalent to that seen in pCUC1:CUC1m plants, whereas the petal morphology defects in rbe-3 were partially rescued in rbe-3 pCUC1:CUC1m plants (Fig. 5M,N; supplementary material Fig. S8). The only synergistic interaction found in rbe-3 pCUC1:CUC1m plants was the increased sepal number compared with the parental pCUC1:CUC1m and rbe-3 single mutants (supplementary material Fig. S8; Table 2; P=0.09 and P=0.009, respectively), which suggests that the extra sepal phenotype in rbe-3 is at least partly due to processes independent of the miR164-CUC pathway. This idea is also supported by the observation that rbe-3 pCUC2:CUC2m also displayed an enhanced extra sepal phenotype compared with the parental rbe-3 and pCUC2:CUC2m single mutants (Table 2; P=0.04 and P<0.005, respectively).
Although not showing strong floral phenotypes on its own, the pCUC2:CUC2m construct could partially compensate for the floral defects of rbe-3 in the rbe-3 pCUC2:CUC2m combination. In the first whorl, the sepal fusion defect of rbe-3 appeared to be ameliorated in the rbe-3 pCUC2:CUC2m plants, indicating a partial rescue of the rbe mutant phenotype (Table 1; P=0.09). In the second whorl, rbe-3 pCUC2:CUC2m plants displayed considerably more normal petals than seen in rbe-3 flowers (Fig. 5O,P; supplementary material Fig. S8). Together, our results demonstrate that RBE activates CUC1 and CUC2 at least in part by modulating the expression of multiple miR164 genes.
In this study, we demonstrated that RBE can bind to the promoter of, and regulate the expression of, MIR164c. RBE also regulates the expression of MIR164a and MIR164b, although this appears to be indirect. The transcriptional regulation of the miR164 genes by RBE appears to be important in defining both the localization and timing of mature miR164 accumulation that, in turn, impacts CUC-dependent boundary specification and concomitant organogenesis.
RBE represents a novel control point in the regulation of miR164-mediated developmental processes
The CUC1 and CUC2 genes are redundantly required for the formation of tissue boundaries as well as for the repression of growth in a variety of tissues. These instances of boundary formation and repression of growth include cotyledon separation during embryogenesis, lateral organ initiation on the flanks of the shoot apical meristem, and leaf margin serration, indicating that a similar mechanism operates to establish boundaries in these different contexts (Aida et al., 1997; Aida et al., 1999; Heisler et al., 2005; Hibara et al., 2006; Laufs et al., 2004; Nikovics et al., 2006; Raman et al., 2008; Takada et al., 2001; Vroemen et al., 2003). In all of these tissues, post-transcriptional regulation of the CUC1 and CUC2 genes by miR164 is a key regulatory mechanism controlling CUC gene expression levels and consequent regulation of growth (Baker et al., 2005; Larue et al., 2009; Laufs et al., 2004; Mallory et al., 2004; Nikovics et al., 2006; Peaucelle et al., 2007; Sieber et al., 2007). However, to date, relatively little is known about how miRNA genes themselves are regulated to effect particular developmental outcomes through the production of a single mature miRNA species. For instance, in Arabidopsis, miR172 is encoded by five genes, and transcription factors binding to promoter regions of individual miR172 genes have been identified (Grigorova et al., 2011; Wu et al., 2009; Yant et al., 2010). However, the mechanisms for coordinately regulating the family of miR172 genes are not yet clear. GIGANTEA (GI) has been shown to regulate overall miR172 abundance, but this regulation appears to occur via GI-mediated processing of miR172, rather than through transcriptional control, and might not be specific to miR172 (Jung et al., 2007).
Direct promotion of the expression of MIR164a by the TCP3 transcription factor has been shown to lead to downregulation of the CUC1 and CUC2 genes during leaf differentiation (Koyama et al., 2007; Koyama et al., 2010). TCP3 also regulates a suite of genes involved in auxin signaling, which in turn provides a mechanism for feedback control on the expression of the CUC genes (Koyama et al., 2010). RBE, by contrast, appears to act in a distinct manner, via the transcriptional control of all three miR164 genes, with MIR164c being directly repressed by RBE during sepal and petal development.
MIR164a and MIR164c have partially redundant functions in that the double mir164ac mutant displays extra sepals, mis-shapen sepals, and chimeric sepal-petal organs not seen in either single mutant (Fig. 4; supplementary material Fig. S5). These phenotypes were also seen in double mutant combinations of mir164c with rbe-3, supporting the idea that RBE positively regulates MIR164a. Furthermore, this result suggests that RBE might have a non-autonomous function in regulating sepal primordium initiation. Our results indicate that MIR164c is directly repressed by RBE in stages 3-5 flowers. Consistent with this result, we observed various petal defects in rbe-3 that can be partially or completely rescued by the loss of MIR164c (Fig. 4). MIR164b transcription is also repressed by RBE during this period, and its premature expression in the rbe mutant probably also contributes to the petal defects we observe. However, it is also clear that RBE acts to regulate many other genes that are likely to impact the development of sepal and petal primordia (supplementary material Table S1). RBE has been reported to negatively regulate the expression of the AGAMOUS (AG) floral homeotic gene and restrict its expression from the second whorl (Krizek et al., 2006). Although we did not identify AG as a significantly downregulated gene in DEX-induced 35S:GR-RBE plants (supplementary material Table S1), potentially the ectopic expression of AG in the second whorl of rbe mutant plants could be a contributing factor to the observed defects in petal primordium initiation and growth.
Our genetic analyses indicate that CUC1 and CUC2 are major effectors of RBE activity, as pCUC1:CUC1m, and to a lesser extent, pCUC2:CUC2m, can suppress many aspects of the rbe mutant phenotype. Through regulating the expression of the miR164 genes, RBE can fine-tune the expression of the CUC genes to precisely define the proliferative capabilities of different domains of cells. Alterations in RBE and/or miR164 function through single and double mutant combinations indicate that one consequence of disrupting normal levels of CUC activity is the formation of ectopic boundaries, mis-specified boundaries, and lack of normal boundary formation. As RBE acts independently of organ identity specification (Takeda et al., 2004), these mutant effects on boundary formation can lead to the formation of primordia that incorporate cells from normally distinct whorls, resulting in chimeric organs. Also, it is clear that miR164 gene activity also is required for CUC-independent organ formation (Baker et al., 2005), suggesting that RBE probably controls other aspects of organ formation, at least in part, through regulating miR164 spatial and temporal expression patterns that regulate CUC-independent processes.
Coordinate control of multiple miRNAs
In contrast to most animal miRNAs, plant miRNAs are usually transcribed from large gene families consisting of multiple miRNA precursors (Li and Mao, 2007). Similar to protein-coding gene families, miRNA gene families, including the miR164 genes, probably originated from gene or genome duplications, and subsequently underwent functional diversification to acquire specific biological functions (Maher et al., 2006). Because miRNA gene families usually produce similar or identical mature miRNAs, the functional specificity of each miRNA gene largely results from its distinct spatial and temporal expression pattern. For instance, the maize miR166 gene family includes nine members that are localized in distinct patterns in the shoot apex, with only some of these genes contributing to the function of miR166 in the specification of the abaxial side of leaves (Nogueira et al., 2009).
Jasinski et al. (Jasinski et al., 2010) have shown that the miR164 genes appear to have evolved in a stepwise manner, with the MIR164b lineage probably pre-dating the radiation of the angiosperms, and a more recent duplication giving rise to the MIR164a and MIR164c gene lineages present in Arabidopsis. Consistent with these evolutionary relationships, MIR164a and MIR164c display partially overlapping expression patterns, whereas MIR164b exhibits a distinct pattern in both early and late stage flowers (Fig. 2; supplementary material Fig. S2). These separate derivations imply that the regulatory control of each miR164 gene also evolved independently. Our observations that RBE regulates the two most closely related Arabidopsis miR164 genes differently supports this idea and implies that these RBE-dependent pathways were either independently recruited to this regulatory role or that the regulation of each gene diversified while maintaining an overall RBE-dependent regulatory logic.
Although RBE appears to directly and negatively regulate MIR164c expression in sepal and petal primordia during early stages of floral development, it is still unclear how miR164 gene expression is modulated in other tissues and at other times. Like RBE, SUPERMAN (SUP) encodes a putative EPF-type zinc finger transcriptional repressor and has been postulated to regulate the formation of the boundary between stamen and carpel whorls in the flower (Sakai et al., 1995). One tantalizing possibility is that SUP, and potentially other related EPF-type zinc finger proteins, act similarly to RBE to regulate miR164 gene expression in order to control CUC activity at other boundary domains.
We thank Drs Elliot Meyerowitz and Patrick Laufs for gifts of seed stocks, and Dr Kathy Barton for a gift of the pJAN33 plasmid.
T.H. was supported in part through a Yale University Brown Fellowship. This project was supported by a National Science Foundation grant [IOS-0817744 to V.F.I.].
Competing interests statement
The authors declare no competing financial interests.