We have analysed the role of a microRNA, miR164, in boundary formation during organ initiation from Arabidopsis meristems. The establishment and maintenance of the boundary domain are controlled by three partially redundant genes, CUP-SHAPED COTYLEDON1 (CUC1), CUC2 and CUC3. We show that miR164 overexpression phenocopies the cuc1 cuc2 double mutant by inducing post-transcriptional downregulation of CUC1 and CUC2 but not CUC3 mRNA levels. Disruption of CUC2 regulation by miR164, either by making CUC2 resistant to the miRNA or by reducing miRNA levels leads to similar enlarged boundary domains. We relate this enlargement to the division patterns of the boundary cells. We propose that miR164 constrains the expansion of the boundary domain, by degrading CUC1 and CUC2 mRNAs.
During pattern formation in plants and animals, groups of cells are divided into domains that acquire different developmental fates. This process requires the establishment of precise gene expression patterns that are maintained despite continuous growth and cell division. A recently discovered class of small RNAs, the microRNAs (miRNAs) involved in gene expression regulation may contribute to this. In particular, plant miRNAs have been proposed to remove the transcripts of important regulators in some daughter cell lineages and thus could participate in the stabilisation of gene expression patterns(Rhoades et al., 2002). To further test this hypothesis, we analysed the role of one miRNA, miR164, during the process of boundary formation around organ primordia in Arabidopsis meristems.
MiRNAs are small, single-stranded RNAs of about 21 nucleotides found in both animals and plants that post-transcriptionally regulate gene expression(for reviews, see Bartel, 2004; Lai, 2003). Animal miRNAs are transcribed as long primary transcripts (pri-miRNAs) that are first processed into hairpin precursors of about 70 nucleotides (pre-miRNAs) and then into mature miRNAs. Although the cleavage affects both strands of the hairpin precursors, only one strand, the mature miRNA, is preferentially accumulated and incorporated into a ribonucleoprotein complex, the miRNP complex(Khvorova et al., 2003; Mourelatos et al., 2002; Schwarz et al., 2003). Interaction of the miRNA with imperfect complementary sequences located in the 3′ untranslated region (UTR) of the target mRNAs leads to translational attenuation. Conversely, plant miRNAs are perfectly or almost perfectly complementary to their targets (Rhoades et al., 2002), and their interaction triggers the cleavage of the mRNA (Han et al., 2004; Kasschau et al., 2003; Llave et al., 2002; Palatnik et al., 2003; Tang et al., 2003; Vazquez et al., 2004; Xie et al., 2003), although examples of translational attenuation have also been reported(Aukerman and Sakai, 2003; Chen, 2004).
In animals, most evidence for miRNA regulation of gene expression results from classical genetic approaches, although potential targets of miRNAs have been recently predicted by bioinformatics(Enright et al., 2003; Lewis et al., 2003; Rajewsky and Socci, 2004). miRNAs were first identified as regulators of the developmental timing in C. elegans (Abrahante et al.,2003; Lin et al.,2003; Reinhart et al.,2000; Slack et al.,2000). Additional evidence suggests that they may be involved in spatial patterning processes. For example, left-right asymmetry during neuronal patterning in C. elegans is controlled by a miRNA(Johnston and Hobert, 2003). In Drosophila, the Bantam miRNA, the expression of which responds to patterning cues, promotes cell proliferation and prevents apoptosis by targeting the pro-apoptotic gene hid(Brennecke et al., 2003). Thus, Bantam may participate in the coordination between patterning events and downstream control of cell death and cell proliferation.
In plants, most of the miRNA targets predicted by bioinformatics are transcription factors involved in the control of development, raising the possibility that miRNAs may play an important role in this process(Rhoades et al., 2002). Organogenesis in plants, in contrast to animals, proceeds throughout their life span as new tissues and organs are continuously produced by meristems. For example, the shoot apical meristem and a related structure, the floral meristem, initiate primordia of lateral organs such as leaves, sepals or stamens. A family of miRNA, miR172 negatively regulates APETALA2-like transcription factors, thus controlling flowering time and floral organ identity (Aukerman and Sakai, 2003; Chen,2004). Another miRNA family, JAW/miR159, which negatively regulates several members of the TCP and MYB transcription factor families is involved in leaf development (Palatnik et al., 2003). miRNAs have also a central role in lateral organ polarisation. Lateral organ polarity is controlled by three members of a homeodomain/leucine zipper transcription factor family, PHABULOSA(PHB), PHAVULOTA (PHV) and REVOLUTA(REV) (Emery et al.,2003; McConnell et al.,2001; Otsuga et al.,2001). These genes and two evolutionary conserved miRNA, miR165 and miR166 predicted to target them, are expressed in complementary domains of the developing lateral organs(Juarez et al., 2004; Kidner and Martienssen, 2004). miRNA-resistant forms of these targets are ectopically expressed in the developing primordia, suggesting that miRNAs normally limit their expression pattern (Emery et al., 2003; Juarez et al., 2004; Kidner and Martienssen, 2004; McConnell et al., 2001; Tang et al., 2003).
Here we have analysed the role of a miRNA, miR164, in the regulation of the boundary domain around developing primordia at the shoot apical and floral meristems. Boundary establishment and maintenance is controlled in Arabidopsis by three partially redundant genes, CUP-SHAPED COTYLEDON1 (CUC1), CUC2 and CUC3 (Aida et al.,1997; Aida et al.,1999; Takada et al.,2001; Vroemen et al.,2003). These three members of the NAC transcription factor family are expressed in the cells forming the boundary domain around primordia where they may repress growth (Aida et al.,1999). A single mutation of either CUC gene has no major effect on boundary formation, whereas double mutants have fused cotyledons reflecting abnormal boundary specification during embryo development(Aida et al., 1997; Vroemen et al., 2003). Later on, the cuc1 cuc2 double mutant phenotype is restricted to the flowers that form partially fused organs(Aida et al., 1997). The absence of a mutant phenotype during the vegetative phase and in the inflorescence stem was proposed to be due to partial redundancy between the three CUC genes identified in Arabidopsis. In addition to their role in boundary specification, the CUC genes are also involved in meristem establishment during embryogenesis. Indeed, the CUC genes promote the expression of the SHOOT MERISTEMLESS (STM) gene, a central determinant of meristem identity (Daimon et al., 2003; Hibara et al.,2003; Takada et al.,2001; Vroemen et al.,2003). Sequence homology suggests that CUC1 and CUC2 mRNAs could be targeted by miR164(Rhoades et al., 2002). Accordingly, their expression levels are increased in Arabidopsisbackgrounds with an impaired miRNA pathway(Kasschau et al., 2003; Vazquez et al., 2004). We show that miR164 targets CUC1 and CUC2 but not CUC3 mRNAs for degradation, in planta. Disruption of CUC2regulation by the miR164, either by making it resistant to the miRNA or by reducing the miRNA level leads to a similar boundary enlargement phenotype. We traced this modification back to the proliferative activity of the boundary cells. Therefore, we propose a model where miR164mediates the degradation of CUC1 and CUC2 mRNAs, and thus limits the expansion of the boundary domain.
Materials and methods
For the 2x35S::miR164A and 2x35S::miR164B constructs,1049 bp and 1021 bp fragments of genomic DNA centred on MIR164A and MIR164B precursors, respectively(Reinhart et al., 2002) were amplified by PCR using primer pairs miR164A-2/miR164A-3 and miR164B-1/miR164B-2, respectively, and cloned between the double 35Spromoter and the 35S terminator from Cauliflower virus of plasmid pLBR19 in the binary vector pGreen0029(Hellens et al., 2000). For the 2x35S::erGFP, the erGFP was excised from pMCB56(Fernandez-Abalos et al.,1998) and further cloned as for the miR164. Mutations were introduced by PCR into the CUC2 cDNA obtained from REGIA consortium. Mutated or wild-type CUC2s were introduced between the alcA promoter and Cauliflower virus 35S terminator of pL4 plasmid and the alcA::CUC2s expression cassettes were introduced into the pEC2 binary vector. The wild-type CUC2 and CUC2-m4 were fused to a HA-tag and inserted between the double 35S promoter and the 35S terminator from Cauliflower virus of plasmid pLBR19 in the binary vector pGreen0029 (Hellens et al., 2000). The STM::ALCR alcA::GUS and STM::ALCR alcA::erGFP constructs were made following the same strategy as described by Deveaux et al. (Deveaux et al.,2003): 4.4 kb of STM regulatory sequences ending 13 bp before the ATG were PCR-amplified using STM-1 and STM-2 and cloned into pLP999 or pLP962. pGreen-based binary vectors were electroporated into Agrobacterium tumefaciens strain GV3101 together with the pSoup plasmid (Hellens et al.,2000), whereas strain C58 was used for pEC2-derived vectors.
Plants were transformed by floral-dip(Clough and Bent, 1998). AlcA::CUC2 STM::ALCR alcA::erGFP lines were generated by retransforming a STM::ALCR alcA::erGFP line. The LFY::ALCR alcA::GUS and LFY::ALCR alcA::GUS alcAH4GFP have been described previously (Deveaux et al.,2003). The M0223 enhancer trap line described by Cary et al.(Cary et al., 2002) comes from the Hasselhoff collection and was provided by the Nottingham Arabidopsis Biological Stock Centre. The hyl1-1(Lu and Fedoroff, 2000) and hen1-5 (Vazquez et al.,2004) mutants were kindly provided by N. Fedoroff and H. Vaucheret, respectively, and the dcl1-9 mutant was provided by the NASC.
Plant growth in vitro or in the greenhouse, and ethanol induction in the greenhouse have been described before(Deveaux et al., 2003) with the exception that 75% ethanol was used for vapour induction instead of 95%. For in vitro induction, 0.1% of ethanol was added to the growth media before pouring the plates.
Total RNA was extracted from inflorescence apices using TRIZOL (Invitrogen)according to manufacturer's instructions.
For miRNA detection, 30 μg of total RNA were separated overnight on a 15% acrylamide, 8 M urea gel and blotted on Hybond-NX membranes using a BioRad semi-dry blotter. Filters were hybridised overnight in Church buffer at 30°C with end-labelled primers, then washed for 1 hour in 2×SSC,0.1% SDS. Blots were reprobed with a 5S RNA probe.
For HMW RNA, 20 μg of total RNA were separated on a 1.5% agarose gel,blotted on nylon membranes and probed with a randomly 32P-labelled DNA fragment specific for CUC2 (from 415 bp after the ATG to the STOP codon).
RT-PCR was carried out as previously described(Laufs et al., 2003) using primers located on two different exons to discriminate between genomic contamination and RT products. Furthermore, the two primers, located on each side of the predicted miR164 cleavage site, selectively amplified only the uncleaved mRNA. Twenty-three PCR cycles were run for CUC1;21 cycles for NAC1 and At5g07680; 19 cycles for At5g61430, CUC2 and CUC3; and 15 cycles for APT. The primers used are indicated in Table 1.
Microscopy and images analyses
miR164 overexpression phenocopies the phenotype of the cuc1 cuc2 double mutant
In order to analyse the role of miR164 during shoot apical and floral meristem development, we generated Arabidopsis lines overexpressing miR164. For this, 1049 bp and 1021 bp of genomic sequences centred on the two loci predicted to code for miR164[MIR164A and MIR164B, respectively(Reinhart et al., 2002)] were cloned under the control of the strong Cauliflower mosaic virus double 35S promoter (2x35S, Fig. 1A) and transformed into wild-type Wassilevskijia (WS) or Landsberg erecta (Ler)ecotypes. Plants carrying the 2x35S::miR164A or 2x35S::miR164B constructs showed similar phenotype modifications and will be called hereafter 2x35S::miR164. Whereas wild-type or control lines expressing erGFP showed free sepals(Fig. 1B, part 1), sepals of 2x35S::miR164 flowers were fused along their margins(Fig. 1B, parts 2,3). Overexpression of the MIR164B locus led generally to stronger defects than the MIR164A locus (Fig. 1B, part 5). Fusion of the sepals along their whole margin prevented petal and stamen expansion (Fig. 1B, parts 3,4). No significant reduction of stamen number was observed (not shown) in the transgenic lines, whereas lines showing a high degree of sepal fusion also had fewer petals(Fig. 1B, part 6). Similar sepal fusion and reductions in petal number were reported for flowers of the cuc1 cuc2 double mutant (Aida et al., 1997). The cuc1 cuc2 double mutants also show stamen fusion. This did not occur when the 2x35S::miR164 constructs were introduced into WS but could occasionally be observed when they were introduced into Ler (not shown), the ecotype in which the original cuc1 cuc2 double mutant was described(Aida et al., 1997).
In addition to the flower phenotype, ∼15% of the 2x35S::miR164primary transformants had embryo patterning defects, such as cup-shaped or partially fused cotyledons (Fig. 1C). The cup-shaped cotyledon phenotype is characteristic for the cuc1 cuc2 double mutant (Aida et al., 1997), while partial cotyledon fusion is infrequently observed in cuc1 or cuc2 single mutants(Aida et al., 1997; Vroemen et al., 2003).
In summary, 2x35S::miR164A and 2x35S::miR164B lines exhibited embryo patterning and floral defects that are characteristic for Arabidopsis lines with reduced CUC1 and/or CUC2 activity.
Northern blot analysis using a probe complementary to miR164revealed a small RNA of ∼21-22 nucleotide whose level was increased in the 2x35S::miR164 lines compared with wild-type or 2x35S::erGFPplants and correlated with the phenotype intensity of the 2x35S::miR164 lines (Fig. 1D, parts 1,2). We next checked if the RNA species we detected corresponded to a single-stranded miRNA or to a double-stranded siRNA. To achieve this, we performed additional northern blots to detect miR164A* and miR164B*, the complementary strands to miR164 that result from RNaseIII processing of miR164A and miR164B precursors, respectively(Fig. 1D, parts 1,3). The accumulation level of miR164B* is below detection level in both wild-type and 2x35S::miR164B plants(Fig. 1D, part 3). Whereas miR164A* could not be detected in wild-type plants, it accumulated in the 2x35S::miR164A line, though at a ∼10-20 times lower level than miR164 (Fig. 1D, part 3). A similar low-level of the miR164A* strand has been reported previously for plant(Reinhart et al., 2002) or animal miRNAs (Lagos-Quintana et al.,2003; Lagos-Quintana et al.,2002; Lau et al.,2001; Lim et al.,2003; Mourelatos et al.,2002). We therefore concluded that bona fide miR164accumulated in the 2x35S::miR164 lines.
In conclusion, miR164 overexpression phenocopied the cuc1 cuc2 double mutants and the severity of the phenotype correlated with the level of miR164 accumulation.
miR164 primarily targets four genes of the NAC family
miR164 was predicted by Rhoades et al.(Rhoades et al., 2002) to target 5 members of the NAC gene family: CUC1 and CUC2, NAC1(At1g56010) that has been implicated in lateral root development(Xie et al., 2000) and two other uncharacterised members (At5g07680 and At5g61430). We analysed by RT-PCR the effects of miR164 overexpression on the steady state accumulation levels of the five predicted targets and of CUC3,a gene partially redundant with CUC1 and CUC2 but lacking a miR164-binding site.
CUC1 and CUC2 mRNA levels in 2x35S::miR164 lines were reduced compared with wild-type plants(Fig. 2A). The reduction could reach 90% of the wild-type level and correlated with the intensity of the floral defect phenotype. Although the CUC3 mRNA accumulation was reduced in the strong lines, the amplitude was lower compared with CUC1 and CUC2. It has been reported that the expression of CUC3 is abolished or reduced in the absence of both CUC1 and CUC2 activities (Vroemen et al.,2003), showing that CUC1 and CUC2 are redundantly required for CUC3 expression. Therefore, downregulation of CUC3 in miR164 overexpressers is likely to be a secondary effect of CUC1 and CUC2 inactivation. The absence of a region complementary to miR164 in CUC3 also supports the hypothesis that CUC3 is not a direct target of miR164.
Similar to CUC1 and CUC2, the mRNAs levels of At5g07680 and At5g61430 were reduced in miR164overexpressers, whereas the mRNA level of NAC1 was not reduced(Fig. 2B). Absence of NAC1 mRNA downregulation suggests that miR164 did not interact with it under the conditions we tested or that the interaction did not trigger transcript cleavage but translational attenuation as shown for other plant miRNAs (Aukerman and Sakai,2003; Chen,2004).
miR164 regulation of CUC2 is essential for plant development
In order to assess in planta the importance of miR164-guided cleavage of CUC2, we modified the CUC2 mRNA to make it potentially resistant to miR164-guided cleavage, without altering the protein sequence. To achieve this, we first introduced four mismatches in the miR164-binding site of CUC2 in addition to the three naturally present (CUC2-m4 in Fig. 3A) and ubiquitously overexpressed this modified CUC2 or the wild-type form using the double 35S promoter in transgenic WS Arabidopsis (Fig. 3B). Most of the lines had wrinkled leaves, regardless of the CUC2 form overexpressed (9/11 2x35S::CUC2 lines and 7/7 2x35S::CUC2-m4lines). Inflorescence size was reduced in six 2x35S::CUC2 lines. A similar phenotype was observed in three 2x35S::CUC2-m4lines. In addition, two CUC2-m4 lines showed a more severe phenotype with extreme reduction of internode elongation, small floral organs and reduced fertility. Reduced growth was reported for transgenic lines expressing CUC1 that, in addition, showed ectopic meristems on the cotyledon surface (Hibara et al., 2003; Takada et al., 2001). We did not observe ectopic meristems, reflecting either different effects of CUC1 and CUC2 or specific response of the ecotypes used as CUC1 overexpressers were in Ler background. The more severe phenotype of the 2x35S::CUC2-m4 lines suggested that miR164 regulation was important during plant development.
To further investigate the role of miR164 regulation of CUC2 during development, we used a strategy that allowed us to induce the expression of different CUC2 forms (wild-type, m1, m4, c1 and c4 in Fig. 3A) in the boundary domain. To achieve this, we cloned the different CUC2 forms under the control of the ethanol regulated promoter alcA and introduced them into a STM::ALCR-alcA::erGFP line. This way, the ethanol-regulated transcription factor ALCR is produced under the control of SHOOT MERISTEMLESS (STM) regulatory sequences and upon ethanol induction of the plant, binds to the alcA promoters and activates simultaneously and in similar domains the expression of erGFP and CUC2 (Fig. 3C). Simultaneous activation of two alcA promoters in similar domains has been demonstrated previously (Deveaux et al., 2003). The STM regulatory sequences we used are active in the boundary domain of apical and vegetative meristems (see precise description of the expression pattern in Fig. S1 at http://dev.biologists.org/cgi/content/full/131/17/4311/DC1),overlapping with the mRNA accumulation pattern of CUC2 in this tissues (Ishida et al., 2000). Therefore, the use of STM regulatory sequences is an alternative enabling us to temporally control the expression of the CUC2 genes in the boundary domain.
To analyse in planta the effects of miR164-binding site mutation,we ethanol-induced from germination onwards the expression of the different CUC2 forms in the boundary domain in 15 randomly selected transgenic lines. Seedling development was normal for all lines tested until the formation of the first leaves. Scoring 10-day-old seedlings revealed that the expression of CUC2-m4 led to severe leaf growth inhibition with absence of any visible leaf in the most extreme case, or two smaller or unequal leaves in the milder cases (Fig. 3D,E). Expression of CUC2-m1 led also to retarded leaf development though in a smaller proportion of transgenic lines and with a milder effect (Fig. 3D,E). Expression of wild-type CUC2 or CUC2 mutated outside the miR164-binding site (CUC2-c1 and CUC2-c4) had no effect on leaf development (Fig. 3E).
We further investigated the developmental effects of the disrupted miR164-binding site by analysing STM::ALCR-alcA::erGFPcontrol lines and STM::ALCR-alcA::erGFP alcA::CUC2-m4 lines that were ethanol-induced for 6 days just after bolting. No modifications of the mature flowers were observable during the first 3 weeks after induction. The mature flowers formed during the beginning of the fourth week were modified as petal number could be reduced while sepal spacing was increased(Fig. 3F).
In conclusion, these results showed that miR164 regulation of CUC2 was essential for plant development. In particular, disruption of the miR164-binding site in CUC2 and miR164overexpression had opposite effects on sepal boundary formation.
The miR164-binding site is required for in plantaregulation of CUC2 messengers
We compared CUC2 mRNA accumulation in STM::ALCR-alcA::erGFP alcA::CUC2-wt and STM::ALCR-alcA::erGFP alcA::CUC2-m4lines. Non-induced 9 day-old seedlings were ethanol-induced overnight, before tissue sampling. This short induction allowed us to compare the direct effects on CUC2 mRNA levels by reducing secondary effects resulting from the modification of meristem organisation and thus of the size of the STM-expressing domain. Variable expression levels were observed between lines, but the average expression in CUC2-m4 lines was about six times higher than in CUC2-wt lines(Fig. 4A). This showed that disruption of the miR164-binding site in CUC2 resulted in higher mRNA accumulation levels. Accumulation of cleavage products was reported for miRNA-mediated regulation of target mRNA levels(Kasschau et al., 2003; Llave et al., 2002; Palatnik et al., 2003). We could detect a cleavage product in the 2x35S::CUC2 but not in the 2x35S::CUC2-m4 lines (Fig. 4B). No such cleavage product of CUC2 could be detected in wild-type, miR164 overexpressers or in the STM::ALCR-alcA::erGFP alcA::CUC2 lines, possibly owing to low representations of the CUC2 transcripts in these lines resulting from their localised expression (result not shown).
Expression of miR164-resistant CUC2-m4 restores sepal separation in miR164 overexpresser lines
We wanted to analyse the contribution to the floral phenotype of miR164 overexpressers of the downregulation of the four miR164 targets we validated (Fig. 2). We tested whether expression of miR164-resistant CUC2-m4 was sufficient to restore the separation of the fused sepals of miR164 overexpressers. Expression of only CUC2-m4 is sufficient to test this hypothesis, as CUC1 and CUC2 have redundant function. We crossed the STM::ALCR-alcA::erGFP alcA::CUC2-m4 line with strong 2x35S::miR164 lines and analysed the mature flowers following transient ethanol induction of CUC2-m4in the STM-expressing domain. One to four flowers with fully separated sepals were formed about 3 weeks after the beginning of a 5-day induction, whereas flowers with fused sepals typical of miR164overexpressers were present below or above the restored flowers or in the absence of ethanol induction (Fig. 5). The restored flowers showed sepal boundary enlargement characteristic of the STM::ALCR-alcA::erGFP alcA::CUC2-m4 lines. This showed that the flower phenotype of miR164 overexpressers could be attributed to CUC2 and CUC1 inactivation.
Abolition of miR164 regulation of CUC2 results in the progressive enlargement of the boundary domain
In order to further characterise the origin of the sepal boundary defects of the STM::ALCR-alcA::erGFP alcA::CUC2-m4 flowers, we took advantage of the temporal control over CUC2-m4 expression provided by the ethanol switch. We ethanol-induced the inflorescences of a STM::ALCR-alcA::erGFP alcA::CUC2-m4 line for 6 days, and followed boundary organisation using erGFP expression under STM regulatory sequences as a boundary domain marker. One day after the start of induction,erGFP expression was unchanged in the STM::ALCR-alcA::erGFP alcA::CUC2-m4 line compared with the control STM::ALCR-alcA::erGFP line (data not shown). After 6 days, in the control line the GFP expressing domain between emerging sepal primordia of a stage 4 flower formed a band two or three cells wide(Fig. 6, part 1), while in the alcA::CUC2-m4 line this band was enlarged up to six or seven cells wide (Fig. 6, part 2). At 12 days, whereas in control flowers at stage 6-7 only one or two cells expressed erGFP between the sepals (Fig. 6, part 3), a group of about 10 cells could be observed in the alcA::CUC2-m4 line (Fig. 6, part 4), showing that the boundaries were enlarged. Therefore,we concluded that expression of miR164-resistant CUC2 under the control of STM regulatory sequences led to progressive enlargement of the boundary domain, resulting in increased spacing between sepals. This enlargement could be amplified by the experimental approach we used: STM expression being activated by CUC2(Aida et al., 1999; Daimon et al., 2003; Hibara et al., 2003; Takada et al., 2001; Vroemen et al., 2003),expression of CUC2-m4 under the control of STM regulatory sequences could lead to a positive regulatory feedback loop between CUC2-m4 and the STM promoter. Therefore, in order to provide additional evidence that the boundary defects observed in the miR164-resistant CUC2 lines were due to the disruption of the miRNA regulation of CUC2, we analysed the boundaries in the dcl1, hen1 and hyl1 mutants that have a general reduction of the miRNAs(Boutet et al., 2003; Chen, 2004; Han et al., 2004; Kasschau et al., 2003; Park et al., 2002; Vazquez et al., 2004). In particular, the level of miR164 is reduced in these mutants and,accordingly, the CUC1 and CUC2 mRNAs levels are increased(Kasschau et al., 2003; Vazquez et al., 2004). dcl1, hen1 and hyl1 mutants show a wide range of developmental defects (Chen et al., 2002; Jacobsen et al., 1999; Lu and Fedoroff, 2000). We confirmed earlier observations that boundaries around sepals are enlarged in dcl1 mutants (Jacobsen et al.,1999; Kasschau et al.,2003) and observed that similar defects occurred in hyl1and hen1 flowers (Fig. 7). We further characterised the sepal boundaries by introducing the STM::ALCR alcA::erGFP in dcl1-9 and hyl1-1mutants. The GFP-expressing domain around the sepals in the mutants was enlarged compared to wild type (Fig. 7). Therefore mutants having reduced miR164 levels exhibited a similar boundary enlargement phenotype as lines expressing the miR164-resistant form of CUC2 under STM regulatory sequences. Interestingly, the initial patterning into boundary and primordium domains is unaffected in both the miRNA mutants and the CUC2-m4lines, suggesting that this step is largely independent of miR164.
The boundary domain is a dynamic structure
We wanted to know if the enlargement of the boundary domain could be driven by the proliferation of the cells forming the boundary. Therefore, we characterised the proliferation patterns within the boundaries of the sepal primordia and compared it to the entire meristem at similar stages. We introduced a HistoneH4::GFP translational fusion under the control of the alcA promoter (alcA::H4GFP) into a STM::ALCR-alcA::GUS line (Fig. 8A, part 1). For comparison, we used a LFY::ALCR-alcA::GUS alcA::H4GFP line that allows expression in the entire floral meristem(Fig. 8A, part 2)(Deveaux et al., 2003). Expression of the H4GFP fusion allowed us to mark the nucleus, to recognize the mitotic figures and consequently to calculate the mitotic index (MI), with an efficiency comparable with DNA staining after plant fixation (see Fig. S2 at http://dev.biologists.org/cgi/content/full/131/17/4311/DC1).
The MI in the sepal boundary was comparable with the MI in the entire meristem and this before, during or shortly after initiation of the sepal primordia showing that proliferation is not globally repressed in the sepal boundary (Fig. 8A, part 3). The sepal boundary was further divided into two domains: a S-S domain corresponding to the boundary between two sepals and a S-M domain corresponding to the boundary between sepal and meristem(Fig. 8B, part 1). The MI in the S-S domain was 63% higher than in the S-M(Fig. 8B, part 2), showing that proliferation is differentially regulated between the different zones of the boundary, with higher proliferation rates between the sepals than between the sepals and the meristem. Interestingly, the S-S domain that shows the highest proliferation rate is also the most affected by the expression of miR164-resistant CUC2-m4. Finally, we wanted to test if divisions in the boundaries could potentially lead to their enlargement. Cells with a division axis parallel to the boundary axis, i.e. that give rise to daughter cells located in the boundary, are not expected to enlarge significantly the boundary domain, whereas divisions that are orientated perpendicular to the boundary axis could induce boundary enlargement. Therefore we measured the orientation of the divisions relative to the boundary (Fig. 8C, parts 1,3). We limited these analyses to the cells of the outer cell layer as their division are only periclinal and therefore the daughter cells remain in the same layer. We did not observe any preferential division orientation of the boundary cells from the outer layer (Fig. 8C, part 3), suggesting that boundary width is not limited by a mechanism of cell division orientation control. This shows that proliferation provides a mechanism by which the sepal/sepal boundary can enlarge and that is controlled by miR164-dependant regulation of CUC1 and CUC2.
We show that miR164 coded by two loci, MIR164A and MIR164B, post-transcriptionally degrades CUC1 and CUC2 mRNAs. Modification of the regulatory relationship between miR164 and the targets, either by increasing or reducing miR164 level or by making the CUC2 target resistant to it,leads to abnormal boundary size regulation. Our cellular analysis of the sepal boundaries shows that they are not maintained via a control of the proliferation patterns but at least in part by miR164-mediated degradation of CUC1 and CUC2 mRNAs, thus demonstrating the involvement of miRNAs in the control of developmental patterns.
By overproducing miR164, we showed that this miRNA reduced the mRNA level of four out of the five predicted targets(Rhoades et al., 2002). This downregulation resulted from miRNA-guided cleavage of the mRNA, as we could detect degradation products of the CUC2 mRNA that were dependent on the presence of a miR164 complementary site in CUC2,confirming previous identification of CUC1 and CUC2 mRNA cleavage products (Kasschau et al.,2003). We did not observe a reduction of NAC1 mRNA level in inflorescences of miR164 overexpressers. Several hypotheses could account for this. NAC1 may not be a real target of miR164. miR164-mediated degradation of NAC1 mRNA could also be developmentally regulated and not occur under the conditions we tested. Alternatively, miR164 may not regulate NAC1 activity via mRNA cleavage but through translational attenuation as generally observed for animal miRNA and for some plant miRNAs(Aukerman and Sakai, 2003; Chen, 2004). Could such a dual mechanism for miR164 be the result of differences in the target sequences? Two or three mismatches are observed between miR164 and the five predicted targets (Rhoades et al., 2002). However, if pairings between G and U are allowed, two mismatches subsist for NAC1-miR164 complexes, whereas only one exists for the four targets for which cleavage is observed. It would therefore suggest that the mode of action of miR164 depends on the extent of its homology with the target, as observed for small RNAs in animals(Doench et al., 2003; Zeng et al., 2003). It must be noted that, although G-U base pairing is possible, our mutational analysis of the miR164-binding site of CUC2 (CUC2-m1) showed that G-U pairing is not functionally equivalent to G-C pairing.
We show that during early phases of sepal boundary development, cell proliferation is not repressed and that there is no strict restriction of cell division orientations. Therefore, transverse cell division that can potentially lead to boundary enlargement can occur unless the boundary identity is rapidly switched off. We provide evidence that miR164-dependent degradation of CUC1 and CUC2transcripts constrains the expansion of the boundary domain resulting from boundary cell proliferation. First, boundary enlargement was observed when a miR164-resistant CUC2 form was expressed in the boundary domain using STM regulatory sequences. Second, similar boundary defects were observed in mutants with reduced miR164 levels. What is the relation of miR164-dependant boundary size regulation with cell proliferation? MiR164 may switch off the CUC1,2 function after division in one of the daughter cells and thus induce different cell identities. Alternatively, the link with cell division could be looser. miR164 may switch off the CUC1,2 function in the outermost boundary cell in response to boundary enlargement, resulting either from the proliferation of this cell or from another cell. In both cases, a boundary cell would reset its identity and adopt either a meristem or a primordium identity. The latter could account for the earlier observation that, during pea leaf development, cells are recruited into the growing primordium from adjacent domains (Lyndon,1970).
miRNAs are evolutionary conserved in both plants and animals. miR164 homologues have been reported for rice and tobacco(Mallory et al., 2002; Reinhart et al., 2002), and could be found in database for poplar and Medicago truncatula. In addition to Arabidopsis, a potential miR164-binding site is present in NAC genes of rice(Rhoades et al., 2002),petunia, Antirrhinum majus, soybean and bean. At least two of them, NAM and CUP, have a similar role to the Arabidopsis CUC genes in petunia and Antirrhinum, respectively(Souer et al., 1996; Weir et al., 2004). Therefore,the mechanism of boundary stabilisation we described for Arabidopsisis likely to be evolutionary conserved. A similar conservation of the miRNA-target function has been described for the control of leaf polarity between Arabidopsis and maize(Floyd and Bowman, 2004; Juarez et al., 2004; Kidner and Martienssen,2004).
We have shown that CUC1 and CUC2 mRNAs are targeted for degradation by miR164, whereas CUC3 is not directly regulated by the miRNA. Why is CUC3 not a target of miR164?A higher level of redundancy seems to exist within the CUC genes in Arabidopsis than in other species(Souer et al., 1996; Weir et al., 2004). Nevertheless, although the CUC1, CUC2 and CUC3 genes have all a role in boundary specification, their contribution is not identical. First, genetic analyses suggest that the contribution of CUC3 to cotyledonary boundaries is more important than those of CUC1 and CUC2 (Vroemen et al.,2003). Second, expression patterns of the CUC1, CUC2 and CUC3 genes differ slightly during embryogenesis(Vroemen et al., 2003). Finally, the CUC2 expression domain is reduced in the embryo of the pin-formed1 mutant, whereas the CUC1 domain is expanded(Aida et al., 2002), suggesting that these two genes differ in their response to the signalling molecule auxin involved in primordia patterning(Reinhardt et al., 2003). It appears therefore that the precise regulation of CUC1, CUC2 and CUC3 involves different mechanisms. In this context, miRNA-regulation is apparently an additional level of control. Besides, it is possible that another, not yet identified, miRNA could regulate CUC3.
Note added in proof
While this paper was under review, Mallory et al.(Mallory et al., 2004)described partially overlapping results.
Supplemental data available online
We thank H. Vaucheret for helpful discussions during this work and comments on the manuscript, S. M. Witiak for correcting the manuscript, J. C. Palauqui for his input in the initial phases of this work. We thank B. Letarnec for taking care of the plants in the greenhouse, O. Grandjean for help with the confocal microscope and S. Domenichini for SEM assistance. We thank H. Vaucheret, N. Fedoroff and the NASC for plant material, E. Belles-Boix, V. Pautot and the REGIA Consortium for providing the original CUC2clone.