Modulation of floral development by a gibberellin-regulated microRNA

Gibberellin (GA) promotes a variety of plant developmental processes by overcoming the repressive effects of the DELLA proteins, a family of nuclear repressors of GA response (Richards et al., 2001). Recent experiments have indicated that GA overcomes repression by DELLA by targeting the DELLA proteins for destruction in the 26S proteasome (Fu et al.,2002; McGinnis et al.,2003; Sasaki et al.,2003). One consequence of this is that lack of DELLA proteins can suppress the phenotypic effects of GA deficiency. For example, the gibberellin-deficient Arabidopsis ga1-3 mutant flowers very late in short-day (SD) photoperiods (Wilson et al., 1992; King et al.,2001), and this phenotype is largely suppressed in ga1-3mutants that also lack the DELLA proteins GAI and RGA(Dill and Sun, 2001; King et al., 2001; Cheng et al., 2004). ga1-3 mutant flowers also exhibit retarded anther development, whilst normal development is restored in ga1-3 mutants that lack the DELLA proteins RGA, RGL1 and RGL2 (Cheng et al.,2004). Thus GA regulates floral development in a manner analogous to the way in which it promotes growth: the restraining effects of DELLA proteins are overcome by GA, most likely via SCF complex E3 ubiquitin ligase-dependent targeting of DELLAs for destruction in the proteasome(Fu et al., 2002; Sasaki et al., 2003; McGinnis et al., 2003; Harberd, 2003).

First identified in barley aleurone cells(Gubler et al., 1995), GAMYB is a GA-specific transcriptional regulator. GAMYB binds specifically to a GA-response element (GARE) in the 5′ regulatory region of GA-activated genes. Binding of GAMYB to the GARE activates the transcription of genes encoding the hydrolytic enzymes that release stored nutrients from the endosperm during seed germination (Cercos et al., 1999; Gubler et al.,1999). Expression of GAMYB (the gene encoding GAMYB) is itself up-regulated by GA (Gubler et al.,1995). In addition, constitutive expression of GAMYB mimics the activating effects of GA application on the transcription of genes encoding hydrolytic enzymes, suggesting that GAMYB plays an important role in the GA signalling pathway in aleurone cells(Cercos et al., 1999). Recent experiments have shown that transgenic overexpression of GAMYB perturbs the development of anthers, indicating that GAMYB can function in cells other than those of the aleurone layer (Murray et al., 2003). Furthermore, loss-of-function mutations in the rice GAMYB gene have recently been shown to abolish GA-mediated induction of the aleurone hydrolase α-amylase, to retard the growth and development of stamens and anthers, and to impair microsporogensis(Kaneko et al., 2004). These latter observations provide genetic proof of the involvement of GAMYB in both aleurone physiology and floral organ development.

Genes encoding proteins closely related in sequence to the barley and rice GAMYB have been identified in Arabidopsis [AtMYB33, AtMYB65and AtMYB101 (Gocal et al.,2001)]. Amongst these AtGAMYBs, AtMYB33 is the most similar to the barley GAMYB (Gocal et al.,2001). The AtGAMYBs have been suggested to be involved in the GA-mediated promotion of flowering in SDs, by activation of the floral meristem identity gene LEAFY(Blázquez et al., 1998; Blázquez and Weigel,2000; Gocal et al.,2001). Activation of LEAFY is known to be important in this process because transgenic over-expression of LEAFY restores a wild-type flowering time to ga1-3 plants grown in SDs(Blázquez et al., 1998). AtMYB33 binds in vitro to a GARE in the LEAFY promoter (a GARE that has been shown to be essential for GA induction of LEAFY) and fails to bind to non-functional mutant derivatives of this element. This suggests that GAMYB plays a key role in LEAFY activation(Blázquez and Weigel,2000; Gocal et al.,2001). Another floral integrator gene, SOC1, has recently been shown to play an important role in the GA promotion of flowering in SD photoperiods (Moon et al.,2003), and it has been suggested that GAMYB and SOC1 both act upstream of LEAFY (Lee et al.,2000; Mouradov et al.,2002; Yu et al.,2002).

To further our understanding of the role of the AtGAMYBs in the regulation of floral development we took advantage of the recent identification of a microRNA (miR159) exhibiting substantial sequence homology with mRNAs encoding AtMYB33, AtMYB65 and AtMYB101 (Gocal et al., 2001; Reinhart et al.,2002). MicroRNAs are single-stranded RNA molecules of 20-22 nucleotides that can cause complementarity-dependent cleavage (e.g. Llave et al., 2002a) or translational inhibition (Chen,2004; Aukerman and Sakai,2003) of target mRNA molecules. More than one hundred distinct small RNAs have been identified in plants, although only a minority of these are microRNAs (Llave et al.,2002b; Reinhart et al.,2002). Most of the predicted targets of these microRNAs are members of transcription factor gene families involved in developmental patterning or cell differentiation(Rhoades et al., 2002; Kidner and Martienssen, 2003). In this paper we describe experiments showing that miR159 directs the cleavage of AtMYB33-encoding transcripts (see also Palatnik et al., 2003). We also show that miR159 sequence is evolutionarily conserved, and that miR159 levels are GA-regulated via a mechanism that is evolutionarily conserved between Arabidopsis and barley and that is dependent on the GA-promoted opposition of the function of DELLA protein GA-response repressors. To further investigate the role of miR159 in vivo, we generated transgenic Arabidopsis plants over-expressing miR159 and examined the floral developmental phenotypes of these plants. We found that elevated expression of miR159 resulted in a delay in flowering in SD photoperiods that was associated with a reduction in the levels of LEAFY transcripts. In addition, previous experiments in barley and rice have identified a role for GAMYB in anther development (Murray et al., 2003; Kaneko et al.,2004). We found that elevated levels of miR159 perturbed anther development and caused a consequent reduction in floral fertility. These observations suggest that miR159 modulates GA-mediated developmental regulation via its effects on GAMYB activity.

Arabidopsis experiments used the Landsberg erectalaboratory strain of Arabidopsis thaliana. The ga1-3, gaiand ga1-3 gai-t6 rga-24 mutations were as previously described(Achard et al., 2003). The plants were grown in a growth cabinet in long days (LDs; 22°C, 16-hours photoperiod) or in short days (SDs; 22°C, 10-hour photoperiod) as indicated. Barley (Hordeum vulgare) floral tissues were obtained from 7-week-old greenhouse-grown plants (16 hour photoperiod). Tobacco(Nicotiana benthamiana) leaves and floral tissues were harvested from 5-week-old greenhouse-grown plants (16-hour photoperiod).

The 35S:MYC-MYB33 construct contained a copy of the AtMYB33 (accession number AF411969) coding sequence. The cDNA was amplified from total RNA by RT-PCR using gene-specific primers(AtMYB33: GCATGTTGAGGCACTTGAGTGGAGTC and TCTGGAATCATAACAGGTAATGTCGG)and cloned into the p35S-2 vector (P. Mullineaux, John Innes Centre). Three myc epitope tags were inserted in phase, upstream of the AtMYB33cDNA. After digestion, an EcoRV fragment containing the construction 35S:MYC-MYB33 was inserted into the plant transformation vector pGreen0029 (P. Mullineaux, John Innes Centre). A mutated version of the AtMYB33 transgene (35S:MYC-mMYB33) was generated by PCR using the Quick Change XL site-directed mutagenesis kit (Stratagene). The mutated mMYB33 primers used were as follows:GGCAGTGAAGCTGGAATTGCCAAGCTTTCAATATTCAGAAACAAC and GTTGTTTCTGAATATTGAAAGCTTGGCAATTCCAGCTTCACTGCC. The nucleotide sequence of the mMYB33 cDNA was as described by Palatnik et al.(Palatnik et al., 2003). The 35S:miR159a construct contained a copy of the precursor of miR159 isoform a (Fig. 1A). A fragment of 228 bp (containing the precursor of 182 bp) was amplified from genomic DNA by PCR using specific primers (miR159a: GAGAAGGTGAAAGAAGATGTAGAGCTCCC and CGATAGATCTTGATCTGACGATGGAAG). These primers were designed to anneal specifically to genomic DNA on either side of the predicted miR159a hairpin precursor. The amplified fragment was then cloned into the p35S-2 vector, and an EcoRV fragment containing the construct 35S:miR159a was then inserted into the pGreen0029 vector. The 35S:GFP construct was made similarly using GFP-specific primers: CGGCACGACTTCTTCAAGAGCGC and GTATTCCAACTTGTGGCCGAGG. The resulting constructs were introduced into Agrobacterium tumefaciens strain GV3101.

Agro-infiltration was performed in Nicotiana benthamiana leaves as described previously (Llave et al.,2002a). 35S:miR159a, 35S:MYC-MYB33, 35S:MYC-mMYB33 and 35S:GFP were introduced into A. tumefaciens and the bacteria injected into N. benthamiana leaves with a syringe. For co-injections of two or three different constructs, bacteria were resuspended in infiltration medium (0.5× Murashige and Skoog salts, 5% sucrose, 0.5 g/l Mes) at OD₆₀₀=1, and incubated for 3 hours at room temperature with 150 μM acetosyringone. Zones of infiltration were harvested at 4 and 5 days after injection for RNA and protein isolations. The 35S:GFP construct was used as a control for the co-Agro-infiltration.

Transformation was performed by floral dipping(Clough and Bent, 1998). A. tumefaciens containing the appropriate constructs were resuspended in infiltration buffer (0.5× Murashige and Skoog salts, 5% sucrose, 0.5 g/l Mes, 0.05% Silwet L77) at OD₆₀₀=1. Transformants were selected by resistance to 50 mg/l of kanamycin sulphate. Plants in Fig. 5A-C, Figs 6 and 7 were homozygous for the 35S:miR159a transgene (T₂ generation); plants in Fig. 5D,E were from segregating populations (self-pollination progeny of 35S:miR159a primary transformants).

Total RNA was extracted using the Trizol reagent (Gibco-BRL) and resuspended in 50% formamide. Low molecular mass and high molecular mass RNAs were isolated by three successive precipitations in PEG₈₀₀₀/NaCl as previously described (Hamilton et al.,2002). 20 μg high molecular mass RNA was separated in a 1%denaturing agarose gel and subjected to blot-hybridisation analysis. Radiolabelled probes were prepared from a 200 bp 3′-end MYB33fragment or from full-length LEAFY, SOC1 and ELF4a open reading frames via random priming reactions (Roche) in the presence of[α³²P]dCTP. 20 μg low molecular mass RNA was separated in a 15% denaturing polyacrylamide gel and subjected to blot-hybridization analysis. Small RNA (miR159a: TAGAGCTCCCTTCAATCCAAAGA; miR167:TAGATCATGCTGGCAGCTTCA) probes were labelled with T4 kinase (Gibco-BRL) in the presence of [γ³²P]ATP. Since miR159 hybridises with miR-JAW[a microRNA that is more divergent from miR159 than each of the miR159 isoforms are from each other; Fig. 1A (Palatnik et al.,2003)], it is likely that the miR159a probe used in our experiments hybridises with all three isoforms of miR159. Each RNA blot analysis was repeated at least twice.

Agro-infiltrated tobacco leaves were harvested and frozen in liquid nitrogen. Total protein was extracted and quantified as described by Achard et al. (Achard et al., 2003). 20μg of proteins were separated by 10% SDS-PAGE and transferred to a nitrocellulose membrane. Immunodetection of the MYC-MYB33 fusion protein was performed using a 2000-fold dilution of anti-cMYC polyclonal antibodies from goat (Santa Cruz Biotechnology, CA) and a 2000-fold dilution of peroxidase-conjugated anti-goat IgG (Santa Cruz Biotechnology, CA). Immunodetection of GFP was performed using a 2500-fold dilution of anti-GFP monoclonal antibodies from mouse (Chemicon International, Temecula, CA, USA)and a 5000-fold dilution of peroxidase-conjugated goat anti-mouse IgG(SouthernBiotech). Signals were detected by chemiluminescence using the ECL Western Blotting Analysis System (Amersham Biosciences).

MiR159 is a microRNA of 21 nucleotides in length derived from a precursor molecule of 182 nucleotides (Reinhart et al., 2002). In Arabidopsis, miR159 is present in three isoforms (Fig. 1A). MiR159 sequence is of near-perfect complementarity to an internal sequence of five Arabidopsis mRNAs encoding three GAMYB-like proteins (AtMYB33,AtMYB65 and AtMYB101) and two additional MYB proteins(Fig. 1B). The Arabidopsis miR159 sequence is also closely complementary to an internal sequence of several GAMYB mRNAs from barley (Hordeum vulgare; HvGAMYB), Oryza sativa, Avena sativa and Lolium temulentum (Fig. 1B). AtMYB33, AtMYB65 and AtMYB101 share substantial homology with HvGAMYB in an N-terminal region corresponding to an R2R3 repeat DNA-binding domain(Fig. 1C). These proteins also contain three additional conserved motifs (Box 1, Box 2 and Box 3) that are characteristic of GAMYB (Gocal et al.,2001). The zone that is complementary between miR159 and the GAMYB-encoding mRNAs is located about 100 nucleotides upstream of the sequence encoding Box 2 (Fig. 1C).

MicroRNAs can regulate gene expression by directing cleavage or by inhibition of translation of target transcripts(Llave et al., 2002a; Chen, 2004; Aukerman and Sakai, 2003). We tested if miR159 directs the cleavage of GAMYB-like mRNAs (see also Palatnik et al., 2003) using an Agrobacterium-mediated delivery system to co-express miR159 and MYB33 target mRNA in N. benthamiana leaf tissue(Llave et al., 2002a)(Fig. 2). Four constructs(35S:miR159a, 35S:MYC-MYB33, 35S:MYC-mMYB33 and 35S:GFP)were inoculated and expressed (Fig. 2A). Whilst basal levels of endogenous miR159 were detected in tissues lacking the 35S:miR159a construct, increased levels of miR159 were detected in tobacco leaves inoculated with this construct, indicating that over-expression and processing of transgenically derived miR159a was successfully achieved (Fig. 2B). Expression of the 35S:MYC-MYB33 construct resulted in detectable levels of MYC-MYB33 transcripts (MYC-MYB33a; Fig. 2B). Co-expression of 35S:MYC-MYB33 with 35S:miR159a resulted in loss of the full length MYC-MYB33a transcript form, and the appearance of detectable levels of a smaller transcript (3′-MYB33b) that hybridized to the MYB33 probe (a 3′ probe fragment, see Materials and methods; Fig. 2B). We attribute this smaller fragment to internal cleavage of the MYC-MYB33 mRNA directed by miR159 (as previously suggested by 5′ RACE-PCR experiments; Palatnik et al., 2003). Consistent with this attribution, co-expression of 35S:MYC-mMYB33 (encoding a MYC-mMYB33 transcript that carries an altered miR159 binding site; see Materials and methods) with 35S:miR159a resulted in detectable transcripts of the MYC-MYB33a size, but not of the 3′-MYB33bsize.

Subsequently, we tested whether miR159 causes a reduction in the level of MYC-MYB33 fusion protein (Fig. 2C). A peptide of ∼83 kDa was recognised by MYC antibodies in a sample from leaves expressing 35S:MYC-MYB33 but not in a sample co-expressing 35S:miR159a and 35S:MYC-MYB33(Fig. 2C). This peptide was of the size expected for the MYC-MYB33 fusion protein encoded by 35S:MYC-MYB33. Taken together, these RNA gel-blot and immunoblot analyses indicate that miR159 can direct the cleavage of transcripts encoding GAMYB, thus down-regulating GAMYB levels.

We next investigated the distribution of miR159 in Arabidopsis,tobacco (Nicotiana benthamiana) and barley. In Arabidopsis,miR159 accumulated predominantly in young seedlings and flowers, was less abundant in rosette leaves, cauline leaves or siliques, and was undetectable in roots (Fig. 3). Interestingly, miR159 was also clearly detectable in tobacco and barley. As in Arabidopsis, miR159 accumulated predominantly in the inflorescence and floral tissues in these species (Fig. 3). These observations indicate that the sequence and developmental regulation of miR159 is highly conserved, despite the considerable evolutionary distance that separates Arabidopsis,tobacco and barley from their last common ancestor.

Since GAMYB functions in GA-mediated gene regulation, we next investigated the possibility that GA might itself regulate miR159 levels. The Arabidopsis ga1-3 mutant is GA deficient and exhibits a dwarf,dark-green phenotype (Koornneef and van der Veen, 1980; King et al.,2001). MiR159 was less abundant in flowers of ga1-3 than in wild-type flowers (Fig. 4A). MiR159 levels in GA-treated ga1-3 plants were similar to those in GA-treated wild-type plants, and higher than in untreated wild-type controls(Fig. 4A). These data indicate that miR159 levels are GA regulated.

GA-regulated processes are usually controlled via the effects of GA on DELLA protein function (Richards et al.,2001). In many cases, DELLA proteins act as repressors of GA-response, and GA acts by targeting the DELLA proteins for destruction in the proteasome (Fu et al.,2002). We therefore tested whether GA enhances miR159 levels by overcoming the effects of DELLA proteins. Firstly, we found that the level of miR159 was reduced in gai mutant plants(Fig. 4A). The gaimutant shares many phenotypic characteristics with ga1-3, except that the gai phenotype cannot be restored to normal with GA(Koornneef et al., 1985; Peng et al., 1997). gai encodes a dominant altered-function mutant GAI protein (gai) that is resistant to the opposing effects of GA and confers a reduced GA response(Peng et al., 1997; Richards et al., 2001). Consistent with this we found that the reduced miR159 levels in gaiwere not restored to normal by GA. Secondly, we investigated the effect of lack of the DELLA proteins GAI and RGA on miR159 levels in ga1-3. Previous experiments have shown that various aspects of the ga1-3phenotype are suppressed in plants lacking GAI and RGA(Dill and Sun, 2001; King et al., 2001; Fu and Harberd, 2003; Cheng et al., 2004). We found that ga1-3 plants lacking GAI and RGA (ga1-3 gai-t6 rga-24)had miR159 levels that were not detectably different from those of wild-type plants (Fig. 4A). Taken together, these observations indicate that the DELLA proteins GAI and RGA repress miR159 levels, and that GA promotes miR159 levels by overcoming DELLA-mediated repression.

We next compared the regulation of miR159 level in Arabidopsiswith its regulation in barley. Barley SLN1 encodes a DELLA protein(Chandler et al., 2002; Fu et al., 2002), and the dominant dwarfing sln1-dwarf allele encodes a mutant protein that is altered in the highly conserved DELLA domain (X. Fu, D. E. Richards and N.P.H., unpublished) (see also Peng et al., 1999; Chandler et al.,2002; Itoh et al.,2002). The effect of this mutation is analogous to that of the Arabidopsis gai allele. We found that miR159 levels were reduced in sln1-dwarf compared to SLN barley(Fig. 4B). Thus, GA-DELLA system regulation of miR159 levels is evolutionarily conserved across the divided lineage that separates Arabidopsis and barley.

We also compared the levels of MYB33 transcripts in the Arabidopsis GA-signalling mutant lines. MYB33 transcript levels were not substantially different in wild type, ga1-3, gai or ga1-3 gai-t6 rga-24 (grown in the absence of exogenous GA; Fig. 4A). However, MYB33 transcript levels were consistently higher in all of these genotypes when grown in the presence of exogenous GA(Fig. 4A). Since MYB33is the target of miR159, it might have been expected that there would be an inverse correlation between miRNA and MYB33 transcript levels. Yet whilst miR159 levels were reduced in ga1-3 (and restored by GA or lack of GAI and RGA) there were no detectable differences in MYB33transcript levels in these various mutant lines. This apparent paradox is considered further in the Discussion.

Recent studies suggest that GA might regulate flowering via GAMYB-dependent activation of the floral meristem identity gene LEAFY(Blázquez and Weigel,2000; Gocal et al.,2001). In Arabidopsis, the GA pathway has a stronger effect on flowering in SD than it does in LD photoperiods(Wilson et al., 1992; Mouradov et al., 2002; Moon et al., 2003). We therefore investigated the effect of elevated expression of miR159 on flowering time in SDs. Four-week-old SD-grown transgenic plants over-expressing miR159a (homozygous for 35S:miR159a) had leaves that were smaller and rounder than wild-type control plants(Fig. 5A). In addition,SD-grown 35S:miR159a plants exhibited a delay in flowering time(Fig. 5B). Wild-type plants bolted then flowered at day 39 (±1.12) whereas transgenic plants flowered at day 53 (±4.66) (Fig. 5C). A delay in flowering time (of varying severity) was observed in plants of ten additional independent transgenic lines (each homozygous for 35S:miR159a), and was heritable in the progeny of these plants (data not shown). 35S:miR159a homozygotes also produced more rosette leaves than did wild-type controls when grown in SDs (37±2.63 against 23±2.38 for wild type; Fig. 5C). In LD photoperiods the effect of 35S:miR159a was much less than in SDs, with LD-grown plants over-expressing miR159a flowering at approximately the same time as wild-type plants (actually 1 day later; data not shown).

In subsequent experiments, segregation of the late-flowering phenotype was followed in two independent transgenic families that were segregating for 35S:miR159a. Only those plants overexpressing miR159a exhibited late flowering in SDs, whilst segregants expressing normal levels of miR159 flowered at the same time as non-transgenic wild-type controls(Fig. 5D). We next determined the levels of MYB33 transcripts, the targets of miR159, in these plants. MYB33 transcripts were detected at substantially reduced levels in plants over-expressing miR159a (and exhibiting a late flowering phenotype), but were not reduced in segregating plants that expressed normal levels of miR159 (Fig. 5E). In addition, the pattern of MYB33 transcript accumulation correlated with that of LEAFY transcripts. LEAFY transcript levels were reduced in SD-grown 35S:miR159a plants, but not in segregants expressing normal levels of miR159 (Fig. 5E). Taken together, these observations suggest that miR159 directs the cleavage of MYB33 transcripts, that the resultant reduction in MYB33 level reduces LEAFY activity, and that the reduction in LEAFY activity delays SD flowering.

Recent experiments have shown that SUPPRESOR OF OVEREXPRESSION OF CO1 (SOC1) is also an important component of the GA-dependent flowering pathway and that integration via SOC1 is necessary for flowering in SDs (Moon et al.,2003). We therefore investigated the relationships between the GA-DELLA system, miR159, MYB33 and SOC1. To begin with, we tested whether the DELLA proteins GAI and RGA regulate SOC1expression. SOC1 transcripts were less abundant in GA-deficient ga1-3 plants than in wild-type plants. In contrast, similar levels of SOC1 transcripts were observed in wild-type and ga1-3 plants treated with exogenous GA (Fig. 4A) (see also Moon et al.,2003). Two further observations indicated that GA regulates SOC1 transcript levels via opposition of DELLA repression. First, SOC1 transcript levels were reduced in gai mutant plants(compared to wild type) and not restored to wild-type levels by exogenous GA(Fig. 4A). Thus the constitutively repressing mutant gai protein confers a reduction in SOC1 transcript levels. Secondly, SOC1 transcript levels in ga1-3 plants lacking GAI and RGA (ga1-3 gai-t6 rga-24) were little different to those in wild-type plants(Fig. 4A). Thus lack of GAI and RGA suppressed the effect of ga1-3 on SOC1 transcript levels, indicating that GA regulates SOC1 levels (and thus flowering time in SD) by opposition of GAI and RGA function.

How does GA-pathway regulation of SD flowering via SOC1 relate to GA-pathway regulation of flowering via miR159/MYB33? To answer this question we determined SOC1 transcript levels in plants from a family segregating for 35S:miR159a. Late flowering plants contained SOC1 transcript levels that were similar to those of non-transgenic wild-type plants (Fig. 5E). Thus an increase in miR159 level has no downstream effect on SOC1transcript levels, demonstrating that the SOC1- and miR159/MYB33-dependent signalling pathways through which GA regulates flowering in SD act independently of one another.

GA and GAMYB are known to be involved in the regulation of anther development (Murray et al.,2003; Cheng et al.,2004; Kaneko et al.,2004). We next investigated the possible role of miR159 in anther development. LD-grown 35S:miR159a homozygotes had smaller cauline leaves than wild-type controls (Fig. 6A). In addition, 35S:miR159a plants had short, sterile siliques (Fig. 6C,D). 35S:miR159a siliques were ∼40% of the length of wild-type siliques and contained no seeds (Fig. 6B; eventually a few fertile siliques were obtained in some plants). In other respects the development of 35S:miR159a homozygotes was indistinguishable from that of wild type.

In order to determine the cause of the infertility, the floral development of 35S:miR159a homozygotes was compared with that of wild-type controls. Except for the anthers, the floral organs of 35S:miR159aplants appeared relatively normal. However, the increase in miR159 levels (and decrease in MYB33 transcript levels; Fig. 6E) characteristic of 35S:miR159a plants was associated with a progressive increase in the size of anthers, and with a darkening of anther colour(Fig. 7A-D). In addition, the 35S:miR159a anthers failed to release pollen(Fig. 7D). Since seed-bearing siliques were recovered following cross-fertilization of 35S:miR159awith wild-type pollen (data not shown) it was clear that the infertility of the 35S:miR159a plants was due to male sterility resulting from failure to release pollen. The anther phenotype characteristic of the 35S:miR159a plants was visible from floral developmental stage 10(Fig. 7E) and was clear until stage 12 (Fig. 7F). Subsequently, the anthers became increasingly dark in colour(Fig. 7B,D). The final stamen lengths in wild-type and 35S:miR159a flowers were indistinguishable(data not shown).

Although the recent discoveries of small RNA species in the cells of diverse organisms have generated much excitement, there are still relatively few cases where small RNAs have been definitively linked to specific plant developmental functions (Chen,2004; Aukerman and Sakai,2003; Emery et al.,2003; Palatnik et al.,2003). In this paper we describe experiments designed to investigate the developmental function of miR159. Our initial interest in miR159 was stimulated by its complementarity with a short region of transcripts encoding proteins of the GAMYB class(Rhoades et al., 2002; Park et al., 2002). The specific function of GAMYB proteins in various aspects of developmental regulation by the phytohormone GA led us to develop the hypothesis that miR159 might also function in GA-regulated developmental processes.

On the basis of the sequence complementarity between miR159 and transcripts encoding GAMYB proteins, it seemed possible that miR159 might direct the cleavage of these transcripts. We tested this possibility, and found that expression of miR159 in N. benthamiana causes a reduction in the level of intact MYB33 transcripts and of detectable MYB33 protein. In addition, expression of miR159 failed to reduce the level of mutant MYB33 transcripts (mMYB33) carrying an alteration in the miR159 complementary region. These results are consistent with the previous observation that MYB33 and MYB65 transcripts are cleaved in Arabidopsis plants in the centre of the miR159 binding site(Palatnik et al., 2003). Furthermore, whereas overexpression of MYB33 has no obvious effect on development in Arabidopsis, overexpression of miR159-resistant mMYB33 transcripts confers dramatic developmental perturbations [e.g. curly leaves (Palatnik et al.,2003)]. These results indicate that miR159-directed cleavage of MYB33 (and other GAMYB-encoding transcripts)(Palatnik et al., 2003) is developmentally relevant. However, it is also possible that miR159 influences development via inhibition of translation of GAMYB-encoding transcripts.

We next investigated the levels of miR159 in Arabidopsis and barley GA-signalling mutant lines. We found that the accumulation of miR159 was positively regulated by GA, and negatively regulated by DELLA proteins(GAI and RGA in Arabidopsis; SLN in barley; Fig. 4). Therefore GA regulates miR159 accumulation via an evolutionarily conserved mechanism that involves opposition of DELLA function. Recent advances have shown that the DELLA proteins, although originally identified as components of the GA signalling pathway, also mediate auxin and ethylene responses(Achard et al., 2003; Fu and Harberd, 2003). Thus DELLA proteins may serve as a means of regulating miR159 levels in response to multiple hormonal signalling inputs. Although it has recently been shown that specific miRNA levels are hormonally regulated in Drosophila(Bashirullah et al., 2003), our results are, to our knowledge, the first demonstration of hormonal regulation of microRNA levels in plants.

Since GAMYB has been previously implicated in the regulation of flowering time in SDs in Arabidopsis(Blázquez et al., 1998; Blázquez and Weigel,2000) and of anther development in barley and in rice(Murray et al., 2003; Kaneko et al., 2004), we determined if transgenic alteration in miR159 levels would perturb these developmental aspects. The timing of the floral transition is of considerable adaptive significance, and hence is subject to multiple interdependent genetic and environmental controls (Mouradov et al., 2002). Genetic and molecular analyses have identified three distinct genetic floral promotive pathways: the photoperiod pathway, the autonomous pathway and the GA pathway; the latter being of particular prominence in SD photoperiods (Simpson and Dean, 2003). Previous data suggested that the GA pathway promotes flowering in SD via GAMYB-dependent activation of the floral meristem identity gene LEAFY(Blázquez et al., 1998; Blázquez and Weigel,2000). Consistent with these ideas, we found that transgenic overexpression of miR159a resulted in a reduction in the levels of MYB33 and of LEAFY transcripts, and a specific delay in flowering in SDs. Despite causing severe developmental perturbations at the vegetative rosette stage, expression of miR159-resistant mMYB33transcripts had no dramatic effect on flowering time in SDs (J. Palatnik and D. Weigel, personal communication). Similarly, treatment of SD-grown wild-type plants with GA promotes flowering time by only a few days (P.A. and N.P.H.;data not shown). Thus, in SD conditions, flowering time is at most slightly advanced by increases in MYB33 activity (due to mMYB33 transcripts or GA treatment) but is markedly delayed by decreases in MYB33 activity (due to overexpression of miR159a).

The GA-deficient ga1-3 mutant also either fails to flower or flowers very late in SD photoperiods(Wilson et al., 1992). Yet the ga1-3 mutant contains reduced, rather than increased, levels of miR159 (Fig. 4A). Thus increased levels of mi159 cannot contribute to the late flowering of ga1-3 in SD photoperiods. Our experiments further investigated the relationship between GA regulation of SD flowering and miR159 by looking at levels of transcripts encoding the floral integrator SOC1. GA regulates SD flowering time both via the GAMYB-LEAFY pathway, and via regulation of SOC1 expression (Fig. 8) (Moon et al.,2003). We therefore investigated the effect of GA-DELLA signalling, and of overexpression of miR159a, on SOC1 transcript levels. We showed that GA regulates SOC1 transcript levels by opposing the negative effects of GAI and RGA(Fig. 4A). We also showed that transgenic overexpression of miR159a had no detectable effect on SOC1transcript levels (Fig. 5E). Thus GA regulates SOC1 via a DELLA-opposition-dependent pathway that does not involve GAMYB (or miR159). It seems probable that SD-grown ga1-3 plants flower late (although miR159 levels are reduced) because the reduction in SOC1 transcripts, also seen in ga1-3,overcomes any promotion of flowering resulting from miR159 reductions.

We also found that transgenic overexpression of miR159a caused the formation of abnormal, sterile anthers. In particular, increased levels of miR159 caused an increase in anther size, a darkening of anthers and a failure to release pollen. These observations are of particular interest since, in barley, increased expression of HvGAMYB has some opposite effects on anther development. Whilst barley plants overexpressing HvGAMYB are also pollen-sterile, these plants have anthers of decreased size and lightened colour (Murray et al., 2003). In rice, lack of GAMYB also perturbs stamen and anther development(Kaneko et al., 2004). Taken together, these various observations suggest that a modulation in GAMYB levels causes a defect in anther development and a consequent reduction in floral fertility.

Overall, the results obtained from our transgenic overexpression experiments support the view that miR159 acts in GA signalling to regulate GAMYB activity, presumably through its ability to direct the cleavage of mRNA molecules encoding GAMYB proteins. However, our experiments also reveal a complexity in the relationship between GA signalling, miR159 and GAMYB levels. The levels of MYB33 transcripts do not obviously differ between wild-type, GA-deficient (ga1-3) or GA-deficient plants lacking GAI and RGA (ga1-3 ga-t6 rga-24), despite the fact that miR159 levels vary between these different genotypes(Fig. 4A). At first sight this observation seems puzzling, since the a priori expectation would be that an increased level of miR159 should result in a reduction in MYB33 transcript levels. However, the levels of another plant microRNA, miR39, actually display a positive correlation (rather than the expected negative correlation) with the levels of its cleavage target(Llave et al., 2002a). These unexpected relationships between microRNA and target transcript levels may result from a failure of gel-blot analyses to reveal underlying heterogeneities in the relative cell-specific distributions of microRNAs and their targets, or may reveal a homeostatic component of microRNA function.

Recent results indicate that miRNAs can work as components in negative feedback regulatory loops (Xie et al.,2003). It is therefore possible that miR159 acts within such a loop, as a general homeostatic regulator of GAMYB function. An illustrative example (Fig. 8) of how such feedback might operate could involve down-regulation of MYB33 activity by miR159 (as shown in this paper), and compensatory up-regulation of miR159 by MYB33. Indeed, potential GARE-like sites have been identified in the putative promoter of the miR159 precursor-encoding gene, suggesting the possibility of feedback regulation of miR159 by MYB33 (P.A. and N.P.H., data not shown).

Of course the above hypothetical example is but one possible route for regulation of levels of MYB33 activity, and concerns miR159 and MYB33 in isolation. The miR159/MYB33 system is also subject to DELLA-dependent regulation. In this case, the fact that MYB33 transcript levels are indistinguishable in ga1-3 compared with ga1-3 gai-t6 rga-24could be explained if absence of GAI and RGA increases both MYB33transcription and miR159 levels. Larger-scale changes to the system (e.g. transgenic overexpression of miR159a) break homeostatic regulation of MYB33 levels and result in detectable changes in phenotype (delayed flowering in SD, perturbed anther development). We therefore propose that miR159 acts as a homeostatic regulator of GAMYB function in GA-regulated plant development. Despite these complexities, the strong evolutionary conservation of both sequence and phytohormonal regulation of miR159 suggests that this microRNA is of adaptive significance to the growth and development of many plant species.

We thank Louise Chappell for technical assistance with miRNA analysis; Lali Sakvarelidze for computing assistance during the creation of the figures; the European Union Research Training Network (INTEGA: HPRN-CT-RT1-2000-00090) and BBSRC (Core Strategic Grant to the John Innes Centre) for funding. We also thank J. Palatnik and D. Weigel for permission to cite unpublished observations.