Genetic analysis reveals functional redundancy and the major target genes of the Arabidopsis miR159 family

Abstract

Currently, there are very few loss-of-function mutations in micro-RNA genes. Here, we characterize two members of the Arabidopsis MIR159 family, miR159a and miR159b, that are predicted to regulate the expression of a family of seven transcription factors that includes the two redundant GAMYB-like genes, MYB33 and MYB65. Using transfer DNA (T-DNA) insertional mutants, we show that a mir159ab double mutant has pleiotropic morphological defects, including altered growth habit, curled leaves, small siliques, and small seeds. Neither mir159a nor mir159b single mutants displayed any of these traits, indicating functional redundancy. By using reporter-gene constructs, it appears that MIR159a and MIR159b are transcribed almost exclusively in the cells in which MYB33 is repressed, as had been previously determined by comparison of MYB33 and mMYB33 (an miR159-resistant allele of MYB33) expression patterns. Consistent with these overlapping transcriptional domains, MYB33 and MYB65 expression levels were elevated throughout mir159ab plants. By contrast, the other five GAMYB-like family members are transcribed predominantly in tissues where miR159a and miR159b are absent, and consequently their expression levels are not markedly elevated in mir159ab. Additionally, mMYB33 transgenic plants can phenocopy the mir159ab phenotype, suggesting that its phenotype is explained by deregulated expression of the redundant gene pair MYB33 and MYB65. This prediction was confirmed; the pleiotropic developmental defects of mir159ab are suppressed through the combined mutations of MYB33 and MYB65, demonstrating the narrow and specific target range of miR159a and miR159b.

Micro-RNAs (miRNAs) are 20- to 24-nucleotide (nt) small RNAs that guide the RNA-induced silencing complex in a sequence-specific manner to target mRNA(s), regulating their expression either through degradation of the transcript or translational attenuation (1). They are derived from longer noncoding RNA precursors known as primary (pri) miRNAs, being processed from these transcripts by RNase III-like enzymes known as DICER-LIKE via multiple cleavage steps (2). Their requirement for development has been well characterized. In Arabidopsis, miRNAs have been shown to play critical roles in stem cell formation, organ identity, leaf polarity, vascular differentiation, and cell division patterns (3). Currently, there are ≈180 known miRNA loci in Arabidopsis, many of which are highly conserved across the plant kingdom (3–5). For instance, the miR159 family has been found in all examined seed-bearing plants (4). In Arabidopsis, this family is encoded by three genes, MIR159a, MIR159b, and MIR159c, located in different regions of the genome (6, 7). As determined by deep sequencing, miR159a and miR159b are highly expressed compared with miR159c (5). Their mature products are 21 nt long, with miR159a and miR159b only differing in sequence at one nucleotide, whereas miR159a and miR159c differ at two nucleotides (6). These sequence differences, together with unknown expression patterns of these individual miRNAs, mean that it is uncertain whether they target similar or distinct genes.

One known target of miR159 in Arabidopsis is MYB33, which belongs to a GAMYB-like family of transcription factors (8). In Arabidopsis, there are seven members in this family, all of which share a conserved putative miR159-binding site (9). Two of these genes, MYB33 and MYB65, function redundantly. This is based on strong sequence similarity, expression patterns, and genetic analysis, where only a myb33/myb65 double mutant displays phenotypic defects (9). MYB33 has been the focus of miR159 regulation. The isolation of miRNA-guided cleavage products for MYB33 (10, 11) and in planta assays (11) have demonstrated that miR159a cleaves MYB33 mRNA. Mutation of the miR159-binding site (without changing the amino acid sequence of the gene) within MYB33, generating the mutant allele known as mMYB33 (10), resulted in dramatic expansion of the expression pattern (9). For instance, expression of the MYB33:GUS reporter gene construct was only detected in anthers and in seeds, whereas the mMYB33:GUS reporter gene has strong expression in root and shoot apices and many floral organs in addition to anthers (anther filaments, carpels, sepals, and receptacles) (9). Furthermore, transgenic mMYB33 plants have pleiotropic developmental defects, having curled/rounded leaves, stunted growth, and altered apical dominance. In contrast, transgenic MYB33 plants have none of these developmental defects, indicating that miRNA control of MYB33 regulation is absolutely critical for proper plant development (9, 10). This phenotype is dramatically different than a loss-of-function myb33/myb65 mutant (9) or plants overexpressing miR159a precursors (35S:miR159a), where the only observable morphological phenotype is male sterility (11, 12).

Currently, only a small number of loss-of-function mutants in miRNA genes have been reported in any organism, which is counterintuitive to the notion that their influence is widespread and that they play pivotal roles in development (13). Genetic redundancy has been proposed as a possible contributing factor to this conundrum. Here, we show that this is the case for the Arabidopsis miR159 family, where only a mir159ab double mutant exhibits pleiotropic development defects. Based on the characterization of the spatial expression pattern of the MIR159 genes, and combined with molecular and genetic analyses, we have found that the major targets of these miRNAs are even more limited in scope than previously had been predicted by using bioinformatics or overexpression strategies.

Results

Genomic Structure of the MIR159 Genes.

The primary transcripts of MIR159a and MIR159b were defined by using 5′- and 3′-end rapid amplification of cDNA ends (RACE) on RNA isolated from imbibed seeds. For MIR159a, a single transcriptional start site was mapped 446 bp upstream of the mature miR159a sequence, an identical position to where the transcription start-site of this gene had been previously mapped (14). In contrast, five different polyadenylation sites were found at the 3′ end from the analysis of an equal number of RACE clones (Fig. 1A), implying that the length of the 3′ end was highly variable. The largest transcript of MIR159a was 806 nt long. Similarly, variable polyadenylation sites were mapped at the 3′ end of MIR159b, with three different sites found from three different clones (Fig. 1B). Furthermore, we found two different transcription start sites, 581 and 358 nt upstream of the mature miR159b sequence. Therefore, the largest possible transcript of MIR159b was 900 nt long. No introns were present in either the MIR159a or MIR159b genes.

Fig. 1.

Characterization and structure of the MIR159 loci. Mapping of the pri-MIR159 transcripts and the T-DNA insertion sites for MIR159a (A) and MIR159b (B). LB, left border; RB, right border; B, Basta-resistant gene. Arrows indicate transcriptional start sites with numbers indicating relative positions of the stem–loop regions and the varying polyadenylation sites. In both instances, the T-DNA loci were tandem inverted insertions because both plant–T-DNA junctions were isolated by using left border primers. (C) Relative RNA levels of pri-MIR159 transcripts as determined by quantitative RT-PCR on RNA prepared from seedlings of wild-type, mir159a, and mir159b plants. (D) RNA gel blot analysis of mature miR159 levels in 72-h imbibed seeds. (E) A phylogenetic tree based on the stem–loop sequences of rice and Arabidopsis MIR159 genes.

Using an RNAfold program (http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi) to predict the secondary structure of the various transcript forms of these two genes, we found considerable variation in their overall structures; however, in each case the stem–loop structures of the pre-miRNAs remained invariant (data not shown). This implied that each transcript isoform could be processed correctly to form a mature miRNA.

mir159ab Double Mutant Plants Have Pleiotropic Developmental Defects.

Searching the SIGnAL database (http://signal.salk.edu/cgi-bin/tdnaexpress), we found transfer DNA (T-DNA) insertional mutants belonging to the SAIL collection (15) that lie within MIR159a (SAIL_430_F11; designated here as mir159a) and MIR159b (SAIL_770_GO5; designated here as mir159b). For mir159a, PCR amplification of T-DNA borders determined that the T-DNA was inserted from nucleotide +14 to +51 relative to the transcriptional start site, therefore lying within the primary transcript of the MIR159a gene but outside of the stem–loop structure (Fig. 1A). The expression of MIR159a had been reduced >6-fold in mir159a plants but not eliminated (Fig. 1C). Because the pre-MIR159a structure may still be present, this allele may only represent a hypomorphic mutation. For mir159b, the T-DNA has inserted within the region encoding the stem–loop structure (Fig. 1B), meaning that any transcript from this allele would be unable to form double-stranded RNA and be processed into a mature miRNA. It is likely that this mutation corresponds to a knockout (null) allele. We examined MIR159b expression in mir159b and found that the level of transcript containing the miRNA portion was more than seven times higher than wild type (Fig. 1C), suggesting that this transcript is not processed, resulting in greater stability and accumulation of this portion of the transcript. Neither mir159a nor mir159b displayed any obvious morphological traits.

A phylogenetic tree generated with the stem–loop regions of all known Arabidopsis and rice MIR159 genes showed that the Arabidopsis MIR159a and MIR159b genes were highly similar, suggesting they may be functionally redundant to one another (Fig. 1E). In F₂ segregating plants of a cross between mir159a and mir159b, ≈1 plant in 16 (12/197; χ² = 0.09; P > 0.99) had a distinctive morphological phenotype. By using PCR genotyping, these plants were confirmed to be mir159ab mutants, and they failed to accumulate detectable levels of mature miR159 (Fig. 1D). Compared with wild-type plants, mir159ab growth was stunted, with an altered habit including reduced apical dominance (Fig. 2A) and curled (hyponastic) leaves (Fig. 2B). Mature siliques of mir159ab plants were significantly shorter than those of wild type (Fig. 2C), indicating reduced fertility and seed set. Seeds were reduced in size and had an irregular shape (Fig. 2D).

Fig. 2.

Phenotypic characteristics of mir159ab plants. (A) The smaller growth stature of mir159ab plants. (B) Curled leaf phenotype (left) compared with wild-type (right). (C) Shorter but fatter fruits in mir159ab plants (left). (D) Smaller, irregularly shaped seeds of mir159ab plants (left) compared with wild type (right).

MIR159a and MIR159b Have Similar Expression Patterns Consistent with MYB33 Repression.

To determine tissue-specific expression of the individual MIR159 genes, we performed quantitative RT-PCR with gene-specific primers against the pri-MIR159 transcripts. MIR159b was expressed in mature seeds and was induced ≈20-fold after 72 h of imbibition (Fig. 3A). Similarly, MIR159a was also present in mature seeds and induced (Fig. 3A), but to a lesser degree (2- to 3-fold). The timing of induction corresponds to the germination of Arabidopsis seeds that occurs between 24 and 48 h of imbibition. MIR159a and MIR159b were also expressed in the shoot apex region and at a much higher level than MIR159c (Fig. 3B). This is consistent with mir159ab displaying a phenotype and suggests that no further redundancy may exist with respect to miR159-mediated processes in these tissues.

Fig. 3.

RNA levels of the MIR159 genes and their targets. (A) Expression of MIR159a (black bars) and MIR159b (white bars) during seed imbibition. (B) Expression of MIR159a (black bars), MIR159b (white bars), and MIR159c (gray bars) in different plant tissues. SA, shoot apical region; IF, inflorescences; SD, 3-day-old imbibed seeds. (C) Expression of MYB33 (black bars), MYB65 (white bars), and MYB101 (gray bars) during seed imbibition. (D) Expression of the GAMYB-like genes in 3-day-old seedlings of wild type (black bars), mir159a (dark gray bars), mir159b (light gray bars), and mir159ab (white bars). (E) Expression of the GAMYB-like genes in wild-type (black bars) and mir159ab (white bars) rosettes. (F) Expression of the GAMYB-like genes in wild-type (black bars) and mir159ab (white bars) inflorescences. (G) Expression of the GAMYB-like genes in wild-type (black bars) and mir159ab (white bars) siliques. Values listed on the right side of graphs correspond to those of MYB101.

To examine the temporal and spatial expression patterns of the MIR159 genes, we generated the reporter gene fusions MIR159a:GUS and MIR159b:GUS, where the regions immediately upstream of the miRNA stem–loop regions were fused to GUS. From the examination of multiple transgenic lines for each construct, we found that MIR159a and MIR159b have near-identical expression patterns (Fig. 4), a fact consistent with their redundancy. They both are strongly expressed in root tips, lateral roots, and the shoot apex region (Fig. 4 B–D and F–H), the latter possibly providing the rationale for the phenotype of mir159ab, because this region is where leaf primordia arise and the architecture of the plant is largely determined (16). Expression is seen in imbibed seeds (Fig. 4 I and J) and inflorescences (Fig. 4 A and E), receptacles, anther filaments, sepals, and carpels, with only subtle differences between MIR159a:GUS and MIR159b:GUS. Expression in these tissues coincides with the regions in which MYB33 was repressed by miR159, as determined by the comparison of the expression patterns of MYB33:GUS and mMYB33:GUS (9). The fact that MIR159:GUS expression is not seen in anthers accounts for the observation that the only tissue in which MYB33 is detected in inflorescences is the anthers (9). Therefore, throughout the plant, both MIR159a and MIR159b are expressed in a temporal and spatial pattern that is totally consistent with the pattern of miR159-mediated MYB33 repression.

Fig. 4.

Spatial expression analysis of MIR159 and target genes. (A–D and I) GUS staining of MIR159a:GUS transgenic plants in inflorescence (A), root tips (B), emerging lateral roots (C), the shoot apex region (D), and seeds (I) imbibed for 24 h. (E–H and J) GUS staining of MIR159b:GUS transgenic plants in inflorescence (E), roots (F), emerging lateral roots (G), the shoot apex region (H), and 48-h-old seedlings (J). Staining of MIR159b:GUS lines was generally much stronger than that of MIR159a:GUS lines (K), MYB101 promoter:GUS in inflorescences (L), MYB33:GUS in wild-type (M), and MYB33:GUS in mir159ab. AF, anther filament; R, receptacle; S, sepal; A, anther; C, carpels.

MYB33 and MYB65 Expression Levels Are Elevated in mir159ab.

We quantified the expression of the seven members of the GAMYB-like family that have a conserved motif that miR159 can potentially cleave (9). Of these genes, MYB33 and MYB65 are strongly expressed throughout the plant (12, 17). The steady-state levels of these transcripts fall during seed imbibition at the time the transcript levels of MIR159a/MIR159b increase (Fig. 3 A and C). To test whether they are under miR159 control, we measured the transcript levels of MYB33 and MYB65 in 3-day-old seedlings of wild-type, mir159a, mir159b, and mir159ab plants (Fig. 3D). Whereas expression was mostly unaffected in the single mir159 mutants, in mir159ab, the levels of MYB33 and MYB65 were ≈3- and ≈5.4-fold higher than in wild type, respectively, indicating that miR159a and miR159b act redundantly in controlling the expression levels of these genes. These studies were extended to rosettes, inflorescences, and siliques (Fig. 3 E–G). In each case, MYB33 and MYB65 have considerably higher transcript levels in mir159ab plants, demonstrating that these genes are deregulated throughout the plant. This is supported by expression of the MYB33:GUS transgene, which was not observable in wild-type seedlings (Fig. 4L) but expressed throughout mir159ab (Fig. 4M).

By contrast, MYB101 shows little or no difference in expression levels between wild-type and mir159ab plants. In siliques and shoot apex regions, MYB101 levels were only 2-fold higher. For the latter, MYB101 transcript levels were ≈100-fold lower than MYB33 and MYB65; therefore, this expression level may not be of physiological significance. In inflorescences and 3-day-old seedlings, MYB101 levels were unchanged. In the case of the inflorescence, an MYB101 promoter:GUS construct (hence without the miR159 target) only shows expression in anthers (Fig. 4K). This supports online Affymetrix data (www.genevestigator.ethz.ch/at/) showing that MYB101 is overwhelmingly expressed in pollen/stamens in the inflorescence (12, 17). This implies that the vast majority of MYB101 transcripts are not located in the same cell types as that of MIR159a and MIR159b and is a likely explanation for why MYB101 levels are not significantly different between wild-type and mir159ab plants in inflorescences.

The transcript levels of the other four GAMYB-like family members (MYB81, MYB97, MYB104, and MYB120) are several orders of magnitude lower than MYB33 and MYB65 [supporting information (SI) Fig. 6]. This is consistent with online Affymetrix data, where these genes are expressed primarily in stamens/pollen, with expression either very low or insignificant in any other part of the plant (12, 17). Of the four genes, only MYB81 has consistently higher transcript levels in mir159ab plants, whereas transcript levels of MYB97 and MYB120 were in fact lower (SI Fig. 6). In most instances, transcript level differences were only 2- to 3-fold, and this may reflect secondary effects due to the different morphologies of mir159ab and wild-type plants rather than miR159 regulation. However, like MYB101, these genes are primarily transcribed in anthers, tissues in which miR159a and miR159b appear to be absent, suggesting that they would only make minor contributions to the mir159ab phenotype.

mMYB33 Plants Can Phenocopy mir159ab.

Previously, mMYB33 transgenic plants (the mMYB33 transgene being under the control of the endogenous MYB33 promoter) were shown to have curled leaves and stunted growth (9), characteristics similar to that of mir159ab. For direct comparison, we grew mir159ab alongside three independent mMYB33 lines that displayed a weak (line 1), intermediate (line 2), and strong (line 3) phenotype and compared their phenotypes throughout development. In all instances, the morphologies of mir159ab and mMYB33 (line 2) plants appear indistinguishable from one another (Fig. 5). This includes the size and shape of the rosettes of plants grown in short days (Fig. 5A) or long days (Fig. 5B). At bolting, the size and shape of inflorescences and siliques appeared identical as did the seeds they set (Fig. 5B). The MYB33 expression levels in these mMYB33 lines were positively correlated with the severity of the phenotype (Fig. 5C). However, mir159ab did not conform to this correlation, reflecting that in addition to MYB33, the level of the redundant gene MYB65 is also higher in mir159ab but remains unchanged in the mMYB33 lines (Fig. 5C). Therefore, it is possible that total MYB33/MYB65 activity is at similar levels in mir159ab and mMYB33 (line 2) plants.

Fig. 5.

Phenotypes of mMYB33 and mir159ab/myb33/myb65 plants. (A) Aerial views of rosettes of 5-week-old plants of wild-type, mir159ab, and three mMYB33 lines grown under short days. (B) Aerial views of 3-week-old rosettes of mir159ab and mMYB33 (line 2) grown under long days. Also shown are siliques, seeds, and mature plants from the same lines. (C) Quantitative RT-PCR of 6-week-old mature plants was used to determine the relative expression of total [(T)] MYB33 levels, endogenous [(E)] MYB33 (using a primer to the miR159 target site that solely amplifies the wild-type MYB33 allele), MYB65, pri-MIR159a, and pri-MIR159b transcripts. (D) Aerial view of rosettes of mir159ab/myb33, mir159ab/myb33/myb65, and myb33/myb65.

We also examined the levels of MIR159a and MIR159b transcripts in the three mMYB33 lines, because it has been hypothesized that these genes may be transcriptionally up-regulated by MYB33, resulting in a regulatory feedback loop (10). However, we found no increase in MIR159a or MIR159b transcript levels in the any of the mMYB33 lines when compared with wild-type plants (Fig. 5C). In addition, we failed to detect evidence of MIR159 down-regulation in the absence of MYB33 and MYB65; MIR159 levels were not decreased in myb33/myb65. Finally, using a primer that discriminated between endogenous and transgenic MYB33, we found that the steady-state levels of endogenous MYB33 levels were not reduced in the mMYB33 lines (Fig. 5C). The fact that both endogenous MYB33 and MYB65 levels did not decrease in these mMYB33 lines again supports the finding that higher miR159 levels are not present in the mMYB33 lines.

myb33 and myb65 Alleles Suppress the mir159ab Phenotype.

All of our data point to MYB33 and MYB65 deregulation being predominantly responsible for the mir159ab phenotype. To confirm this, we crossed the myb33 and myb65 alleles into the mir159ab background. A mir159ab/myb33 triple mutant displayed a milder phenotype than that of mir159ab, where growth was less stunted and leaf curling was less severe (Fig. 5D). Moreover, in a mir159ab/myb33/myb65 quadruple mutant, all phenotypic characteristics of mir159ab were suppressed, and the mutant appeared to be identical to myb33/myb65 (Fig. 5D). This reversion of the mir159ab traits in mir159ab/myb33/myb65 demonstrates that MYB33 and MYB65 are solely responsible for the phenotype exhibited by mir159ab plants. Finally, because the phenotype of mir159ab/myb33 reflects only deregulated MYB65 activity, this triple mutant confirms that MYB65 regulates similar processes to that of MYB33 in the shoot. However, in the mir159ab background, MYB33 and MYB65 are no longer redundant; their effects have now become additive.

Discussion

Bioinformatics approaches (3), overexpression strategies (12), and isolation of miR159-cleavage products (10, 18) together predicted that the closely related Arabidopsis MIR159 genes could regulate seven GAMYB-like genes. Through the characterization of Arabidopsis loss-of-function mir159 mutants, along with genetic and molecular analyses, we have shown that the predominant role of miR159a and miR159b is to redundantly control just two of these genes, the redundant gene pair of MYB33 and MYB65. This demonstrates a greater functional specificity than previously thought and excludes other regulatory mechanisms, such as targets with low complementarity, that exist in animals.

Currently, there are very few examples of loss-of-function mutants in plant miRNA genes, with only mutations being reported in the miR164 family (19–22). This scarcity has been thought to be due to their small size and/or potential genetic redundancy, because most miRNAs are members of small- to medium-sized gene families (3). Our findings are consistent with the latter, where the single mutants, mir159a and mir159b, failed to display a phenotype, suggesting that they are fully redundant to one another. This implies that neither MIR159a nor MIR159b are limiting in controlling target gene expression, and this was shown by the fact that neither MYB33 nor MYB65 expression levels increased in the mir159a or mir159b single mutants (Fig. 3D). Furthermore, because only the mir159ab double mutant displayed a phenotype under our growth conditions, this indicates that just a single copy of one wild-type allele of either MIR159a or MIR159b is sufficient to carry out miR159 function, implying that miR159 is produced in a substantial excess.

Of the seven GAMYB-like genes, we have demonstrated that the deregulation of the redundant gene pair of MYB33 and MYB65 is responsible for the mir159ab phenotype. There are several lines of evidence supporting this. MIR159a:GUS and MIR159b:GUS are expressed exclusively where MYB33 was being repressed, as determined by analysis of the spatial expression patterns of MYB33:GUS and mMYB33:GUS (9). For instance, in inflorescences, the only tissue in which MIR159a:GUS and MIR159b:GUS did not overlap with mMYB33:GUS was in anthers, the sole tissues in which MYB33 is expressed (9). These cotranscriptional domains of miR159a/miR159b and MYB33 (and presumably MYB65) explain why mir159ab has global developmental defects, which is in stark contrast to transgenic plants overexpressing a 35S:miR159a transgene that does not lead to any severe morphological defects other than in anthers (11, 12). Constitutive miR159 expression from a 35S promoter would have little impact, because the transcriptional domains of MYB33/MYB65 are covered by endogenous miR159a/miR159b. Furthermore, these overlapping transcriptional domains imply that transportation of miR159a or miR159b is not required for them to repress MYB33, and this is in agreement with other miRNA systems where MIRNA transcription matched precisely to the site of its action (23, 24).

Consistent with these cotranscriptional domains, MYB33 and MYB65 transcript levels accumulate to 3- to 10-fold higher throughout mir159ab plants. In contrast, the other five GAMYB-like genes are predominantly transcribed in anthers and pollen, tissues in which miR159a and miR159b appear to be absent. For MYB101, this finding was demonstrated by the anther-specific expression of a MYB101 promoter:GUS construct. For MYB104, the finding is supported by the lack of cleavage products recovered in wild-type Arabidopsis (10). Hence, this anther/pollen specificity would explain why the expression of these other five GAMYB-like genes do not dramatically increase in mir159ab and why 35S:miR159a plants are male-sterile (11, 12). Furthermore, a pollen-specific gene called DUO1 that belongs to a different class of MYB transcription factors also contains a functional miR159-binding site; when expressed as an miR159-resistant version under the constitutive 35S promoter, it produces severe developmental defects (18). However, this gene, like the five anther-specific GAMYB-like genes, does not contribute to the mir159ab phenotype; it appears that miRNA regulation is largely redundant to the transcriptional control of this gene.

The strongly overlapping expression patterns of mMYB33:GUS and the MIR159:GUS reporter genes suggest that their transcription is controlled by a common regulator, and previously it has been found that they are both induced by gibberellin (11). Furthermore, these overlapping patterns could be explained by a proposed regulatory feedback mechanism where the expression of MYB33 induces the transcription of miR159 (11). However, the fact that the steady-state transcript levels of MIR159a and MIR159b were not elevated in the mMYB33 transgenic lines goes against this possibility. Supporting this, endogenous MYB33 and MYB65 steady-state transcript levels were not lower in the mMYB33 lines, indicating that mature miR159a/miR159b levels have remained unchanged.

By using overexpression strategies and transcriptome analysis, it has been shown that plant miRNAs appear to have only a limited number of targets that they cleave (12). Our loss-of-function strategy suggests that miR159a/miR159b predominantly regulates MYB33 and MYB65, whereas the other predicted targets are predominantly transcribed in tissues where the miRNAs are absent. This scenario is similar to the few examples of miRNA mutants characterized to date. In plants, it was shown that for the mir164abc loss-of-function mutant, only two of the targets of miR164 were likely to account for the majority of the phenotypic changes in mir164abc plants (22). In animals, although lin-4 and let-7 are predicted to regulate many genes, either mutant can be suppressed through the mutation of single target genes (25, 26). One explanation for these observations is that the miRNAs and their targets are transcribed in adjacent but mutually exclusive expression zones, where it is thought that the role of the miRNA is to provide genetic buffering to ensure accuracy to gene-expression programs (27). Similarly, miR159a and miR159b may have a dual role in which they (i) cleave transcripts of MYB33 and MYB65 in tissues in which they are cotranscribed and (ii) ensure that other targets with nonoverlapping transcriptional domains are restricted to those tissues. This may explain the presence of miR159 target sites in the GAMYB-like genes that are apparently not targeted by miR159a or miR159b. Alternatively, the presence of these putative target sites may be required for cleavage by miR159c or the closely related miR319 family; recently, they have been shown to have activity against the GAMYB-like genes, although they are only very minor regulators of MYB33 and MYB65 (18). However, it cannot be ruled out that they are major regulators of the other five GAMYB-like genes. It will be of interest to examine what selective disadvantage mir159ab/myb33/myb65 plants have now that this highly expressed (5) and highly conserved MIR159-MYB regulatory component has been removed.

Materials and Methods

Determination of the pri-MIR159 Transcripts.

To determine the pri-MIR159 transcripts, 5′ and 3′ RACE reactions were performed with first-strand cDNA synthesized on RNA isolated from imbibed seeds. A GeneRacer kit (Invitrogen Life Technologies, Carlsbad, CA) along with nested PCR primers for MIR159a cDNA and MIR159b cDNA (SI Table 1) were used to amplify the 5′ and 3′ cDNA ends. The phylogenetic tree of the stem–loop regions was constructed by using ClustalW on the program at www.ebi.ac.uk/Tools/clustalw (28).

Isolation and Genotyping of T-DNA Insertional Mutants.

T-DNA mutants were found on the SIGnAL “T-DNA Express” Arabidopsis Gene Mapping Tool (29) and were from the Syngenta Arabidopsis insertion library (15). Amplification by using the following gene-specific primers (SI Table 1) detected the wild-type alleles: 159a-5 and 159b-3 gave an 884-bp fragment; 159b-5 and 159b-3 gave a 707-bp fragment. To detect the mutant T-DNA alleles, gene-specific primers were combined with the T-DNA-specific primer LB3 in the following combinations: for the mir159a-1 allele, 159a-5 and LB3 to give a 210-bp fragment and 159b-5 and LB3 to give a 530-bp fragment.

Expression Analysis.

RNA was prepared from Arabidopsis tissues by using a cetyltrimethylammonium bromide (CTAB) procedure (30). Total RNA (100 μg) was digested with 10 units of RQ1 RNase-free DNase (Promega, Madison, WI) for 15 min at 37°C, then cleaned by using RNeasy Plant columns (Qiagen, Valencia, CA). Five micrograms of this RNA was then used to synthesize cDNA in a 20-μl reaction using SuperScript III (Invitrogen). cDNA was diluted to 100 μl, and then 1 μl was used in 20-μl PCRs in 1× SYBR Green JumpStart Taq ReadyMix (Sigma, St. Louis, MO) and 1 μmol of each primer. Specific primers used to quantify each Arabidopsis gene are listed in SI Table 1, and the expression of each gene was normalized with Cyclophilin (At2g29960). All measurements represent the average of three replicates with error bars representing the standard error of the mean (SEM). For the MYB genes, two sets of primers were designed for each gene, either spanning the cleavage site or not. Differences between primer pairs were minimal, and results from only one of the pairs is shown (Fig. 3). For the MIR159 gene, one primer was located in the stem–loop region and another was located 3′ to the stem loop. All MIR159 primers fell within the limits of the shortest transcripts as defined in Fig. 1. Analysis of mature miR159 was carried out as described in ref. 31. Oligoprobes for miR159a and U6 were end-labeled with T4 polynucleotide kinase (Promega).

Generation of Binary Vectors and Transgenic Plants.

For pMIR159a:GUS, upstream sequences (≈1.7 kb) of the MIR159a stem loop were PCR-amplified from Arabidopsis (Columbia) genomic DNA, with the primers mir159a-13 and mir159a-14 (SI Table 2) and cloned into the HindIII/Sal1 sites of pBI 101.1. For pMIR159b:GUS, upstream sequences (≈2.4 kb) of the MIR159b stem loop were amplified with the primers mir159b-6 and mir159b-7 and cloned into the HindIII/Xba1 sites of pB1 101.1. For pMYB101 promoter:GUS, the primers 101Pro-5′ and 101Pro-3′ were used to amplify ≈2.3 kb immediately upstream of the start codon of the MYB101 and cloned into the HindIII/Sal1 sites of pBI 101.1. All amplified DNA was sequenced and confirmed to be correct. Vectors were transformed into Agrobacterium tumefaciens (GV3101) and then transformed into Arabidopsis by using the floral dip method (32). GUS staining was performed as previously described (9).

Abbreviations

miRNA

micro-RNA

T-DNA

transfer DNA

pri

primary.