A Collection of Target Mimics for Comprehensive Analysis of MicroRNA Function in Arabidopsis thaliana

1MicroRNAs (miRNAs) are a class of small RNA (sRNA) molecules that has recently emerged as a key regulator of gene activity. In plants, miRNAs are released from larger precursors (pri-miRNAs) in the nucleus mainly, by DICER-LIKE1 (DCL1). The resulting sRNA duplex is methylated and translocated to the cytoplasm where it can be loaded into an RNA-induced silencing complex (RISC) that includes a member of the ARGONAUTE (AGO) family as catalytic component. The RISC can then recognize mRNAs containing sequences complementary to the loaded miRNA. In plants, cleavage of the target mRNA is an important mechanism for plant miRNA action, but there are also direct effects on protein accumulation, as reported for many animal miRNAs –.

2The spatio-temporal expression pattern of miRNA genes is regulated to a large extent at the transcriptional level, and different members of a miRNA family can have distinct, specialized expression domains –. An additional layer of regulation in miRNA action has been reported by Franco-Zorrilla and colleagues. IPS1 (INDUCED BY PHOSPHATE STARVATION 1) encodes a non-coding RNA with a short motif that is highly complementary to the sequence of miR399, which like IPS1 is involved in the response to phosphate starvation –. In contrast to regular miRNA target sites, the IPS1 sequence contains a three-nucleotide insertion in the center, corresponding to the position where normally miRNA-guided cleavage takes place, and this bulge in the miRNA/target pair prevents endonucleolytic cleavage of IPS1 transcripts. This results in sequestration of RISCmiR399, leading to a reduction of miR399 activity. A similar phenomenon, negative regulation of small RNA activity by a partially complementary mRNA, has been recently described in bacteria as well,.

3MiRNA target mimicry can be exploited to study the effects of reducing the function of entire miRNA families. Simultaneous inactivation of all miRNA family members by constructing multiply mutant lines has so far been achieved for only two relatively small families,. Plant target mimics are conceptually similar to miRNA sponges, used to reduce miRNA activity in animals. MiRNA sponges are transcripts containing multiple miRNA binding sites that compete with endogenous target mRNAs, thereby reducing the efficiency of the corresponding miRNA. Although in animals perfect-match miRNA binding sites seems sufficient to sequester miRNAs, such optimal sites would be generally cleaved in plants, and they would not succeed in sequestering the miRNA-loaded RISC. Consistent with this, plants overexpressing non-modified versions of miR156 and miR319 target genes show much milder phenotypes than plants expressing the corresponding target mimics,,. Modifications of the miRNA binding site that prevent cleavage but still allow miRNA binding are therefore required to reduce miRNA activity in plants.

4Here, we present a collection of transgenic plants expressing artificial target mimics designed to knockdown the majority of Arabidopsis thaliana miRNA families. One fifth of these lines have obvious morphological defects, which is in the same range as the approximately 10% of miRNA knockouts that caused phenotypic abnormalities or lethality in Caenorhabditis elegans. We found a clear correlation between the evolutionary conservation of plant miRNA families and their effect on aerial plant morphology.

5We generated artificial target mimics for 73 different families or subfamilies of miRNAs and expressed them in Arabidopsis thaliana plants under the control of the constitutive 35S CaMV promoter. As described, we modified the 23 nucleotide, miR399-complementary motif in IPS1. The different constructs, and the corresponding transgenic lines, are named “MIM”, followed by the numeric identifier of the targeted miRNA family or subfamily. We targeted all miRNA families reported in miRBase (http://microrna.sanger.ac.uk/sequences/index.shtml) and ASRP (http://asrp.cgrb.oregonstate.edu) at the beginning of 2007, plus some of the miRNAs described subsequently. The majority of the analyzed families have only been described in Arabidopsis thaliana and Arabidopsis lyrata,. The remaining families are shared with other angiosperms, and less than a quarter has been detected in non-flowering plants, including gymnosperms, ferns or mosses,,,. A complete list of MIM constructs, and the primer pairs used to generate them, can be found in Table S1. For miRNA target predictions, see,, unless stated otherwise.

6A single artificial target mimic could be designed for most miRNA families. The mature miRNAs produced by members of the miR169 and miR171 families differ slightly, and different target mimics were designed for these subfamilies. Two target mimics were also designed for the miR161 family, which produce two mature miRNAs that have only partially overlapping sequences, and that target similar subsets of the PPR gene family. Conversely, some miRNA families have very similar sequences and overlapping in vivo targets (e.g., miR159/319, miR156/157 and miR170/171a), and artificial target mimics might not be able to unambiguously discriminate between different miRNAs.

7In some cases, the sequence of the bulge in the miRNA/target mimic pair had to be modified. For example, maintaining the original central sequence of IPS1 in MIM172 could have reconstituted a cleavage site for miR172. Consistent with such modifications being important, plants expressing the appropriately modified version of MIM172 showed an altered phenotype (see below), whereas plants expressing an initial version of MIM172 in which a putative miR172 cleavage site was present (MIM172cs) did not. Moreover, plants expressing a MIM172 version with only a single-nucleotide mismatch corresponding to position 11 of the mature miRNA (MIM172sn) did not show any abnormal phenotype either, suggesting that the three-nucleotide bulge is required for target mimic activity (Figure 1).

8We generated at least 20 independent transformants for each of 75 separate constructs. Of these, 15, targeting 14 different families, caused reproducible phenotypes in the shoot system of the plants, which are described below. Phenotypic alterations were consistent across most, if not all, independent transformants examined for each construct. An example of the phenotypic variation among primary transformants is shown in the histograms in Figure 1. An overview of all lines with morphological defects is given in Table 1, together with the main target genes of the corresponding miRNA family and a list of other taxa in which they can be found. The phenotypes of MIM156 and MIM319 plants have been briefly described before,. All miRNA families whose inactivation resulted in visible phenotypical alterations are conserved among the angiosperms, and most of them are also found in non-flowering plants.

9MIM156 and MIM157 plants (Figure 2) had reduced leaf initiation rates, such that they flowered at about the same time as wild type, but with only two or three true leaves. This phenotype is similar to what is seen in plants carrying non-targetable versions of SPL9 or SPL10, two of the miR156/157 targets, and opposite of plants overexpressing miR156b or spl9 spl15 double mutants, –. In addition, these plants had bent, spoon-shaped cotyledons. The few rosette leaves were characterized by serrated margins, indicating adult leaf identity, consistent with a role of miR156 and its targets in controlling phase change.

10MIM159 plants had extensive pleiotropic defects, and similar phenotypes were observed in most MIM319 lines. These plants had reduced stature, with rounder, upward curled leaves (Figure 2), shorter stem internodes, and smaller flowers with short sepals, reduced petals and anthers that did not develop completely. More severe MIM319 lines were progressively smaller, had warped leaves and lacked well-developed petals (Figure 3A). Stem elongation was often completely suppressed (Figure 3B). Most plants had reduced fertility, and this phenotype was particularly severe in MIM319 plants, for which only a few viable seeds could be recovered after they were grown for prolonged periods at 16°C in long days. Both vegetative and floral phenotypes reminiscent of MIM159 defects have been reported for plants that express non-targetable forms of miR159 target genes, and in plants doubly mutant for miR159a and miR159b. In particular, upward curled leaves have been observed in plant expressing non-targetable forms of MYB33, which can be targeted both by miR159 and miR319. Milder MIM319 lines showed different leaf defects, with leaves curled downward (Figure 2). This is consistent with what has been reported for plants that express non-targetable forms of TCP2 and TCP4, which are both exclusive miR319 targets, suggesting that target mimics can at least partially discriminate between these two miRNA families.

11Serrated and hyponastic leaves were seen in MIM160 plants (Figure 2), in agreement with the phenotype of plants that express non-targetable versions of ARF10 or ARF17, two of the three miR160 targets,. In addition, MIM160 plants were smaller than wild type. Compared to other constructs, fewer transformants were recovered, consistent with the known requirement of miR160 for seed viability or germination.

12A different type of leaf serration was caused by MIM164 (Figure 2), similar to what has been reported for plants expressing a non-targetable version of CUC2, one of the miR164 targets, and for plants lacking one of the miR164 isoforms, miR164a. While expression of MIM160 affected the entire leaf, with the serrations being regular and jagged, MIM164 caused mainly serration of the basal part of the leaf, with more irregular and rounded sinuses and teeth (Figure 3C). Although carpel fusion defects have been described for plants lacking miR164c, the carpel defects in MIM164 plants seemed to be different, with ectopic growths forming at the valve margins (Figure 3D), resembling those seen in the cuc2-1D mutant, in which a point mutation affects the miR164 complementary motif in CUC2. In some cases, this tissue could develop into adventitious pistil-like structures (Figure 3E).

13Rounder leaves with an irregular surface, which appeared to be hollowed out between the main veins, were caused by MIM165/166. Younger leaves tended also to be cup-shaped (Figure 2). Targets of miR165/166, including the transcription factor-encoding genes PHAVOLUTA and PHABULOSA, control leaf polarity, and dominant mutations that disrupt the miRNA target site in these genes cause severe alterations in leaf morphology –.

14A substantial delay in flowering was observed in MIM167 plants, which flowered with 20.8±4.2 (mean ± standard deviation; n = 30) leaves in long days, compared to 13.0±0.9 rosette leaves in wild-type plants (Figure S1A and Figure S2). These plants had in addition twisted leaves (Figure 2), as well as defects in the maturation of anthers (Figure 3F) and in the development and shattering of seeds, which often remained attached to the dehiscent siliques (Figure 3G), resulting in reduced seed production and germination (not shown). This is consistent with what has been observed in plants that express a non-targetable form of the miR167 target ARF6 or ARF8. Such plants have smaller leaves and are often sterile due to defects both in ovule and anther development. Effects on flowering time have not been previously associated with miR167,, and the late-flowering phenotype of MIM167 plants reveals a new role for this miRNA family.

15Two constructs were used to downregulate different subfamilies of miR169 family, whose main targets are HAP transcription factors. MIM169 was designed for miR169a, b, c, h, i, j, k, l, m and n, and MIM169defg for miR169d, e, f and g. Both target mimics reduced the size of transgenic plants (Figure 2).

16MiR170 and miR171 target a group of SCARECROW-like transcription factor genes, and both MIM170 and MIM171A plants had round, pale leaves (Figure 2), as well as defective flowers, with sepals that did not separate properly, resulting in reduced fertility (Figure 3H and 3I). Expression of target mimics against the b and c members of the miR171 family did not confer any phenotype, suggesting less important roles for these two miRNAs.

17MIM172 plants were also late flowering, with 20.0±3.5 (n = 30) rosette leaves in long days (Figure S1B), consistent with the flowering time phenotype of plants that have increased expression of miR172 targets,,. In addition, leaves of MIM172 plants appeared to be somewhat narrower than those of wild type, and mildly curled downward, and severe MIM172 lines presented reduced apical dominance (not shown). In contrast to plants that express a non-targetable version of AP2, flowers of MIM172 plants were normal. These differential effects could be due to the particularly high levels of miR172 levels during early flower development.

18MiR393 targets a small group of auxin receptor genes. MIM393 plants had mild defects in leaf morphology, with narrow leaves that were curled downward (Figure 2). Leaf epinasty is often associated with high auxin levels, and is consistent with an increase of auxin signaling caused by downregulation of miR393 activity.

19Finally, epinastic leaves were observed also in MIM394 plants (Figure 2). MiR394 is predicted to target a gene encoding an F-box protein.

20Artificial target mimics are thought to sequester their target miRNAs, presumably by stably binding to miRNA-loaded RISCs. To obtain additional evidence for such interactions, we embedded a functional MIM159 site in the 3′-UTR of a triple- Enhanced Yellow Fluorescent Protein (EYFP) reporter; stable recruitment of RISCmiR399 to the mimic site could be expected to interfere with EYFP translation. In 80% of MIM159 expressing T1 plants, as in control plants, the EYFP transgene was completely silenced. In the remaining 20%, we detected EYFP signal that was strongly reduced in the region where MIR159 genes are known to be expressed (Figure 4A). In addition, these plants presented the typical phenotypic defects of MIM159 plants, confirming that the EYFP:MIM159 construct functions properly as a target mimic.

21RISCmiRNA sequestration in turn should relieve target genes from miRNA-dependent regulation, resulting in increased levels of the encoded protein. In agreement with such a scenario, activity levels of a genomic MYB33:GUS reporter were markedly increased in MIM159 plants (Figure 4A). In analogy with EYFP:MIM159, reporter activity was increased in the tissues expressing MIR159 genes, as expected.

22Sequestration of RISCmiR399 by the natural target mimic IPS1 prevents miR399-guided cleavage of PHO2 mRNA, thus increasing PHO2 mRNA levels. To assess the effects of artificial target mimics on the levels of mRNA of miRNA target genes, we tested them by reverse transcription followed by quantitative PCR (qRT-PCR) in a subset of MIM lines. We preferentially analyzed organs in which miRNA abundance was high according to the ASRP database,, or organs with major phenotypic alterations in MIM lines. Two independent lines were tested for each construct. Among the miRNA targets, we chose ones known to induce phenotypic defects when expressed as non-targetable forms,, and ones that show altered expression in miRNA biogenesis mutants,,. PCR products spanned the miRNA target sequence, allowing quantification of the attenuation in slicing activity by the corresponding miRNA. Surprisingly, in most cases there were no major changes in target transcript levels (Figure 4B and Figure S3).

23For comparison, we examined the expression of the same miRNA target genes in seedlings of several mutants impaired in small RNA biogenesis and function, including dcl1-100, se-1, hyl1-2 and ago1-27, and in plants overexpressing viral silencing suppressors that are known to counteract the action of the small RNA machinery, including P1/HC-Pro, P0, P19 and p21 –. In most cases, the changes seen in MIM lines correlated with those seen in miRNA biogenesis mutants. Stronger effects were observed only in dcl1-100 plants (Figure 4C). These results are consistent with what has been observed in microarray studies of miRNA biogenesis mutants, including other dcl1 alleles, se and hyl1,.

24As in animals, inhibition of translation is an important component of miRNA function in plants,,. To test whether artificial mimics impact miRNA effects independent of changes in target transcript accumulation, we monitored the protein levels produced by CIP4, a gene that is regulated by miR834 through translational inhibition,. In MIM834 lines, CIP4 levels were appreciably increased, while CIP4 mRNA levels were unchanged (Figure 4D). Direct effects on protein translation could explain the absence of a clear correlation between target mRNA levels and plant phenotype in plants expressing artificial target mimics.

25Finally, we investigated the levels of mature miRNAs in plants expressing artificial target mimics. In all MIM lines we examined, levels of the targeted miRNA were decreased, suggesting that unproductive interaction of RISCmiRNA with a decoy affects miRNA stability (Figure 4E). Although such an effect has not been observed in case of the endogenous IPS1-miR399 interaction, a similar reduction in small RNA levels triggered by a target mimic has been reported in bacteria,.

26We have generated a collection of transgenic plants expressing artificial target mimics designed to reduce activity for most of the known miRNA families in Arabidopsis thaliana. Inhibiting the function of 14 out of 71 miRNA families with target mimics led to morphological abnormalities. All of these families belong to the more abundant and widely conserved miRNA families, which were the first ones to be discovered (Table 1). This agrees with results from experiments in which miRNAs were overexpressed, miRNA target genes were mutated, or miRNA genes were inactivated by conventional knockouts [reviewed in 63]. Together, these findings are consistent with the scenario of frequent birth and death of miRNA genes, with only a few becoming fixed early on during evolution because they acquired a relevant function in plant development,. More recently evolved, species-specific miRNAs could instead play a role in adaptation to certain abiotic or biotic challenges, or have no discernable function at all. Some miRNAs are known to regulate physiological traits, and they do not cause morphological abnormalities under standard benign conditions,,. Such conditional effects would have escaped our screen, as would have defects in the root system of the plant. Moreover, compared to expression of non-targetable forms of miRNA target genes, or miRNA loss-of-function mutants, the defects of MIM plants were often weaker. Examples are the absence of an altered floral phenotype in MIM172 plants, which is seen in plants that express a non-targetable version of AP2 under the control of normal regulatory sequences, or the extra-petals phenotype seen in mir164c mutants, but not in MIM164 plants. Another caveat is that some miRNAs might be required for embryonic development, in which case only lines with relatively weak expression of the artificial target mimic might have survived. Such limitations could be overcome by tissue-specific or inducible expression of target mimics. On the other hand, while artificial mimics increase levels of individual miRNA target genes less strongly than what can be achieved by expression of miRNA-resistant forms, mimics have the advantage that they affect all targets simultaneously. Apart from translational regulation –, another possibility for the absence of a clear correlation between phenotypic severity and change in mRNA levels of miRNA targets could be that many miRNAs affect their targets only in a small set of cells. In these cases, assaying expression in whole organs would obscure the effects of miRNA downregulation on mRNA levels.

27It has recently been suggested that plant miRNAs could also repress the translation of target mRNAs that have only limited sequence complementarity, as often happens in animals. Support for the existence of miRNA binding sites with reduced complementarity in plants comes from an analysis of miR398, which regulates COPPER SUPEROXIDE DISMUTASE (CSD) genes. Certain mutations in the miR398 complementary motif site reduced the effects of miR398 on CSD mRNA, but not on protein levels. We have shown that mimic-like sites, when introduced into the 3′-UTR of a protein-coding gene, not only are active in sequestering the targeted miRNA, but can also reduce protein levels produced by the mRNA linked in cis. This reduction likely occurs at the translational level, since mimic sites are not subject to miRNA-dependent slicing. This observation opens an intriguing scenario in which mRNAs containing mimic-like sites, or possibly other sites with reduced complementarity to miRNAs, are regulated by miRNAs exclusively through translational inhibition. A further level of complexity is added by such sites reducing the effects of an miRNA on other mRNA with a sliceable miRNA targeting motif, similarly to what has been recently proposed in animal systems.

28Nevertheless, as pointed out before, miRNA overexpression and knockout of major target genes normally produce very similar phenotypes, and these are generally the opposite of what is seen in plants with reduced activity of the miRNA. These observations are supported by our finding of extensive similarities between phenotypes caused by target mimics and by expressing resistant forms of individual targets. We conclude that, at least for the instances in which developmental defects could be observed, target genes with extensive complementarity likely account for the majority of miRNA effects, but that in certain cases targets regulated solely through translational inhibition via diverged target sites might be important as well.

29Plants were grown on soil in long days (16 h light/8 hours dark) under a mixture of cool and warm white fluorescent light at 23°C and 65% humidity. The se-1, ago1-27, and hyl1-2 and dcl1-100 mutants have been described –. MIM834 plants were grown on MS media plates supplemented with 1% sucrose for 14 days in long days at 23°C. Plants overexpressing viral proteins Hc-Pro, P0, P19 and P21 were a kind gift from the Carrington lab.

30Artificial target mimics were generated by modifying the sequence of the IPS1 gene. All target mimics constructs were placed behind the constitutive CaMV 35S promoter in the pGREEN vector conferring resistance to BASTA. For the MYB33:GUS reporter, a MYB33 genomic fragment was PCR amplified, cloned into the TOPO-PCR8 Gateway vector (Invitrogen), and recombined through LR clonase reaction into pGWB433 to generate a GUS translational fusion. The MIM159 construct was introduced into three independent MYB33-GUS T2 lines. A MIM159 site was placed in the 3′-UTR of a triple-EYFP sequence linked to a fragment encoding a nuclear localization signal (NLS) and driven by a CaMV 35S promoter. Constructs were introduced into A. thaliana (accession Col-0) plants by Agrobacterium tumefaciens-mediated transformation.

31Nine-day-old seedlings from three independent T2 lines for all the GUS reporter backgrounds were fixed in acetone 90%. GUS activity was assayed as described.

32Total RNA was extracted from 11-day old seedlings and 30-day old inflorescences (47 days for the MIM172 lines), using TRIzol Reagent (Invitrogen). For dcl-100, 13-day old seedlings were collected, to obtain a similar developmental stage compared to the other plants. For real time RT-PCR, two biological replicates with tissue pooled from 8 to 10 plants were assayed from two independent MIM lines per miRNA family or subfamily. Complementary DNA was produced with the RevertAid First Strand cDNA Synthesis Kit (Fermentas), using as starting material 4 µg of total RNA that had been treated with DNase I (Fermentas). PCR was carried out in presence of SYBR Green (Invitrogen) and monitored in real time with the Opticon Continuous Fluorescence Detection System (MJR). Oligonucleotide primers are given in Table S2. Small RNA blots were performed on the same RNA used as template for real time RT-PCR, with DNA oligonucleotides as probes.

33Proteins were extracted from four MIM834 lines using a Tris buffer (50 mM Tris pH 7,5; 150 mM NaCl; 1 mM EDTA; 10% [v/v] Glycerol; 1 mM DTT; 1 mM Pefablock and 1 complete protease inhibitor cocktail [Roche]). Protein concentration was measured using a commercial Bradford assay (BioRad). 50 µg of raw protein extract per sample were resolved on an 8% acrylamide gel. Blotting and antibody incubation were performed as described, except that the secondary antibody was incubated for 8 hours at 4°C. Two biological replicates from 4 independent lines were analyzed. Band intensity was measured using the ImageJ software (http://rsbweb.nih.gov/ij/).