Comprehensive prediction of novel microRNA targets in Arabidopsis thaliana

Abstract

MicroRNAs (miRNAs) are 20-24 nt long endogenous non-coding RNAs that act as post-transcriptional regulators in metazoa and plants. Plant miRNA targets typically contain a single sequence motif with near-perfect complementarity to the miRNA. Here, we extended and applied the program RNAhybrid to identify novel miRNA targets in the complete annotated Arabidopsis thaliana transcriptome. RNAhybrid predicts the energetically most favorable miRNA:mRNA hybrids that are consistent with user-defined structural constraints. These were: (i) perfect base pairing of the duplex from nucleotide 8 to 12 counting from the 5'-end of the miRNA; (ii) loops with a maximum length of one nucleotide in either strand; (iii) bulges with no more than one nucleotide in size; and (iv) unpaired end overhangs not longer than two nucleotides. G:U base pairs are not treated as mismatches, but contribute less favorable to the overall free energy. The resulting hybrids were filtered according to their minimum free energy, resulting in an overall prediction of more than 600 novel miRNA targets. The specificity and signal-to-noise ratio of the prediction was assessed with either randomized miRNAs or randomized target sequences as negative controls. Our results are in line with recent observations that the majority of miRNA targets are not transcription factors.

Figure 1.

Distribution of mfe values of previously validated miRNA targets (Supplementary Table S2), given as percentage of a perfect-match hybrid, calculated with RNAhybrid.

Figure 2.

Calculated miRNA:mRNA hybrid structures for selected examples of novel miRNA targets from our prediction. Twenty-six structures of miRNA:mRNA hybrids predicted with RNAhybrid are presented. The miRNA binding site of the target mRNA is shown on top, the complementary miRNA as the bottom strand, calculated mfe values are given to the right (kcal/mol). AGI designation and, if applicable, the name of the gene as well as the designation of the miRNA (miRBase) are shown to the left.

Figure 3.

Analysis of gene ontology (GO) annotation terms for molecular function categories. The percentage of GO annotation terms for each category was normalized to the percentage of this category in the whole genome (black bars), which was set to 1. Novel predicted miRNA targets found with our approach are given as grey bars, previously predicted/validated targets are given as white bars. Hatched bars show the distribution among GO annotation terms of all miRNA targets predicted in this work, i.e. novel predicted and previously predicted/validated targets.

Figure 4.

Detection of mature miRNAs in protoplasts. Protoplasts from Arabidopsis AT7 or tobacco BY-2 cell suspension cultures were transfected with plasmids harboring the precursors of (A) ath-MIR161, (B) ath-MIR156h, (C) ath-MIR414, (D) ath-MIR159a, (E) ath-MIR172a and (F) ath-MIR395b under the control of the 35S promoter. Total RNA was extracted and Northern blots to detect mature miRNAs were prepared from denaturating polyacrylamide gels. In each lane, 20 μg of total RNA was loaded from transfected (+) or untransfected (−) protoplasts. A positive control for transfer and hybridization, consisting of a DNA oligonucleotide with the same sequence as the corresponding mature miRNA, was included in all experiments (only shown in A, C and E). U6snRNA was used as loading control. An RNA oligonucleotide of 21 nucleotides in length was used as size marker. The position corresponding to 21 nucleotides is indicated.

Figure 5.

Five out of ten targets for four different miRNAs that were confirmed by validation experiments. MYB101 and MYB125 have been predicted previously (28). MYB101, MRG1 and ACS8 were predicted with our stringent parameters that we finally applied, MYB125 and GAE1 were predicted using less stringent parameters. During progress of our work, MYB101, MYB125 and MRG1 were also validated by others (24,31,55,56). Total RNA was extracted from young inflorescences of Arabidopsis wild-type plants or from AT7 protoplasts that were co-transfected with plasmids harboring the cDNA of the target and the precursor of the corresponding miRNA under the control of the 35S promoter. cDNAs were synthesized after ligation of an RNA linker to 5′RNA ends retaining a phosphate group. After amplification by two nested PCR reactions, DNA fragments were cloned and sequenced. Each panel shows part of the target mRNA (top) with the corresponding miRNA annealed to it (below). The name and Arabidopsis gene identifier for the target mRNA as well as the designation of the miRNA are given to the left. Watson–Crick base pairs are indicated with vertical dashes, G–U base pairs are indicated with a colon. Arrows indicate the cleavage sites and the corresponding relative numbers of analyzed 5′ RACE products.

Figure 6.

Five out of ten targets for four different miRNAs that were not confirmed by validation experiments. These targets were predicted using less stringent parameters than those that were finally applied. Therefore, they are not listed in the results of our final prediction. They served as experimental controls for the parameters of our final prediction. Experiments were done in parallel with those shown in Figure 5. MYB97 has been predicted previously (19). Hybrid structures calculated by RNAhybrid are shown, and the name and Arabidopsis gene identifier for the target mRNA as well as the designation of the miRNA are given to the left. The mfe (kcal/mol) of each hybrid is given to the right.