Arabidopsis lyrata small RNAs: transient MIRNA and small interfering RNA loci within the Arabidopsis genus

Abstract

Twenty-one-nucleotide microRNAs (miRNAs) and 24-nucleotide Pol IV-dependent small interfering RNAs (p4-siRNAs) are the most abundant types of small RNAs in angiosperms. Some miRNAs are well conserved among different plant lineages, whereas others are less conserved, and it is not clear whether less-conserved miRNAs have the same functionality as the well conserved ones. p4-siRNAs are broadly produced in the Arabidopsis genome, sometimes from active hot spot loci, but it is unknown whether individual p4-siRNA hot spots are retained as hot spots between plant species. In this study, we compare small RNAs in two closely related species (Arabidopsis thaliana and Arabidopsis lyrata) and find that less-conserved miRNAs have high rates of divergence in MIRNA hairpin structures, mature miRNA sequences, and target-complementary sites in the other species. The fidelity of miRNA biogenesis from many less-conserved MIRNA hairpins frequently deteriorates in the sister species relative to the species of first discovery. We also observe that p4-siRNA occupied loci have a slight tendency to be retained as p4-siRNA loci between species, but the most active A. lyrata p4-siRNA hot spots are generally not syntenic to the most active p4-siRNA hot spots of A. thaliana. Altogether, our findings indicate that many MIRNAs and most p4-siRNA hot spots are rapidly changing and evolutionarily transient within the Arabidopsis genus.

Figure 1.

Less Conserved MIRNAs Are Often Species Specific, Weakly Expressed, and Encoded by Single Loci. (A) Identification of homologous MIRNA loci (left panel) and families (right panel) in A. thaliana and A. lyrata. Only loci/families that passed the expression criteria in at least one species were considered. MC, more conserved; LC, less conserved. (B) Cumulative distributions of the number of paralogous loci per miRNA family. (C) Cumulative distributions of the number of sequencing reads per A. thaliana miRNA family (based on nine sRNAseq data sets totaling ∼1.6 × 10⁷ reads; Table 1). (D) As in (C) for A. lyrata miRNA families (based on ∼4.8 × 10⁷ reads; Table 1).

Figure 2.

Predominant Lengths and 5′ Nucleotides Produced by A. thaliana and A. lyrata MIRNA Hairpins. (A) Proportions of sRNAseq reads of the indicated lengths from more conserved (MC) families. Families are grouped according to the most abundant small RNA length, as indicated below the chart. Percentages indicate the percentage of families dominated by small RNAs of the indicated length for a given species. (B) As in (A) for less conserved (LC) families. (C) As in (A) for 5′ nucleotides of reads from MC families. (D) As in (A) for 5′ nucleotides of reads from LC families.

Figure 3.

Less Conserved miRNAs Diverge More between A. thaliana and A. lyrata Than Do More Conserved miRNAs. (A) A sketch showing the five regions of MIRNA hairpins that were analyzed. For convenience, the mature miRNA is shown on the 5′ arm, although in reality it can be either on the 5′ arm or 3′ arm. (B) Average sequence divergence between more conserved A. thaliana and A. lyrata MIRNA hairpins. Both 5′-arm and 3′-arm mature miRNAs were tallied and displayed together. To account for differences in lengths among the population of hairpins, the five regions were each scaled to seven bins. Bars indicate the standard errors of the means. (C) As in (B) for less conserved MIRNAs. (D) As in (B) for each nucleotide position within mature miRNAs from more conserved families. (E) As in (D) for less conserved families.

Figure 4.

Less Conserved MIRNAs Tend to Be Processed Imprecisely. (A) Scatterplot of A. lyrata versus A. thaliana MIRNA processing precisions for more conserved MIRNAs. Green lines show the precision value of 0.25, which we used as a cutoff for determining miRNA-like expression patterns (Meyers et al., 2008). (B) As in (A) for less conserved MIRNAs.

Figure 5.

Targets of Less Conserved miRNAs Are Difficult to Identify and Inconsistent between A. thaliana and A. lyrata. (A) Cumulative distributions of the number of miRNA families with the indicated target prediction scores. The lowest scoring prediction (i.e., the most confident prediction) for each family was used. MC, more conserved; LC, less conserved. The gradient of shading indicates increasingly less confident predictions, beginning at a score of 3. (B) miRNA target predictions by family. The number of predicted targets found only in A. thaliana (Ath), only in A. lyrata (Aly), or syntenic homologs predicted in both species are shown. Families without any predicted targets in either species are omitted, as are families that were expressed only in a single species. (C) Sliced targets confidently found by degradome sequencing. Sliced targets were those that were found in both biological replicate degradome libraries for the given species. (D) As in (B) for degradome-confirmed targets.

Figure 6.

24-Nucleotide RNA Expression and Hot Spots Frequently Differ between A. thaliana and A. lyrata. (A) Log₂ ratios of observed to expected overlaps between 24-nucleotide small RNA occupied 1-kb bins for all pairwise comparisons between various A. thaliana and A. lyrata small RNA samples. F1-F4, floral; Sq1, silique; Se1-Se2, seedlings; L1-L2, rosette leaves. (B) As in (A) for 21-nucleotide small RNA occupied bins. (C) Overlaps between the top 100 24-nucleotide small RNA expressing 1-kb bins for all pairwise comparisons between various A. thaliana and A. lyrata small RNA samples. (D) As in (C) for the top 100 21-nucleotide small RNA expressing loci. (E) Percentages of A. thaliana 21- and 24-nucleotide small RNA occupied bins that were ambiguously aligned in the A. thaliana–A. lyrata whole genome alignment. Dashed line indicates the percentage of the entire A. thaliana genome that was ambiguously aligned. (F) As in (E) for the top 100 A. thaliana 21- and 24-nucleotide small RNA hot spots. (G) Fractions of 24-nucleotide small RNAs that mapped to annotated genes, compared with the overall fraction of genomic nucleotides overlapping with gene annotations in A. thaliana and A. lyrata. A. thaliana sRNAseq data were the combination of all nine sRNAseq libraries, and A. lyrata sRNAseq data were the combination of all three sRNAseq libraries (Table 1).