The mirtron pathway generates microRNA-class regulatory RNAs in Drosophila

Abstract

The canonical microRNA (miRNA) pathway converts primary hairpin precursor transcripts into approximately 22 nucleotide regulatory RNAs via consecutive cleavages by two RNase III enzymes, Drosha and Dicer. In this study, we characterize Drosophila small RNAs that derive from short intronic hairpins termed "mirtrons." Their nuclear biogenesis appears to bypass Drosha cleavage, which is essential for miRNA biogenesis. Instead, mirtron hairpins are defined by the action of the splicing machinery and lariat-debranching enzyme, which yield pre-miRNA-like hairpins. The mirtron pathway merges with the canonical miRNA pathway during hairpin export by Exportin-5, and both types of hairpins are subsequently processed by Dicer-1/loqs. This generates small RNAs that can repress perfectly matched and seed-matched targets, and we provide evidence that they function, at least in part, via the RNA-induced silencing complex effector Ago1. These findings reveal that mirtrons are an alternate source of miRNA-type regulatory RNAs.

Figure 1. Essential Characteristics of Mirtrons and Pri-miRNAs

(Top) A typical pri-miRNA transcript contains a “lower stem,” which mediates recognition and cleavage by Drosha (blue arrows). The resulting pre-miRNA hairpin is then cleaved by Dicer-1 (green arrows) to yield a ∼22 nt duplex with ∼2 nt 3′ overhangs. Depending on the miRNA precursor, either the 5′ or the 3′ hairpin product is preferentially transferred into an effector Ago complex; here, it is depicted as the 5′ arm. This diagram was modeled after Han et al. (2006). (Bottom) A typical mirtron locus lacks the “lower stem” found in pri-miRNAs but is instead constrained by primary nucleotide motifs that mediate their recognition and cleavage by the splicing machinery (i.e., 5′ GURAGU splice donors, 3′ polypyrimidine tracts, and CAG [or much less frequently UAG] splice acceptor sites). Once the spliced mirtron is debranched, it can adopt a pre-miRNA-like hairpin structure and be cleaved by Dicer-1. In contrast to canonical miRNA hairpins, mirtron hairpins are strongly biased to exhibit preferential stability of small RNAs from their 3′ arms.

Figure 2. Evolutionary Characteristics of Mirtrons

(A) Example of a well-conserved mirtron locus, mir-1003. Alignments and conservation data were produced by the UCSC Genome Center (http://genome.ucsc.edu/). Gold tracks depict the positions of a subset of sequenced small RNAs mapping to this locus, the blue track depicts the exon/intron structure of its host CG6695, and the black tracks at bottom depict nucleotide conservation of this region across 12 Drosophilid species. Greater height of the “conservation” track reflects deeper sequence conservation. The mir-1003 mirtron is highly constrained, with perfect conservation of the miR-1003 seed (positions 1–8, red box). Down-stream of the seed, positions 9–13 have undergone significant divergence. Note also that the terminal loop region (red arrow) exhibits accelerated divergence relative to the mirtron hairpin arms. (B) Example of a poorly conserved mirtron locus, mir-1008. Its sequence is only preserved among melanogaster subgroup species and thus arose sometime in the last <10 million years. Despite its rapid evolution, this mirtron still exhibits accelerated divergence in the terminal loop region (red arrow).

Figure 3. Distinct Temporal and Spatial Expression of Endogenous Mirtrons

Northern blots were prepared using RNA from 0–24 hr embryos (E), third-instar larvae, and 0–2 day pupae (LP), adult males and females (A), and S2 cells (S2) and probed with antisense LNA probes to the 3′ ends of several mirtrons. Endogenous ∼21∼24 nt RNAs and ∼55–70 nt precursors were detected in all cases. Blots were stripped and reprobed for 30 nt 2S rRNA as a loading control. RNA sizes were judged with reference to a Decade RNA marker (Ambion) run in parallel. Lengths of the mirtron hairpins inferred from intron boundaries are mir-1003 (56 nt), mir-1010 (71 nt), and mir-1008 (57 nt).

Figure 4. Structure-Function Analysis of Mirtron Biogenesis

(A)—(D) On the left are four constructs for mirtron expression. In (A), ∼400 nt pri-mirtrons containing their flanking endogenous exons (gray boxes) were cloned into the 3′ UTR of UAS-DsRed. In (B), ∼55-70 nt mirtron hairpins were cloned between coding exons for DsRed and 2xmyc, thereby separating them from endogenous exonic context. In (C), using construct B as a template, a G -> C point mutation in the first base of the 5′ splice donor was introduced (asterisk and arrow). In (D), the mature ∼22 nt mirtron product was substituted into a miRNA precursor based on pri-mir-6-1 and then inserted into the 3′ UTR of UAS-DsRed. On the right, northern analysis of mirtrons cloned into constructs (A)—(D), as depicted on the left, and activated in S2 cells using ub-Gal4. Lanes 2 and 6 show that miR-1003, miR-1008, and their associated mirtron hairpins were efficiently generated from mirtron genomic fragments using construct (A), (compare with lanes 1 and 6 for endogenous mirtron expression; note that the blot has been underexposed relative to blots in Figure 3). miR-1003 and miR-1008 and their mirtron hairpins were also readily expressed from construct B (lanes 3 and 8), demonstrating that endogenous flanking exons are dispensable for entry into the mirtron pathway. Point mutation of the splice donor site demonstrated that splicing is necessary to generate the mirtron hairpin (lanes 4 and 9). Mature miR-1003 could also be expressed by reprogramming a canonical pri-miRNA (lane 5). Since the mir-6-1 hairpin is 63 nt, the hybrid pri-mir-6-1/mir-1003 hairpin is slightly longer than 56-nt-long mirtron for miR-1003 (as indicated by arrows). Blots were stripped and reprobed for 2S rRNA as a loading control.

Figure 5. Mirtron Biogenesis Involves a Hybrid Pathway that Couples Splicing and Dicing

(A) S2 cells were treated with the indicated dsRNAs and transfected with ub-Gal4 and UAS-DsRed-small-RNA plasmids as noted. Total RNA was then extracted and analyzed for small-RNA expression by northern blot; these were stripped and reprobed to detect 2S rRNA as a loading control. Lanes 1–5: processing of UAS-DsRed-mir-1. Drosha knockdown decreased the amount of hairpin pre-miR-1 and mature miR-1 (lane 2). Ldbr (lariat-debranching enzyme) knockdown had little effect (lane 3). Dicer-1 knockdown caused a profound increase in pre-miR-1 hairpin (lane 4), while Dicer-2 knockdown had little effect (lane 5). Lanes 6–10: processing of endogenous mir-1003 mirtron. Its accumulation was largely unaffected by Drosha knockdown (lane 7), but Ldbr knockdown abolished the accumulation of mirtron and mature forms of miR-1003 (lane 8). Knockdown of Dicer-1 (lane 9) but not Dicer-2 (lane 10) depleted mature miR-1003 and resulted in the accumulation of its precursor mirtron. Lanes 11–15 depict processing of UAS-DsRed-mir-1003. Its genetic requirements were similar to those of its endogenous counterpart. Lanes 16–18: processing of UAS-DsRed-pri-mir-6-1/mir-1003. Expression of miR-1003 from a pri-miRNA backbone was similar to that of miR-1, in that it was now strongly dependent on Drosha (lane 17) and appeared not to require Ldbr (lane 18). Lanes 19–23: processing of a UAS-DsRed-mir-1010 mirtron construct. Similar genetic requirements were seen as for the mir-1003 mirtron. (B) Mirtron biogenesis requires intron debranching in intact animals. Larvae carrying da-Gal4 and UAS-Ldbr^RNAi exhibited a marked decrease in the levels of mir-1010 mirtron and mature product compared to wild-type Canton S larvae. The same blot was stripped and reprobed for miR-1, which showed only minor alteration in its steady-state level. The reduction in miR-1010, as normalized to miR-1, was 67 ± 2% SD (see Figure S3). (C) Mirtron maturation requires Exportin-5 and loqs. Treatment with either of two nonoverlapping dsRNAs against Drosophila Exportin-5 (Exp5-N, Exp5-C) induced substantial loss of mirtron hairpins and small-RNA products for endogenous mir-1003 (lanes 1–3) and mir-1008 (lanes 5–7). Additional Exportin-5 knockdown data are reported in Figure S4. Treatment with loqs dsRNA increased the steady-state level of endogenous mir-1003 mirtron hairpin and decreased its mature form (lane 4); similar results were obtained for endogenous mir-1008 mirtron (lane 8). (D) Endogenous ∼22 nt small RNAs derived from mirtrons mir-1003 and mir-1010 associate with endogenous Ago1. Coimmunoprecipitation (coIP) assay from 0–10 hr embryos using control mouse α-T7 or mouse α-Ago1. Fifteen percent of the RNA input and supernatants, twenty percent of the protein input and supernatants, and one hundred percent of the IP fractions were loaded. Ago1 lane is a western blot; other lanes are northern blots for the indicated small RNAs. The absence of miRNAs from control IPs, along with absence of 2S rRNA from Ago1 IPs, provided evidence for specific recovery of mirtron-derived miRNAs in the Ago1 IP fraction (an arrow marks the relevant lanes).

Figure 6. Mirtrons Repress Both Perfect-Match and Seed-Match Targets

(A) Luciferase sensor assays in S2 cells. Sensors contain two antisense copies of miR-1003, miR-1004, and miR-1010 cloned downstream of renilla luciferase in psiCHECK2; this vector also carries a control firefly luciferase gene. miR-1003 and miR-1004 share the same seed (boxed), and mutant sensors are mispaired at the 3 nt positions highlighted. Sensor plasmids were cotransfected with ub-Gal4 and UAS-DsRed or UAS-DsRed-mirtron plasmids as indicated (see key, inset). Luciferase activity was expressed as the mean ratio of the experimental renilla/firefly luciferase sensor value in the presence of DsRed + mirtron relative to DsRed alone. These data were pooled from quadruplicate transfections, and error bars represent the standard deviations. The following relevant comparisons exhibited p < 0.0001 (equal variance Student’s t test): miR-1003 sensor versus miR-1003 mut sensor (lane 5 versus lane 9), repression of miR-1003 sensor by mir-1003 or mir-1004 (lane 5 versus lane 6 or 7), repression of miR-1004 sensor by mir-1003 or mir-1004 (lane 12 versus lane 13 or 14), and repression of miR-1010 sensor by mir-1010 (lane 19 versus lane 22). Lanes 1–4: empty sensor. DsRed-mirtron expression had little intrinsic effect on psiCHECK2-luciferase activity. Lanes 5–8: miR-1003 sensor. This sensor was strongly repressed relative to empty vector (compare lane 5 with lane 1). Expression of mir-1003 (6) and mir-1004 (7) but not mir-1010 (8) mirtron constructs further reduced its activity. Lanes 9–11: miR-1003 mut sensor. Mutation of its seed-complementary region elevated its basal expression, indicating relief from repression by endogenous miR-1003. Such mutations also abolished its response to mir-1003 and mir-1004 mirtron-expression constructs. Lanes 12–15: miR-1004 sensor. This sensor was robustly inhibited by mir-1004 (14), mildly suppressed by mir-1003 (13), and largely unaffected by mir-1010 (15) mirtron-expression constructs. Lanes 16–18: miR-1004 mut sensor. Mutation of its seed-complementary region eliminated its response to ectopic mir-1003 and mir-1004. Lanes A19–22: miR-1010 sensor. This sensor was specifically repressed by mir-1010 (22) but not mir-1003 (20) or mir-1004 (21) mirtron-expression constructs. (B) Ago knockdowns in S2R+ cells, followed by transfection with miRNA and mirtron-expression constructs and sensors. The miR-1010 mi sensor contains centrally placed mismatches, as indicated, to render it a miRNA-type sensor. These data were pooled from two independent sets of quadruplicate transfections (n = 8), and error bars represent the standard deviations. The following comparisons exhibited p < 0.00001 (equal variance Student’s t-Test): derepression of nerfin sensor by Ago1 dsRNA (lane 1 versus lane 2) and derepression of miR-1010 mi sensor by Ago1 dsRNA (lane 4 versus lane 5). Lanes 1–3: Ago1 dsRNA, but not GFP or Ago2 dsRNA, derepressed the nerfin 3′ UTR sensor in the presence of ectopic miR-279. Lanes 4–6: Ago1 dsRNA, but not GFP or Ago2 dsRNA, derepressed the 4xmiR-1010 mi sensor in the presence of ectopic miR-1010. (C—H) GFP expression is shown in grayscale in panels (C), (E), and (G) and as a merge (in green) with DsRed/miR-1004 expression (in red) in panels (D) (F), and (H). In (C) and (D), GFP-miR-7 sensor activity, which can be completely abolished by ectopic expression of miR-7 (Lai et al., 2005; Stark et al., 2003), was unaffected by expression of mir-1004. In (E) and (F), GFP-miR-1004 sensor activity was strongly repressed (asterisk) in cells that express the mir-1004 mirtron. In (G) and (H), the GFP-miR-1003, seed-paired, sensor was weakly repressed (asterisk) in cells that express mir-1004.

Figure 7. Model for the Convergence of the Mirtron and Canonical miRNA Pathways

The canonical miRNA pathway initiates with the recognition and cleavage of pri-miRNA transcripts by the Pasha/Drosha complex to yield pre-miRNA hairpins. Our data support the existence of an alternate pathway in which short introns with hairpin potential are spliced and debranched to yield mirtron hairpins. Both pre-miRNA and mirtron hairpins are exported from the nucleus by Exportin-5 and cleaved by Dicer-1/loqs to generate ∼22 nt RNA duplexes. One strand, the active miRNA, is transferred to an Ago complex and guides it to repress fully complementary or seed-matched target transcripts.