Dicer-TRBP complex formation ensures accurate mammalian microRNA biogenesis

Abstract

RNA-mediated gene silencing in human cells requires the accurate generation of ∼22 nt microRNAs (miRNAs) from double-stranded RNA substrates by the endonuclease Dicer. Although the phylogenetically conserved RNA-binding proteins TRBP and PACT are known to contribute to this process, their mode of Dicer binding and their genome-wide effects on miRNA processing have not been determined. We solved the crystal structure of the human Dicer-TRBP interface, revealing the structural basis of the interaction. Interface residues conserved between TRBP and PACT show that the proteins bind to Dicer in a similar manner and by mutual exclusion. Based on the structure, a catalytically active Dicer that cannot bind TRBP or PACT was designed and introduced into Dicer-deficient mammalian cells, revealing selective defects in guide strand selection. These results demonstrate the role of Dicer-associated RNA binding proteins in maintenance of gene silencing fidelity.

Figure 1

Structure of the Dicer–TRBP interface. (A) Cartoon representation of the primary sequence of Dicer and TRBP with brackets indicating the interacting domains. (B) Overlaid backbone cartoon and surface representations of the Dicer partner-binding domain (PBD) and the third dsRBD of TRBP. (C) Front and back views with interfacial residues shown. Dicer residues mutated to abrogate TRBP and PACT binding are shown in pink with resulting residues indicated in parentheses. See also Figure S1.

Figure 2

A single surface of Dicer binds TRBP or PACT. (A) A homology model of PACT₃ based on the TRBP₃ structure reveals conservation of interfacial residues (white, identical or similar; orange, dissimilar). Mutated Dicer residues are shown as in Figure 1C. (B) Sequence alignment comparing TRBP and PACT in human and mouse. Black, identical residues; grey, similar residues; orange, dissimilar residues; dots, TRBP residues located within 5 Å of Dicer in the crystal structure. (C) A view of interfacial contacts between TRBP and Dicer, showing Dicer residues targeted for mutation as in 1C. (D) MBP-tagged TRBP or PACT was used to pull down WT Dicer or Dicer_mut, demonstrating the latter protein's lack of affinity for dsRBP partner proteins. See also Figure S2.

Figure 3

Roles of Dicer partner proteins TRBP and PACT in miRNA biogenesis. (A) Correlation of strand selection behavior (scored as log〈5′ arm coverage/3′ arm coverage〉 among Ago2-associated miRNA) between WT MEF (X axis) and rescue conditions (Y axis) with either WT Dicer or Dicer_mut demonstrates the importance of TRBP and PACT in maintaining fidelity of strand selection. In the Dicer_mut condition, an increased deviation from the diagonal is observed due to impaired strand selection fidelity. Labels denote the 14 miRNA duplexes most dramatically affected (to >1 standard deviation) by the loss of TRBP and PACT recruitment to Dicer. 23 candidate duplexes varying in strand selection behavior by more than one standard deviation between MEF and WT rescue conditions (shown in grey) are not considered based on the possibility that they are behaving aberrantly due to the Dcr^-/- MEF context. The blue open circle represents miR-132. (B) Cluster analysis of miRNA lengths observed in WT MEF cells or Dicer knockout cells rescued with either WT Dicer or Dicer_mut, revealing an increased propensity for formation of 22 nt products instead of 21 nt products when Dicer can recruit TRBP and PACT for certain sensitive miRNAs (pink boxes). Ordering is based on results obtained in the WT Dicer rescue condition. See also Figure S3.

Figure 4

Impact of TRBP on Dicer cleavage position of miR-30e. (A) In vitro dicing of pre-miR-30e shows a dependence on TRBP for consistent cleavage of a 24 nt product. (B) Quantification of isomiR distributions resulting from dicing assays of pre-miR-30e. Error bars represent standard deviation based on triplicate experiments. (C) Schematic showing the effect of altered dicing on pre-miR-30e and ramifications for downstream Ago2 loading. See also Figure S4.

Figure 5

The absence of inter-domain interactions in TRBP. ¹H-¹⁵N HSQC spectra of individual TRBP dsRBD constructs (dsRBD1, red, top; dsRBD2, green, middle; TRBP₃ which includes dsRBD3, blue, bottom) show no pronounced chemical shift perturbations when overlaid with a spectrum of full-length TRBP containing all three dsRBDs (black), indicating the absence of inter-domain interactions. Asterisks mark peaks assigned via their diagnostic intensity to the C-termini of the dsRBD1 or dsRBD2 constructs, expected to be shifted in the spectrum of the full-length TRBP due to change in chemical environment. Brackets mark regions wherein the full-length TRBP spectrum is displayed at a slightly lower threshold to show low-intensity peaks. See also Figure S5.

Figure 6

Mechanisms of Dicer partner proteins in the context of the miRNA biogenesis machinery. (A) The human Dicer architecture (as determined by electron microscopy) is colored according to functional domains (PAZ, pink; RNase IIIa/b, yellow; helicase, green), with the Dicer–TRBP interface structure determined in the present crystallographic work shown in dark green (Dicer_PBD) and cyan (TRBP₃). NMR results suggest that the two N-terminal RNA-binding domains of an extended TRBP can readily access an RNA bound near the paired RNase III active sites of Dicer. (B) Models for how Dicer partner proteins contribute to isomiR formation (top) and strand selection fidelity during transfer of a Dicer product duplex to Argonaute (bottom). See also Figure S6.