Phase Transitions in the Assembly and Function of Human miRISC

Abstract

miRISC is a multi-protein assembly that uses microRNAs (miRNAs) to identify mRNAs targeted for repression. Dozens of miRISC-associated proteins have been identified, and interactions between many factors have been examined in detail. However, the physical nature of the complex remains unknown. Here, we show that two core protein components of human miRISC, Argonaute2 (Ago2) and TNRC6B, condense into phase-separated droplets in vitro and in live cells. Phase separation is promoted by multivalent interactions between the glycine/tryptophan (GW)-rich domain of TNRC6B and three evenly spaced tryptophan-binding pockets in the Ago2 PIWI domain. miRISC droplets formed in vitro recruit deadenylation factors and sequester target RNAs from the bulk solution. The condensation of miRISC is accompanied by accelerated deadenylation of target RNAs bound to Ago2. The combined results may explain how miRISC silences mRNAs of varying size and structure and provide experimental evidence that protein-mediated phase separation can facilitate an RNA processing reaction.

Keywords: Argonaute; GW182; TNRC6; miRISC; miRNA; microRNA; molecular condensation; phase separation.

Figure 1. Structure of the trp-binding region in Ago2

(A) Cartoon representation of Ago2 illustrates the position of the trp-binding region. (B) Close up view of the trp-binding region. Fo-Fc tryptophan omit map, contoured at 2.5 σ, (green mesh) shows well-ordered indole side chains of three bound tryptophan molecules. (C) Surface representation of the region illustrates the three trp-binding pockets. Curved lines indicate approximate distances between adjacent pockets. See also Figure S1 and Table S1.

Figure 2. Dissection of interactions between Ago2 and TNRC6B

(A) Linear schematic of TNRC6B domain structure. MBP fusion of the ABD used in pull down assays, and sequence context of motif I (W623 and W634) are indicated. (B) Coomassie stained SDS gel showing pull down of ABD variants by wild type (WT) Ago2. AllA is an ABD variant in which all trp residues have been replaced with alanine. Ago2, wild type ABD variants, and AllA ABD variants are indicated by single, double, and triple asterix, respectively. (C) Schematic of trp-binding pockets in Ago2 with residues mutated to disable individual pockets indicated in blue (K660S, P590G, and R688S mutations in pockets 1, 2 and 3, respectively). (D) Pull down of ABD variants by Ago2 with individual trp-binding pockets inactivated. (E) Pull down of ABD by Ago2 with combinations of trp-binding pockets inactivated. (F) Schematic of ABD construct with sequence context motif II (W875, W896, and W910) indicated. (G) Pull down of W875, W896, and W910 AllA-ABD variants by WT Ago2, and (H) by trp-binding pocket Ago2 mutants. See also Figure S2.

Figure 3. Ago2 drives phase separation of the TNRC6B-ABD

(A) Initial observation of ABD-Ago2 phase separation. Solutions containing the TNRC6B ABD become turbid upon introduction of Ago2. Solutions of the AllA-ABD mutant remain clear. (B) Turbidity, measured by absorbance of 480 nm light, of solutions containing Ago2 (20 µM), ABD (40 µM), or ABD + Ago2 (40µM and 20µM, respectively) plotted as a function of temperature. (C) Turbidity versus temperature for ABD samples (10 µM) with various concentrations of Ago2. (D) Light microscopy images of mixtures of ABD (20 µM) and Ago2 taken at room temperature (~23 °C). Scale bar, 10 µm. (E) Time-lapse images showing fusion of three adjacent ABD-Ago2 droplets (left), and confocal microscopy images showing fusion of ABD-Ago2 droplets containing Alexa-488 labeled ABD (right). (F) Fluorescence microscopy images from a FRAP experiment in which an entire ABD-Ago2 droplet was bleached. 10% of Ago2 molecules were labeled with TMR, and ~10% ABD molecules were labeled with Alexa Fluor 488. (G) FRAP recovery curves for three ABD-Ago2 droplets with error bars indicating SEM. All droplets were formed in 100 mM KOAc and 20 nM NaCl. See also Figures S3, and Movies S1 and S2.

Figure 4. Phase separation of miRISC

(A) Fluorescence microscopy images showing Ago2 (TMR labeled) promotes phase separation of full length TNRC6B (Alexa Fluor 488 labeled) in vitro. (B) Spherical miRISC droplets formed in vitro coalesced into larger clusters over time. (C) Representative fluorescence microscopy images from miRISC FRAP experiments. (D) FRAP recovery curves for three miRISC droplets with error bars indicating SEM. (E) Live cell images showing fusion of two GFP-TNRC6B cytoplasmic foci in HEK 293 cells over time. (F) Representative live cell images of GFP-TNRC6B cytoplasmic foci FRAP experiments. (G) FRAP recovery curves for three GFP-TNRC6B cytoplasmic foci with error bars indicating SEM. (H) Representative live cell images of GFP-TNRC6B/mCherry-Ago2 cytoplasmic foci FRAP experiments. (I) FRAP recovery curves for four GFP-TNRC6B/mCherry-Ago2 cytoplasmic foci with error bars indicating SEM. See also Figure S4, and Movies S3, S5 and S6.

Figure 5. Biochemical characterization of Ago2-TNRC6B droplets

(A) Ago2-TNRC6B droplets can be separated from the bulk solvent by centrifugation. Cartoon schematic of procedure (left), and images of droplets in input and supernatant fractions (right). (B) Droplets recruit full-length TNRC6B, Ago2, and miRNA target RNAs. TNRC6B (~1 mM, partially purified) was mixed with Ago2 (0.5 µM) loaded with either let-7 or miR122, and a ³²P-labeled let-7 target RNA (8xlet7 target, ~3 nM). After centrifugation, supernatant and pellet fractions were analyzed by Coomassie stained SDS PAGE (right, top panel) and phosphorimaging of a denaturing gel (right, bottom panel). (C) Ago2 remains active in the separated phase. TNRC6B (~ 1 µM) was mixed with Ago2-miR122 (250 nM) and a ³²P-labeled target RNA (~0.5 µM) with perfect complementarity to miR122 in the absence of divalent cations. After centrifugation MgCl₂ (3 mM) was added to the separated phase. Target RNA was extracted and analyzed by denaturing PAGE and phosphorimaging (right panel). (D) TNRC6B-Ago2 droplets recruit other miRISC components. TNRC6B (40 nM) was mixed with Ago2 (40 nM) and soluble lysate from HEK 293 cells (OD₂₆₀ ~3). Input, supernatant, and pellet fractions were analyzed by Western blot (right panel). See also Figure S5.

Figure 6. Efficient target deadenylation in miRISC droplets

(A) Schematic of experiment. (B) Deadenylation of a target RNA by miRISC. Ago2 (40 nM, final concentration), loaded with either let7 or miR122 (negative control), was mixed with a ³²P-5'-cap-labeled target RNA harboring binding sites for let-7 and a 114 nt. poly(A) tail in the presence of soluble lysate from HEK 293 cells (OD₂₆₀ ~3), with and without additional TNRC6B (~300 nM, partially purified). After a 15-minute incubation, supernatant and pellet fractions were isolated by centrifugation and target RNA was extracted and analyzed by denaturing PAGE and phosphorimaging. (C) Deadenylation timecourse. Reactions containing Ago2-let7 (40 nM) and soluble HEK 293 lysate (OD₂₆₀ ~3), with and without exogenous TNRC6B (~300 nM, final concentration) were fractionated at various times, and analyzed by denaturing gel. (D) Estimation of deadenylation rates. Target RNA bands in (C) were quantified and fraction of total intact RNA (A₁₁₄) for +/− TNRC6B conditions was plotted as a function of incubation time. Data were fit with a first order decay, yielding A₁₁₄ half-lives of 40 and 4 minutes for plus and minus exogenous TNRC6B, respectively. Plotted data are the average of three independent experiments with SEM indicated as error bars. See also Figure S6.

Figure 7. Inducing miRISC phase separation accelerates target deadenylation

(A) PEG 8000 promotes TNRC6B-Ago2 phase separation. Images of droplets formed from Alexa-488 labeled TNRC6B (~20 nM) and Ago2 (200 nM) in the presence and absence of 5% (w/v) PEG 8000. (B) Effects of PEG 8000 on target deadenylation reactions. 8xlet7 deadenylation reactions containing combinations of Ago2 (20 nM), HEK 293 lysate (OD₂₆₀ ~1.5), and exogenous TNRC6B (~30 nM) were treated with 5% (w/v) PEG 8000, separated into supernatant and pellet fractions, and analyzed by denaturing PAGE. (C) Effect of PEG 8000 on deadenylation rates. Reactions containing Ago2-let7 (20 nM), HEK 293 lysate (OD₂₆₀ ~1.5), and exogenous TNRC6B (~30 nM) were treated with 5 % (w/v) PEG 8000, separated into supernatant and pellet fractions, and analyzed by denaturing PAGE at various times. (D) PEG 8000 accelerates deadenylation. Bands corresponding to target RNA species in (C) were quantified and fraction of intact target (A₁₁₄) plotted as a function of time. Plotted data are the average of three independent experiments with SEM indicated as error bars. See also Figure S7.