Multivalent recruitment of human Argonaute by GW182

Summary

In miRNA mediated gene silencing the physical interaction between human Argonaute (hAgo) and GW182 (hGW182) is essential for facilitating the downstream silencing of the targeted mRNA. GW182 can interact with hAgo via three of the GW/WG repeats in its Argonaute-binding domain: motif-1, motif-2 and the hook motif. The structure of hAgo-1 in complex with the hook motif of hGW182 reveals a “gate”-like interaction that is critical for GW182 docking into one of hAgo1’s tryptophan binding pockets. We show that hAgo-1 and -2 have a single GW182-binding site and that miRNA binding increases hAgo’s affinity to GW182. With target binding occurring rapidly, this ensures that only mature RISC would be recruited for silencing. Finally, we show that hGW182 can recruit up to three copies of hAgo via its three GW motifs. This may explain the observed cooperativity in miRNA-mediated gene silencing.

Introduction

microRNAs (miRNAs) are short non-coding RNAs, 20–24 nucleotides in length, that are loaded onto Argonaute (Ago) proteins to form the RNA-induced silencing complex (RISC). Base-pairing complementation then guides the RISC complex to its target mRNA and silencing of the target mRNA is achieved by multiple pathways that ultimately lead to degradation of the target mRNA (Jonas and Izaurralde, 2015). GW182 is a key player in miRNA-mediated gene silencing. It was first identified in patients with motor and sensory neuropathies (Eystathioy et al., 2002) and was later described to accumulate in GW-bodies/P-bodies where it co-localized with components of the mRNA degradation pathway (Eystathioy et al., 2003). GW182 physically bridges Argonaute proteins and downstream decapping, deadenlyation and mRNA degradation complexes (Fabian and Sonenberg, 2012; Liu et al., 2005a; 2005b; Meister et al., 2005; Rehwinkel et al., 2005). In humans, there are three GW182 paralogs: GW182/TNRC6A, which we will refer to as human GW182 (hGW182), TNRC6B and TNRC6C. They each share a similar primary structural arrangement consisting of an unstructured N-terminal/Ago binding domain and a C-terminal/silencing domain (composed of PAM2 and RRM domains). In humans, the Ago-binding domain (ABD) contains multiple glycine-tryptophan (GW/WG) repeats which mediate the interaction between hGW182 to all human Argonautes (Lazzaretti et al., 2009; Lian et al., 2009; Takimoto et al., 2009; Till et al., 2007), while the silencing domain mediates the recruitment of mRNA degradation machineries such as the CC4R-NOT complex and PABPC1(Braun et al., 2011; Chekulaeva et al., 2011; Fabian et al., 2009). In C. elegans the GW182 orthologs, AIN- 1 and AIN-2, are shorter and consist of only three domains: N-terminal domain, Mid domain that mediates the interaction with ALG-1 and ALG2 and a C-terminal domain (Kuzuoğlu-Ozturk et al., 2012).

The ABD of hGW182 contains more than 30 GW/WG repeats, yet only three GW containing motifs, motif-1, motif-2 and the hook motif, have been shown to mediate the interaction with human Argonautes (Lazzaretti et al., 2009; Lian et al., 2009; Takimoto et al., 2009; Till et al., 2007). While the biological function of the three GW182 proteins in humans are believed to be at least partially redundant (Landthaler et al., 2008), it is not clear whether the multiple GW motifs on a single GW182 protein can simultaneously interact with a single or multiple Argonautes, and whether such an interaction might contribute to cooperative gene silencing of adjacent miRNA binding sites (Kloosterman, 2004; Saetrom et al., 2007).

Here, we present the crystal structure of human Argonaute-1 (hAgo1) in complex with the hook motif of hGW182 providing the structural basis for the interaction between these two key components of the miRNA silencing machinery. We also show that guide RNA binding to hAgo1 and 2 greatly enhances binding of hGW182, and that while hAgos can only bind a single GW motif, hGW182 can enlist multiple Argonaute proteins simultaneously.

Results

GW motifs bind a single site on Argonaute

Previous studies pointed at three regions in the Ago-binding domain (ABD) of hGW182 that can mediate binding to human hAgo2 (Behm-Ansmant et al., 2006; El-Shami et al., 2007; Till et al., 2007). Therefore, we focused on these motifs named motif-1 (residues 455–494) motif-2 (residues 730–773) and the hook (residues 821– 841) (Takimoto et al., 2009; Till et al., 2007). Each of these motifs includes tandem GW/WG repeats that are spaced 9–20 residues apart (Figure 1A and Figure S1A). We expressed and purified each of the isolated GW/WG motifs of hGW182 (Figure S1B) and measured binding to human Argonaute-1 (hAgo1) and human Argonaute-2 (hAgo2), using isothermal titration calorimetry (ITC) (Figure 1B and 1C). Both hAgos co-purified with endogenous RNA from SF9 cells (Elkayam et al., 2012; Faehnle et al., 2013). Label-free binding by ITC showed a tight binding affinity of all GW motifs to both RNA-loaded hAgo1 and hAgo2. The resulting dissociation constants (K_d) were: 118 nM for motif-1, 46 nM for motif-2, and 234 nM for the hook to hAgo2-RNA (Figure 1B). Similar dissociation constants were observed for hAgo1 with dissociation constants of 96 nM, 78 nM and 383 nM for motif-1, 2 and the hook, respectively. (Figure 1C). Since guide RNAs bind very tightly to hAgo2, on the order of 1 nM (Elkayam et al., 2016), the contribution of RNA-hAgo dissociation would be negligible compared to the equilibrium between hAgo and the GW/WG motifs, and was therefore not taken into account. We should note that binding of hGW182 (TNRC6A) motif-1 to hAgo2-RNA is approximately 16 times tighter than a previously measured binding constant for the homologous region of TNRC6B (Pfaff et al., 2013) (with 72% identity) (Figure S1A) binding to hAgo2-RNA (118 nM vs 1.87 μM). The previous measurement was done by fluorescence polarization (FP) using a longer fragment of TNRC6B, and somewhat different purification protocols for both components.

Figure 1.

hGW182 Argonaute-binding motifs binding to hAgo2 and hAgo1. (A) Schematic domain organization of hGW182 (TNRC6A). The different Ago-binding motifs within the Ago-binding domain (ABD) are colored in yellow. (B–C) Isothermal titration calorimetry (ITC) analysis of the interaction between hGW182 Ago-binding motifs and hAgo2 (B) hAgo1 (C). In all experiments the ITC cell was filled with either hAgo2 or hAgo1 and the different hGW182 Ago-binding motifs were titrated as the ligands. Affinities are reported as dissociation constants (Kd) ± standard errors calculated from the fit. (D–E) Fluorescence Polarization (FP) binding experiments of FITC-labeled hGW182 hook motif with hAgo2 (D) and hAgo1 (E) showing increased affinity of the hook motif to both hAgo2 and hAgo1 that are loaded with endogenous RNA (solid blue line) compared to RNA-free hAgo2 and hAgo1 (dashed red line). Dissociation constants (Kd) were calculated by fitting data from three different experiments and are shown as the average ± standard deviation. See also Figure S1.

The molar binding ratio for all three motifs to both hAgo1 and hAgo2 determined from the ITC experiment is 1:1, suggesting a single binding site on hAgo1 and hAgo2 for each of the motifs. In addition, motif-2 showed the highest affinity for both hAgo1 and hAgo2. The tryptophan repeats in motif-2 are spaced by 20 amino acids (aa) compared to 9–10 aa in the case of the hook and motif-1, hinting at a possible advantage for the longer linker between the two tryptophan residues in hAgo binding or perhaps a slightly different mode of binding.

hGW182 binds guide RNA loaded hAgo with increased affinity

In miRNA-mediated gene silencing in humans, binding of Argonaute to GW182 facilitates the downstream effector step by promoting translational repression, decapping, deadenylation, and mRNA degradation (Fabian and Sonenberg, 2012; Jonas and Izaurralde, 2015). However, recruitment of unloaded Argonaute might prove counter-productive, and would sequester GW182 unnecessarily. Therefore, we examined the effect of guide RNA binding to Argonaute on the hGW182-Argonaute interaction. Taking advantage of our RNA-free hAgo1 and hAgo2 purification method (Elkayam et al., 2012; Faehnle et al., 2013), we measured the binding constants of RNA-free hAgo1 and hAgo2 and endogenous RNA-loaded hAgo1 and hAgo2 to the N-terminal fluorescently-labeled hook of hGW182 in a fluorescence-polarization (FP) binding assay. The K_d measured for guide-loaded hAgo1 and hAgo2 were comparable to those obtained by ITC (within a factor of 2 for hAgo1 and 3 for hAgo2) (Figures 1D and 1E). However, the hook had 5–8 times higher affinity to guide-loaded hAgo1 and hAgo2 compared to the RNA-free forms. We speculate this might be due to the reduced conformational flexibility conferred by guide RNA binding to hAgo2 (Elkayam et al., 2012).

Structure of hAgo1-guide RNA-hGW182 hook ternary complex

To investigate the structural basis for the observed binding affinities of the hGW182 hook to hAgo, we screened hAgo/hook complexes for crystallization. Previously published structures of hAgo2-RNA complexes included either phenol or tryptophan as an additive to promote crystallization (Elkayam et al., 2012; Schirle et al., 2014; Schirle and MacRae, 2012; Schirle et al., 2016). The phenol and tryptophan occupied two pockets in hAgo2 that were previously shown by genetic studies to be important for GW182 binding (Boland et al., 2011; Eulalio et al., 2009). Since tryptophan/phenol may interfere with GW182 binding, we selected hAgo1 as the prime candidate for structural studies with the hook because previously determined structures of hAgo1 did not require tryptophan/phenol for crystallization (Faehnle et al., 2013; Nakanishi et al., 2013).

The crystal structure of hAgo1 in complex with endogenous RNA and the hook motif of hGW182 was determined to 2.8 A resolution (Table 1). The overall structure of the ternary complex is practically identical to the previously solved binary complexes of hAgo1 with either endogenous RNA from SF9 or with let-7 miRNA (Faehnle et al., 2013). The overall root-mean-square-deviation (rmsd) is 0.33 A between the ternary and binary complexes and 0.11 A between the RNAs in these complexes, the largest movement being that of the PAZ domain (rmsd=0.27A), probably due to different crystal packing in the two structures. Overall, the hGW182 hook motif resembles a thumb and forefinger pinching the PIWI domain of hAgo1 with W828 and W838 occupying the two W-binding pockets, and the N-terminus of the peptide protruding into the MID-PIWI-L2 interface (Figure 2A and S2A).

Table 1.

Data collection and refinement statistics

	hAgo1-GW182 hook
Data collection
Space group	P21212
Cell dimensions
a, b, c (Å)	93.75, 136.86, 86.05
α,β,γ (°)	90.0, 90.0, 90.0
Resolution (Å)	2.83 (2.84–2.83)*
Rsym or Rmerge	7.9 (54.9)
I /σ(I)	16.3 (2.5)
Completeness (%)	99.9 (100)
Redundancy	4.0 (4.1)
Refinement
Resolution (Å)	2.83
No. reflections	27069
Rwork /Rfree	20/24.5
No. atoms
Protein	6539
RNA	215
Water	4
B-factors
Protein	56.5
Ligand/ion	87.3
Water	36.4
R.m.s. deviations
Bond lengths (Å)	0.004
Bond angles (°)	1.06

^*Values in parentheses are for highest-resolution shell.

Figure 2.

The structure of the hGW182 hook motif in complex with hAgo1-miRNA. (A) Cartoon representation of the overall structure of the hAgo1-miRNA-hGW182 hook motif complex. The hGW182 hook (gold) is bound to the PIWI domain (purple) with the two tryptophans (in red) anchored in the hydrophobic GW-binding pockets (GWBP). The N- and C-termini of the peptide are marked. (B) A close-up of hAgo GW-binding pockets (GWBP) with the bound hGW182 hook motif. GWBP1 residues are colored in green and GWBP2 residues are colored in pink. The PIWI domain is colored in purple and the hook in gold. See also Figure S2 and Table 1.

The two W-binding pockets in the PIWI domain are 14.5 A apart and the connecting linker spans a distance of ~35 A along the surface of the PIWI domain of hAgo1 (Figure S2B.). The interaction with the hook is confined to the W-binding pockets with no visible interactions to any of the other residues of the hook. We refer to the binding pocket for W828 as GW-binding pocket 1 (GWBP1) and the binding pocket for W838 as GW-binding pocket 2 (GWBP2). Both binding pockets are lined with hydrophobic residues (Figure 2B). GWBP1 is comprised of side chains from three loops at the bottom of the central β-sheet of the PIWI domain, as well as from an α-helix, α15, positioned between the two binding sites: P588, A618, F651, T655, F657.

The GWBP2 pocket is nestled between α15 and α16, with side chains from these helices and adjacent loops, I649, Y652, L692 and Y696 lining the pocket. The aliphatic portions of the side chains of K658 and E693 also line the pocket, with K658 stacking against the tryptophan indole ring. The special positioning of these two residues will be discussed below.

The hook residues, W828 and W838 are in a different conformation compared with the hAgo2-tryptophan complex (Schirle and MacRae, 2012) (Figure S2B). Specifically, the W838 side chain is tilted by 30° with the carboxylic and amino groups flipped, consistent with the N-C directionality observed in this structure, compared to the directionality observed in the earlier structure (Pfaff et al., 2013; Schirle and MacRae, 2012). The W828 side chain is also tilted by 40° and is pulled back by the polypeptide chain bringing it closer to P588 and F657, while the backbone of free tryptophan protrudes into the space occupied by S826 of the hook (Figure S2B).

Curiously, K658 and E693 from GWBP2 form a tight salt bridge with a distance of 2.2 A between the two side chains that locks the W838 indole ring inside the binding pocket (Figure 3A). This salt bridge is present in all previously determined hAgo2 structures (Figure 3C and 3D), probably because either tryptophan or phenol occupies the GWBP2 site. The presence of tryptophan/phenol would appear to induce a “gate” closure stabilized by the salt bridge between K658 and E693 to complete the GWBP2 pocket. Indeed, these residues are swung away from each other in the structures of hAgo1 determined in the absence of the hook, tryptophan or phenol (Faehnle et al., 2013). This leaves the gate open with the pocket available for binding (Figure 3B). Moreover, sequence alignment of various Argonaute proteins revealed these two residues are highly conserved among GW-interacting Argonautes such as all four human Argonautes, D. melanogaster Ago1, and C. elegans Alg1 and Alg2, while extremely variable among non-GW interacting Argonautes such as the D. melanogaster Ago2 and the human, mouse and Drosophila PIWIs (Figure S3). We therefore speculate that the “gate” interaction between E693 and K658 plays an important role in GW binding to Argonaute proteins.

Figure 3.

The open and closed hGW182-binding “gate”. (A) hAgo1 GWBP2 gate residues form a tight salt bridge upon binding of the hGW182 hook, thus closing the gate. (B) Open gate conformation in the absence of hGW182 hook binding as in the structure of hAgo1 with endogenous RNA (PDB ID 4KRE). (C) Similar “gate” closure is observed in the presence of tryptophan in the structure hAgo2 in complex with tryptophan (PDB ID 4OLB)(D) and in the presence of phenol in the hAgo2-miR20a complex (PDB 4F3T). See also Figure 3.

The integrity of the GW binding pockets is critical for binding

To investigate the role of residues that form the W-binding sites, we used the structure of the hAgo1-RNA-hook ternary complex as a guide in designing hAgo1 and hAgo2 binding-pocket mutants. Mutants were tested for hook binding in a FP assay. When we mutated the hydrophobic residues in GWBP1 and GWBP2 to a charged residue, the interaction between the hook and either hAgo1 or hAgo2 was drastically reduced, or even completely abolished (Table S1). Interestingly, we often observed single point mutants in one site are sufficient to prevent binding to the hook completely, regardless of the integrity of the other pocket. For example, mutating I649 to an arginine, which is the corresponding amino acid in many of the PIWI clade proteins (HsHiwi, MmMiwi, DmPiwi and DmAub, see Figure S3), caused complete loss of hook binding for both hAgo1 and hAgo2. Surprisingly, mutation of I649, F651, L692, or Y696 to the shorter aliphatic residue alanine also had a substantial negative impact on hook binding (Table S1). We suggest that the geometry of the hydrophobic residues within GWBP1 and 2 are required to devise effective GW-binding sites. This could also be partially due to limited flexibility of the W residue, in the context of a polypeptide chain, to penetrate deep enough into the pocket in order to maintain van der Waals contact with the shorter aliphatic alanine. Moreover, hAgo binding is also lost when mild GWBP2 mutants (such as K658A and E693A) are combined together with GWBP1 mutants (Figures 4A and 4B), which suggest some level of synergy between the two GW-binding sites.

Figure 4.

Mutational analysis of the GW binding pockets of hAgo1 and hAgo2. (A) Fluorescence Polarization (FP) binding experiments of FITC-labeled hGW182 hook to hAgo2. (B) Same as (A) but with hAgo1. GW binding pockets mutants shows drastic decrease in hGW182 hook binding when residues from both binding pockets are mutated simultaneously. Data shown are from three different experiments and presented as the average± standard deviation. The lines for fitted curves with Rsquare<0.7 were omitted from the figure. (C) FP binding experiments of hAgo2 “gate” residue mutants with FITC-labeled hGW182 showing a substantial decrease in binding compared to wild-type. (D) Similar experiments for gate residue mutants of hAgo1. Dissociation constant (Kd) were calculated by fitting data from three different experiments and are shown as average ± standard deviation. See also Table S1.

Finally, we tested the role of the “gate” residues, K658, and E693, by mutation to alanine as well as swapping between the two. As expected, mutation of the gate to alanine resulted in a drastic reduction (~10 fold) or complete loss of binding of both hAgo1 and hAgo2 to the hGW182-hook (Figures 4C and 4D). Unexpectedly, when we swapped the gate residues (K658E and E693K), to preserve the salt bridge, binding is also nearly abrogated. We think this might be due to the loss of the stacking interaction between K658 and W838 of the hook (Figures 3A and 2B), which would not be maintained by the shorter glutamate residue. Taken together, our structure-based mutational analysis identified the critical residues that form the two GW-binding pockets. In addition we identify a movable “gate” in GWBP2 important for hook binding.

Argonaute has a single GW182 binding site

Inspection of hAgo1 and hAgo2 structures (Elkayam et al., 2012; Faehnle et al., 2013) did not reveal additional GW-binding pockets. In addition, ITC binding data described above implies that hAgo1 and hAgo2 interact with each motif at a 1:1 molar ratio (Figures 1B and 1C). This suggests a single binding site on Argonaute for all of the motifs. To rule out the possibility of multiple GW-binding sites on hAgo1 and hAgo2, other than the site identified in the crystal structures, we performed competition assays. hAgo1 and hAgo2 were incubated with a fluorescently-labeled hook prior to titration of increasing concentrations of unlabeled hook or unlabeled motif-1 or motif-2 of hGW182. As predicted for a single binding site, both motifs 1 and 2 as well as the unlabeled hook compete off the fluorescently-labeled hook (Figures 5A and 5B). To validate this observation, we measured the K_d of the hook motif in the presence of increasing concentrations of motif-2 in a label-free ITC experiment. Here, either hAgo1 or hAgo2 were premixed with a fixed concentration of hGW182 motif-2, acting as a competitor, and subjected to titrations of the hook. The ITC titration curves were fit using a competitive binding model. In the presence of 5 μM motif-2, the hook binds to hAgo2 with 7 times lower affinity (Figure 5C) and 15 times lower in the presence of 10 μM motif-2 (Figure 5D) compared to the K_d in the absence of competitor (Figure 1D). This demonstrates that motif-2 has an inhibitory effect on the hAgo2-hook interaction by competing for the same binding site. These results are consistent with a single hGW182-binding site on hAgo1 and hAgo2 as shown in Figure 1B. We should note that a decrease in affinity could also be explained by an allosteric inhibitory effect that motif-2 might have on hook binding. These data would have to be modeled using a non-competitor titrant with an increase in the number of binding sites. However, this data could only be fitted accurately using the competitive binding model (Sigurskjold, 2000) where the number of binding sites remained unchanged (N=1.1 and N=0.95 for 5 and 10μM, respectively).

Figure 5.

All Three Argonaute binding motifs of hGW182 compete for the same binding site on both hAgo2 and hAgo1. (A) FP binding experiments of FITC-labeled hook motif binding to hAgo2 with increasing concentrations of motif 1 (red), motif 2 (green) and unlabeled hook (purple) of hGW182. (B) Same as (A) but for binding to hAgo1. (C) Isothermal titration calorimetry (ITC) analysis of the interaction between the hGW182 hook motif and hAgo2 in the presence of 5 μM hGW182 motif 2. (D) Same experiment as in (C) but in the presence of 10 μM motif 2. The decrease in binding affinity of the hook motif to hAgo2 correlates with the increasing concentration of motif 2. Affinities are reported as dissociation constants (Kd)± standard errors calculated from the fit using a competitive binding model.

Different GW motifs bind human Argonaute independently

Next, we decided to investigate the role of multiple Ago-binding motifs at the N-terminus of hGW182, given the singular GW-binding site on Argonaute. To examine the role of the tryptophan residues located in each of these motifs, we designed a hGW182 construct that contained all three Ago-binding motifs (residues 455–841, Figure S4A). We mutated the tandem tryptophan residues in each of the motifs to alanine and measured hAgo2 binding using a pull-down assay (Figures 6A, S4B and S4C). Surprisingly, tandem mutations of both tryptophan residues within a single motif did not significantly reduce hAgo2 binding. Moreover, combinatorial tandem tryptophan mutations reduced binding by no more than 10% for motif-1/hook and motif-2/hook combinations, while combining motifs-1 and 2 reduced binding by almost 30%, indicating a possible link between these two sites (Figure 6A). Only when we mutated all six tryptophan residues (those in motif-1 and motif-2, and the hook), did we observe loss of hAgo2 binding (Figures 6A and S4B). This very moderate reduction in binding when tandem tryptophans in one or two of the motifs were mutated could be the result of either residual interaction with the other residues in each of the three motifs or that a single intact motif can maintain binding to hAgo2.

Figure 6.

Tryptophan residues in all three motifs of hGW182 mediate binding to hAgo2. (A) Pull-down assays using the ABD of hGW182 as bait were used to capture hAgo. Point mutations of the tandem tryptophans of each of the motifs show similar binding levels as wild type. A mutant where all six tryptophans were changed to alanine showed nearly complete loss of binding (right most bar), (B) Pull-down assays of hAgo2 with deletion mutants of the ABD of hGW182 with either one, two or all motifs were removed. All deletions affected binding to Ago2, Elimination of all three motifs (right-most bar) had a similar effect on hAgo2 binding to the six-tryptophan point mutant. Data shown is from three different experiments and are shown as average ± standard deviation. (C-H) Analytical size exclusion chromatography of hAgo2-RNA with different deletions of hGW182 ABD. Complexes were formed using a similar pull-down protocol as in figure 6B and were injected onto a superdex 200 column. The UV signal at 280nm was used to monitor elution volumes of the complexes (red) and the individual components: hAgo2-RNA (dashed black line), and ABD (blue). The different ABD deletion constructs used are full length (residues 455–841) (C), Δ1 (D), Δhook (E), Δ1 and Δ2 (F), Δ1 and Δhook (G) and Δ2 and Δhook (H), and noted for each chromatogram. See also Figure S4 and Table S2.

To test this, we deleted each of the motifs entirely and measured hAgo2 binding. Deletion of each motif individually reduced binding by 30–50%, while a decrease of 80–85% was observed for constructs with any combination of two motifs deleted (Figures 6B and S4C). Deletion of all three motifs decreased binding by more than 95%, similar to the reduction in binding observed with tandem tryptophan mutation in all three motifs. Indeed, a previous report has shown that while tryptophan residues are exquisitely important for hAgo2 binding, the flanking residues can contribute weakly to the interaction (Pfaff et al., 2013) explaining the differences in hAgo2 binding between tryptophan point mutations and motif deletions. We should note that these pull-down binding assays are qualitative and are therefore not sensitive enough to detect small changes in affinity that might result from weak interactions between hAgo2 and the flanking residues to the tryptophans. In addition, while we did not observe such interactions in the hAgo1-hGW182 hook complex, we cannot rule out additional interactions between the flanking residues in motif-1 and motif-2 of hGW182 and hAgo. We postulated that since hAgo2 binding to GW182 could only be eliminated by either mutation of all six tryptophan residues or truncation of all three motifs, coupled with the fact that the three motifs bind only a single site on hAgo1 and hAgo2, it is likely that multiple copies of hAgo could assemble on the different motifs of hGW182.

A single hGW182 can bind multiple Argonautes using all three Ago-binding motifs

Intrigued by the hAgo-hGW182 pull-down results, we assembled hAgo2- RNA/hGW182 (ABD, aa 455–841) ternary complexes and determined the molar ratio by analytical size exclusion chromatography (SEC) on a Superdex200 column. Remarkably, we noticed the hAgo2-RNA-ABD complex elutes earlier than the predicted mass for a 1:1:1 complex, which would result in a MW of ~150 kDa. The elution volume (~12 mL) was consistent with a 300–400 kDa particle based on a comparison to known MW weight standards (Figure 6C). We therefore speculated that multiple hAgo2-RNA complexes bind concurrently to the different motifs on the ABD of hGW182.

To test this hypothesis, we measured the elution volumes of hAgo2-RNA in complex with different truncations (Δ455–494, Δ730–772, Δ821–842) of the ABD of hGW182 by (SEC) (Figures 6D–6H). Complexes with single-motif truncations elute at larger volumes (~13 mL) compared with the native ABD (Figures 6D and 6E). An additional shift to an even larger elution volume (~13.5 mL) occurred with a truncation of any combination of two motifs (Fig 6F, 6G and 6H). Since hAgo2-RNA elutes at a similar volume to the triple-motif truncation of the ABD, we employed SEC on a Superdex 75 to monitor the interaction of the two proteins. Indeed, we observed no interaction between hAgo2 and the triple-motif truncation of hGW182- ABD (Figure S4D), a result that is consistent with the pull-down assays described above (Figure 6B).

We hypothesized that each motif in the GW182-ABD could bind a single Ago-miRNA complex concurrently. We therefore calculated the predicted molecular weight of the different complexes, examined these using SEC, and plotted the partition coefficient (K_av) obtained from the experiment as a function of the predicted MW of the complex (Figure 7A). As shown in the plot, the different hGW182-ABD truncations were clustered according to the number of available Ago-binding motifs. While the intact ABD correlated with binding of three hAgo2-guide RNA complexes, deletion of a single motif correlated with binding of two hAgo2-RNA complexes and the double motif truncations correlated with binding of a single hAgo2-RNA complex per hGW182 fragment (Figure 7A).

Figure 7.

hGW182 ABD can utilize all three GW-binding motifs to recruit multiple copies of Argonaute. (A) Molecular weight calibration curve of different hGW182 ABD-hAgo2-RNA complexes. The partitioned coefficient (Kav) was calculated based on the elution volumes presented in figure 6C and plotted against the predicted log MW of the various complexes. (B) Isothermal titration calorimetry (ITC) analysis of the interaction between hGW182 ABD and three hAgo2 molecules. The ITC cell was filled with hGW182 ABD and hAgo2 was titrated as the ligand. Affinities are reported as dissociation constants (Kd)± standard errors calculated from the fit using the sequential binding model equation. (C) Four selected 2D averages showing three hAgo2 linked together in the presence of hGW182. The number at the lower left corner in each panel refers to the number of raw particles that contributed to the class average. The dimensions of the hAgo2 structure (4F3T) was measured in two different orientations (longest and shortest) and used as a reference to estimate the size of each of the sub-particles. See also Figure S5.

Next, we measured the binding of multiple hAgo2-RNA complexes to the ABD of hGW182 by ITC. As expected, and in accordance with our previous results, the ratio between the hAgo2-RNA complex and hGW182 is ~3:1 (Figure 7B). Though we were not able to use a simple mathematical model to perform curve fitting due the multiple binding events of three hAgo2-RNA complexes to three different sites on the hGW182-ABD, the overall titration curve could be readily separated into higher-and lower-affinity sequential binding events (Figure 7B). The first event consists of a single hAgo2-miRNA complex that binds with relatively high affinity (7 nM), followed by a second binding event of two additional Argonaute-miRNA complexes with much lower affinity (1 μM). Considering the similar binding affinities to the individual motifs shown earlier (Figures 1B), it is not clear whether the first binding event corresponds to a hAgo2-RNA complex binding to a particular GW-motif or to any of the three. We imagine the first high-affinity binding event involves any one of the three Ago-binding motifs. Following this, further hAgo2-RNA recruitment is more limited with reduced affinity, ensuring that only hAgo2-RNA complexes that are in close proximity, and occupying the same transcript would bind. This would promote binding of a single GW182 to multiple Argonautes-guide RNA complexes that are in turn bound to a single target mRNA.

The complex that corresponds to three hAgo2-guide RNA complexes and a single hGW182-ABD has a calculated MW weight of ~330 kDa. In contrast, the unbound ABD of hGW182 is only ~40 kDa and is predicted to be mostly unstructured and highly flexible. We reasoned that such a complex would have multiple conformations, hence recalcitrant to crystallization either alone or in complex with an Ago-miRNA complex. Therefore, to validate the molecular assembly of the hAgo2- RNA-hGW182 complex we used negative stain single-particle electron microscopy (EM) to visualize the complex. We purified the ABD of hGW182 that includes all three Ago-binding motifs in the presence of a 3-fold molar equivalent of hAgo2-RNA (Figure 6C). We identified intact particles that appeared to be composed of three Argonaute molecules. We compared each of these apparent “sub-particles” to the radius of the known hAgo2 structure in various orientations. These “sub-particles” (hAgo2-RNA) appear to be bound by the ABD of hGW182 (Figure 7C). From the complete 2D class averages, it is also clear that hGW182 binds up to three hAgo2 molecules, although we also observed some bound to only two hAgo2 molecules (Figure S5). The complex clearly adopts multiple states and conformations, probably due to the high flexibility of the hGW182 ABD. The presence of multiple conformations precluded a meaningful 3D reconstruction of the complex since the latter relies on a relatively uniform and homogeneous particle conformation. However, we believe that the intrinsic flexibility of the ABD would in fact be beneficial for accommodating binding of GW182 to multiple Argonautes occupying differently spaced miRNA binding sites on the same transcript (Grimson et al., 2007; Saetrom et al., 2007).

Discussion

The interaction between GW182 and Argonaute is critical for mediating miRNA gene silencing in animals (Behm-Ansmant, 2006; Jakymiw et al., 2005; Liu et al., 2005a; Rehwinkel et al., 2005). It is mediated by multiple GW repeats located on the ABD of GW182, which bind the PIWI domain of Argonaute. Despite the presence of many GW repeats on the ABD domain of GW182, only three pairs were shown to mediate the interaction with human Argonaute proteins (Lazzaretti et al., 2009; Lian et al., 2009; Takimoto et al., 2009; Till et al., 2007). However, it is still not clear what is the exact correlation between the different motifs and whether they all bind Argonaute concurrently or whether they have a redundant role in binding Argonaute.

Here, we provide a comprehensive analysis of the interaction between human GW182 and human Argonaute 1 and 2 using structural, biophysical and biochemical methods. Our binding studies of the single GW motifs to hAgo1 and hAgo2 show that all three motifs bind with high affinity, despite a varying spacer length between the tryptophan residues. We reason that the plasticity of the linker between the two tryptophan residues could provide the flexibility needed when binding to the different human Argonaute proteins. In fact, this feature has been exploited as a investigative tool to pull-down Argonautes from different species such as Drosophila, Mouse and Arabidopsis using one of the GW motifs of the human GW182 family member TNRC6B (Hauptmann et al., 2015). Moreover, Argonaute loaded with guide RNA has significantly increased affinity towards GW182 compared to the RNA-free Argonaute, indicating that Argonaute recruitment by GW182 is more likely to occur when Argonaute is loaded with guide-RNA and primed for target recognition as was indicated by early studies of hGW182- Argonaute association (Baillat and Shiekhattar, 2009). It makes sense that hGW182- mediated gene silencing will only take place upon formation of the RISC-target mRNA complex, considering target mRNA recognition happens almost instantly for guide-loaded Argonaute (Wee et al., 2012) and might occur prior to hGW182 binding.

We further characterized the Ago-GW interaction by solving the structure of the hAgo1-hook complex and identified the key residues in hAgo1 and 2 that mediate the interaction with hGW182. While some of these residues have been predicted to be important for the interaction between GW182 and Argonaute proteins (Boland et al., 2011; Eulalio et al., 2009; Schirle and MacRae, 2012) our structure-guided mutational analysis provides an in-depth analysis of the interaction and identified additional key residues that are important for GW binding that are evolutionary conserved amongst GW-interacting Argonaute proteins. Indeed, a recent report pointed at the importance of the integrity of residues from both binding pockets of hAgo2 for the interaction with GW182 in cell based tethering assays (Kuzuoğlu-Ozturk et al., 2016). We also showed that both hAgo1 and 2 can accommodate a single tandem-GW motif and that this site cannot bind multiple GW motifs simultaneously. It is likely that this trait is shared among other GW binding Argonautes considering the highly conserved PIWI domain. In contrast, GW182 can bind up to three Argonaute proteins by exploiting all three GW/WG binding motifs to form a stable yet flexible complex. Earlier studies have pointed to the cooperative nature of adjacent miRNA binding sites on their target mRNA silencing (Doench et al., 2003; Pillai et al., 2004). These sites are required to be within 7–40 bases apart to maintain their cooperative effect (Grimson et al., 2007; Saetrom et al., 2007) and can also sustain cooperativity even when targeted by different miRNA families (Kloosterman, 2004). Here, we provide the structural basis for how multiple Argonautes bound to adjacent miRNA target sites might recruit hGW182. We reason that the “unstructured” flexible regions separating the different Ago binding motifs within GW182 provide the plasticity needed to accommodate the variable distances between potential cooperative miRNA binding sites, while at the same time preventing steric hindrance between two adjacent Argonautes on neighboring sites. In addition, and as we show here, GW182 preference toward guide-RNA loaded Argonautes (Figures 1D and 1E) ensures that only guide-primed Argonautes will be recruited by GW182 where RISC is most likely already assembled on its target mRNA (Wee et al., 2012). This GW182 mediated interaction of the RISC with multiple adjacent sites on the same transcript can result in longer dwelling on the target transcript as was suggested recently (Denzler et al., 2016).

Finally, by having multiple Ago binding motifs, GW182 can also recruit different members of the Argonaute family or Argonautes loaded with different miRNA allowing GW182 to interact with multiple Argonaute/miRNA species that occupies adjacent sites while still maintaining their cooperative silencing effect (Broderick et al., 2011; Baillat and Shiekhattar, 2009).

STAR methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Leemor Joshua-Tor (leemor@cshl.edu).

METHOD DETAILS

Protein expression and purification

hAgo2 and hAgo1 were expressed and purified as previously described (Elkayam et al., 2012; Faehnle et al., 2013). All GW182 clones used in this study are listed in Table S2. Briefly, human GW182/TNRC6A (Uniprot ID Q8NDV7-2) motif 1 (aa 455–490) motif 2 (aa 730–773) and the hook (aa 820–841) were cloned as His₁₀-Sumo fusion proteins into a pET28a vector and expressed in E. coli BL-21 cells. All other GW182 fragments were cloned into the pFL vector of the Multibac expression system as N-terminal Strep-Sumo fusions and were expressed in SF9 cells. GW182 fragments were purified with an affinity column using NiNTA or StrepTactin, followed by on-bead tag removal by Utd1 (for BL-21 expression) or TEV (for SF9 expression). All proteins were then subjected to size exclusion chromatography using Superdex 200 16/60 in 50mM Tris pH 8.0, 100 mM KCl and 5 mM DTT.

Crystallization, Data collection and Processing

hAgo1 loaded with endogenous RNA from SF9 cells was mixed at a 1:1.5 ratio with a fragment encompassing the hGW182 hook (aa 820–841) and loaded onto a Superdex 200 10/300 column pre-equilibrated with 10 mM Tris pH 8.0, 100 mM NaCl and 10 mM DTT. The hAgo1-hook complex was then concentrated to 5 mg/ml and crystallized using the sitting drop vapor diffusion method by mixing 1 μL of the complex with 1 μL reservoir solution comprised of 21.2% (w/v) PEG3350 and 0.1 M di-ammonium tartrate. Crystals appeared after ~5 days and were fully grown after 13 days. Crystals were cryoprotected by briefly transferring them into a reservoir solution supplemented with 30% (v/v) ethylene glycol and flash frozen in liquid nitrogen.

Data were collected to 2.8 A at beamline 8.2.2 at the Berkeley Center for Structural Biology (BCSB) at the Advanced Light Source (ALS). Diffraction data were indexed, integrated and scaled using autoPROC (Vonrhein et al., 2011). The structure was solved by molecular replacement in PHASER (McCoy, 2007) using the hAgo1 structure as a search model excluding the RNA and water molecules from the search model (Faehnle et al., 2013) (PDB ID:4KRE). The molecular replacement solution was rigid-body refined in PHENIX followed by simulated annealing refinement prior to manual correction in COOT (Emsley et al., 2010). Final TLS refinement of the model was done using PHENIX with manually selected TLS groups (version 1.10.1–2155) (Adams et al., 2010). The final structure was refined to a R_work/R_free of 20%/24.4%, with two Ramachadran outliers. Data collections and refinement statistics are shown in Table 1.

Coordinates and structure factors for the hAgo1-RNA-hGW182 hook complex have been deposited in Protein Data Bank under accession code PDB 5W6V Structure figures were generated using PyMol (The PyMOL Molecular Graphics System, Version 1.3 Schrodinger, LLC.).

Isothermal calorimetry (ITC)

All ITC experiments were carried out in phosphate buffered saline (PBS) using a MicroCal iTC 200 (Malvern Instruments, Malvern UK) at 25°C. Data were analyzed using Origin 7.0 (OriginLab, Northhampton, MA) software. Fluorescence polarization (FP) The N-terminal fluorescein amidite (FAM) labeled hGW182 hook fragment (aa 820- 841) was chemically synthesized (Genescript, Piscataway, NJ). 5 nM of the labeled peptide was titrated with increasing concentrations of RNA-free or RNA-loaded hAgo1 or hAgo2 in 50 mM Tris pH 8.0, 100 mM KCl and 10 μg/ml BSA. Data were acquired using the Synergy 4 plate reader and analyzed using Gen 5 software (BioTek, Winooski, VT).

Pull-down assays

For the pull-down assays, 1 pmol of purified Strep-Sumo tagged wild type GW182 Ago binding domain (ABD) (residues 455–841) or the indicated (W->A) point mutations were mixed with 5 pmol (1:5 ratio) untagged hAgo2-RNA and incubated for 30 minutes at room temperature. The mixture was then applied to 20 μL StrepTactin affinity resin (IBA BioTAGnology) pre-equilibrated with PBS and incubated for an additional 10 minutes. The resin was washed 3 times with 150 μL of each of the following three buffers: PBS; PBS supplemented with 1 M KCl (high salt wash) and 50 mM Tris pH 8.0 (no salt buffer). The final wash was done with PBS before eluting with PBS supplemented with 2.5 mM desthiobiotin. The samples were separated on SDS-PAGE and blotted on a nitrocellulose membrane and tested for the presence of hAgo2 and Strep-Sumo hGW182 proteins using anti-hAgo2 (ABnova) and anti-strep antibodies (IBA BioTAGnology). The results were quantified using GeneTools software (Synoptics, Cambridge, England) and analyzed using Prism 6 software (GraphPad).

Analytical Size Exclusion Chromatography

2 nmols of purified hAgo2-RNA and different truncations of hGW182 fragments were injected onto a Superdex 200 10/300 column that was pre-equilibrated with 10 mM Tris pH 8.0, 200 mM KCl and 5 mM DTT. Elution volumes of the complexes were monitored by absorbance at 280 nm. The partition coefficient was calculated using the formula: K_av = (V_e−V_o)/(V_t−V_o). The partition coefficients vs log molecular weight (MW) were plotted using Prism 6 (GraphPad).

Negative stain single-particle electron microscopy (EM)

The hAgo2-RNA-hGW182 (aa 455–841) complex was diluted to a concentration of 0.05 mg/ml, and was applied to a glow discharged carbon-coated EM grid, and negatively stained with a small droplet (5 μl) of 1% uranyl acetate aqueous solution. Data were collected on a Gatan 4k x 4k CCD camera in a JEOL2010F TEM operated at a high tension of 200 kV and at a magnification of 50,000 (2.12 A/pixel). Particles were auto-picked using the EMAN2.1 gauss method with a box size of 128 pixels (Ludtke, 2016). The raw images were corrected for the contrast transfer function (CTF) effect. To facilitate particle sorting which uses only low-resolution information the final dataset was mean-shrunk by a factor of 2, to 4.24 A/pixel, leading to a reduced particle box size of 64 pixels. Particles that did not belong to any of the well-defined classes are considered “bad” and were rejected for further analysis. The remaining particles were pooled and subjected to another round of reference-free classification, leading to the final class averages as presented in Fig. 7 and Fig. S5.

QUANTIFICATION AND STATISTICAL ANALYSIS

Isothermal calorimetry experiments were repeated three times and representative results were shown. Fluorescence polarization and in vitro pull-down experiments were repeated three times. Fluorescence polarization binding data were plotted and fit using GraphPad Prism 6.

DATA AND SOFTWARE AVAILABILITY

The structure of the hAgo1-RNA-hGW182 hook complex has been deposited in the protein database bank (PDB) under ID code 5W6V.

The raw images for supplemental figure 4 can be found in: http://dx.doi.org/10.17632/7ctzn47d46.1