Helix‐7 in Argonaute2 shapes the microRNA seed region for rapid target recognition

Abstract

Argonaute proteins use microRNAs (miRNAs) to identify mRNAs targeted for post‐transcriptional repression. Biochemical assays have demonstrated that Argonaute functions by modulating the binding properties of its miRNA guide so that pairing to the seed region is exquisitely fast and accurate. However, the mechanisms used by Argonaute to reshape the binding properties of its small RNA guide remain poorly understood. Here, we identify a structural element, α‐helix‐7, in human Argonaute2 (Ago2) that is required for speed and fidelity in binding target RNAs. Biochemical, structural, and single‐molecule data indicate that helix‐7 acts as a molecular wedge that pivots to enforce rapid making and breaking of miRNA:target base pairs in the 3′ half of the seed region. These activities allow Ago2 to rapidly dismiss off‐targets and dynamically search for seed‐matched sites at a rate approaching the limit of diffusion.

Introduction

Argonaute proteins use small (~22 nucleotide) RNAs as guides to identify complementary sites in messenger RNAs (mRNAs) targeted for repression and degradation. In human cells, most Argonaute molecules are loaded with microRNAs (miRNAs), an abundant class of small RNAs (Lagos‐Quintana et al, 2001; Lau et al, 2001; Lee & Ambros, 2001). miRNAs function in diverse physiological processes, with more than half the protein‐coding genes in humans carrying evolutionarily conserved miRNA recognition sites (Lewis et al, 2005; Friedman et al, 2009). Misregulation of miRNA activity has been implicated in many diseases, including type 1 and type 2 diabetes, heart failure, and several forms of cancer, including breast, lung, and prostate cancer (Lujambio & Lowe, 2012; Ghai & Wang, 2016; Zhang & Schulze, 2016).

In order to carry out its molecular function, Argonaute must effectively distinguish mRNAs matching its guide from non‐target RNAs, which vastly outnumber even the most abundant miRNA targets (Garcia et al, 2011). Biochemical data suggest that Argonaute achieves this feat by modifying the properties of its guide RNAs such that, upon binding to Argonaute, small RNAs behave more like RNA‐binding proteins than free RNAs (Haley & Zamore, 2004; Ameres et al, 2007; Wee et al, 2012; Salomon et al, 2015). When bound to Argonaute, guide (g) nucleotides g2–g8 (termed the “seed” region) pair with complementary target RNAs 50–200 times faster than the annealing rates of equivalent naked RNAs (Salomon et al, 2015). Similarly, seed‐paired target RNAs stay associated with the Argonaute–miRNA complex > 20 times longer than the expected lifetime of equivalent naked RNA duplexes (Salomon et al, 2015). The Ago2–miRNA complex is also much more sensitive to mismatches and G:U wobble base pairs with target RNAs than predicted by base pairing thermodynamics alone (Salomon et al, 2015).

Although functional studies clearly demonstrate that Argonaute reshapes the binding properties of its nucleic acid guides (Haley & Zamore, 2004; Ameres et al, 2007; Wee et al, 2012; Chandradoss et al, 2015; Jo et al, 2015; Salomon et al, 2015), exactly how reshaping occurs is poorly understood. Pioneering structural studies of prokaryotic Argonaute proteins revealed that Argonaute pre‐organizes the seed region in a helical conformation and thereby reduces the entropic cost of forming a duplex structure with complementarity targets (Ma et al, 2005; Parker et al, 2005, 2009; Wang et al, 2008a,b). However, subsequent functional studies found that target binding is far more sensitive to mismatches in the 5′ end of the seed than in the 3′ end, indicating that Argonautes do more than simply pre‐organize the seed, which would pre‐pay entropic hybridization costs to a similar extent for all seed nucleotides (Salomon et al, 2015). Indeed, structures of eukaryotic Argonautes show the seed is only partially pre‐organized in a helical conformation, with base stacking interrupted by a kink between nucleotides g6 and g7 (Nakanishi et al, 2012, 2013; Schirle & MacRae, 2012; Faehnle et al, 2013). This kink appears to be promoted by an α‐helix (helix‐7 in human Ago2), which inserts between the nucleobases of g6 and g7 (Fig 1A and C). Helix‐7 also creates a steric barrier to base pairing beyond g5 in these crystal structures. In contrast, crystal structures of the Ago2–miRNA–target complex show helix‐7 shifts to dock into the minor grove of the guide:target duplex upon stable seed pairing (Fig 1B and D).

Figure 1. Helix‐7 dynamically interacts with the miRNA seed region.

1. Cartoon representation of the Ago2–miRNA crystal structure. Ago2 is colored gray except for helix‐7, which is yellow. miRNA guide colored red.
2. Crystal structure of the Ago2–miRNA–target RNA ternary complex. Target RNA is colored blue.
3. Close‐up view of the unpaired complex shows helix‐7 breaks nucleobase stacking in the miRNA seed region by intercalating between g6 and g7.
4. Close‐up view of the Ago2–miRNA–target complex shows helix‐7 docks into the minor groove of the guide:target duplex, directly contacting base pairs at positions g6 and g7.

Considering the close proximity of helix‐7 to the miRNA seed region and its apparent dynamic nature, we hypothesized that helix‐7 may be a structural element that shapes the binding properties of miRNAs bound to Ago2. We tested this hypothesis by examining the RNA‐binding properties and structure of Ago2 helix‐7 mutants. Our results show that, despite its close proximity to the seed, disruption of helix‐7 has very little effect on the affinity of Ago2 for a seed‐matched target RNA. However, the dynamics of binding are significantly altered, with a reduction in both target binding and target release rates. Helix‐7 mutants also display extended dwell times on off‐targets, fail to discourage non‐Watson‐Crick pairing in the 3′ end of the seed, and are 10 times more sensitive to inhibition by off‐targets than wild‐type Ago2. Structural analysis revealed that helix‐7 mutants fail to position the 3′ end of the seed for pairing to incoming target RNAs. The cumulative results indicate Argonaute reshapes the binding properties of its guide by creating a two‐part seed that is fixed in an A‐form conformation on the 5′ end and dynamic and flexible on the 3′ end. 3′ end dynamics are modulated by helix‐7 such that Argonaute can interrogate potential target sites with speed and fidelity.

Results

Helix‐7 mutants retain the ability to bind target RNAs

Based on static Ago2 crystal structures, we previously suggested that helix‐7 may be involved in monitoring guide:target pairing, displacing mispaired target RNAs, and/or establishing high‐affinity interactions with seed‐paired targets (see Fig 1; Schirle & MacRae, 2012; Schirle et al, 2014). To test these hypotheses, we first assessed whether or not helix‐7 is essential to the folding and overall structure of Ago2. We generated recombinant Ago2 mutants in which amino acid residues Met364 and Ile365 on helix‐7 were changed to alanine (MI‐AA), or the entire helix, composed of residues L356–T368, was replaced with a penta‐glycine linker (Δhelix‐7). The helix‐7 mutant proteins accumulated to levels comparable to those of wild‐type Ago2 when expressed in Sf9 cells and could be loaded with a single‐stranded guide RNA and purified using an immobilized complementary oligonucleotide (Flores‐Jasso et al, 2013; Fig EV1A). Purified samples of wild‐type and Δhelix‐7 Ago2 were stable at temperatures ≤ 55°C and displayed overlapping ultraviolet circular dichroism (UV CD) spectra (Fig EV1B–D). Thus, all results indicate that disruption or removal of helix‐7 does not prevent Ago2 from folding into a near‐native conformation.

Figure EV1. ∆Helix‐7 Ago2 adopts a near‐native fold.

1. SDS–PAGE of purified wild‐type and Δhelix‐7 Ago2 samples. The Δhelix‐7 mutant was indistinguishable from wild type during recombinant expression and purification.
2. Cartoon schematic of heat inactivation experiment. Ago2–miR122 complexes were heated to various temperatures for 10 min, then cooled on ice, and then measured for endonuclease activity at 37°C.
3. Denaturing PAGE of a miR‐122 target cleaved by wild‐type and ∆helix‐7 Ago2–miR122 complexes heated to indicated temperatures.
4. UV CD spectra of wild type (dashed black line) and ∆helix‐7 (blue line) at 25°C are nearly superimposable.

We next asked whether helix‐7 is necessary for binding a seed‐paired target RNA. We prepared purified samples of wild‐type and Δhelix‐7 Ago2 loaded with miR‐122 as the guide RNA. We then measured the affinities of the two complexes for a target RNA with complementarity to the full seed region (g2‐g8) of miR‐122 using a filter‐binding assay (Wong & Lohman, 1993). The target RNA was purposely kept short (12 nt) with the goal of minimizing the possibility of unpredicted guide:target interactions (Fig 2A). As observed previously (Schirle et al, 2014), wild‐type Ago2 bound the target with a dissociation constant (K _d) of 0.28 ± 0.03 nM. The MI‐AA and Δhelix‐7 Ago2 mutants bound the target with affinities close to that of wild type (K _d = 0.33 ± 0.02 nM and 0.46 ± 0.03 nM, respectively). No Ago2 sample bound an RNA unrelated to the miR‐122 guide to a measurable extent in our assay conditions (data not shown). We conclude that helix‐7 does not contribute substantially to the overall affinity of Ago2 for a seed‐matched target RNA.

Figure 2. Helix‐7 is not required for target recognition.

1. Plot of g2‐g8 target RNA bound (0.1 nM) versus Ago2–miR122 concentration for wild‐type (WT), helix‐7 double point mutant (MI‐AA), and Δhelix‐7 Ago2. Average values from at least three independent experiments ± standard deviation (SD) are plotted. Top panel shows target RNA paired to seed region of the guide RNA (miRNA‐122).
2. Dissociation of a ³²P‐labeled target RNA (0.1 nM) from the Ago2–miR122 complex (1 nM) was monitored in the presence of unlabeled target RNA (100 nM). Fraction of the target RNA bound to Ago2–miR122 is plotted as a function of time for WT and Δhelix‐7 Ago2. Average values from at least three independent experiments ± SD were fit to single‐exponential decays.

Helix‐7 accelerates target release

In the absence of target RNAs, Ago2 adopts a conformation in which helix‐7 inserts between the nucleobases of g6 and g7 (Fig 1C). This observation suggests that helix‐7 may facilitate release of target RNAs by displacing seed‐paired targets and stabilizing the unpaired conformation of the guide. We tested this hypothesis by comparing release rates of seed‐paired target RNAs from wild‐type Ago2 and helix‐7 mutants.

Ago2–miR122 complexes were incubated with ³²P‐labeled target RNAs complementary to the miR‐122 seed, allowed to reach equilibrium, and then mixed with a 1,000‐fold excess of unlabeled target RNA. The labeled RNA remaining bound to Ago2 was then separated from unbound RNA at various times using a filter‐binding apparatus. The fraction of target RNA bound to wild‐type Ago2 was plotted as a function of time and fit to a one‐phase exponential decay, yielding observed release rates (k _off) of 0.52 ± 0.03 min⁻¹ (Fig 2B). This value is similar to reported k _off rates for target RNAs seed‐paired to let‐7 (0.18 min⁻¹) and miR‐21 (0.22 min⁻¹) in wild‐type mouse Ago2 (Salomon et al, 2015). In contrast, the seed‐paired target was released from the Δhelix‐7 Ago2 at an observed rate of 0.025 ± 0.008 min⁻¹, which is more than 10 times slower than release from wild‐type Ago2. We conclude that the disruption of helix‐7 represses the release of target RNAs from Ago2.

Helix‐7 minimizes dwell time on off‐targets

Previous studies demonstrated that Ago2 rapidly binds off‐targets, with imperfect seed complementarity, but then quickly dissociates (Chandradoss et al, 2015; Salomon et al, 2015). We used an established single‐molecule assay for Ago2–target interactions (Chandradoss et al, 2015) to determine how helix‐7 contributes to the ability of Ago2 to dismiss off‐targets. Briefly, we immobilized an acceptor (Cy5) labeled target RNA on a passivized surface and flushed in Ago2 loaded with donor (Cy3) labeled miRNA (Fig 3A and B). The binding of Ago2–miRNA to its target leads to FRET between the donor and acceptor fluorophores, which we measured using total internal reflection microscopy (Fig 3C). This approach allowed observation of transient interactions between Ago2–miRNA and off‐target RNAs, only partially matching the miRNA seed, which were difficult to detect in bulk filter‐binding assays.

Figure 3. Helix‐7 minimizes dwell time on off‐targets.

1. Cartoon schematic of our single‐molecule FRET assay.
2. Guide:off‐target pairs examined listed with increasing values of N (the length of the matched region between miRNA and target RNA). N = 7 is the fully seed‐paired target.
3. Representative fluorescence time traces (with a resolution of 100 ms for wt, 300 ms for MI‐AA and Δhelix‐7) show docking and dissociation of wild‐type (top), MI‐AA (middle), and Δhelix‐7 (bottom) Ago2–miRNA at single spots in the microfluidic chamber.
4. Dwell time histograms obtained for N = 4, 5, and 6, with wild‐type Ago2; histograms were fit with single‐exponential decay. The first column of the data was not included in the analysis to avoid potential artifacts arising from the limit of the time resolution. The bin size is 0.5 s for N = 4 and 1 s for N = 5 and 6.
5. Bar plot showing the dependence of dwell time on N for wild‐type Ago2.
6. Dwell time histogram for different N's with the MI‐AA helix‐7 mutant. The dwell time histogram for N = 6 fit with double‐exponential decay (green; R ² = 0.98); the dwell time for N = 6 with single exponential fit was 15.2 ± 3.41 s, R ² = 0.96.
7. Bar plot for dwell times of MI‐AA Ago2 as a function of N.
8. Dwell time histogram for different N's with the ∆helix‐7 mutant. The dwell time histograms for N = 5 and N = 6 were fit with double‐exponential decays (green; R ² = 0.97 for N = 5 and R ² = 0.96 for N = 6); the dwell time for N = 5 and N = 6 with single exponential fit was 9.1 ± 0.91 s (R ² = 0.86) and 22.3 ± 5.08 s (R ² = 0.90), respectively.
9. Bar plot for dwell times of ∆helix‐7 Ago2 as a function of N.

Data information: For all bar plots (E, G, and I), error bars are the SD of three independent experiments carried out on different days. The dashed red lines indicate the observation time limit (300 s), which is constrained by photobleaching.

As seen previously, wild‐type Ago2 displayed short dwell times (Δτ ≤ 2 s) on a variety of off‐targets with partial seed complementarity (Fig 3D and E). Even an off‐target matching g2–g7 nucleotides of the seed (N = 6) was released > 150 times faster than the fully seed‐matched target (N = 7, Δτ > 300 s; Fig 3E). [Note that dwell times longer than 300 s could not be accurately measured due to photobleaching effects (Chandradoss et al, 2015)]. The helix‐7 mutants also quickly dismissed an off‐target with complementarity to guide nucleotides g2–g4 (N = 3), but remained bound substantially longer to off‐targets with increased complementarity toward the 3′ end of the seed (Fig 3F–I). In the most extreme case, the N = 6 off‐target, MI‐AA and Δhelix‐7 Ago2 remained bound for an average of 25.1 and 49.9 s, respectively (Fig 3G and I), which is 13‐fold and 25‐fold longer than wild‐type Ago2. Measuring dwell times for variants of the N = 6 off‐target, with single mismatches at different positions, indicated that helix‐7 is most important for recognizing mismatches to the 3′ end of the seed (Fig EV2A–C). We also noted that dwell time histograms for off‐targets with mismatches restricted to the seed 3′ end (N = 5 and N = 6) deviate substantially from single exponentials in the Δhelix‐7 data, indicating heterogeneity in the population of binding events (Fig 3H). The combined results suggest that helix‐7 modulates guide:target interactions in the 3′ end of the seed, potentially involving a conformational change, and thereby allows Ago2 to efficiently distinguish miRNA targets from off‐targets.

Figure EV2. Dwell times of a spectrum of off‐target sequences.

1. Sequences of guide RNA (green) paired to mismatched target RNAs (red).
2. Dwell time histograms for wild‐type and ∆helix‐7 Ago2 on the off‐target RNAs.
3. Bar plot of average dwell times for indicated off‐target RNAs.

Helix‐7 prevents G:U wobble pairing at positions g6 and g7

Single‐molecule dwell time measurements (Figs 3 and EV2B) for off‐targets suggested that helix‐7 may help control pairing to the 3′ end of the seed such that non‐canonical guide:target interactions are minimized. Supporting this idea, we previously noted that helix‐7 directly contacts positions 5–7 of the guide:target duplex minor groove (Fig 1D), and speculated that helix‐7 may thereby help Ago2 distinguish between perfectly and imperfectly seed‐matched RNAs (Schirle et al, 2014). To test this hypothesis, we examined how G:U wobble pairs in the second half of the seed region affect target affinity in wild‐type and Δhelix‐7 Ago2.

Introducing a G:U wobble at position g8 did not have a significant impact on target affinity for wild‐type or Δhelix‐7 Ago2 (Fig 4A). Tolerance for a wobble pair at g8 is consistent with the observation that Ago2 does not directly contact the g8:t8 base pair (Schirle et al, 2014). In contrast, including an additional G:U wobble at g7 decreased target affinity of the wild‐type Ago2 > 80‐fold (K _d ~25 nM). The effects of the g7 wobble on Δhelix‐7 Ago2 were less pronounced, decreasing affinity by about 12‐fold (K _d = 6.9 ± 1.2 nM; Fig 4B). Furthermore, adding another G:U wobble, at position g6, led to a pronounced decrease in affinity for the wild‐type Ago2, which bound the g6‐g8 wobble off‐target with a > 300 times lower affinity than the seed‐paired target (K _d > 100 nM; Fig 4C). Δhelix‐7 Ago2, on the other hand, showed essentially no change in affinity (K _d = 5.7 ± 0.7 nM for the g6–g8 wobble target). Both wild‐type and Δhelix‐7 Ago2 bound an off‐target with g6‐g8 mismatches poorly (Fig 4D and E), though we note that the Δhelix‐7 mutant appeared to have a higher affinity (K _d ~30 nM) than wild type (K _d > 100 nM). We conclude that helix‐7 contributes to fidelity of target binding by preventing Ago2 from forming non‐Watson‐Crick base pairs at positions g6 and g7.

Figure 4. Helix‐7 discourages non‐Watson‐Crick pairing at g6 and g7.

• A–C
  Plots of target RNA (0.1 nM) bound to Ago2 versus Ago2–miR122 (wild type and Δhelix‐7) concentration for targets with an increasing number of G:U wobble pairs in the seed 3′ end.
• D
  Plot of target RNA (0.1 nM) bound to Ago2 versus Ago2–miR122 concentration for a target with mismatches at positions g6–g8.
• E
  Bar plots of the K _d of wild‐type (left) and Δhelix‐7 (right) Ago2–miR122 for indicated target RNAs. The dashed red lines indicate the upper limit (100 nM) for which a dissociation constant can be determined under our experimental regime.

Data information: Average values from at least three independent experiments ± SD are plotted.

Helix‐7 dampens inhibition by non‐target RNAs

Slow release of imperfectly paired targets by the ∆helix‐7 mutants suggested that a major function of helix‐7 might be to expedite miRNA target searches by helping Ago2 avoid off‐targets. To explore this idea experimentally, we measured Ago2 binding to a seed‐paired target RNA in the presence of increasing concentrations of non‐target RNAs. We first measured the effects of a synthetic RNA oligonucleotide with two partial seed matches to miR‐122 (Fig 5A). The off‐target competitor inhibited target binding by both wild‐type and ∆helix‐7 Ago2. However, the half‐maximal inhibitory concentration (IC₅₀) for wild‐type Ago2 was 10 times greater than ∆helix‐7 Ago2 (135 nM versus 13 nM). Similar results were observed when we repeated the experiment using total RNA isolated from Sf9 cells as the competitor (IC₅₀ = ~115 and 12 μg/ml for wild‐type and ∆helix‐7 Ago2, respectively; Fig 5B). We conclude that the removal of helix‐7 leads to a significant increase in susceptibility to inhibition by non‐target RNAs.

Figure 5. ΔHelix‐7 Ago2 is sensitive to inhibition by off‐target RNAs.

1. Ago2–miRNA122 (1 nM) was incubated with a ³²P‐labeled, seed‐matched target RNA (0.1 nM) in the presence of increasing concentrations of unlabeled a competitor RNA (top panel). Fraction target RNA bound to Ago2 is plotted as a function of off‐target competitor RNA concentration.
2. Fraction of target RNA bound to Ago2 in the presence of increasing concentrations of unlabeled total cellular RNA.

Data information. Average values from at least three independent experiments ± SD are plotted.

Helix‐7 accelerates target binding

The observation that Δhelix‐7 Ago2 releases seed‐paired targets slower than wild‐type Ago2 was also intriguing because equilibrium binding experiments demonstrated that mutation or removal of helix‐7 does not have a significant impact on target affinity (Fig 2B). The observed K _d is the ratio of the target release rate, k _off, and target binding rate, k _on. We therefore predicted that, in addition to reducing k _off, removal of helix‐7 must also reduce k _on. We tested this prediction by comparing target binding rates of wild‐type and Δhelix‐7 Ago2.

We first measured Ago2–miR122 target binding in bulk solution. Various concentrations of the Ago2–miR122 complex were mixed with ³²P‐labeled target RNAs matching the seed region of miR‐122. RNA bound to Ago2 was separated from unbound RNA at various times using a filter‐binding apparatus. Even at the lowest protein concentration used, the wild‐type Ago2 bound the target RNA faster than we could measure using this technique, with an observed binding rate (k _on,obs) ≥ 4.9 ± 1.0 min⁻¹, which corresponds to a k _on value of ≥ 4.3 × 10⁹ M⁻¹ min⁻¹ (Fig 6A and B). In contrast, gradual target binding by the MI‐AA and Δhelix‐7 Ago2 mutants was readily observed (Fig 6A). Measuring the binding rate at various Ago2–miR122 concentrations, we calculate k _on values of 1.2 ± 0.06 × 10⁹ M⁻¹ min⁻¹ and 3.4 ± 0.02 × 10⁸ M⁻¹ min⁻¹ for MI‐AA and Δhelix‐7, respectively, using this experimental setup (Fig 6B).

Figure 6. Helix‐7 mutants display reduced target binding rates.

1. Ago2–miR122 complexes were mixed with a ³²P‐labeled seed‐matched target RNA; bound and free RNAs were then separated using a filter‐binding apparatus at various times. Representative time course for target RNA (0.1 nM) binding to wild‐type, MI‐AA, and Δhelix‐7 Ago2 (1 nM). Average values from at least three independent experiments ± SD are plotted.
2. Observed binding rates (k _on,obs) plotted as a function of Ago2–miR122 concentration. Average values ± SD are plotted.
3. Arrival time distribution of single Ago2–miRNA complexes binding immobilized targets in a microfluidic chamber. The value is the average of three independent measurements.

We next used a single‐molecule assay to directly observe target binding. We introduced purified Cy3‐labeled Ago2–miRNA complexes into a microfluidic chamber and measured the time of arrival on immobilized target RNAs. The observed binding rate k _on,obs was determined by fitting the arrival time distribution to a single‐exponential growth curve, $A(1-e^{-k_{\mathrm{on}}t})$ (Fig 6C). Using this assay, the observed binding rate of wild‐type Ago2 was 3.3 ± 0.35 × 10⁸ M⁻¹ sec⁻¹ (2.0 ± 0.021 × 10¹⁰ M⁻¹ min⁻¹), while the k _on,obs for ∆helix‐7 was fivefold lower, 0.67 ± 0.05 × 10⁸ M⁻¹ sec⁻¹ (4.0 ± 0.03 × 10⁹ M⁻¹ min⁻¹). It is noted that these rates are about an order of magnitude faster than k _on values determined by bulk assays. We attribute the discrepancy to differences in target sequence and structure and/or physical differences in the two experimental regimes. Supporting this idea, the k _on,obs for wild‐type Ago2 determined by single‐molecule measurements closely matches the single‐molecule k _on value for mouse Ago2 (3.9 × 10⁸ M⁻¹ sec⁻¹) determined previously (Salomon et al, 2015). Additionally, shuttling between adjacent target sites, which requires rapid rebinding of Ago2 after release from a target site, is substantially reduced in the ∆helix‐7 mutant (Fig EV3A–D). Thus, for all assays employed, we find the binding rate for helix‐7 mutants is 5–10 times slower than wild type. We conclude that the removal of helix‐7 reduces the rate at which Ago2 can establish stable interactions with seed‐paired target RNAs.

Figure EV3. Helix‐7 facilitates lateral diffusion.

1. Schematic of target RNA containing two seed‐matched sites separated by a 15 nt linker. Guide RNA (green) is shown paired to site 1.
2. Representative fluorescence time traces and FRET efficiencies for wild‐type (top) and Δhelix‐7 (bottom) Ago2. FRET changes arise from a single Ago–miRNA complex shuttling between the two adjacent binding sites (Chandradoss et al, 2015).
3. Bar graph comparing wild‐type and Δhelix‐7 Ago2 shuttling rates. k _{shuttling,(obs)} = 0.28 and 0.006/s for wild‐type and ∆helix‐7 Ago2, respectively. For wild‐type Ago2, 86–90% of each stable binding event included at least one shuttle between site 1 and site 2. For ∆helix‐7 Ago2, only 13–16℅ of binding events displayed any shuttling. The error bars are the SD of three independent experiments.
4. Dwell time histogram for high FRET and low FRET states for wild‐type (top) and ∆helix‐7 (bottom) Ago2.

Helix‐7 pre‐organizes the 3′ end of the seed for target pairing

To shed light onto how Ago2 achieves high on rates for target binding, we sought to visualize how helix‐7 mutants display the seed for target pairing. Crystals of ∆helix‐7 Ago2 were difficult to grow. However, the MI‐AA Ago2 mutant crystalized readily in conditions similar to those used previously with wild‐type Ago2 (Schirle & MacRae, 2012; Appendix Table S1). The overall conformation of MI‐AA Ago2 closely matches wild type (RMSD of 0.521 Å for 684 equivalent Cα atoms), and thus, major differences between mutant and wild‐type structures are restricted to the area surrounding helix‐7 (Fig 7A). Notably, helix‐7 in the mutant structure is tilted 18° away from the guide RNA, relative to the wild‐type conformation (Fig 7B). This shift is not accompanied by a corresponding change in the position of the PAZ domain, which moves as a single rigid body with helix‐7 in wild‐type Ago2 (Schirle et al, 2014). We also note that the crystallographic temperature factors in the mutant helix‐7 are substantially higher than in wild type, such that more than half of the mutant helix‐7 could not be modeled with confidence. Additionally, residues adjacent to helix‐7 in the L1 stalk (S189–G194) and loop connecting helix‐7 to the PAZ domain (K355–T361) are disordered. These observations indicate that the MI‐AA mutations increase the mobility of helix‐7, likely by uncoupling its motions from the PAZ domain and L2 stalk.

Figure 7. Structure of a helix‐7 Ago2 mutant.

1. Superposition of wild‐type (white) and MI‐AA (gray) Ago2. Helix‐7 colored yellow; ordered region of the guide RNAs colored in red.
2. Close‐up view shows helix‐7 tilts away from the seed region in the MI‐AA mutant.
3. Side‐by‐side comparison of the seed region in wild‐type (left) and MI‐AA (right) Ago2 crystal structures.

The shift in helix‐7 is accompanied by a rearrangement of guide nucleotides in the 3′ end of the seed (Fig 7C). The largest observable change is nucleotide g7, which is switched from the anti conformation observed in wild‐type Ago2 to the syn conformation in the mutant structure. Notably, it is not possible for g7 to engage a complementary target nucleotide via Watson‐Crick pairing in this conformation. Additionally, nucleotide g8 is completely disordered in the mutant, and there is a minor shift in the position of the g6 nucleobase such that stacking with g5 is slightly reduced. In contrast, seed nucleotides g2–g5 in the mutant structure are nearly identical to wild‐type Ago2. We conclude that disruption of helix‐7 specifically leads to a structural disorganization of the 3′ half of the unpaired seed.

Discussion

The finding that removal of helix‐7 reduces both target binding and release rates by Ago2, without changing overall target affinity, leads us to propose that helix‐7 acts like a catalyst for seed pairing. We suggest that helix‐7 accelerates the target binding step by enabling lateral diffusion, orienting guide nucleotides g6–g8 for pairing to incoming target RNAs, and directing pairing through association with the minor groove of the guide:target duplex (Fig 1). Helix‐7 also places hydrophobic groups adjacent to the Watson‐Crick edges of g6 and g7, and thus may facilitate target pairing at these positions via desolvation (Rozners & Moulder, 2004). These positive effects on target binding appear to be balanced by the ability of helix‐7 to break guide:target interactions by intercalating Ile365 between g6 and g7, thereby creating a stable seed conformation that is unable to pair to target RNAs beyond g5. Thus, the opposing activities of helix‐7 reduce the energy barrier involved in making and breaking guide:target base pairs in the 3′ end of the seed.

Understanding the functions of helix‐7 allows us to propose a refined model for the structure of the miRNA seed region and mechanism of seed pairing. We suggest that the seed can be viewed as two functional domains. The 5′ domain (g2–g5) is held by Ago2 in a near A‐form conformation both before and after pairing to target RNAs (Schirle et al, 2014). Thus, the 5′ domain behaves similarly to a locked nucleic acid (Braasch & Corey, 2001), making it well suited for establishing initial base pairing contacts to potential target RNAs (Chandradoss et al, 2015; Salomon et al, 2015; Klein et al, 2017). The 3′ end of the seed (g6–g8) is flexible and moves between kinked and A‐form conformations. Movements of the 3′ domain are governed by helix‐7, which increases the rates of target pairing and unpairing, as discussed above. The tendency of helix‐7 to break target pairing is likely mitigated by helix‐7 stably docking into the minor groove of the guide:target duplex (Fig 1D). Thus, Watson‐Crick base pairing, which provides an A‐form duplex for docking, is favored over all types of non‐canonical interactions. The existence of distinct functional domains within the seed was first predicted from single‐molecule measurements that showed mismatches to g2–g5 are more detrimental to target binding rates than mismatches to g6–g8 (Chandradoss et al, 2015; Salomon et al, 2015).

Sequence homology indicates that helix‐7 is a structural element common to diverse eukaryotic Argonaute proteins (Schirle & MacRae, 2012). Additionally, the recent crystal structure of Piwi (a distinct clade of the Argonaute family) from the silkworm demonstrated that Piwi proteins also contain an α‐helix analogous to helix‐7 (Matsumoto et al, 2016). As seen in Ago2, the helix in silkworm Piwi (Siwi) directly contacts the guide RNA seed region. However, there are notable differences between Siwi and Ago2. Compared to Ago2, the position of the seed region is shifted in Siwi, such that the helix disrupts stacking between g5 and g6, as opposed to disrupting stacking between g6 and g7 observed in Ago2 (Matsumoto et al, 2016). We therefore suggest that Ago and Piwi proteins likely use similar target searching mechanisms, but predict that the requirements for establishing stable guide:target pairing may be different between the two protein families.

Ago2 finds target mRNAs with remarkable speed and fidelity (Ameres et al, 2007; Wee et al, 2012; Chandradoss et al, 2015; Salomon et al, 2015). Target searches are conducted in a crowded cellular environment wherein the vast majority of RNAs present are not targets (Garcia et al, 2011). Because Ago2 can only discriminate between targets and off‐targets through direct binding, we suggest that a critical aspect of Ago2 function is minimizing time spent interrogating non‐target RNAs. Our results suggest that swift association and disassociation with potential target sites, catalyzed by helix‐7, allows Ago2 to nimbly and deftly identify its targets, such that miRNA silencing may occur on a biologically meaningful timescale.

Materials and Methods

miR‐122 oligonucleotides

Guide RNA:

5′‐Phosphate‐rUrGrGrArGrUrGrUrGrArCrArArUrGrGrUrGrUrUrUrG‐3′

Capture Oligo:

5′‐Biotin‐mUmCmUmCmUmGmCmUmAmAmCmCmAmUmGmCmGmAmA mCmAmCmUmCmCmAmUmCmUmCmUmGmC‐3′

Competitor DNA:

5′‐Biotin‐GCAGAGATCAAGTGTTCGCATGGTTAGCAGAGA‐3′

Northern Blot Probe:

5′‐rArArArCrArCrCrArUrUrGrUrCrArCrArCrUrCrCrArArA‐3′

Binding Assay Target RNAs:

2‐5: 5′‐rArArArUrGrUrCrUrCrCrUrA‐3′

2‐5 G‐U: 5′‐rArArArGrUrGrCrUrCrCrUrA‐3′

2‐6: 5′‐rArArArUrGrArCrUrCrCrUrA‐3′

2‐6G‐U: 5′‐rArArArGrUrArCrUrCrCrUrA‐3′

2‐7: 5′‐rArArArUrCrArCrUrCrCrUrA‐3′

2‐7G‐U: 5′‐rArArArGrCrArCrUrCrCrUrA‐3′

2‐8: 5′‐rArArArArCrArCrUrCrCrUrA‐3′

4‐8: 5′‐rArArArArCrArCrUrGrGrUrA‐3′

4‐8G‐U: 5′‐rArArArArCrArCrUrUrUrUrA‐3′

Binding Assay Seed‐Matched Competitor RNA:

5′‐rArArArCrUrCrCrArArArArCrUrCrCrArArA‐3′

Protein expression

Full‐length human Ago2 was cloned into the pFastBac HT A plasmid for expression using the Bac‐to‐Bac (Invitrogen) baculovirus expression system and overexpressed in Sf9 insect cells, as described previously (De & Macrae, 2011). Both the Ago2 M364A‐I365A mutant and the Ago2 ∆helix‐7 mutant were created using the QuikChange (Agilent) site‐directed mutagenesis kit. In the ∆helix‐7 mutant, α‐helix‐7 (residues 358–368) was replaced with a penta‐glycine linker. Serine residues subject to phosphorylation during protein expression were changed to either alanine or aspartate in all Ago2 constructs, as described previously (Schirle & MacRae, 2012). Sf9 cells were infected with virus for 60–72 h at 27°C and harvested by centrifugation.

Ago2–miR122 preparation

Sf9 cell pellets were resuspended in a lysis buffer (50 mM monosodium phosphate, pH 8.0, 300 mM sodium chloride, 0.5% Triton X‐100, 0.5 mM tris[2‐carboxyethyl] phosphine hydrochloride (TCEP)) and lysed in a Dounce homogenizer (Fisher Scientific) using 15 strokes of pestle B. Lysate was centrifuged to pellet cellular debris, and the clarified lysate was applied to 1.5 ml Ni‐NTA resin (Qiagen) and incubated on a nutator at 4°C for 1.5 h. Contaminating proteins were removed by washing resin with one hundred column volumes of wash buffer (50 mM Tris, pH 8.0, 300 mM sodium chloride, 7.5 mM imidazole, 0.5 mM TCEP), then with thirty column volumes of calcium‐supplemented wash buffer (50 mM Tris, pH 8.0, 300 mM sodium chloride, 7.5 mM imidazole, 5 mM calcium chloride, 0.5 mM TCEP). Contaminating RNAs were removed by treatment with micrococcal nuclease for 1 h at room temperature, followed by washing resin with thirty column volumes of wash buffer. Bound proteins were eluted from resin by addition of three column volumes of elution buffer (50 mM Tris, pH 8.0, 300 mM sodium chloride, 300 mM imidazole, 0.5 mM TCEP). Ethylene glycol tetraacetic acid (EGTA), pH 8.0, was added to the eluate to a final concentration of 5 mM to chelate any remaining calcium ions. The synthetic guide RNA was then added to the nickel‐purified Ago2 and the resulting Ago2–RNA mixture was dialyzed overnight at 4°C in wash buffer to remove excess imidazole.

Ago2 molecules loaded with miR‐122 were then isolated using the Arpón method, as described previously (Flores‐Jasso et al, 2013). Excess biotinylated competitor DNA was removed by incubation with neutravidin resin for 30 min at room temperature. Ago2–miR122 complexes were then dialyzed overnight at 4°C into 2‐(cyclohexylamino)ethanesulfonic acid (CHES) buffer (10 mM CHES, pH 9.0, 100 mM NaCl, 0.5 mM TCEP).

Purified Ago2–guide complexes were brought up to 20% (v/v) with glycerol, aliquoted, frozen in an ethanol dry‐ice bath, and stored at −80°C. Before using in biochemistry assays, aliquots were thawed on ice.

Quantification of purified guide‐loaded Ago2

miR‐122 levels in Ago2–miR122 preparations was determined by quantitative Northern blotting. Ago2–miR122 samples were boiled in 1× formamide loading buffer (47.5% formamide, 0.0125% (w/v) bromophenol blue, 0.0125% (w/v) xylene cyanol, 0.0125% sodium dodecyl sulfate [SDS], 0.25 mM ethylenediaminetetraacetic acid [EDTA]), and resolved alongside known amounts of miR‐122 by 14% denaturing polyacrylamide gel electrophoresis. RNA was transferred to Hybond‐NX membrane (Amersham, GE Healthcare Life Sciences), cross‐linked to the membrane with 125 mM 1‐methylimidazole, pH 8.0, and 150 mM 1‐ethyl‐3‐(3‐dimethylaminopropyl)carbodiimide hydrochloride (EDC; Pall et al, 2007) and probed with a complementary ³²P‐5′‐radiolabeled probe in hybridization buffer (750 mM sodium chloride, 75 mM sodium citrate dihydrate, pH 7.0, 20 mM monosodium phosphate, pH 7.2, 7% SDS, 2× Denhardt's solution, 16 μg/ml salmon sperm DNA). After washing with low‐stringency wash buffer (450 mM sodium chloride, 45 mM sodium citrate dihydrate, pH 7.0, 5% SDS) and high‐stringency wash buffer (150 mM NaCl, 15 mM sodium citrate dihydrate, pH 7.0, 1% SDS), guide and standard signals were visualized using a Typhoon FLA 9500 phosphorimager (GE Healthcare Life Sciences) and bands quantified using ImageQuant TL 8.1 software (GE Healthcare Life Sciences).

Protein concentrations in purified Ago2–miR122 samples were determined by Bradford assay (Bio‐Rad) using bovine serum albumin as standards.

CD spectroscopy

Ago2 concentration was determined by measuring the UV absorbance. Spectra were recorded at 25°C at a total protein concentration of 8.5 μM in 10 mM potassium phosphate buffer, pH 8, and 100 mM KCl using a 0.2 cm cell in an AVIV model 202 CD spectrometer. The molar ellipticity for the protein solutions was calculated using the total concentration of amino acid residues present.

Equilibrium binding assays

Equilibrium target binding assays were carried out in a manner similar to described previously (Schirle et al, 2015). Wild‐type and mutant Ago2–miR122 samples of various concentrations (0–40 nM) were incubated with ~0.1 nM 5′ ³²P‐radiolabeled target RNA in binding reaction buffer (30 mM Tris pH 8.0, 100 mM potassium acetate, 2 mM magnesium acetate, 0.5 mM TCEP, 0.005% (v/v) NP‐40, 0.01 mg/ml baker's yeast tRNA), in a reaction volume of 100 μl, for 45 min at room temperature. Using a dot‐blot apparatus (GE Healthcare Life Sciences), protein–RNA complexes were captured on Protran nitrocellulose membrane (0.45‐μm pore size, Whatman, GE Healthcare Life Sciences) and unbound RNA on Hybond Nylon membrane (Amersham, GE Healthcare Life Sciences). Samples were applied by vacuum and then washed with 100 μl of chilled wash buffer (30 mM Tris pH 8.0, 0.1 M potassium acetate, 2 mM magnesium acetate, 0.5 mM TCEP). Membranes were air‐dried, and signals from radioactive decay were visualized by phosphorimaging. Quantification was performed using ImageQuant (GE Healthcare Life Sciences). Dissociation constants were calculated using Prism version 6.0 g (GraphPad Software, Inc.), with the one‐site specific binding formula:

$$F=\frac{B_{\max }[E]}{[E]+K_{\mathrm{d}}}$$

where F = fraction bound, B _max = maximum bound fraction of ligand, [E] = enzyme concentration, and K _d = apparent equilibrium dissociation constant.

Target dissociation assays (bulk)

Ago2–miR122 samples (1 nM) were equilibrated with 0.1 nM 5′ ³²P‐radiolabeled target RNA in binding reaction buffer for 45 min. After equilibration, a “zero” time point 100 μl aliquot was applied to a dot‐blot apparatus under vacuum, then chased with 100 μl of ice‐cold wash buffer. After addition of excess unlabeled target RNA (final concentration of 100 nM), 100 μl aliquots taken at time points from 0.5 to 45 min and applied to a dot‐blot apparatus under vacuum, then chased with 100 μl of ice‐cold wash buffer. Membranes were air‐dried, and signals were visualized by phosphorimaging. Quantification was performed using ImageQuant, and dissociation rates were calculated by fitting data to a one‐phase exponential decay using Prism.

Fluorescent RNA preparation

The variants of the hsa‐let‐7a guide and target RNAs with amine modification (amino‐modifier C6‐U phosphoramidite, 10‐3039, Glen Research) were purchased from STPharm (South Korea). The guide and target strands were labeled with donor (Cy3) and acceptor (Cy5), respectively, using the NHS‐ester form of Cy dyes (GE Healthcare). The labeling efficiency was ~100%.

A target strand was ligated to a biotinylated polyuridine RNA (U₃₀) as follows. A target RNA strand (200 pmole) was mixed with a polyuridine RNA strand (U₃₀, 200 pmole) that had 5′ phosphate and 3′ biotin. This mixture was annealed with a DNA splint (600 pmole) in TE buffer with 100 mM NaCl by rapid heating to 80°C and slow cooling down (−1°C/4 min in a thermal cycler). The annealed constructs were ligated using 3 μl T4 RNA ligase2 (Ambion, 5 U/μl), 3 μl 0.1% BSA (Ambion), 5 μl 10× ligation buffer (Ambion), and 19 μl H₂O at 37°C overnight. After ethanol precipitation, the ligated RNA strands were purified on a denaturing (8 M urea) 12.5% polyacrylamide gel.

Single‐molecule sample preparation

Most single‐molecule preparation was similar to as reported previously (Chandradoss et al, 2015; Schirle et al, 2015). A microfluidic chamber was incubated with 20 μl Streptavidin (0.1 mg/ml, Sigma) for 30 s. Unbound Streptavidin was washed with 100 μl of buffer T50 (10 mM Tris–HCl [pH 8.0], 50 mM NaCl buffer). The fifty microliters of 50 pM acceptor‐labeled mRNA construct were introduced into the chamber and incubated for 1 min. Unbound labeled constructs were washed with 100 μl of buffer T50. The effector complex was formed by incubating 50 nM purified recombinant hAgo2 with 1 nM of donor‐labeled hsa‐let‐7a miRNA in a buffer containing 50 mM Tris–HCl [pH 8.0] (Ambion) and 110 mM NaCl (Ambion) at 37°C for 20 min. For binding rate (k _on) measurements, an immobilized capture oligo (Flores‐Jasso et al, 2013) was used to purify the effector complex away from Ago2 molecules loaded with co‐purifying cellular RNAs and guide RNAs not loaded into Ago2 prior to single‐molecule experiments. An imaging buffer for single‐molecule FRET was added before the mixture was injected to a microfluidic chamber. The final concentration of the imaging buffer consists of the 0.8% dextrose (Sigma), 0.5 mg/ml glucose oxidase (Sigma), 85 μg/ml Catalase (Merck), and 1 mM Trolox ((±)‐6‐hydroxy‐2,5,7,8‐tetramethylchromane‐2‐carboxylic acid, 238813, Sigma). The experiments were performed at room temperature (23 ± 1°C).

Single‐molecule FRET

Ago2 single‐molecule dwell time measurements were performed as described previously (Chandradoss et al, 2015; Schirle et al, 2015). Briefly, target RNAs bearing a Cy5 dye and a 3′ biotin were immobilized on a polymer(PEG)‐coated quartz surface in the microfluidic chamber (Chandradoss et al, 2014) of a prism‐type total internal reflection fluorescence microscope. Ago2 was loaded with a guide miRNA containing a Cy3 dye. The resulting complex was introduced into the microfluidic chamber and Cy3 molecules were excited with a 532‐nm diode laser (Compass 215M/50mW, Coherent). Fluorescence signals of Cy3 and Cy5 were collected through a 60X water immersion objective (UplanSApo, Olympus) with an inverted microscope (IX73, Olympus). Laser scattering was blocked by a 532‐nm long‐pass filter (LPD01‐532RU‐25, Semrock). The Cy3 and Cy5 signals were separated with a dichroic mirror (635 dcxr, Chroma) and imaged using a EM‐CCD camera (iXon Ultra, DU‐897U‐CS0‐#BV, Andor Technology).

Single‐molecule data acquisition and analysis

Single‐molecule data were acquired and analyzed as previously (Chandradoss et al, 2015; Schirle et al, 2015). CCD images of time resolution 0.1 or 0.3 s were recorded, and time traces were extracted from the CCD image series using IDL (ITT Visual Information Solution). Colocalization between Cy3 and Cy5 signals was carried out with a custom‐made mapping algorithm written in IDL. The extracted time traces were processed using Matlab (MathWorks) and Origin (Origin Lab).

Binding rate (k _on) determination was carried out in a manner similar to described previously (Schirle et al, 2015). Binding rates were determined by first measuring the time between introduction of Ago2–miRNA into a microfluidic chamber and the first Ago2–miRNA docked to a target RNA; and then fitting the time distribution with a single‐exponential growth curve, $A(1-e^{-k_{\mathrm{on}}t})$. The analysis of binding rates requires precise determination of effector complex concentrations. The concentration of guide pre‐loaded WT and Δhelix‐7 Ago2 was estimated by non‐specifically adsorbing a sub‐nanomolar concentration of the samples on a positively charged poly‐L‐lysine surface as follows. After 5 min of incubation with 20 μl 0.01% poly‐L‐lysine (P4707, Sigma), the chamber was washed with 100 μl of buffer T50 (10 mM Tris (pH 8.0), 50 mM NaCl). After washing, WT or Δhelix‐7 Ago2 pre‐loaded with Cy3‐labeled guide was introduced into the microfluidic chamber. After 5 min of incubation, unbound substrate was washed away with 100 μl of imaging buffer and images were obtained from at least 10 fields of view. The same procedure was repeated for Cy3‐labeled Ago‐free guide, which stock concentration is accurately known. By comparing the number of fluorescence spots from WT and Δhelix‐7 Ago2 with that from free guide, the concentrations of WT and Δhelix‐7 Ago2 were determined.

Dissociation rates were estimated by measuring the dwell time of binding events. Each dwell time distribution was fit by either a single‐exponential decay curve (Ae ^−t/Δτ) or a double‐exponential decay curve ($A_1{e}^{-t/\mathrm{\Delta }{\mathit{\tau }}_1}+A_2{e}^{-t/\mathrm{\Delta }{\mathit{\tau }}_2}$). In case of a double‐exponential decay, the percentages of Δτ ₁ and Δτ ₂ populations are determined by A ₁Δτ ₁/(A ₁Δτ ₁ + A ₂Δτ ₂) and A ₂ τ ₂/(A ₁Δτ ₁ + A ₂Δτ ₂), and the average dwell time is determined by ($A_1\mathrm{\Delta }{\mathit{\tau }}_1^2+A_2\mathrm{\Delta }{\mathit{\tau }}_2^2$)/(A ₁Δτ ₁ + A ₂Δτ ₂).

The k _{shuttling(obs)} was estimated by dividing a total number of alternating events (high FRET → low FRET; low FRET → high FRET) by total bound time (Δτ_total). If FRET alternation ended with dissociation, all transitions except the last one were counted. For example, if a molecule alternated between “high FRET” → “low FRET” → “high FRET” → “low FRET” before dissociation, the number of alteration were counted as three in this case. Δτ_total is estimated by summation of the bound time of individual molecule (Δτ_total = ΣΔτ_i, where i is the molecule). The alternation between two FRET states was analyzed using a Matlab algorithm that distinguishes two FRET states using a threshold above shot noise.

Target association assays (bulk)

Ago2–miR122 samples (1 nM) were mixed with 0.1 nM 5′ ³²P‐radiolabeled target RNA in binding reaction buffer and 100 μl aliquots taken at 0–45 min. Samples were immediately applied to a dot‐blot apparatus under vacuum and then chased with 100 μl of ice‐cold wash buffer. Membranes were air‐dried and signals visualized by phosphorimaging. Quantification was performed using ImageQuant, and association rates calculated using Prism, with the following formula:

$$F=\frac{B_{\max }[E]\left(1-e^{-(k_{\mathrm{on}}[E]+k_{\mathrm{off}})t}\right)}{[E]+K_{\mathrm{d}}}$$

where F = fraction of target bound, B _max  = calculated maximum number of binding sites, [E] = enzyme concentration, k _on  = association rate, k _off = dissociation rate (as calculated from the one‐phase decay equation), and K _d = apparent equilibrium dissociation constant.

Competition assays (bulk)

Ago2–miR122 samples (1 nM) were mixed with 0.1 nM 5′ ³²P‐radiolabeled target RNA and either 0–50 μg/ml Sf9 total cellular RNA or 0–300 nM seed‐matched competitor RNA in binding reaction buffer for 45 min at room temperature. Samples were applied to a dot‐blot apparatus under vacuum and then chased with 100 μl of ice‐cold wash buffer. Membranes were air‐dried, and signals were visualized by phosphorimaging. Quantification was performed using ImageQuant, and K _i values were calculated using the one‐site K _i fit in Prism.

Crystallization and data collection

Ago2 M364A‐I365A loaded with cellular small RNAs was purified from clarified lysate using standard nickel‐affinity purification, as described previously (Schirle et al, 2014). Crystals were grown at 20°C using hanging drop vapor diffusion by iterative rounds of seeding. Drops contained a 0.8:0.8:0.2 ratio of protein (3 mg/ml), reservoir solution (12% PEG 3350, 0.1 M phenol, 12% isopropanol, and 0.1 M Tris, pH 9.0), and seeds from fragments of small Ago2 crystals grown in the same condition. Crystals were harvested with nylon loops and soaked in reservoir solution containing cryoprotectant (25% ethylene glycol, 12% PEG 3350, and 0.1 M Tris, pH 9.0) before cryo‐cooling by immersing in liquid N₂. Diffraction data were collected under cryogenic conditions on Beamline 24‐ID‐E at the Advanced Photon Source (APS) and were processed using XDS and Scala (Kabsch, 2010; Winn et al, 2011).

Structure determination

The Ago2 M364A‐I365A helix‐7 mutant diffraction data were refined against the wild‐type Ago2 structure (PDB ID: 4OLA) after removal of the guide RNA and residues 358–368 using the PHENIX graphical user interface (Adams et al, 2010). The model was inspected and rebuilt using Coot (Emsley et al, 2010) and submitted to XYZ coordinate, TLS, and B‐factor refinement with secondary structure restraints and optimization of stereochemistry weighting using PHENIX. Model building and refinement continued iteratively until all interpretable electron density was modeled. Water molecules were identified automatically in Coot (2F_obs‐F_calc map, above 1.8 σ, and between 2.4 and 3.2 Å from hydrogen bond donors or acceptors) and by manual inspection of electron density maps. All structure figures were generated in PyMOL (Schrödinger, LLC).

Data deposition

Atomic coordinates for the MI‐AA Ago2 crystal structure have been deposited in the Protein Data Bank (accession code 5WEA).

Author contributions

SMK and NTS prepared Ago2 samples. SMK conducted bulk biochemistry experiments. SDC conducted single‐molecule experiments. NTS determined the structure of the helix‐7 mutant Ago2. CJ and IJM directed and oversaw research. All authors contributed to various parts of the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Review Process File

The EMBO Journal (2018) 37: 75–88